ENVIRONMENTAL HEALTH SERIES
Air Pollution
and
Water Supply
SYMPOSIUM
ENVIRONMENTAL MEASUREMENTS
Valid Data and
Logical Interpretation
U. S. DEPARTMENT OF HEALTH
EDUCATION, AND WELFARE
Public Health Service
-------
SYMPOSIUM
ENVIRONMENTAL MEASUREMENTS
Valid Data and
Logical Interpretation
Sponsored by
Division of Air Pollution
and
Division of Water Supply and Pollution Control
September 4-6, 1963
Co-chairmen
JOHN S. NADER
Laboratory of Engineering and Physical Sciences, DAP
and
E. C. TSIVOGLOU
Technical Services Branch, DWS&PC
ROBERT A. TAFT SANITARY ENGINEERING CENTER
Cincinnati, Ohio
U. S. DEPARTMENT OF HEALTH
EDUCATION, AND WELFARE
Public Health Service
July 1964
-------
The ENVIRONMENTAL HEALTH SERIES of reports was established to report
the results of scientific and engineering studies of man's environment: The community,
whether urban, suburban, or rural, where he lives, works, and plays; the air, water, and
earth he uses and re-uses; and the wastes he produces and must dispose of in a way
that preserves these natural resources. This SERIES of reports provides for professional
users a central source of information on the intramural research activities of Divisions
and Centers within the Public Health Service, and on their cooperative activities with
state and local agencies, research institutions and industrial organizations. The general
subject area of each report is indicated by the two letters that appear in the publication
number; the indicators are
AP Air Pollution
WP Water Supply and Pollution Control
AH Arctic Health
EE Environmental Engineering
FP Food Protection
OH Occupational Health
RH Radiological Health
Triplicate tear-out abstract cards are provided with reports in the SERIES to
facilitate information retrieval. Space is provided on the cards for the user's accession
number and key words.
Reports in the SERIES will be distributed to requesters, as supplies permit. Re-
quests should be directed to the Division identified on the title page or to the Publications
Office, Robert A. Taft Sanitary Engineering Center, Cincinnati, Ohio 45226.
Public Health Service Publication No. 999-AP-15
(or No. 999-WP-15)
-------
PREFACE
The rapid development of air and water quality management programs within the
Public Health Service and elsewhere has brought into sharper focus the many complex
problems involved in obtaining valid environmental data from which to draw the most
useful and valid conclusions. The availability of continuous measurement and recording
devices, as well as the electronic computer, has made it possible to attempt the solution
of increasingly complex environmental health problems that are associated with our ex-
panding modern society. Complexity for its own sake is not a useful goal, however, and
before we embrace the newer complex measurement and computational schemes we should
take stock by deciding what it is we really wish to accomplish. Only thus can we rationally
select the most suitable measurement system for a specific problem.
Although the problems associated with measurement systems are not unique to the
environmental health field, some of the current needs of the Division of Air Pollution and
the Division of Water Supply and Pollution Control of the Public Health Service led to
this Symposium on Environmental Measurements. The Symposium Committee, in consider-
ing how best to approach the total problem, found it most susceptible to analysis by iso-
lating each major operational step in the measurement system: sampling, detecting,recording,
validating, interpreting, and drawing conclusions. This classification of operational steps
provided the basic topics for General Sessions that would lead to better understanding of
the operations common to diverse applications in environmental fields. We hoped in
particular, by the very arrangement of the Symposium program, to emphasize that no
measurement system can be any better than the weakest of the operational steps.
Separate afternoon sessions were designed to explore the specific application of the
operational steps to investigations of air and water environments.
It was hoped that this program orientation would enhance our understanding of the
whole task of conducting a measurements program, and that it would thereby benefit
pollution control and technical administrators, as well as researchers and scientists in the
environmental field. It is anticipated that, following this orientation, there may well result
a series of svmposia, held from time to time, to discuss in more depth specific operational
aspects of measurement programs. The purpose of the Symposium, therefore, was to
provide comprehensive orientation; it was intended more to raise questions than to
provide solutions.
THE CHAIRMEN
-------
SYMPOSIUM COMMITTEE
Division of Air Pollution
Division of Water Supply
and Pollution Control
John S. Nader
Laboratory of Engineering
and Physical Sciences
CO-CHAIRMEN
Dr. E. C. Tsivoglou
Technical Services Branch
COMMITTEE MEMBERS
Robert A. McCormick
Laboratory of Engineering
and Physical Sciences
Charles E. Zimmer
Laboratory of Engineering
and Physical Sciences
Richard O'Connell
Technical Services Branch
Alfred W. Hoadley
Basic Data Branch
-------
CONTENTS
Page
Welcome: J. E, Flanagan 1
Session 1 General
C, S. Draper: Information Engineering New Frontier of Technology 5
G. W. Anderson: Objectives of Measurement Systems 11
Session 2 General
P. W. MacCready: The Design of Measurement Systems 21
A. Goetz: Parameters 29
W. J. Youden: Sampling and Statistical Design 35
L. Bollinger: Transducers 41
Session 3 General
P. K. Stein: Classification Systems for Transducers and Measuring Systems 65
G. C. Gill: Data Validation 85
R. S. Green: The Storage and Retrieval of Data for Water Quality Control
a Summary 101
Session 4 Measurements of Air Environment
J. S. Nader: Data Acquisition Systems in Air Quality 107
H. E. Cramer: Data Acquisition Systems in Meteorology 125
0. Balchum: Data Acquisition Systems in Physiology 140
Discussion: Data Acquisition Systems 159
Session 5 Measurements of Water Environment
S. S. Baxter: Data Acquisition Systems in Water Supply 163
P. DeFalco, Jr.: Data Acquisition Systems in Water Quality Control 173
W. Isherwood: Data Acquisition Systems in Hydrology 179
J. J. Gannon: The Interpretation and Analysis of Hydrological Data 187
Session 6 General
J. C. Bellamy: Data Display for Analysis 213
G. W. Brier: Techniques for Data Analysis 227
D. W. Pritchard: Interpretations and Conclusions 235
L. A. Chambers: Summation - 245
Session 7 Measurements of Air Environment
R. I. Larsen: Determining Basic Relationships between Variables 251
G. W. Brier: Interpretation of Trends and Cycles 265
L. D. Zeidberg: Data Interpretation (Air) Drawing Conclusions 273
Discussion: Interpretations and Conclusions 285
Session 8 Measurements of Water Environment
H. B. N. Hynes: The Interpretation of Biological Data with Reference to
Water Quality 289
W. Stumm: Chemistry of Natural Waters in Relation to Water Quality 299
G. A. Rohlich: Data Interpretation (Water) Drawing Conclusions 325
-------
Joseph E. Flanagan, Jr.*
Acting Director
Robert A. Taft Sanitary Engineering Center
WELCOME
I think it is fair to say that the opportunity to open a session like this ranks very
high among the very numerous and pleasant duties that fall to the Director of the Taft
Sanitary Engineering Center. I am not going to give you a discourse on the Center.
You can draw many conclusions about us just from this particular program. Since this
is a center for multiple environmental health programs, it is particularly appropriate that
we are able to sponsor a meeting of this type. I think that the most significant point
about this session is that it is sponsored by two of our operating divisions, probably the
first time that this has happened. This particular symposium is a direct outgrowth of
communications and rather constant associations between scientists of the Air Division
and of the Water Division. Representatives of our third, fourth, and fifth divisions are
also present at the meeting today, and I trust that everyone will get a good bit from it.
One of the peculiar situations about opening meetings is that you are supposed to say
a word of "welcome." 1 will never forget attending a session similar to this where one
of the rather high-level administrators at a school stood on a platform like this and
said, "I've been told that I'm supposed to welcome you. Well of course, you're welcome,"
he said, "but it just seems like an undue interference with the program for me to stand
here and continue along this line."
I had the privilege this morning of having breakfast with the gentleman who is
going to give the keynote address. About the only thing that I am going to say is that
if the rest of this program stands up to what I think this keynote address is going to be,
after meeting this chap and chatting with him informally, I think we are all due for a
worthwhile experience. So, welcome to Cincinnati nice to have you with us.
' Now Associate Director, Department of Environmental Health, American Medical
Association, Chicago.
-------
SESSION 1: General
Chairman: Keith S. Krause
Chief, Technical Services Branch
Division of Water Supply and Pollution Control
U. S. Public Health Service
-------
Dr. Charles S. Draper
Head, Department of
Aeronautics and Astronautics
Director, Instrumentation Laboratory
Massachusetts Institute of Technology, Cambridge
SUMMARY
Information engineering is described as the region of human activity that deals pro-
fessionally with the conception, design, building, testing, manufacture, and operation of
components and systems to sense physical quantities and from these as inputs, to generate
operating commands for the machines and the organizations that serve the needs and
desires of mankind. Automatic information systems are now recognized as necessities
throughout the realms of science, business, industry, and transportation. The evolution
of this important frontier of today's technology is described here along with the various
elements that contribute to such a system. A broad and accurate knowledge of the
environment has become essential for the health, economic welfare, and general progress
of the human race. From an engineering standpoint the transmission, processing, in-
dicating, and recording signals that represent the environmental information have been
established; the difficult problems remaining involve the balancing of benefits from
results to be expected against funds and other resources that must be made available.
INFORMATION ENGINEERING
THE NEW FRONTIER OF TECHNOLOGY
Information engineering is described as the region of human activity that deals pro-
fessionally with the conception, design, building, testing, manufacture, and operation of
components and systems to sense physical quantities and from these as inputs, to
generate operating commands for the machines and the organizations that serve the
needs and desires of mankind. The devices that provide for these purposes by sensing,
transmitting, processing, and applying information are called instruments. The complex
of instruments used to meet the information-handling requirements of a particular set
of circumstances is the instrumentation for the given situation^ The over-all technology
of instrumentation is the sum total of knowledge, engineering, devices, resources,
facilities, manpower, and services that are directed toward the realization of means to
fulfill the information requirements of civilization.
Instrumentation is based on components and subsystems designed for sensing,
communicating, processing, and using information. Today, the complex of industry and
business devoted to providing instrumentation for the United States involves operations
at the billion-dollar-per-year level; a level that has grown by several orders of magni-
tude during the last 4 decades. This expansion has naturally been accompanied by
the addition of many thousands of scientists, engineers, and technicians to the ranks
of those who are concerned with instrumentation as a profession. The end is not yet
even if we recognize the revolutionary growth that has already occurred, instrumentation
is still so far from exhausting requirements and possibilities that it must be classed
among the pioneering areas of human activity. Because of its universal importance for
other areas and its very great remaining growth potential, instrumentation may reason-
ably be considered as a most important frontier region of today's technology.
This special situation, which is associated with the means for handling information,
Draper
-------
does not come from any recent discoveries. Rather, the increasing importance of t e
field depends on an accelerating shift from human operators to inanimate equipment
that provides revolutionary new features in capacity, speed, reliability, accuracy,^ size,
weight, operating cost, and general utility. Recent improvements in these characteristics
have surely expended the feasibility regions of instrumentation and brought into a clear
focus some situations that were largely unrecognized until the middle years^ of this
century. The situations in question generally follow the model of those occurring m a
human body as it reacts to external stimuli. This response involves power level actions
from bone and muscle structures as they follow the command signals generated by the
brain on the basis of signals from the sense organs. These commands are sent over
nerve paths to muscles that respond by actions suited to the situation represented by
the sensor signals. The correctness of these responses is determined by the sensors and
brain that work together in comparing actual positions and motions with intended
positions and motions. Deviations of actual from desired conditions appear as feedback
signals within the human information system. These error signals cause the command
signals to change in ways that bring about corrective changes in the muscle actions.
Behavior of this kind, which applies feedback information for control purposes, has been
present in all high-level animals since the beginning of their existence. It is interesting
to note that feedback control has only recently appeared in the governors, the con-
trollers, and the servomechanisms that are now essential parts of substantially all
operating systems.
Until a new era started some 200 years ago, the progress of mankind was pri-
marily concerned with increasing the power of organizations and equipments. For
example, in ancient times, ships became bigger and stronger with propulsion by larger
sails and more oars. Such ships did not, however, become useful until their actions
were controlled by orders from commanders and pilots. Regarded as parts of over-all
systems, the human being in charge functioned by sensing information, combining it with
essential facts and plans through reasoning processes and applying the resulting de-
cisions to the power systems involved by means of command signals.
The powered effector subsystems for carrying out such commands have never existed
in nature. It has always been necessary for men to conceive, design, and build effector
systems. On the other hand, the information subsystem, always a necessary part of
any over-all system and effectively the "mirror image" of the power subsystem, could
depend on the senses, brain, and nerves of some human being to care for all its essential
functions. For thousands of years, this availability of human information subsystems
that could be easily matched to power subsystems much stronger than any individual
prevented attention from being directed toward a clear realization of the essential
role played by information in all the devices of technology. For example, from the
beginning of navigation, ships were controlled by the judgment and skill of a single
man. Armies of great size were commanded by general officers using disciplined men
in a proper organization as the means of control. Perhaps communications were slow,
but given enough time, military machines could usually be made to follow the plans
of their accepted leaders.
It is true that ships were wrecked and armies were lost by failures of their informa-
tion systems to cope with difficult situations, but on the whole human beings met the
needs of such systems well enough until the age of mechanized power began in the
eighteenth century. First, it was the substitution of automatic gear for the uninspiring
job of manipulating steam valves on the basis of piston position in a pumping engine
then it was the use of a centrifugal governor to keep engine speed constant by changing:
INFORMATION ENGINEERING
-------
steam flow, a chore that human operators could not have performed well in any case.
Regulators for pressure, temperature, and voltage followed the automatic timing devices
for engines, affording relief from simple repetitious tasks. Such devices became common-
place during the first half of the twentieth century, and are still being developed in
the direction of more sophistication and higher performance.
This same time period saw, among other advances, the realization of complex
electric power systems depending on accurate adjustment of voltage and frequency, the
building of very large high-speed ships, the use of high-performance aircraft, the
development of ballistic missiles, and the design of vehicles for travel through space.
The requirements laid on information systems by these new devices forced the necessary
performance well beyond the capabilities of human beings. The factors introducing
difficulties include complexity, accuracy, speed of response, length of working periods,
reliability, and environments too severe for comfort or in some cases even for survival
of human beings. Modern supersonic aircraft still carry pilots but provide many
radiation sensors, automatic adjustments, and booster devices to assist with information
system functions. Ballistic missiles with their one-way missions and space ships with no
men aboard actually force information system designers to use only inanimate elements.
The firmly demonstrated fact that in-production self-contained guidance systems of this
kind can receive, process, and apply information well enough to produce hits at great
ranges is only one of many proofs that the era of the automatic information system is
not only beginning but is already well on its way.
Evidence is all around us that information systems are now recognized as neces-
sities throughout the realms of science, business, industry, and transportation. The
worldwide credit card organizations and airline reservation services could not exist
without very rapid, accurate, and reliable collection, transmission, and processing of
information. At least one airline uses an information system in which an agent, in San
Francisco for example, receiving a request for a reservation punches keys to send a
signal across the country to a central computer in New York State. Results from this
computer, in terms of signals sent over telephone lines, return information on seat
availability within a few seconds. This example merely illustrates one of the ways that
information systems are revolutionizing the operations of modern society. Only the
surface has been scratched as yet; very wide regions remain to be explored and ex-
ploited by able individuals dedicated to the professional practice of information engineer-
ing. Today this field truly belongs to the frontiers of technology.
ELEMENTS OF INFORMATION ENGINEERING
Information engineering is concerned with applying scientific knowledge, professional
education, experiences, judgment, initiative, and perseverance in the use of natural
resources, facilities, available funds, and the capabilities of technology so that we can
realize information systems and their components able to meet stated specifications
within given limits. Briefly, the engineer undertakes to produce certain practical results
under the restrictions of existing circumstances. His particular stock in trade is state-of-
the-art technology, imagination concerning future developments, and the know-how to
build components and techniques into satisfactorily working systems. In fact, he has
many well-developed and extensive segments of technology at his disposol. These com-
ponent technologies are concerned with devices and systems that fall into five principal
classes:
1. Sensors, the devices that receive states of physical quantities as inputs and produce
signals representing these states as outputs.
Draper
-------
2. Communication systems that transmit signals among information subsystems.
3. Coupling systems that modify output signals from one subsystem so that these signa s
are suitable for inputs to other subsystems.
4. Computing systems that receive one or more independent signals as inputs, carry
logical operations, and produce outputs that represent information derived from the
inputs.
5. Display and recording systems that provide direct visual indications and records of
signals and the information with which these signals are associated.
Each of these five categories is now the basis for a more or less distinct area of
over-all technology. Some of these component technologies are very large and broad
within themselves, while others are smaller but still have magnitudes of some conse-
quences in the world of industry.
For example, communication surely includes all the techniques of wired telegraphy,
telephones, radio, and radar. Indeed, certain information systems may include trans-
continental telephone lines, submarine cables, and million-mile radio links from earth
to space vehicles. On the other hand, communication within a system may involve no
more than a few inches of wire.
Couplers concern a relatively small number of specializing companies that deal with
such components as amplifiers, transformers, digital-to-analogue converters, data storage
systems, etc.
Display and recording devices have many forms, with analogue and digital presenta-
tions ranging from lines on high-speed cathode ray tubes and electroluminescent figures
to typed numbers and inked points or curves on paper sheets. Equipment of this kind is
available from a number of companies that are able to provide standard arrangements
and to meet the needs of special situations.
Computing systems are the basis for a major industry supplying both analogue and
digital computers that operate over a wide range of speeds, capacities, and complexities.
It is certain that substantially any data-processing requirements of practical importance
can be fulfilled by currently available techniques. The differences among various designs
lie in capacity and speed for given weights and sizes, matters that have all received and
are still receiving great attention. Very large computer installations exist today for solving
complex problems, and very small units to deal with the complex but specialized
situations associated with missies in flight are also in production. Performance is
generally adequate, but the premiums for lighter, faster, and more reliable equipment
are so great that developments can be expected to .continue for a long time into the
future.
Because electrical signals are especially suitable for representing any kind of data
with low power levels, can be transmitted in many ways over short and long distances,
and are easily adapted for rapid processing, electronic techniques are very widely applied
in information system designs. Pneumatic, hydraulic, and mechanical principles may be
and have been used in computers, but recent tremendous developments of electronic de-
vices such as the widely used magnetic transistors and magnetic tape recorders will surely
continue to force information systems toward electronics.
SENSORS
Sensors are instruments that respond to states of physical quantities as inputs and
INFORMATION ENGINEERING
-------
deliver representations of these states as their outputs. For example, a mercury-in-glass
thermometer is a sensor for temperature as its essential input and produces the length
of a mercury column index as its output. When reference marks are placed near the
index so that "higher" and "lower" readings may be qualitatively distinguished, the
sensor is an indicator. If a scale having figures related to a series of systematically
placed reference marks is used to associate a number with each state of the input, the
sensor becomes a measuring instrument. When the sensor output is a signal of a kind that
has a series of states uniquely associated with correponding states of the input, the
sensor become a signal generator.
In practice, sensors may simultaneously serve each of the three output functions.
A sensor may have an on and off light for the purposes of indication and a scale and
pointer arrangement for measurement, and at the same time, may produce signals
representing input states. Indications have various forms including index positions,
"flag" exposures, lights, color gradations, etc. Measurements, by definition, always
involve numbers, but output signals may have many forms. Continuously varying
pressures, gas or liquid flow rates, current levels, voltage magnitudes, mechanical dis-
placements, and other configurations are all used as sensor output signals. In recent
years, discrete pulses of fluid or electricity have come to be widely used as signals.
These digital signals have the great advantage that they can be used as direct inputs
to modern computers and data-processing systems. It is to be expected that as time
goes on all information systems will be based on such signals.
Sensors are often called transducers because their functions involve "transducing,''
that is, carrying power from one region to another region. The term "sensor" is pre-
ferred because it stresses the fact that information, not power, is the essential factor in
the primary function of a sensor.
Of all the elements that make up information systems, sensors have the longest and
most honorable background in history. Devices to measure physical quantities really
helped very much to start modern science by providing sound experimental information
to supplement, and often to replace, the pure philosophy that scholars had inherited
from ancient Greece. Always, careful description of measuring equipment and proof of
accuracy formed substantial parts of any scientific paper. For purposes of this kind
each instrument was conceived and built to perform a specific task, usually one of
laboratory measurement. There was no consistent pattern of concepts, terminology, or
design. Outputs usually took the form of numbers read from a scale and index combi-
nation by observers who wrote down readings as they appeared. High accuracy for the
slow changes and completely static condition was the performance objective, rather than
ability to handle many inputs at high speed. Much attention was devoted to compensating
out or correcting for errors caused by environmental effects. Very fine instruments were
developed for research purposes. In general, these devices were not well suited for
reliable operation under severe service conditions.
With the development of ships, stationary power plants, automotive vehicles, and
aircraft, indicators and measuring instruments emerged from the laboratory and became
integral working components of information systems. About 1930 the universal pro-
cedure of regarding theory and practice for each instrument type- as a special isolated
section of science began to be replaced by methods in which concepts and notation
were made consistent and adapted to the purposes of information systems having arbi-
trary levels of complexity. This process has continued until the engineering oi sensors
is now a well-established part of system technology, with production components avail-
able that meet written specifications on the basis of routine inspection operations.
Draper
-------
Certainly, it is now practical to engineer information systems in which the recep
of data by sensors is a well-defined and reliable aspect of operation. A wide range
sensors is available from manufacturers who list them as catalogue items. ,
requirements of a particular situation cannot be met by production instruments,
whole background of science and technology is available for use in designing specia
instruments.
To meet the needs of present-day information collection, research work, and con-
trol operations, multiple sensing units are formed into patterns that serve the purpose
of making simultaneous. observations over an extended field of physical quantities. Ine
complex of worldwide meteorological stations is an example of a system that uses
multiple sensors to cover pressure, temperature, humidity, wind velocity, etc., for wide
geographical areas.
In general, indications, measurements, and signal generator outputs, all of low or
high quality as required, are well provided for by existing sensors or by principles that
can be embodied in practical instruments if this is necessary. The present frontier lies
not in sensors but in systems for collecting, processing, and applying information from
areas of significant size for understanding and control of essential conditions.
ENVIRONMENTAL INFORMATION SYSTEMS
Broad and accurate knowledge of the environment has become essential for the
health, economic happiness, and general progress of the human race. Adequate informa-
tion on the conditions in water supplies, lakes, rivers, and ocean shores is necessary for
the prevention of illness. Data on watershed situations, rain and snow, are required for
the control of floods by the operation of storage reservoir spillways. Atmospheric condi-
tions and wind patterns over wide areas must be systematically known for weather
prediction purposes. Smog from air pollution must be known if the public health is
to be properly protected and undesirable conditions brought under control.
Problems of sensors to indicate, measure, and generate signals for collecting any
amount of information on the environment are surely not difficult. From the standpoint
of technology, the transmission, processing, indicating, and recording signals that repre-
sent the environmental information are all matters of established engineering. The
interpretation of results in terms of safety and control measures may be subjects of
some controversy, but should not hold up major decisions.
The really difficult problems involve the balancing of benefits from results to be
expected against funds and other resources that must be made available. At the present
time, information systems to sense and interpret environmental data in terms suitable
for controlling reactions are just beginning to demonstrate their usefulness; however,
it is only a question of time before complete coverage networks will send environmental
information to central computers and display centers from which fast and effective
decisions may be made on proper reactions to correct or control undesirable situations.
The next few years will surely be exciting and interesting for both scientists and engineers
who carry the responsibility of dealing with the human environment.
10 INFORMATION
-------
Dr. Gaylord W. Anderson
Director, School of Public Health
University of Minnesota, Minneapolis
SUMMARY
The basic reason for measuring the factors of man's environment is to determine
the magnitude of the various external forces and, insofar as possible, the effect
these forces have on man. Examples are cited that point out the difficulties en-
countered in establishing valid associations between environmental variables and human
disease. Unfortunately, many of the measurements used to evaluate the magni-
tude of environmental hazards are no more reliable than are certain data on the
occurrence of diseases that we attempt to attribute to these hazards. As we approach
the problem of environmental measurements, our objectives must be twofold: We must
seek ways of measuring the magnitude of a vast array of environmental variables that
may conceivably have a bearing on human health, and we must attempt to measure the
development of human disease so that we may correlate these findings with the results
of environmental measurements.
OBJECTIVES OF MEASUREMENT SYSTEMS
The World Health Organization in its charter has defined public health as "a
state of complete physical, mental and social well-being and not merely the absence
of disease or infirmity." If we accept this forward looking definition of public health,
as have over 100 nations in joining the Organization, then is it obvious that in con-
sidering man's environment and attempting to establish valid bases for measurement
we must concern ourselves with a vast array of factors that, through their effect on the
environment, may adversely affect the physical, mental, and social well-being of mankind.
The purpose of our conference is to attempt to determine to what extent and in what
manner we may measure some of these factors and determine their significance for man.
Undoubtedly, we could find as many different definitions of environment as there
are registrants at this conference, and no one of us could rightfully claim for his
definition superiority over that of his colleagues. For my purposes and for the purposes
of this discussion, I like to think of environment as the sum of all the external biological,
chemical, and physical forces that surround man and therefore may influence his body
processes or his behavior. Under this broad concept, you and I can be considered as
part of each other's environment for we can spread infection to each other, or through
our personalities or behavior can not only irritate one another but can at times actually
jeopardize human life.
The usual connotation of the term sets aside the human factor, however, and concerns
itself with the physical, chemical, and non-human biologic forces that surround man.
Included in the latter are not only the pathogenic forms of animal and plant life but
also those animals that serve as reservoirs of infection and those insects that serve as
vectors. For the purposes of this conference, however, we may focus our attention on
the physical and chemical factors of the environment and set aside the biologic and
human elements. Yet, we cannot in truth set those aside for, even though we may limit
our discussion to a consideration of measuring air, water, or ionizing radiation we must
remember that to a very high degree it is man and his behavior that have introduced
Anderson 11
-------
hazards into those media. Industrialization resulting from human discoveries has resu
in pollution of our environment. Economic forces have resulted in demands for pro
the production of which results in environmental hazards not alone to the wor
the consumer, but also to the general public. Sociological and political forces emana in
in part from technological developments have created suburbanization with the resu an
creation of an entirely new set of environmental hazards. Cultural forces, expresse
in hum in behavior, are in many parts of the world of fundamental importance in po
tion of the environment, for example, the disposal of human excreta in such a manner
as will endanger the water that man must drink or in which he must work. Thus, while
we may, in a conference such as this, limit our discussions to the fairly specific problems
of measuring some of the physical, chemical, or microbial hazards of our environment,
we must never forget that behind these factors are vast economic, sociologic, cultural, and
political forces that, in the last analysis, are responsible for endangering the environment
and, in the years ahead, will in all probability add to this danger. To a certain degree,
we who are concerned with the control of the environment must resemble the fabled
King Canute, who is reputed to have attempted to sweep back the tide from the
beaches with his broom. Unquestionably, our brooms of control, while less regal, will
have more effect, yet we must never forget that these other forces are more fundamental,
that they are part of an inevitable social evolution that is far less easily controlled or
regulated than are the tangible components of our environment.
Keeping in mind this concept of environment, we may now ask ourselves why we
wish to measure various factors of our environment, what factors we wish to measure,
what conclusions we may wish to derive from those measurements, and finally, how we
shall proceed with such measurements. This latter I shall leave to others far more
competent mathematically than I can ever hope to be; yet from the point of view of an
epidemiologist, I may have the temerity to make a few suggestions.
Our basic reason for measurement is to determine the magnitude of the various
external forces in man's environment and, so far as possible, the effect that these forces
have on man. Ideally, therefore, we are seeking to measure both cause and effect,
variables that are interrelated, in that the causal force, if sufficiently great, may be
harmful to man, who is exposed to the force. If this cause-and-effect relationship is
known or can be established, we can then presume that the effect will increase or
decrease to the extent that we are able to alter the magnitude of the cause, other
secondary or related factors being equal and constant. Unfortunately, however, we can
rarely assume that these other factors are static for the social and biologic relationships
between man and his environment are conditioned by a vast array of changing variables.
In the laboratories of physical science, it is often possible to study the effect of a single
variable by stabilizing all other components of a reaction system; yet in the biologic world,
and especially in the study of man in his normal community environment, we can rarely,
if ever, keep all but one variable constant.
May I illustrate this point with a simple example from the realm of the infectious
diseases, and notably typhoid fever, the control of which stands as a magnificent monu-
ment to environmental sanitation? The spread and development of any infectious dis-
ease may be considered as consisting of six components, the etiologic or causative agent,
the reservoir of infection, the escape from the reservoir, the transmission to a new host,
the entry into this host, and the susceptibility of the host. We may easily measure the
extent of contamination of water, a well-known and easily performed measurement, and
may also count the number of typhoid cases and deaths. It would be simple to correlate
these two measurements and arrive at certain very satisfying conclusions, which
12 OBJECTIVES OF MEASUREMENT SYSTElMs
-------
might, however, be erroneous because such a correlation would overlook such variables
as the size of the reservoir and the susceptibility of the host.
The accepted measurement of safety of water for human consumption has been the
so-called Treasury Standards, going back to the era when the Public Health Service was
a part of the Treasury Department. These standards are based on the presence or
absence of gas-forming organisms in detectable quantity in various quantities of water,
it being correctly assumed that these gas-formers are of intestinal origin. The standards
are thus a measure of sewage pollution, not of contamination with typhoid organisms,
the detection of which has been technically impossible with a satisfactory degree of
accuracy. In the establishment of these standards and the acceptance of water that did
not have more than a certain amount of demonstrable sewage pollution, there was an
assumption that this same water did not have enough typhoid organisms to be dangerous
to the consumer in the quantities he might reasonably be expected to drink. In other
words, these standards were based on an assumed ratio between the numbers of typhoid
and of colon bacilli in the sewage of a given community. If the sewage pollution did
not exceed a certain level, we could assume that the number of typhoid bacilli was below
the danger level. Experience showed this assumption to be correct.
Yet, as time has passed and as we have observed the trends in water pollution and
typhoid incidence, we have been forced to recognize that these relationships no longer
hold. The number of colon bacilli that an individual or a unit population contributes to
sewage has remained essentially stationary, since we are dealing with organisms that are
normal and invariable inhabitants of the human intestine. On the other hand, the
reservoir of typhoid, chiefly in the form of carriers, has declined strikingly as the carriers
have died off faster than they have been replaced. Thus, while the number of gas-
formers per unit quantity of sewage may have remained unchanged, the number of
typhoid organisms has declined and the older ratio between gas-formers and typhoid
bacilli is no longer valid. In other words, so far as the risk of typhoid is concerned, we
may safely drink water that contains far more colon organisms than could have been
safely consumed 20 to 50 years ago. Our epidemiologic experience confirms this. In
former days, sewage pollution of water resulting in a short-lived community-wide outbreak
of diarrhea was invariably followed by typhoid. In recent years, we have had innumer-
able instances of community-wide diarrhea due to accidental sewage contamination of
water; yet typhoid has not ensued, simply because the reservoir of typhoid carriers has
dwindled to a point at which the number of typhoid bacilli in the sewage is too small
to cause disease. Thus the old standard, while still defensible on aesthetic grounds as
a matter of common decency and possibly still as a measure of other pathogens that
may contaminate the water, is no longer a valid measurement of the safety of the water
as a vehicle for the spread of typhoid, even though the public, through its continuing
lack of exposure to typhoid, is more susceptible today than in former years.
This very change in susceptibility may, however, have an opposite effect in a
situation in which the number of organisms available for ingestion may be no greater
than in former years. As one examines the records of food-borne outbreaks of typhoid
during the past half century, one notes that although the number of such outbreaks
has declined the attack rate among those assembled for a meal prepared by a carrier
has become progressively greater. The decline in number of outbreaks can be attributed
to the reduction in the number of carriers while the increased attack rate is due to the
fact that with the declining incidence of cases and prevalence of carriers fewer persons
have been latently immunized through repeated small doses of organisms. Any standard
that might have been developed to determine the number of typhoid organisms in food,
Anderson 13
-------
even if accurate, would have afforded little clue to the human effect, unless at the same
time we had taken cognizance of the susceptibility of those who were to eat the oo^.
Thus, mere measurement of the number of typhoid bacilli in food would not
given a true picture of the hazard of consumption of the food.
Two more examples will suffice to point out some of the problems
measurements as indicators of hazards to human health. It is a wel s
that milk from cows infected with Q-fever contains large numbers of the causative
rickettsiae. Equally well established is the failure of pasteurization carried out in
accord with present standards to kill certain strains of the Q-fever organisms, with the
result that persons drinking such milk not only can but actually do develop a rickettsiai
infection as manifested by the development of specific antibodies. On first thought,
therefore, one might logically conclude that our criteria for satisfactory pasteurization
should be changed to require a higher temperature so as to kill these organisms. Yet,
in spite of incontrovertible evidence of infection as a result of drinking milk containing
viable rickettsiae, there is no evidence that clinical illness has resulted. There are
reasons for believing that although serious and even fatal illness may result from
inhalation of very small numbers of the Q-fever organisms ingestion in even large
amounts does not produce disease but only latent unimportant infection, which may
actually be beneficial in that it may possibly immunize the individual against illness
if at a later date organisms are inhaled. Measurement of the number of organisms in
milk may therefore have little meaning or significance; yet great importance can be
attached to any measurements of the number of organisms suspended in the air, their
survival in the air, and the physical forces that govern their dispersion into the air.
Quantitative studies of the production of aerosols may have tremendous significance in
the understanding of this as well as of several other infectious diseases and of various
conditions attributed to inhalation of chemical agents.
For a third example, I should like to turn to poliomyelitis, a situation in which we
unfortunately find an amazing amount of unreliable data, both as to occurrence of the
disease and causative environmental factors. Much has been published regarding trends
in the incidence of the disease and innumerable attempts made to correlate these ap-
parent trends with various forces that lend themselves to easy though not always highly
accurate measurement. The sad fact is that we have few reliable statistics that can be
used to determine trends of this infection. Morbidity or even mortality data are of
little value, because diagnostic criteria have changed tremendously from the era when
only the severely paralyzed were counted to a later period when the much more numerous
non-paralytic infections were included in the report data but not clearly separated from
the paralytic cases. The situation has become even more complex of recent years as we
have come to recognize that a high proportion of the non-paralytic cases clinically
diagnosed as poliomyelitis are in fact infections with other viruses, the true nature of
the infection being determinable only by laboratory procedures that may be expensive
of both time and personnel and hence are not routinely performed. Even more disturbing
from the statistical standpoint is the fact that certain other viruses may produce
paralytic conditions clinically indistinguishable from those of true poliomyelitis. Only
in recent years have we had practical laboratory procedures that will identify the true
etiology of such infections, and even today there is no universal use of such tests in
diagnosis. Thus, even though we may make exact measurements of environmental factors
such as improvement in various aspects of community sanitation, correlation of these
findings with those of poliomyelitis incidence would be of little value because of the
highly inaccurate nature of such incidence data.
This very problem may well be the basis for no small amount of controversy in the
14 OBJECTIVES OF MEASUREMENT SYSTEMS
-------
years immediately ahead of us. It is common knowledge that with the suburbanization
of our large city population vast numbers of persons previously served by public water
supplies and sewerage systems have moved to areas where reliance is placed on individual
wells and septic tanks or cesspools. That the waste products may drain into the wells
is amply shown by the amount of household detergent in the water from such wells.
Currently, we are seeing the widespread use of oral vaccines for immunization against
poliomyelitis, vaccines containing living attenuated organisms that pass through the
intestinal tract and are given off in the feces for variable periods of time. Although the
survival of these in water and sewage has not been well studied, it is conceivable that
they may appear in various water supplies. Could we infer therefore that by demonstra-
tion of such virus or by any measurement of its concentration in water we could
establish a hazard that required the development of new control measures? I doubt it,
since we would be dealing with a virus designed for human ingestion as an immunizing
agent and therefore safe. In fact, were we to demonstrate that the excreted oral vaccine
passed into the untreated water of such wells, we might conceivably look upon it as a
valuable means of immunizing certain persons who had failed to take the vaccine for
their own protection. All of this may not occur. I introduce the possibility merely to
point up the need for proper interpretation of data, however exactly they may be
determined.
Unfortunately, many of the measurements used to evaluate the magnitude of en-
vironmental hazards are no more reliable than are certain data on the occurrence of
diseases that we attempt to attribute to these factors. I am reminded of a study, un-
fortunately published and made the basis of editorial comment, in a reputable medical
journal that purported to show an inverse correlation between the hardness of water
and the development of coronary disease. Since this correlation purportedly existed,
the inevitable conclusion was advanced that softening of water was undesirable as it
carried with it an adverse effect on the cardiovascular system of the user and might
therefore be a factor in the high mortality rate from heart diseases. One is accustomed
to accept cardiovascular morbidity and mortality with a certain degree of caution
because of diagnostic difficulties; yet one can expect a reasonably exact measurement
of the hardness of water. Yet, as one examined the data, one learned to his unbelieving
astonishment that the author had taken for each state in the Union a single numerical
value that was supposed to represent the hardness of water throughout the state. Not only
had he ignored geological factors and the resultant great variations in the hardness of
the public water supplies within a given state, but he had also ignored movements of
population, assuming that the individual throughout his life had been under the in-
fluence of the hardness of the water of the state in which he drew his last breath.
It is almost incredible to me that data so obviously unreliable should be published
and made the basis of conclusions, both by a scientific journal and by such eminently
reliable sources of medical news and opinion as our well-known popular news digests.
Even more amazing and naive is the statement in the article that the data have superior
reliability and significance because they have been processed by an electronic computer.
I realize full well that in my attempt to point out some of the difficulties in estab-
lishing valid associations between environmental variables and human disease I may
have appeared to be purely destructive. If such be the illusion that I have created, I
must beg your forgiveness; although I recognize the unavoidable difficulties, I equally
recognize the importance of such correlations whenever they are valid. In all too many
instances, we are presented with environmental variables, the significance of which is still
problematical. We can easily recognize the potential significance of carcinogens in the
air, in water, or even in food, even though we cannot as yet assess their true role or
Anderson 15
-------
importance. We recognize the fact that the concentration of these carcinogens is »
creasing. At what point, if at all, do they become significant as factors in the eve
ment of human cancers? Certainly, we need th» most exact measurements and iden i ^
tion of these carcinogens, not only to help determine their significance but a s
measure the need for and efficacy of control programs. Similarly, in our tima e
evaluation of the deleterious effects of radiation, we must have precise basic measure-
ments, even while we are in the stage of speculation and controversy as to the significance
of these data.
Currently we are embroiled in a heated public controversy as to the significance of
various pesticide, fungicide, or herbicide contaminants of our food and water supplies.
A certain female journalist has unduly alarmed the general public with her speculations
as to significance, but we know that we did not need to wait for her to recognize the
potential hazards attendant upon the use of these chemicals. It is important that we
use all possible means to establish the best possible base lines for measurement of the
real import of various chemical concentrations and not be forced to rely on the specula-
tive dire predictions of the journalist. Even though some of the base lines that we now
attempt to establish may later be found to have little value, we must in our limited
knowledge establish as many of these as possible so that we may ultimately select those
of greatest value. I include here the baselines that in one way or another may measure
the degree to which there are natural or man-made chemical, biological, or physical
contaminations of our environment. Concurrently, there rests with us in the realm of
medicine a responsibility for developing or improving upon the standards for the
measurement of the development, the incidence, and the prevalence of disease. In this
latter, you from the environmental field must share with us in the realm of medicine in
recognizing that disease and ill health are not simple processes subject to single en-
vironmental or metabolic influences, but rather represent a complex interplay of various
forces, some destructive, some protective, but all conditioning the ultimate response of
the human body.
In our attempt to establish valid environmental factors in the causation of disease,
we must not be led into the false assumption that every component of an environmental
control program must be evaluated on the basis of a demonstrated relationship to human
illness. We must never forget that quite apart from its pathogenic significance an en-
vironmental variable may be of import as it affects that vague something that we call
the sanitary culture of the community, common decency, or even the aesthetics of
human life. The mere fact that I cannot demonstrate a disease relationship for polluted
bathing beaches or swimming pools does not alter my reluctance to go swimming in
sewage. That I cannot show valid morbidity or mortality statistics as to the significance
of excess noise does not make me any less reluctant to live in the cacophony of
bedlam. A wise court has long since ruled in a decision on nuisances that demonstra-
tion of ill effect is not requisite and that abatement of an environmental nuisance may
be required if such renders habitation uncomfortable or interferes with the normal
enjoyment of human life. Ideally, we may hope that in as many situations as possible our
control measures may be based on exact cause-and-effect measurements, but we must
never forget that there are intangibles that contribute to the sanitary culture, the peace
of mind, the standard of living of mankind.
As we approach consideration of this problem of environmental measurements our
objectives seem quite clear. On the one hand we must seek ways of measuring the
magnitude of a vast array of environmental variables that may conceivably have a
bearing upon human health, being still uncertain as to the relative importance or signifi.
16 OBJECTIVES OF MEASUREMENT SYSTEMS
-------
cance of many of these variables. On the other hand, we must attempt to measure the
development of human disease so that these findings may be correlated with the results
of environmental measurements. In both cases, we must strive for the most precise
possible measurements, yet constantly recognizing that statistical correlation does not
in itself mean demonstration of a cause-and-effect relationship. Like the pilot of the
ancient world beset with mythical maritime dangers in the Straits of Messina, we must
attempt to sail the difficult and treacherous course between Scylla and Charybdis,
Scylla the multiheaded monster that snatches at and feasts upon fragmentary evidence
of no validity and the whirlpool Charybdis that engulfs our imprecise data, churns them
through the whirling intricacies of modern electronics, and spews forth the pieces in
the form of false conclusions. Wise and skilled is he who safely and successfully sails
between Scylla and Charybdis.
Anderson 17
-------
SESSION 2: General
Chairman: Dr. August T. Rossano, Jr.
Research Professor
Department of Civil Engineering
University of Washington
-------
Dr. Paul B. MacCready, Jr.
President, Meteorology Research, Inc.
Altadena, California
SUMMARY
Principles common to air pollution measurement systems (chiefly meteorological) are
outlined, with emphasis on the need for balance between statistical methods and the
understanding of physical relationships. Four representative types of measurement-
forecast systems are described. Development of a system and application of diverse
measurement techniques are exemplified in a hypothetical field study of flow and
turbulence regimes at a western hazards site.
DESIGN OF MEASUREMENT SYSTEMS
INTRODUCTION
A measurement system for basic studies in air pollution in the atmosphere or for
operational uses can take on one of an infinite variety of forms determined by the program
aims, resources in money and people, location, meteorological situation, and the state of
the instrumentation art. Since all these subjects cannot be treated meaningfully here,
this paper will examine only the basic principles common to most systems and describe
a specific project that illustrates some of the principles. Some knowledge of the funda-
mental principles will permit any given system to be viewed in a useful perspective.
The author's background in the subject has derived primarily from many field programs
in various weather regimes and terrains, field programs that often had operational aims
but necessarily involved some basic physical studies. Therefore in this paper field
research systems will be emphasized. The term "system" is here used in a broad sense
that means instrumentation, its use, and the handling of the resulting data.
Public Health Service Publication No. 1022, the Proceedings of the National Con-
ference on Air Pollution, is highly recommended as a general reference to this subject,
especially the section on "Applying the Measuring and Monitoring Know-How.''
SOME GENERALIZATIONS
DATA ACQUISITION
The director of the program using a field measurement system will invariably want
more data than any reasonable system is capable of giving him.
ANALYSIS
Nevertheless, usually the data that are obtained cannot be properly assimilated.
This is especially true in research projects, for the data treatment is not routine. The
data reduction may be easy, but its analysis is not. Virtually every project could benefit
materially from more analysis. A reasonable compromise to aim toward on some research
programs is to split the funds equally between the instrumentation-field phase and the
data reduction-analysis phase.
INSTRUMENTATION SYSTEM
Often the absolute accuracy of measurement is not very important. Making the
MacCready 21
-------
pertinent measurement at the appropriate place may be more vital, such as being su
primary wind measurement pertains to the dominant flow. In many studies a ne
of crude wind recorders will be more useful than one precise unit. In some cases
absolute accuracy of 1°C in temperature is unnecessary, although in other cases
accuracy of 0.01°C in temperature difference is desired.
STATISTICS
The output of most studies or operations is statistical data. The accuracy or useful-
ness of these data is much improved if the correct sort of physical understanding was
involved in deriving the data. This balance between statistical methods and physical
understanding is of great importance, and it will be emphasized later in this paper.
One main point is that although statistics are involved in typical meterological studies,
statistics often constitute only a blunt research tool. Another primary point is that the
use of statistics decreases as physical understanding increases. The field of meso-
meteorology has advanced considerably recently, as have meteorological instrumentation,
data processing methods, and turbulence-diffusion relationships; it is now generally
possible to interpret the movement and spread of pollution somewhat quantitatively from
standard synoptic data. Thus statistical pollution estimates can be refined by physical
inputs.
SOME FACTORS IN AN INTEGRATED
METEOROLOGICAL SYSTEM
SAMPLING
Meteorological variables and pollution variables cannot be ascertained completely in
all three space dimensions and time, but rather samples are taken. The sample is con-
sidered to represent the variable over a larger range of time or space. If the system
sampling design is good, the samples can be truly representative. The sample may be,
for example, measurement of wind or pollution at one place at one moment, or the
same measurment averaged over a long period; for typical diffusion or pollution studies
the latter is more likely to be representative, except for rapidly varying quantities.
AIR MOTION
The movement of pollutants is usually considered to consist of two parts, the mean
transport of material and the spread of the material into lower concentrations by means
of turbulence. Thus the measurement system must illuminate these two parts. Most
commonly the mean flow data would be normal surface wind measurements and tracking
of balloon ascents. The turbulence data that is desired actually describes the diffusing
power of the air. This diffusing power depends both on the turbulence and on the type
of pollution release; a small individual cloud puff is treated by different equations than
are used for a continuous point source. The complete relationship between turbulence
and diffusion is complex and not adequately understood, but some significant simplifica-
tions have been developed recently. These simplifications are based on measuring the
turbulence as direction fluctuations of a fast response direction vane in the vertical or
horizontal, and processing the analog signal with one or more electronic filtering devices
("sigma" meters) to show the energy over particular broad wavelength ba d V
certain cases, such as the diffusion of puffs, only one "sigma" meter is need d A
rate of cloud spread is simply proportional to the square of the meter 'j-
least to the accuracy required in most studies). The turbulence measure M mg
ements can be
22
DESIGN OF MEASUREMENT SYST
-------
made from ground-based equipment or even from aircraft. This turbulence measurement
approach is supplanting the older method of inferring the turbulence from measuring
the wind and temperature profiles. The turbulence method measures the important param-
eter directly rather than indirectly, and also it can be used in complex terrain situations
where mean profiles are less informative.
In summary, measurement techniques are available now to define the velocity field
that carries the pollutants. The measurements show the mean flow of the air and also
its diffusing power.
POLLUTANT OR TRACER
In actual pollution work the pollutant is the tracer. Its source is somewhat known.
It can be picked up on the ground and sometimes in the air. A tracer can also show
what concentrations can be expected at a spot downwind from a particular source, and in
addition a tracer is often useful in filling in the picture of just how the material moves
from source to destination. The tracer can be a cloud of identifiable particles. Among
the many available particulate or gaseous tracers, the most commonly used is zinc cadmium
particles of about 2 microns mean diameter, which fall less than 1 meter per hour
through the air. The particles are collected by filter or impactor and usually counted
by fluorescence under ultraviolet light. Tracing can be done to distances in the hundreds
of miles.
An excellent tracer is oil-fog smoke, for by a single visual or photographic observa-
tion it can show the entire course of a diffusing mass of air. Although it may be
deemed a qualitative tracer technique, in many cases it may actually be more quantitative
than particle tracer methods because it can provide many tests simply during one experi-
ment. Particles are measured only at fixed collection points, but the smoke is observed
wherever it goes. Smoke from standard generators can sometimes be detected as far
as 5 miles from the release point.
A versatile new tracer method is the use of radar to track mylar super-pressure
balloons, which float at approximately a constant-density level. Tracking for periods
longer than a day has been successful in Los Angeles pollution studies.
ENVIRONMENTAL METEOROLOGICAL DATA
Most pollution meteorological studies hope to provide techniques for forecasting
pollution factors from the standard weather data supplied by the U. S. Weather Bureau.
During the study the gross environment features are noted, but also the smaller links
are examined that connect the flow and diffusion to these environment features
important links such as surface roughness, topography, radiation, and stability.
DATA REDUCTION AND ANALYSIS
The details of the data reduction and analysis depend so much on the specific prob-
lem that only a few generalizations are appropriate here. As the automatic weather
station concept undergoes continual development, the data acquisition tends to be handled
digitally. This puts the data in a convenient form for automatic data reduction. Other
data can readily be converted to the digital format, and thus digital data handling can
be employed throughout some projects. Automatic data handling is, of course, desirable
for routine monitoring programs, and even for research projects it makes analysis
easier because more of the pertinent data can be economically provided to the analyst.
MacCready 23
-------
The "sigma" meter represents data handling that in many cases is proba y^ ^
done by analog methods. The "sigma" meter is actually an electronic an_7°^.e °more
running mean of direction variances. Digital techniques for this one task are ^
expensive than the analog method, and are also less suitable because the digitizing
vane angle is usually done with a resolution (say 1°) that introduces appreciable errors
into a "sigma" calculation for weak turbulence conditions.
THE AIM OF A MEASUREMENT SYSTEM
The following list suggests representative types of operational systems:
1. Completely Automatic "Present Picture." This system monitors the three-dimensional
field of wind flow and turbulent diffusing power, and can thus present at any moment
the picture of the transport and diffusion of a pollutant cloud released at any given
point. Since measurements cannot be made everywhere, some empirical extrapolations
of data from a few points are actually used.
2. Completely Automatic "Future Picture." This forecasts the "Present Picture," and
then derives the flow and diffusion picture. Thus this system must incorporate
meteorological forecasting factors such as stability, radiation, topography, and pressure
trends. This is the most desired system, but has the basic limitation inherent in any
forecasting system in the present state of the forecasting art.
3. Practical Category Type of "Present Picture." Here the most common wind flow
and turbulence fields are categorized into a small number of types, and the rules for
cloud transport and diffusion are summarized for these types.
4. Practical Category Type of "Future Picture." The flow and turbulence field cate-
gories are forecast by theoretical-empirical relationships to available standard mete-
orological parameters. Then the categories are interpreted, as above, in terms of
rules for cloud transport and diffusion. This system is the most practical in the
average case. It still has the forecasting limit inherent in any meteorological
forecast. It is suitable for compiling statistics on the diffusion climatology of a region.
An example of a pollution research program is given later in this paper. It represents
an example of the category types of system outlined in 3 and 4.
DEVELOPMENT OF A MEASUREMENT SYSTEM FOR
DIFFUSION MONITORING OR FORECASTING
The key to the development of an operationally useful system lies in basing the
system on the simplest possible factors that are dominant in determining the mean flow
and turbulence. Thus the success of the system will depend on some physical studies
at least a small field research program is necessary. The system development might follow
this course:
1. Decide on the aim of the project or system.
2. Design a tentative operational system, considering likely meteorological factors
and taking account of the accuracy obtainable within the economic framework
provided.
3. Perform a research program to develop the factors for the operational s
4. Redesign the final operational system.
5. Keep refining the system as more weather history becomes available
24
DESIGN OF MEASUREMENT
-------
A SPECIFIC EXAMPLE
The following example of a hazard study illustrates some of the principles inherent
in any study or measurement system. The example is a fictitious composite of several
real projects, but the results have some general applicability.
The site is considered to be a 25-mile by 25-mile square in the mountain-desert region
of the western United States. Toxic materials may be released at two points: one at the
top of a 1000-foot ridge, the other halfway up the west face of a 3000-foot ridge to the
east of the lower ridge. The problem is to establish the concentrations at which this
material might reach the valley east of the high ridge, for the predominant west wind
conditions.
The data system consisted of:
1. A portable 10-foot meteorological tower located at the release site or along the
course of movement of the tracer, giving mean wind, turbulence, temperature, and
temperature gradient.
2. A similar fixed tower on the lower ridge.
3. Four wind stations.
4. A light plane that records altitude, temperature, IR ground "temperature," hu-
midity, turbulence, and rate of climb, with an observer making notes on a
magnetic tape recorder and photographing smoke plumes.
5. 40 rotoroid-particle samplers (and 3 filter samplers for backup calibrations) in
two lines crosswind to the flow at distances of 3 to 10 miles from the release point.
6. Various generators to dispense fluorescent tracing material (uranine dye for rapid
assessment but shorter distances; zinc cadmium sulfide for longer distances), and
two oil fog generators.
7. A pibal wind station.
8. Several wind recording systems already at the site (and having provided previous
data records).
9. USWB station records at distances of about 25 miles outside the site.
A total of about 30 quantitative tracer tests were made, each one (except for night-
time releases) being accompanied by visible smoke releases. In addition, smoke alone
was released about 15 times. The smoke releases were of great value in this project
because they show so conveniently the mesoscale flow patterns in this complex terrain.
In the most convective situations the quantitative tracer samplers on the ground would
provide virtually no information; in these situations the visible smoke gave the complete
explanation for the lack of counts on the ground.
The tests all took place during one 3-week period, and yet the results can be applied
fairly well to other seasons of the year since the flow and turbulence regimes fit rather
well into a few identifiable categories. The categories also show a relationship to flows
studied on other projects involving waves and turbulence in mountainous terrain.
Figure 1 summarizes the four distinct flow and diffusion categories.
The categories are defined as follows, with the wind speeds referring to velocities
at ridge-top levels.
MacCready 25
-------
a) Wind-associated.
(1) with sunny conditions, wind medium to medium-strong.
(2) with cloudy conditions, wind medium.
b) Strong wind-associated.
(1) with sunny conditions, wind strong.
(2) with cloudy conditions, wind medium-strong.
c) Convective, wind light.
d) Semi-convective.
(1) with sunny conditions, medium-light.
(2) with cloudy conditions, wind medium-light.
e) Stable, nighttime conditions, low wind speed.
(a) Wind-Associated
(b) Strong Wind-Associatpd
Wind Flow
Smoke from Continuous Source
Semi -Convective
Convective
(c) Convective or (d) Semi-Convecti
e (e) Stable, drainage (night)
Fig. 1 Schematic Patterns of Flow and Diffusion Categories
The main categories are primarily determined by wind speed, ^th some stab;lity
" y bution of these
estimated
is defined somewhat quantitatively for a gv n weh , ^
predicted to the extent that winds and cloudTness c7n t ior ^^'^ hazard *»
statistics can be derived with the help of USW hazard climatology
help of USWB records
26
DESIGN OF MEASUREMENT SYSTEMS
-------
There are of course more complications than are shown here, such as considerations
on wind direction, how to define cloudiness, statistical significance of the estimates, etc.
Nevertheless, this study illustrates how practical answers can be derived from a modest
program in which the dominant factors are measured.
The program benefited greatly from the versatility of the aircraft, which did stability
soundings and turbulence regime plotting, provided a good vantage point for smoke
observations, did some of the tracer pickup, and helped to establish gross vertical air
motions by vertical velocity measurements and by horizontal plots of potential temperature.
Programs in less complex terrain could rely more on ground-based and tower-based
equipment.
The categories that came out of this field study could not have been derived from
sampler data by statistical analysis, because there were too few tests. It was essentially
the physical interpretation of the data that yielded the significant results. The most
complete hazard presentation should be given in statistical terms, however. After the
sort of preliminary study shown here, in some circumstances it would be warranted to
repeat the experiments in meteorological situations representing the greatest hazards,
and thus build up statistical data for proper presentation in terms of statistical
significance and extreme values.
MacCready 27
-------
Dr. Alexander Goetz
Associate Professor of Physics
California Institute of Technology, Pasadena
SUMMARY
Air and water as gaseous and liquid components of the environment are considered
essential ingredients for human, animal, and plant life ingredients that are also acted
upon by these live forms. Air and water are evaluated in terms of chemical and physical
parameters relating to their occurrence in the natural regenerative and degradative cycle
and to their physiological assimiliation. Particulate pollutants and reactive gases are
discussed. Emphasis is given to the physical and chemical characteristics of aerosols
and their potential role as pollutants of environmental significance.
PARAMETERS
All forms of life exist by continuous interaction with the liquid and gaseous com-
ponents of their environment, i.e. water and air, for both represent the indispensable
vehicles for nutrition and metabolism. One could be tempted to term the function of
the vehicles catalytic, but this would be incorrect for, unlike true catalysts, both water
and air are gradually degenerated by supporting live forms, regardless of size and
complexity, a fact that is the base of all waste problems. The large difference in the
physical and chemical constitution of these two vehicles is reflected in the way they
serve and are required by specific live units.
Man's average rate of passing the gaseous vehicle is by mass about 3 times that of
the liquid (about 7.5 kg or 17 Ib of air versus 2.5 kg or 5.5 Ib of water in 24 hours),
which means that the demand for air on a volume scale is 3000 times greater. Accordingly
the type of degeneration caused by the passage through the organism is very different
for both vehicles, and it is vastly greater for water than for air.
This environmental degradation by life as such is corrected by natural regeneration,
i.e. by processes of neutralization or dispersion, which again differ principally from one
another. Water is regenerated by drainage into the oceans, evaporation with subsequent
condensation in fogs and clouds, redelivery by rainfall, and subsequent nitration through
soil. The atmosphere is regenerated by diffusion, convection, washout through rain,
interaction by plant metabolism (C02) and probably on a major scale by photo-
chemical oxidation (volatilization) of airborne organic constituents.
These natural processes had kept the equilibrium between live matter and its
environment on our planet for more than one billion years until this balance was upset
by the technological age with a huge variety of artificial energy conversion processes
of growing magnitude and the consequence of dense habitation centers, both causing
rates of environmental degeneration that rapidly exceed nature's janitorial capacity.
This situation requires corrective action by man, as an answer to the question:
"To be or not to be . . ." in the face of the growth rate of these technological en-
deavors. This action must be effected with minimal economic penalty and thus can
succeed only if guided by detailed definition and systematic evaluation of environmental
parameters to arrive at standards of general validity.
This evaluation not only consists of the chemical and physical definition of each
Goetz 29
-------
single pollutant type, as derived from its physiological tolerance limits, but must also
critically consider the validity of each such limit in the presence of other pollutant
types. Their presence may shift the tolerance for a specific agent substantially m either
direction if its effect is synergistically attenuated or intensified.
But even assuming that a system of perfect parameter definitions has been derived
from the tolerance pattern, the major problem remains in correlating this system -with
the indications of the sensor devices available, for it cannot be taken for granted that
they always represent quantitatively truly or even approximately the reaction of
the live organism. Consequently the critical selection of the sensor-types and the
judicious correlation of their data represents a major task upon which rests the ultimate
success of the effort.
As is to be expected, the parameters pertaining to water and air reflect the principal
differences between liquid and gaseous matter and their modes of physiological
assimilation.
Contaminants confined to a liquid are much more readily identified and removed
at the source than are contaminants dispersed in air.
The lowest tolerance limits for both media are of the same order, a 10-1*, which
represents in water a mass concentration of 10 u,g/liter and is comparable with the
molar concentration of 1 pphm in air. Accordingly the same magnitude is required for
the maximal sensitivity of the senor devices. These limits are exceeded by several orders
for radiological and microbiological pollutants, for the latter because they are potentially
self-propagating in many water supply sources but not in the atmosphere. For these
pollutants the sensitivity threshold reaches about one billion times further, for the device
must sense, e.g. one E. coll cell in 100 milliliters (equivalent to 10-8 ^g/liter).
The most obvious difference between water and air is the possibility that water can
be delivered to the consumer from supply centers where permanent surveillance and cor-
rective action is applied. The water parameters encompass a wide variety of pollutants,
present in the dissolved or colloidally suspended state, and their low tolerance limits
apply particularly to traces of toxic metals and metalloids, such as arsenic, lead, mercury,
manganese, etc.
The highly developed parameter definitions of today for water resources are the
result of experience gained during the last 80 years. It is quite interesting to follow
the trends of this period: first efforts were to eliminate the visible "dirt," and then to
progressively define and control the less and less obvious pollutant traces. This appears
to be analogous to our present early state of parameter definition for the air, because
it was only about 20 years ago that atmospheric pollutants were generally realized as
visible "dirt" in the form of smokes, fumes, and dusts.
Since then we have gone a long way, so that today's pattern of air pollution param-
eters may be summarized as follows. Principally the parmeters discriminate between
contaminants that are temporary and those that are permanent, a distinction that to
some degree depends on the rates of atmospheric regeneration as determined by the
prevailing micrometeorological conditions. Paniculate matter of the larger size classes is
considered temporary airborne material because their fallout rates limit their suspension
to a few hours. It is difficult to exactly define the lower size limit, because the r te of
precipitation depends largely on particle shape and density. In terms of V
(Stokes') diameter the borderline can be placed at 5 to 10 microns i J- -j
particle masses of 10~9 to 10-10 gram.
30
PARAMETERS
-------
Paniculate matter of smaller sizes must be considered as a practically permanent
component, because the fallout rate becomes negligible compared to the prevailing
velocities of vertical atmospheric convection. This category encompasses the range of
aerocolloidal particulates (aerosols). It could be expected that their sizes range down-
ward to those of the individual gas molecules, but present experiences indicate that under
normal conditions the lower limit is at about 0.1 micron (in terms of kinetic diameters),
since physically definable particulates in natural and polluted air decline rapidly in
number below 0.2 micron.1 This evidence appears to be contradicted by nuclear counts,
obtained by adiabatic expansion and subsequent condensation of HQ0 in a supersaturated
atmosphere, because these nuclear counts indicate in general much higher numbers of
smaller particles with sizes far into the millimicron range.2 The explanation for this
fundamental discrepancy may be that this procedure yields "snapshots" of unstable
molecular agglomerates with brief statistical existence. One may also postulate that two
principal factors cause the absence of stable particles in the two decades above the
molecular sizes: the rapidly increasing mobility of smaller particles, which promotes
coagulation to larger units and the chemical instability of extremely curved surfaces
(Kelvin relation) .s, * Both facts would predict a very brief existence for individual
particles smaller than 0.1 micron.
Experience has shown that the mass distribution over the total size range is in
first approximation constant, which means that the same air volume carries for instance
about 1000 times more particulates in the range between 0.2 and 0.3 micron than
between 2 and 3 microns.
At the same mass contribution the smaller fraction would present 10 times the
particle surface area and mobility of the larger fraction and hence would have about a
100-fold larger chance of interaction with reactive traces in the gas phase surrounding
them. This consideration indicates the inadequacy of defining this pollution parameter
by mass concentration. As a matter of fact, a truly representative method for assaying
the density of aerocolloidal matter appears still to be missing, for it is only partially
accomplished by procedures of light-scattering and impaction.
Another form of permanent contaminants, much easier to define, is those in molecular
dispersion, i.e. gaseous additives to the normal atmospheric constituents. Among these
one has to discriminate between two classes: the chemically inert (like C02 and CO)
and those that can react with other contaminants in the airborne state and form different
substances of potentially lesser tolerability. Typical examples of the latter type, though
certainly not the only ones, are SO,, and certain organic hydrocarbons. Their modifica-
tion by oxidation reactions requires the simultaneous presence of two additional factors:
photons in the spectral range of about 320 to 420 millimicrons, as supplied by sun
radiation to the lower atmosphere, and ozone (03) or oxides of nitrogen NOX.
According to the well-known pattern5' 6 NOX acts in photoactivation as a catalytic
activator of the atmospheric oxygen in the presence of gaseous traces that can be
subject to oxidation. This leads to the conversion of SO, to H2S04, which forms
particulates and thereby adds active condensation centers. Similarly NOX produces
a large variety of photochemical oxidation products (oxidants) from a variety of
hydrocarbons. Many so-resulting oxidants, particularly those of larger molecular
weight (C5), have the tendency to accumulate on existing particulates,7 thereby
causing their growth to many times the original size. This process in turn produces
visibility reduction and may cause synergistically intensified irritation, i.e., the typical
smog reaction. The particulates present at the time of the reaction thus play a
role similar to that of the condensation nuclei for water vapor in the formation of fog.
Goetz 31
-------
Detailed studies of such aerocolloidal matter have definitely indicated that these
particulate accumulations are not permanent, as evidenced by their gradual shrinkage
upon additional irradiation or moderate heat exposure.8 This suggests that the aero-
colloidal pollutants represent intermediate oxidation products, which are gradually con-
verted into more volatile stable end-products such as C02, H2O, and probably NH3OH.
Comparative investigations of aerosols present as "hazes" in unpolluted air (rural, torest,
ocean) have indicated this metastability of the particulates to be in a qualitative sense
a general property, the main differences from the smog aerosols being the much lower
concentrations and the absence of irritant capacity, most likely because of the different
nature of the hydrocarbon traces yielded by vegetative life.9
For the parameter definition of aerocollodial pollution the general pattern of forma-
tion and decay of photoactivated irritants postulates the importance of the initial
reaction rate. This pattern should thus depend, under identical conditions of irradiation
exposure, on the concentration ratio of the hydrocarbon reactant (HC) to the oxidation
catalyst (NOX), because an excess of the latter should accelerate the decay of the
oxidants into their final (neutral) oxidation products. Recent systematic tests10 appear
to support this prediction. The irritation response (which closely parallels the oxidant
production) was increased for the same NOX level with the concentration of HC, while
the increase of NOX beyond 1 ppm decreased the response for the same HC level.
Similarly, the oxidant formation depended over a wide range on the ratio HC/NOX, since
for the same HC level it decreased when more NOX was available and vice versa.
This brief resume, although barely outlining the complexity of these reactions, should
serve to enumerate and define the parameters pertaining to this type of environmental
pollution.
Because the reaction rate in highly diluted systems is slow, obviously time must be
available for the irradiation exposure of a given air mass this means that local condi-
tions of low atmospheric regeneration rates (inversion) are prone to produce such
reactions, particularly if sun irradiation is not frequently impaired. Other micro-
meteorological factors that can affect the photochemical chain reaction rate are, as to
be expected, temperature and relative humidity.11, I2
Depending on such local conditions, the presence of photochemical activators (NOX)
becomes of primary significance whenever they coexist with hydrocarbon traces. In the
enormous variety of this parameter the molecular structure and weight of the organic
compounds appears to be of decisive importance; about these factors much too little is
known. It is certain, however, that unsaturated compounds of olefinic and/or aromatic
structure represent reaction partners of high avidity. It has also been shown that the
tendency toward accumulant formation on existing nucleating particulates, i.e. aerosol
formation, increases for analogous hydrocarbons (e.g. aliphatic olefins) with their chain
length and the asymmetry of their double bonds.13
In addition to these reactions, trace pollutant aerosol exists, no doubt, as an in-
dependent parameter. The particle number, size-distribution, and type (submicron
particulates) characterize the aerosol particles that serve as potential centers of
reactant accumulation and reaction-promoting catalysts. Aerosols characterized in this
manner have received rather little attention in the past, even though the exhaust of auto-
motive traffic represents a prolific source of particulate matter (lubricants, etc.) for the
metropolitan air mass. This is amply evidenced by the fraction precipitated on the
surface of the traffic lanes and by the benzene-soluble components of filter deposits.
In a certain sense, the very well-known SO, parameter may be considered in the
32 PARAMETERS
-------
same category. While the irritation records show no statistical coincidence with the
S02 content,14 S02 is well known as a cause of aerosol formation, because its photo-
chemical oxidation by activating agents produces aerocolloids of high stability and thereby
reactive nuclei for the organic accumulants.
The S02 parameter requires critical coordination with the coexisting pollutant
components in the air mass under consideration. As an isolated pollutant in the absence
of reaction partners, S0p does not appear to be of major significance.
Finally, another parameter type that appears worth mentioning is sensed by the
nuclei-counting devices, which indicate the number of centers for the condensation of
water vapor under supersaturation artificially induced. Experience shows that a correla-
tion between permanent particles and the frequency of such centers is difficult to
establish, if it can be established at all. On the other hand, the nuclei counts seem
to indicate the potential reactivity of a system prior to irradiation exposure, i.e. prior
to the onset of photoactivation, which gradually declines with the completion of the
reaction, i.e. the depletion of reactants. Thus it may be possible to develop from nuclei
counting a new parameter that describes the potential activity of the air mass and
therefore the likelihood that its chemical composition will be altered by subsequent
interreaction of the components.
In summary attention may be drawn once more to the aerocolloidal phase, although
it is but an insignificant mass-fraction of the biosphere (>- 10~7) and thus is not yet
accessible to specific chemical analysis, it represents probably a most significant environ-
mental parameter. This parameter is unique, since it does not refer to a defined substance
but rather to the "micromorphological" constitution of the air we breathe.
Originating continuously from atmospheric interaction with the planet surface,
these submicron particulates modify the statistical pattern of the gas phase by their
highly dispersed surface area: they act as centers for H20 condensation (fog and cloud
formation), they also can accumulate more permanent reaction products (hazes), radio-
active molecules, etc., and they are important in the photochemical reaction pattern of
organic contaminants (smogs). Metropolitan activity changes locally the natural aero-
colloid supply, largely by number and chemical constitution. Growth of such nucleating
particulates lowers the visibility range by light-scattering or attenuation. The statistical
coincidence between irritation and certain aerosol patterns emphasizes the importance of
the aerosol parameter. Its detailed definition and application to environment control
appears as a primary task of research and development.
REFERENCES
1. A. Goetz and 0. Preining, The Aerosol Spectrometer and its Application to Nuclear
Condensation Studies, Am. Geophysical Union "Physics of Precipitation", Geo-
physical Monograph No. 5, 164-182, 1960.
A. Goetz, The Physics of Aerosols in the Submicron Range, Internal. Symposium
on "Inhaled Particles and Vapours", Oxford, England, Pergamon Press 1961,
295-301.
2. C. Junge, Atmospheric Chemistry, Advances in Geophysics, 4, 1-108 1958. Acad.
Press Inc., N. Y.
H. W. Georgii, Probleme und Stand der Erforschung des atmospharischen Aerosols,
Berichte des Deutschen Wetterdienstes No. 51, 1958 (A comprehensive bibliography
of aerosol literature.)
Goetz 33
-------
3. G. Zebel, Zur Theorie der Koagulation elektrisch ungeladener Aerosole, Kolloid-
Zeitschrift, 156, 102-107, 1958.
G. Zebel, Zur Theorie des Verhaltens elektrisch geladener Aerosole, Kolloid-
Zeitschrift, 157, 37-50, 1958.
4. C. Orr, Jr., F. K. Kurd, and W. J. Corbett, Aerosol Size and Relative Humidity,
J. Colloid Science, 13, 472-482, 1958.
A. Goetz, 0. Preining, and H. J. R. Stevenson, Synergistic Properties of Aerosols,
Preliminary Report U.S.P.H.S. Grant, Sept, 1958-Dec. 1960.
5. A. J. Haagen-Smit, C. E. Bradley, and M. M. Fox, Ozone Formation in Photo-
chemical Oxidation of Organic Substances, Ind. Engin. Chem., 45, 2086-2089, 48,
1884-1887, 1956.
6. P. A. Leighton, Photochemistry of Air Pollution, Academic Press, New York, 1961.
7. A. Goetz, W. Stoeber, T. Kallai, U.S.P.H.S. Grant Progress Report RG 6743, 1961.
A. Goetz and R. Pueschel, Photochemical Aerosol Formation as a Nucleation
Phenomenon, A.C.S. Meeting, New York, September 1963.
8. A. Goetz, O. Preining, and T. Kallai, The Metastability of Natural and Urban
Aerosols, Rev. Geofisica Pura e Applicata Milano, 50, 67-80, 1961.
9. F. W. Went, Organic Matter in the Atmosphere, and its Possible Relation to
Petroleum Formation, Proceedings of Nat. Academy of Sciences, 46, 212-221, 1960.
H. W. Georgii, Nitrogen Oxides and Ammonia in the Atmosphere, J. of Geophysical
Res., July 1963 (in press).
10. M. W. Korth, A. H. Rose, Jr., and R. C. Stahman, Effects of Hydrocarbon to Oxides
of Nitrogen Ratios on Irradiated Auto Exhaust, Part I, Ann. APCA Meeting,
Detroit, Mich., June 1963 (Preprint No. 63-19).
11. J. N. Pitts and J. H. Sharp, Effects of Wavelength and Temperature on the Primary
Processes of Nitrogen Dioxide, presented at 142nd Meeting, Am. Chem. Soc.,
Atlantic City, N.J., September 1962.
12. L. A. Ripperton and W. J. Jacumin, Effect of Humidity on Photochemical Oxidant
Production, presented at Los Angeles, Calif., A.C.S., Div. of Water and Waste
Chemistry, April 1936 and 56th Ann. Meeting APCA, Detroit, Michigan, June 1963.
13. A. Goetz, Methods for Measuring Particle Composition in Photoactivated Aerosols,
56th Ann. Meeting APCA, Detroit, Mich., June 1963 (in press).
14. Technical Progress Report of the Los Angeles County Air Poll. Control District 1962.
34 PARAMETERS
-------
Dr. William J. Youden
Consultant, Applied Mathematics Division
National Bureau of Standards
U. S. Department of Commerce, Washington, D. C.
SUMMARY
The effective use of statistical analysis in sampling work is closely tied up with the
use of statistically designed sampling plans. Good sampling plans can be designed only
by people closely familiar with the features that characterize the area to be sampled.
To get the most out of any statistical consultant, we should supply him with results
obtained in a small exploratory survey that will give quantitative information regarding
the sampling and testing errors. Examples of several basic sampling schemes are
presented.
SAMPLING AND STATISTICAL DESIGN
INTRODUCTION
The problem of obtaining a sample that will represent adequately an area or popula-
tion of interest appears in many forms. Many census and public opinion studies are
based on a small fraction of the population. Evaluation of an ore body is made by use
of a limited number of borings. The assessment of import duties on wool is based on
samples taken from only a small fraction of the number of bales in the shipment. The
quality control of manufacturing processes rests on the inspection of samples taken during
manufacture. Different as these settings appear, certain considerations are common to
all of them.
One of the most basic, and frequently violated, principles is that of the principle
of random selection of the samples. The price of adopting a convenient method of
sampling at the sacrifice of a random selection of samples is to vitiate statistical evalua-
tion of the data and to modify the setting of probability limits to the estimates obtained.
Whenever systematic sampling is employed on the grounds of convenience, there is no
escaping the necessity for first demonstrating that the results check those obtained by
random sampling.
SOME GENERAL CONSIDERATIONS REGARDING SAMPLING
By and large there would be no sampling problem if unlimited time and resources
were available. But time and resources are usually limited, and the investigator faces
the challenge to use these limited resources as effectively as possible.
Almost always the element of cost enters into the problem. The inherent value of
the samples or the cost of getting them, together with the cost of testing the samples,
influence the choice of the sampling scheme. Often provision must be made for keeping
tab on the sampling. And always there is the inescapable need first, for a careful prior
specification of just what is the region of interest, and second, for a prior decision on
the statistical procedures that will be employed. Any sampling investigation that is
undertaken with the idea of settling these problems after the sampling is done might as
well never be started. In some cases a statistician may be consulted. I wish I could
promise you that this was an easy way to obtain a sampling scheme appropriate for the
Youden 35
-------
investigation at hand. There is really no easy way, because much depends on the
sampling difficulties of the area. Accessibility, uniformity or heterogeneity of the area,
and pronounced natural subdivisions within the whole area all exercise a considerable
effect on the choice of the sampling plan. The statistician, lacking such detailed informa-
tion, may make some useful general suggestions but he may miss opportunities to fit
the sampling scheme to the problem. So really the statistician and investigator should
work together, perhaps even explore together the geography or other relevant features
of the region. The investigator must pass on the feasibility of suggestions. He should
be prepared to give some idea of how accurate he wants his answers to be and he should
have available a limited amount of preliminary data to provide information on both sam-
pling and testing errors. All of this is really a minimum for devising a program tailored to
fit the immediate problem. This preliminary information often returns many-fold any
outlay in time and effort expended before undertaking the final program.
EXAMPLES OF EXPERIMENTAL DESIGN IN SAMPLING
Whole volumes have been written on the topic of sampling. No good would be
served in trying to abstract works devoted to the theory and practice of sampling. The
available time offers the opportunity to discuss some examples of sampling problems,
particularly from the viewpoint of the statistical design of experiments. The design of
experiments grew up in the setting of investigations in which the experimenter had many
of the important variables virtually under complete control. Thus an experimenter
studying the effect of light on the growth of plants could construct an isolated universe
wherein he could control the quality, the intensity, and the duration of the light; the
temperature; and many other variables that he deemed relevant to his problem. Here
the scientist virtually creates the population that he wishes to study, and of course his
conclusions are pretty well restricted to this population. Controlled experimentation
permits an enormous gain in efficiency compared with investigations of natural popula-
tions that necessarily have to be sampled.
There are many problems that cannot be suitably simulated on a laboratory scale,
and there is no recourse from the necessity of studying on location, as it were, the
phenomenon of interest. The effort is then directed to an examination of the area with
a view to ascertaining the actual state of affairs that exists there. Generally speaking
this involves a series of point inspections on samples taken at certain points in the
region of interest. One sampling technique is to lay down a grid of points or plots,
like a checkerboard. The area is divided into rectangles by an equal number of north-
south and east-west lines. The spacing between the lines is dictated by the number
of samples that can be collected and tested with the assigned resources. Many such
programs have been followed. The results permit easy visual representation on a map
by drawing contour lines. Often duplicate samples are taken at each point in order
to throw light on the adequacy of the sampling technique.
Sometimes little is known about a region, and there is a real necessity to sample
the region in such a way that all concerned would accept the sampling as fair. As an
example I mention a rather large rectangular slab of concrete on which a large housing
unit was to be constructed. The question was raised as to whether or not the slab
met specifications, and this could be determined only by boring cores to be tested.
Coring is expensive, and even if it were not, no one wants to honeycomb the foundation
with holes. The obvious approach was to lay out crisscrossing lines, say three in each
direction and take cores, nine in all, at the points of intersection. This is the checker-
board scheme just mentioned.
36 SAMPLING AND STATISTICAL DESIGN
-------
-o
-o
C
E
F
A
B
H
I
G
D
D
G
A
B
I
E
C
F
H
E
D
H
F
C
B
A
I
G
F
I
G
E
H
D
B
A
C
G
F
I
C
D
A
H
B
E
H
A
E
I
F
G
D
C
B
I
H
B
G
A
C
E
D
F
An alternative approach was suggested that introduced a random element and
included parts of the slab closer to the edges. The slab was divided into a 9 by 9
rectangular checkerboard. For ease of presentation here the 81 small rectangular sub-
areas are presented in the form of a Latin Square:
A B
B C
C D
D H
E G
F I
G F
H E
I A
The first nine letters of the alphabet are used to designate the nine subareas in the
top row. The same nine letters are used in every row subject to the restriction that when
all nine rows have had letters assigned there will be a complete set of letters in every
column, i.e., crosswise of the slab.
The Latin Square has been used in agricultural experimentation for 40 years. The
idea is that if n, plant varieties (or fertilizers, or sprays, or other items under test)
designated A, B, . . . , G are to be intercompared, it is essential to give every one a fair
chance at the available environment. The available area is subdivided into n rows
and n columns. Any treatment, such as C, for example, samples every row and every
column, and this puts it on a par with any other treatment. It was found that the accur-
acy of the comparisons was greatly enhanced by this device of arranging for equality
of opportunity for the several treatments. This means that, if in fact all treatments are
identical, the average for any one letter should check the average for any other letter.
Therefore, from the viewpoint of anyone trying to obtain a representative value for the
area, one letter should be as good as any other. Consequently it should suffice to
sample just the subareas associated with a particular letter chosen at random. There
are a very large number of ways of constructing Latin Squares so that a further element
of randomness is also present.
In the actual cement slab problem both interested parties were perfectly willing to
abide by the result of samples selected in this manner. This is the real test that the
sampling scheme is inherently fair. If more samples are needed, a larger square could be
used. Rather than enlarge the square it would be better to choose a second letter, also
at random. If the averages for two letters, say D and H, check each other, there is
Youden
37
-------
convincing evidence that the sampling is satisfactory. This suggests that the size of the
Latin Square should be % or Ys or other fraction of the total number of samples
contemplated. In fact if samples are collected from three letters, the test results on
samples from two of the letters may check so well that no tests are needed on samples
for the third letter. Incidentally, the samples of any one letter should be identified aa
to the row and column sampled. Labeling the sample locations makes it possible
to compare the average of the samples in the north half of the square with the average
for the south half. A similar east-versus-west comparison may be tried.
Twenty-five years ago I was enlisted by a soils man in a study to ascertain the
variation in pH of a particular soil type. I devised a sampling scheme that has some
interesting features. The approach was simple enough. Two samples were taken ten
feet apart. From the midpoint of a line joining these two spots a distance of 100 feet
was paced off in an arbitrary, i.e., random direction. Here a second pair of samples
was taken. Starting midway between the two pairs a distance of 1000 feet was paced
off and another matching set of four samples taken. This set of eight samples was
designated a ''station.'' Several such stations were established at intervals of two miles.
Table I (Table III in the original publication2), shows how the difference between samples
depended upon their separation. The interesting thing about this early publication was
the noting that, given such preliminary data, a more efficient allocation of the samples
could easily be devised.
Table 1. Difference in pH of Duplicate Samples of Culvers Gravelly Silt Loam.
Distance between
duplicate samples
10 feet
100 feet
1000 feet
1-3 miles
0-2
Av. diff.
0.14
0.18
0.26
0.36
inch layer
Max. diff.
0.44
0.84
0.69
1.32
2-6
Av. diff.
0.11
0.20
0.25
0.28
inch layer
Max. diff.
0.49
0.53
0.81
1.05
The table shows plainly that samples taken close together agreed more closely than
samples separated by a considerable distance.
Sometimes the heterogeniety of an area may be quickly demonstrated by a succession
of paired samples strung out in line like this:
Pair 1 2 3 4 --- n
The two samples in each pair of results can be added to give a sum u and subtracted
to give a difference d. If there is no trend the a's and the d's should have the same
variance. If there is a trend the a's should vary more among themselves than the d's.
We may calculate 2d2 and 2ja2 (2a) 2/n with re and n-1 degrees of freedom, respectively.
Divide 2^d2 by «, call the quotient D. Divide the other quantity by (n-1) and call the
quotient A. Then the variance ration, F = A/D, may be evaluated by the standard
statistical table for F. Large values for F are evidence of heterogeniety.
A somewhat similiar test may be made to ascertain whether a series of single samples
taken in sequence (either along a line or in time) exhibit only random variation. Let
the observed results be x^ x2, ..., xn. Obtain the successive differences d1 == x - x ;
d2 = x, - x3; . . . , da_1 =xn_1 -xn. Square and sum these differences: 5d2 = D2. Calcu-
late S2 = 2 (* x) 2. That is, we take the difference between each x and the mean x
38 SAMPLING AND STATISTICAL DESIGN
-------
and sum the squares of these differences from the mean. In the absence of any trend
the theoretical expected value for the ratio D2/S2 is exactly 2. If there is a trend along
the space or time line, it is natural for adjacent points to be more alike than separated
points. Here we would expect the ratio D2/S2 to be reduced below 2 because two
successive samples give only a small chance for the trend to manifest itself. Tables for
evaluating this ratio are given by Bennett1 and in an excerpt from them in Reference 3.
If the ratio drops to unity with as few as ten samples, there is evidence at the conventional
5 percent level for a trend.
One widely used sampling device is that of stratification. If the entire region may
clearly be subdivided into subregions that are relatively homogeneous within each region,
then the allocation of samples can take advantage of this feature with a decided improve-
ment in the formation obtained for a given number of samples. This procedure is well
known and will not be discussed here.
I cannot hope to provide even a preliminary list of experimental designs that ap-
pear suitable for sampling. There is much to be said for giving the imagination free
rein in designing a program instead of limiting oneself to a few standard approaches.
REFERENCES
1. C. A. Bennett, Ind Eng. Chem. 43, 2063, 1951.
2. W. J. Youden, Contrib. Boyce Thompson Institute 9, 59, 1937.
3. W. J. Youden, Science 120, 627, 1954.
Youden 39
-------
Dr. Loren E. Bellinger
Assistant Professor, Department of Aeronautical
and Astronautical Engineering
Ohio State University, Columbus
SUMMARY
Transducers are devices that can be actuated by waves from one or more trans-
mission systems or media and that can supply related waves to one or more other
transmission systems or media. Transducers are used to transform one physical
phenomenon to another; for example to convert pressure disturbances to related electrical
signals.
Major emphasis is placed on transducers used to measure pressure, temperature,
and flow rates of various fluids. Transducers used to ascertain chemical composition
are discussed also.
The basic principles of operation of the various transducers are given, and the
inherent limitations and sources of error are discussed.
TRANSDUCERS
INTRODUCTION
Before any discussion of transducers, their limitations, and their inherent errors,
it is well to consider exactly what is meant by "transducer."' By one very general defini-
tion, transducers are devices that can be actuated by waves from one or more transmission
systems or media and that can supply related waves to one or more other transmission
systems or media.
On the input side a transducer can convert, for example, a nonelectrical quantity
into an electrical signal. A specific example is that of pressure actuating a strain-gage
type of transducer that then delivers an electrical output signal, which is some function
of the input pressure, to an amplifier or some type of "black box." On the other end of
the "black box," an output transducer can change the electrical signal into a non-
electrical quantity, such as the position of a pointer on a meter.
In this paper primary emphasis is on input transducers. Measurements of pressure,
temperature, flow rates, and chemical composition are considered. Transducers are
needed to sense these primary variables and to change them into corresponding electrical
signals, ordinarily, so that appropriate measurement or control can be effected.
Ideally, transducers should respond instantaneously. That is, for a step-function
input, the output signal should follow the input variable without distortion of amplitude,
frequency, or phase. This concept, of course, is an ideal. No transducer satisfies these
requirements over the complete range of feasible input variables. Over a particular,
limited range many transducers follow the input variations quite well. Thus, the time
element must be considered when transducers are used for measurement.
In some cases there is more concern about the rate of change of the variable than
about the magnitude of the change. When the transducer does not follow the input
variable exactly, a lag in response results. Lag is the dropping behind or retardation of
the output signal in comparison to the input signal. Although the lag of the system
may be high, the over-all error could be small. That is, if the error is considered as
Bellinger 41
-------
the deviation of the actual output signal from the ideal output signal, integrated over a
long period of time, then the error could be extremely small even though the lag is
great. If dynamic error is considered, however, then at any one particular time the
error can be very high because of lag. Two types of error should be considered: static
error and dynamic error. In a sense, the static error is a deviation of the output signal
from the true value of a static variable. Obviously, static error should be held to a
minimum. Dynamic error is the deviation of the output signal from its true value when
the input signal is varying.
One other major point that should be considered is reproducibility. For repeated
measurements of one fixed value of the input variable, reproducibility is a measure of
how closely the same output value can be obtained. A high degree of reproducibility
is most desirable.
There are many different types of transducers. So that some of these may be
examined in detail, the scope of this paper has been limited. The primary transducers
considered herein are those utilized to measure pressure, temperature, flow rate, and
chemical composition.
PRESSURE
INTRODUCTION
In our world of expanding technology, pressures, both static and stagnation, must
be measured in flow fields to satisfy the needs of various industries, scientific and
engineering laboratories, and the armed services. Many times the measured pressure
can be converted into suitable signal form so that automatic control and regulation
systems can be employed advantageously. Before various pressure transducers are
discussed, fluids and fluid properties should be clearly understood.
"Fluid" is a comprehensive term that includes two of the three basic categories
into which all physical materials are classified generally: solids, liquids, and gases. A
fluid can be either a gas or a liquid. Vapors, if considered as a separate classification,
are fluids, too. Selection of a specific type of pressure transducer for a particular
application depends upon such factors as range, accuracy, frequency response, location
of the detector and indicator, reliability, simplicity, availability, fluid temperature, fluid
velocity, fluid corrosiveness, adaptability to automatic control, and cost.
The pressure of a fluid, p, is the force per unit area exerted by the fluid on each
bounding surface. Within a flowing fluid the pressure may change from location to
location because of friction, expansion, contraction, and so forth. At any particular
point in a fluid at rest, the pressure acts equally in all directions. Furthermore, the
pressure force acts normal or perpendicular to each surface.
In pressure measurement, the force that acts on a known area must be ascertained.
In general, pressure is measured by use of two different scales. One scale is absolute
in that it is the actual total pressure that acts on a body or surface. When this scale
is employed, zero total pressure directly implies an absolutely perfect vacuum. The other
scale is relative in the sense that only the pressure above or below the local atmospheric
pressure is measured. The barometer is an example of an absolute pressure gage; the
conventional Bourdon gage ordinarily uses a relative scale. Analytically, the absolute
pressure is the sum of the gage pressure in the vessel and the local value of the
atmospheric pressure.
42 TRANSDUCERS
-------
MANOMETERS
One of our oldest transducers for the measurement of pressure is the common
liquid-column manometer, in many respects the simplest, most direct, and most accurate
of all of our pressure-measuring instruments. Unless special manometers are used, the
pressure range that can be covered is not great. At very high pressures, manometers
become unwieldy; however, they can be used to measure small differential pressures at
very high line pressures with great accuracy. Generally, the term manometer is applied
universally to a pressure-measuring device that uses a liquid as the measuring medium.
There are two principal types of manometers: the U-tube and the well-type.
The simplest manometer consists of a tube of glass or some other transparent
material that is bent into the shape of a U. Both legs are filled approximately half-full
with a liquid. A modification of the conventional manometer is the inclined tube, in
which the manometer tubes are inclined from the vertical for detection of smaller pressure
differentials. The fluids normally used for manometers are mercury, oil, or water;
different fluids can be used to achieve special effects.
When one looks across the fluid in the manometer, it is obvious that the surface of
the liquid is not flat. With some liquids the curvature, the meniscus, is considerable.
Liquids that wet the wall of the tubes produce a concave meniscus. Liquids with high
surface tension, such as mercury, do not wet the tube wall and produce a convex
meniscus. In determinations of pressure difference, a fixed position of the meniscus must
be used from one reading to the next. The top of the meniscus is usually read on
mercury-filled manometers..
When the U-tube manometer is used, two measurements must be made: the position
of the fluid in each leg must be determined. Well-type manometers require the reading
of only one leg. In effect, the well-type manometer is a U-tube manometer in which the
volume of the second leg is very large. Therefore, the fluid level in the well does not
change appreciably as the fluid moves up the vertical column. Corrections in the
manometer scale can be made to account for the slight change in elevation of the
well-leg.
When manometers are used for pressure measurements, values are commonly given
in inches of water or inches of mercury, which are units of pressure. When a column
of liquid is subjected to a pressure, the equation describing the equilibrium condition is
where F represents the forces. In other words, the pressure force acting on the liquid
column is balanced by the force arising from gravity acting on that portion of liquid
above the meniscus of the lowest leg. By Newton's second law of motion, it can be
shown that the difference in pressure acting on the fluid column is
AP = phg/gc
Since g equals gc numerically, the equation for the balance of forces can be written as
AP = pt
Therefore, for a specific type of manometer fluid, the pressure can be expressed in terms
of the height of this particular column of liquid.
To change the pressure range without increasing the height of the manometer
tubes, the manometer fluids can be changed. For example, if the pressure range of a
Bellinger 43
-------
water-column manometer must be increased by a factor of 12, the resultant water column
could be unreasonably high. If mercury is substituted for water, approximately the same
height of manometer tubing can be used because of the change in specific gravity.
Since the density of a liquid is a function of temperature, the temperature must
be given when pressure is quoted in terms of liquid head. A reference temperature of
0°C is used commonly for mercury, and 3.9°C (39°F, the temperature at maximum
density) for water. Although the pressure may not have been measured at these specified
temperatures, the height of the column of liquid at the desired reference temperature can
be calculated by use of the ratio of densities at the actual temperature and the reference
temperature. Since the pressure is the same at either temperature, the height of the
liquid column at the reference temperature is given as
_ P actual
nref -- n actual
It is important to consider that the manometer measures pressure in fundamental
units of length and mass. Few other pressure-measuring devices are so basic. By
use of various types of oils (silicone, octoil, etc.), the density can be decreased greatly
to achieve a greater column height for a given pressure change. "When the pressure
change is small, very low-density liquids are employed. One of the difficulties with this
type of arrangement is that the meniscus of the oil often is difficult to determine.
Capillary effects can cause error when the diameter of the manometer column is
too small. In general, the tube diameter should be not less than 10 millimeters for
mercury. For water and other fluids that wet the surface of the tubing, the diameter
can be somewhat smaller.
With some varieties of well-type manometers a movable well holder is used so that
the position of the liquid in the vertical column can be changed. The meniscus of
ordinary dater in an manometer is difficult to detect. By the addition of fluorescent dyes
such as fluorescein or eoscein together with a mild detergent to facilitate surface wetting,
the available light is collected and concentrated at the meniscus, which is thereby easier
to detect. These additives do not stain the glass tube as do ordinary inks or dyes. Some
special manometer fluids, in particular some of the oils, must not be used for measuring
pressures of oxygen or oxidizing compounds because of possible chemical reactions.
Special devices, such as check valves and traps, can be used with manometers to prevent
the liquid from being blown over when excessive pressure is applied accidentally.
PITOT AND PITOT-STATIC TUBES
Generally, almost any combination of mechanical tubes arranged to determine static
or stagnation pressures is called a Pilot tube. Strictly, Pilot used the lube lo determine
stagnation pressures only. Actually there are three basic types of lubes:
1. Pilot tube a tube, generally cylindrical, poinled directly upstream to measure
the stagnation pressure.
2. Static tube a square-ended tube whose longitudinal axis is perpendicular to
the slream lines of the fluid flow, to sense the static pressure.
3. Pilot-static tube a combination, usually coaxial, of a Pilot tube and a slatic
tube, used to measure stagnation and static pressures at one local region
44 TRANSDUCERS
-------
In many applications of Pilot-static tubes, the difference between the two pressures
is determined directly by a differential-pressure indicator. The difference between the
stagnation and static pressure can be measured more accurately with a differential-
pressure manometer or similar device than with two independent sensors used to measure
each pressure separately. Pilot, static, and Pitot-stalic tubes are used extensively to
measure static and stagnation pressures in wind and water tunnels, on aircraft and
marine vessels, and in ducts lhat carry flowing fluids.
The relationship of static, or stagnation pressures, or bolh, lo ihe olher variables makes
it possible to calibrate the dials of Pilot-type instruments in terms of the desired variables
instead of ihe pressures thai are aclually measured. Also, by use of suitable linkage
elemenls in ihe indicator, functions related to the ratio of slatic to stagnation pressures
can be delermined indirectly.
The velocity of a fluid can be obtained by use of a Pitol-slalic lube from ihe
relation
2 AP
where u is ihe average velocity and Ap is ihe difference belween ihe stagnation and
static pressures.
Pilot lubes are fabricated in many differenl physical configurations, each of which
has inherent properties thai musl be understood, for a particular application. Ordinarily
il is assumed lhal the flow field is one-dimensional, thai is, ihe velocity is a function of
one dimension only. When the flow is almost one-dimensional (where the velocity varia-
tions in the other two mutually orthogonal directions are small), the error involved by
assuming one-dimensional flow is negligible for most applications. Considerable caution
musl be used, however, when ihis assumption is made.
If the incompressible, one-dimensional form of Bernoulli's equation for the dynamic
pressure is solved, ihe resull is
Dynamic Pressure = q = % pV2 = p° - p
where q is dynamic pressure and V is ihe average (veclor) velocity. When the flow
is not one-dimensional, the kinetic energy lerm q can be wrillen
q = A Va pV2
where A is 3- correction faclor and V is ihe average velocity over the flow area. The
average velocily V is a veclor quantity composed of ihe veclor sum of the three ortho-
gonal velocity components, u, v, and w in ihe x, y, and z direclions, respectively. Note
lhat q is ihe kinetic energy of the fluid per unit volume. If somelhing is known aboul
ihe variation of the velocity as a function of coordinate system, then a theoretical value
for A may be calculated. If this value can be found, the fluid velocity can be delermined
more accurately from ihe measuremenl of dynamic pressures in multidimensional flow
fields.
Suppose, for example, lhal a fully established laminar flow field exisls in a circular
pipe. The velocily dislribulion is parabolic. Setting up and evaluating ihe various integrals
results in a value of 2 for A- Thus, ihe velocities calculated from dynamic pressure
measurements would be too large by the faclor V2-
Bolliiiger 45
-------
BOURDON TUBES
The Bourdon tube is usually elliptical in cross-section. Ordinarily it is coiled into
a spiral or helix or into a C-shape. In any of the many variations of Bourdon tubes,
the free end of the tube moves when pressure is applied internally, and the tubes tend
to straighten when the internal pressure is increased. The general tendency is to form
a straight, cylindrical tube. When first-order effects are considered, the motion of the
free end of the Bourdon tube is directly proportional to the change in internal pressure.
Therefore the output function of the device is essentially linear.
The pressure range for Bourdon tubes is from 30 inches of mercury vacuum to
100,000 pounds per square inch pressure. Although many improvements have been
made on the basic Bourdon tube, the principle is still the same. Round hollow tubes of
suitable material and dimension are flattened to give an elliptical cross-section and then
bent into the shape of a C. A tip is sealed onto the free end and the other is connected
to a socket that permits connection to the pressure source. With suitable linkage elements,
a rack and pinion, and a rotating pointer, the Bourdon tube deflects and causes the
pointer to move as pressue is applied.
For a good Bourdon-tube pressure gage, the tube material must be of high quality,
with good spring characteristics. Errors can arise from hysteresis in the metal used in the
Bourdon tube, from poor material used in the linkage element or in the rack and pinion
assembly, and from friction. A diaphragm type of seal can be applied to separate the
Bourdon tube from corrosive fluids. Non-corrosive fluids within suitable temperature
limitations can be connected directly to the Bourdon tube. If the pressure is pulsating,
precautions must be taken to prevent excessive wear or damage to the rack and pinion
of the Bourdon-tube gage. Usually some type of pulsation dampener is used to smooth
out the pulsations. The inertia of the system limits the frequency response to a low value.
The so-called master or test gages are those that have been fabricated to very high
standards of accuracy and can be used for calibration of other gages. It is not uncommon
for test gages to be accurate to within 0.25 percent. Temperature variations tend to
affect Bourdon-gages too. In some high-precision gages, a bimetal, compensated move-
ment and a hand-calibrated dial are utilized. Accuracy of 0.2 percent or better can be
obtained.
RESISTANCE GAGES
Pressures can be measured by transducers with pressure-dependent resistance char-
acteristics. Some variable-resistance pressure transducers have movable contacts; others
use continuous-resolution devices. Often the pressure force is converted into an electrical
signal by the stretching or compression of a wire (e.g., strain-gage type transducers)
or by movement of a sliding contact across a coil of resistance wire, which changes the
electrical resistance in the output circuit. Numerous mechanical designs are employed:
the resistance element may be a coiled wire, a tapped resistance wire, or a continuous
single wire. Carbon strips, an electrolyte, or some liquids, such as mercury, can be
employed.
CAPACITANCE GAGES
By use of a movable and fixed metal plate, a variable capacitance gage can be
utilized to measure pressure. When a pressure is applied, thj capacitance is changed
because the distance of separation between the two plates is modified. When a suitable
46 TRANSDUCERS
-------
AC carrier voltage is applied across these plates and fed into an appropriate circuit
(usually some form of bridge circuit), an output signal that is a function of the pressure-
can be obtained. Capacitance gages can yield fairly good transient response. This type
of transducer does suffer from temperature effects unless special low-expansion metals,
such as invar, are employed. If the capacitance probe is water-cooled, the effects from
temperature changes can be minimized.
The air gap between the movable plate and the fixed plate is small, for example,
0.003 inch. The displacement during application of pressure is approximately 1/10
that value. A dielectric other than air for example, mica may be substituted
between the plates. For reproducibility of data, the two plates must be kept parallel.
Some special capacitance probes have natural frequencies as high as 500 kc, but their
use is limited.
PIEZOELECTRIC GAGES
When a force is applied to certain types of crystals along specific planes of stress,
the crystal produces an electrical charge. When the crystal is appropriately coupled
to the pressure system, an electrical output signal can be obtained simply by allowing
the deformation force (pressure) to act on the crystal. Electrical contacts are made
to the crystal, and the delivered charge, which is a function of the pressure, is measured.
Typical crystals are quartz, tourmaline, ammonium dihydrogen phosphate, barium titanite,
and Rochelle salts. Quartz crystals, either natural or synthetic, are often used because
they allow very low electrical leakage and permit the measurement of slowly varying
pressures.
The output signal must be fed into an extremely high-impedence amplifier to
decouple the crystal effectively. The charge produced per unit pressure is low. Usually the
input resistance to the amplifier ranges high in megohms. Some type of electrometer circuit
is employed ordinarily, in which case the input resistance is usually higher than 109
ohms. The quartz pressure transducers are very useful in measuring transient pressure
waves that have very fast rise times. With special care it is possible to measure accur-
ately the transient pressure of a shock wave having a rise time in the microsecond region.
About the only practical method of displaying the output from such a device is to use
an ocilloscope with high-frequency response. The trace can be photographed with a
camera.
TEMPERATURE MEASUREMENTS
INTRODUCTION
Temperature is an intensive and not an extensive quantity. No unit temperature
interval can be applied successively to measure any other temperature interval, as can be
done in the measurement of such quantities as length or mass. The size of the degree
on one part of the scale, no matter how well defined, can bear no relation to the size
of the degree on any other part of the scale.
Temperature scales based on different thermometric substances or thermometric
properties differ fundamentally. The difference between two scales that differ only in
function chosen is superficial because the conversion from one scale to another is merely
a matter of calculation. If the same basic fixed points are used, the scales will necessarily
agree at these points but not at others. For example, two scales, both based on the
apparent expansion of mercury in glass, will differ unless the type of glass used is
identical.
Bolliiiger 47
-------
LIQUID-IN-GLASS THERMOMETERS
One of the simplest temperature-measuring devices is the common liquid-in-glass
thermometer, in which mercury is often used. The basic principle is the use of the
volumetric expansion of mercury as a function of temperature as a means of indicating
temperature. The glass thermometer or glass tube has a bulb formed by a glass en-
velope, which contains the mercury deposited in a metal or glass well at the bottom.
When'heat is applied to the thermometer, it is transferred through the wall into the
mercury. As the mercury expands the column rises in the capillary tube.
Temperatures can be measured by calibration of the position of the mercury in
the glass tubing as a function of temperature. The expansion and contraction of the
glass envelope must be considered when the calibration marks are etched on the glass.
Some thermometers are made for partial immersion, usually 3 inches, or for total
immersion. The scales on the tubes ordinarily are calibrated for one or the other
condition. By shaping of the glass stem, magnification can be incorporated for easier
determination of the position of the mercury meniscus.
The space above the mercury column generally is filled with pure nitrogen under
pressure. The gas above the liquid mercury tends to minimize breaking of the mercury
thread when the thermometer is handled roughly. Also, the increased gas pressure above
the mercury raises its boiling point.
Pointing of quality glass thermometers consists of placing file marks on the stem.
A five-point thermometer is calibrated at five fixed points.
It is essential that the bore of the glass thermometer is uniform and that the mercury
is pure. Readability is improved by use of color contrasts such as black, yellow, ruby
glass, white glass, etc. An enormous ratio of bulb volume to bore volume gives good
precision but not necessarily good accuracy. The finest test-grade glass thermometers can
be read to within approximately 0.02° with engraved graduations of 0.1°.
Since glass ages regardless of precautions, the stability of a themometer is affected
somewhat with age. Elasticity is the primary property of concern. Exposure of the bulb
to much higher or lower temperatures than those for which the thermometer was designed
can upset the aging process of the glass. Manufacturers allow for the aging process in
design of the thermometer.
The response time of the liquid-in-glass thermometer is one of the longest among
ordinarily used temperature-measuring devices. Calibration can be affected if the bulb
volume changes with time. The change is generally less than 0.1°C for a good grade of
glass if it has not been used above 150°C. Hysteresis can be noted when thermometers
are heated to or above 150 °C and then cooled and checked. The thermometer will read
low because the volume has increased. Many times the thermometer will return to its
original calibration in a few days. Nitrogen at 1 atmosphere pressure above the mercury
is used for measurement of temperatures up to 300°C; 20 atmospheres of nitrogen are
used for temperature measurements as high as 550 °C. The softening point of glass
must be considered.
For differential temperature measurements, the Beckman thermometer can be em-
ployed. The expansion chamber of an ordinary thermometer is enlarged so that mercury
can be poured into it from the main reservoir. The range is usually from 35° to
+300°C; a differential range of 5°C with readability down to 1/100, or over 1/1000°C
can be obtained. The scale lengths are available up to approximately 30 centimeter
48 TRANSDUCERS
-------
In many measurements the temperature difference is more important than the absolute
temperature.
BIMETALLIC THERMOMETERS
By coupling two metals that have different rates of expansion with temperature,
the temperature can be measured by observing the deflection of the free end of the
combined strip. The bimetallic thermometer is a rugged and simple device for the
indication of temperature. The accuracy is not high.
The bimetallic strip can be made in the form of a straight cantilever beam; a change
in temperature causes the free end to deflect, and this movement can be calibrated.
Generally the deflection is nearly linear with temperature. In other transducers, the
bimetallic strip is wound in the form of a helix; one end is fastened permanently
to the case while the other is attached to a pointer on a dial. Commercial bimetallic
thermometers generally cover the range from 40° to +425 °C.
RADIATION PYROMETRY
Radiation and absorption is a universal process of heat transfer. Radiant energy
travels from a source or a radiator until the energy is absorbed by the medium in which
it is traveling or is intercepted by an object. Energy may be transferred from one body
to another by the process of radiation and absorption even though there is no material
in the space between the bodies. Upon interception, the energy is partly reflected,
partly absorbed, and partly transmitted.
All bodies emit radiant energy at a rate that increases with temperature and is
independent of the neighboring bodies. A "black body'' is a body that absorbs all
radiation incident upon it and reflects or transmits none. A black body is an ideal
radiator. It emits, at any specified temperature, in each part of the spectrum the maxi-
mum energy obtainable per unit time from any radiator as a result of temperature alone.
Often it is convenient and desirable to measure the temperature of the surface of a body
by means of the neutral radiant energy emitted from it. One need not make any con-
nection to the body or be in close physical proximity to it. Measurements can be made
on moving bodies, corrosive liquids, and distant objects at high temperatures. Radiant
flux is the rate of flow of energy from a radiator.
Let P = radiant flux
U = radiant energy
Then P = dy (ergs/sec or watts)
dt
Let R = Radiance of a source = (ergs/ (sec cm2) )
dA
e = emitted from source
The radiance of an actual source is related to that of a black body by the total
emittance £.
_ R
£~ Rb
where b refers to a black body and e is also called the emissivity of the body. It is a
measure of the deviation from a perfect radiator. One of the pertinent radiation
characteristics is stated as Kirchhoff's Law.
Bellinger 49
-------
The emittance E of a non-black body is equal to the total absorbtance a for radia-
tion from a black body at the same temperature. £ = a and EX = aX where A is a par-
ticular wavelength. No material is a true black body. Some solid bodies can be con-
verted into artificial black bodies by drilling them with a small hole or a wedge. By
use of radiation from the hole or the wedge, black-body radiation can be approached.
By the Stefan-Boltzmann total radiation law, Rb =o- T*, temperature can be determined
by measuring the total radiation from a black body or a gray body, which is one that
deviates by a known and constant amount for any wavelength from a true black body.
The indication or deflection of an instrument depends on the surrounding temper-
ature T0 and the temperature of the body T. Thus, D = C^ (T4 - T04) where C^ depends
on the instrument used and the physical arrangement of the heated body.
Total radiation transducers are useful for measurements at low temperatures. To
provide sufficient output, the radiation detector can be formed of a collection of thermo-
couples, connected in an additive arrangement known as a thermopile. At high temper-
atures, say above 600°C, where the object glows visibly, an optical pyrometer can be used.
Often the disappearing-filament type of optical pyrometer yields good results. The
radiation from a black body or known gray body is focused with an optical system.
The observer sees the radiator and a heated wire filament, which is located within the
pyrometer. The filament is a tungsten wire, heated with a battery; a series variable
resistor is included so that continuous changes can be made in the temperature of the
filament.
By sighting on the black body hole, one sees both the hole and the heated filament.
The current through the filament is adjusted until its color temperature matches that
of the black-body radiator. Then, when the image of the filament is moved slightly in
the pyrometer across the hole, the filament tends to disappear when the color temper-
atures have been matched. The temperature reading is obtained by examining the scale
connected to the variable resistor that controls the current through the filament wire.
In turn, this scale and filament color temperature are calibrated against a standard
tungsten lamp whose radiation characateristics are well known.
Accuracy of measurement depends on how well the observer is trained and on
the quality of calibration of the pyrometer against a standard tungsten lamp source.
The National Bureau of Standards calibrates optical pyrometers over the temperature
range from 800° to 4200°C The uncertainty of calibration varies from 3° at the gold
point to 40° at 4200°C. Also, the standard tungsten strip lamps, which are used as
sources of known brightness temperature over the range from 800° to 2300°C, can be
calibrated.
Another type of pyrometer employs two selected wavelengths to obtain temperature
measurements. The optical disappearing-filament-type pyrometer uses a single filter whose
wavelength usually is centered at 6500 angstroms. With the ratio pyrometer, two wave-
lengths are chosen and a ratio is formed from the two output signals. This ratio, which
is a unique function of temperature over a wide range, is calibrated as a function of
temperature.
Some experiments are being conducted with a photoelectric pyrometer, which re-
places the human-eye detection system with a photomultiplier tube to make brightness
matches between the subject and the pyrometer lamp. This instrument should eliminate
the variability in calibration caused by the observer's lack of precision.
50 TRANSDUCERS
-------
RESISTANCE THERMOMETERS
Since the resistance of most metals is temperature dependent, a thermometer can
be made by winding a resistor with a selected metal. Then, by accurate measurement of
the resistance, the temperature can be determined. The resistance bulb can be used
to measure the absolute temperature because the resistance of the wire in the coil
depends directly on temperature. Resistance thermometers can have high sensitivity;
that is, the change of resistance per degree is appreciable. There is a maximum tempera-
ture limit, however, above which the resistance bulb cannot be used.
The wire material must not undergo any phase changes during the temperature
excursion, or its characteristics will be changed and the calibration altered. Normally
resistance thermometers are relatively large compared to thermocouples or thermistors.
Resistance thermometers are made from a variety of materials, often nickel with platinum
and copper. The temperature-resistance curve of nickel is non-linear. The shape of the
thermometer varies greatly depending upon application. For accurate readings, all
portions of the resistance thermometer must be at the same temperature.
The resistance of the thermometer is measured frequently with one of a number of
bridge circuits. In some cases the resistance bulb can be made an integral part of a
bridge circuit that incorporates the slide wire of a chart or indicating-type recorder.
Platinum is often selected because of its excellent reproducibility from 260° to 1100°C.
Nickel is limited generally to use at temperatures below 300°C. Below 260°C, platinum
becomes a super conductor.
The wire in the resistance thermometer generally is doubled upon itself to preclude
inductive effects. The wire must be free of supports to minimize heat conduction losses.
Protection tubes, either metal or ceramic, can be used if required. Metal tubes often are
filled with a dry gas, such as air or nitrogen, at approximately 0.5 atmosphere pressure
at room temperature. Some pressure should be maintained in the tube to increase the
rate of heat transfer and thus improve the response time. Accuracy of approximately
0.001 °C can be obtained without extreme care. Some care must be taken with the leads
coming from the resistance element, since they can affect the accuracy because of the
effect of the temperature gradient on their resistance.
THERMISTORS
One of the newer transducers for temperature measurements is the thermistor, a
resistor that is extremely sensitive to temperature. Thermistors show a high negative
coefficient of resistance as a function of temperature. It is not uncommon to find
a semiconductor material that changes its resistance by a ratio of 107: 1 over the temper-
ature range from 100° to +450°C. The use of thermistors at temperatures above 450°C
and below 180° is uncommon.
The normal resistance value of thermistors at ambient temperatures varies widely;
sometimes it is only a few ohms, sometimes a few megohms. Many semiconductive
materials can be used to fabricate thermistors. These include some of the metal oxides
and a number of mixtures. Thermistors can be made in extremely small sizes and in
odd shapes. The method of measuring the resistance of a thermistor must be selected
carefully because current from the resistance-measuring device can change the junction
temperature and thereby give a false value because of the great sensitivity of the junction
resistance to temperature.
The sensitivity of resistance to temperature varies widely, but values ranging from
Bellinger 51
-------
1 to 5 percent per degree centigrade near ambient temperature are not uncommon.
These values often increase at low temperatures, and decrease at high temperatures.
Thermistors are useful in temperature control devices because of their high sensitivity.
THERMOCOUPLES
In 1821 Seebeck discovered than an electric current will flow in a closed circuit
-when two dissimilar metals are used and the temperature of one junction is hotter than
that of the other junction. In 1834 Peltier discovered that when a current flows in one
direction across the junction of two dissimilar metals, heat is absorbed and the junction
is cooled. If the direction of current is reversed, the junction is heated instead of being
cooled. This phenomenon is reversible.
The heat developed in the junction is a function of the first power of current
rather than the conventional I2R Joule heating, which is irreversible. The Peltier
heat depends only on the pair of metals chosen and is independent of the form and
dimensions, whereas Joule heating is a function of form and dimensions. The amount
of current that flows as a result of the junction of dissimilar metals depends primarily
on the temperature difference, the choice of metals, and other factors, including the
total resistance of the circuit.
If an open circuit is used, the potential difference that will exist between the terminals
will depend on the temperatures at both ends of the couple, but not on the shape or
the dimension of the conductors. When two metals are placed in contact, electrons
diffuse across the boundary continuously until an electric field is established whose
force opposes the transfer of more electrons, thereby establishing an equilibrium condi-
tion. This output voltage is a function of the temperature difference and the absolute
temperature of the cold junction. Since the voltage of a single junction cannot be
measured alone without introducing additional junctions, at least two junctions exist
in any practical thermocouple circuit.
Thomson deduced that the Peltier effect was not the only reversible heat effect,
but that there is a reversible effect within the conductor itself when there is a tempera-
ture gradient and a current. Later he proved it experimentally.
In summary, the following effects exist:
1. Seebeck Effect Electric current flows in a closed circuit if two dissimilar con-
ductors are used when the temperature of one junction is higher than that of the
other junction. The Seebeck effect is the sum of the Peltier and Thomson effects.
2. Peltier Effect Electric current flowing in one direction across the junction
of two dissimilar metals causes heating or cooling. The amount of heating or cooling is
directly proportional to the quantity of current. When the direction of electric current
flow is reversed, the heating and cooling effects are reversed. Peltier heat depends on the
type of metal. The amount of heat is independent of material form and dimension.
3. Thomson Effect The Peltier effect is not the only reversible heat effect.
Thomson concluded that there must be a reversible effect within the conductor itself if
there is a temperature gradient along the metal conductor. The temperature rises for
cadmium, silver, and zinc for a particular direction of electric current. For the same
direction, the temperature drops for iron and nickel. The temperature change is zero
for lead.
In any simple thermocouple circuit consisting of two junctions and two wires a
minimum of four voltages exist; two Peltier emf's appear at each junction and t
52 TRANSDUCERS
-------
Thomson emf's appear along each wire. Ordinarily the Peltier voltage is less than 0.1
volt. At low temperatures the voltage output is extremely low, in the microvolt region.
Originally, most thermocouple tables were prepared with lead as the reference metal
because lead has a zero Thomson coefficient. Nearly all modern tables use platinum as
the reference because it melts at a much higher temperature.
In conventional thermocouple usage, one of the thermocouple junctions is maintained
at a fixed reference temperature, which ordinarily is a well-prepared ice bath. One
cannot prepare an ice bath by placing ice cubes in water and expect to have a good,
stable, known reference temperature. Pure water must be used, with the proper amount
of air saturation. The ice must be crushed well and be in good contact with water,
because the definition of the ice point depends upon equilibrium between ice and water.
The temperature of the cold junction is known as the reference temperature. For
special applications liquid hydrogen or liquid nitrogen may be used instead of ice to
provide the reference temperature, and at high temperatures the boiling point of liquid
sulphur may be used. Under these conditions data are difficult to interpret because
separate calibration curves must be employed, since the thermoelectric power changes
with absolute temperature. Thermoelectric power, which is a misnomer, is the rate of
change of voltage with respect to temperature, that is, dE/dT.
If the temperature of the hot junction is raised sufficiently high without the metal
melting, ordinarily a neutral temperature can be reached. At this point, the voltage
no longer rises with increasing temperature; the slope of the voltage-temperature curve
iz zero. A further increase in temperature causes a decrease in output voltage; finally
an inversion temperature is reached, at which the output voltage is zero. Still higher
temperatures cause the output voltage to reverse polarity. Ordinarily thermocouples are
never used even as far as the neutral temperature.
The number and type of themocouple materials are manifold. Conventional types
are listed in many handbooks; these data are readily available. For special high-
temperature applications, however, one must use some peculiar metals. Platinum and
platinum-rhodium alloys can be used at temperatures up to approximately 1800 °C
with care. Above that temperature pure tungsten, tungsten alloys, rhenium, and other
refractory materials, including tantalum, molybdenum, and iridium can be employed.
Reproducibility is not good, however, and the thermocouples are difficult to calibrate.
Many of these metals are extremely sensitive to oxidation at high temperatures.
Within ordinary temperature ranges, say several hundred degrees above and fifty
degrees below the freezing point of water, temperatures can be measured very accurately
with thermocouples. Adequate low-level measuring devices, of course, must be employed.
By use of extremely small-diameter wire to form the junctions, very high-speed transient
responses are achieved, sometimes in the microsecond range.
Three basic laws apply to thermoelectric circuits.
1. Law of homogeneous circuits. An electric current cannot be maintained in a
circuit composed of a single homogeneous metal, regardless of the cross-sectional area,
by the application of heat alone.
2. Law of intermediate metals. If a number of different thermocouple junctions
exist in a circuit and the entire circuit is maintained at one temperature, the algebraic
sum of the thermal voltages will be zero.
3. Law of intermediate temperatures. The electromotive force generated by a.
Bellinger 53
-------
,
at T and T3 and the same thermocouple with junctions at ! ana 12.
FLOW
INTRODUCTION
Almost anything that provides resistance to flow can be made into a flowmeter.
The pressure drop across the flow resistance can be calibrated in terms of flow rate.
For general usage the calibration must be repeatable, the transducer must be sufficiently
sensitive, the flowmeter must withstand the action of corrosive fluids, and the flowmeter
must give adequate frequency response.
Most flow measurements are made by inferential techniques; that is, pressures or
positions are measured and the flow rates are inferred. Although almost any kind of
restriction can be used for a flowmeter, it is desirable to use one for which published
coefficient data are available so that the pressure drop can be predicted for a given flow
rate without calibration of the flowmeter.
HEAD FLOWMETERS
The basic principle of head flowmeters is the conversion of energy from one form
to another by the primary element. The conversion from kinetic energy to potential
energy is made in flowmeters for liquids; the liquid is essentially incompressible. An
average flow measurment is obtained with the common head meter because of the
difficulty of making point measurements. When the flow rate of a liquid is measured,
the pressure difference across the head meter is a function of the velocity, density, and
viscosity of the flowing stream. It is generally assumed that the fluids are homogeneous;
otherwise most of the existing coefficient data would not be applicable. Nonhomogeneous
fluids give considerable difficulty in interpretation. "When the fluid is a compressible gas,
the internal energy of compression must be considered in the energy conversion.
A number of primary elements can be employed to convert some of the energy from
kinetic to potential. One of the most common is the thin-plate, square-edged orifice. It is
easy to install, to inspect, and to replace if damaged or to substitute if a change in
flow rate is required for a particular value of differential pressure. The orifice can be
reproduced readily, although appreciable care must be taken in fabrication of the plate.
Beyond certain minimum ^sizes, the characteristic coefficient depends primarily on the
ratio of the diameter of the orifice to the diameter of the pipe. Tables of coefficient data
are readily available in the literature.
In general, there are four basic types of orifice plates, each designed for particular
applications. The most common is the concentric type, in which the bore is located
concentrically with the inside of the pipe. This type is used often when the fluids are
clean and the gases contain little or no liquid. If the gas carries some liquid or solid
material, a vent or drain hole can be provided.
In some applications the liquid contains a large quantity of undissolved gases or
the gas to be measured contains a considerable number of condensible components that
are carried along in the pipe. For these conditions an eccentric orifice plate can be
utilized. The hole is located tangent to one wall of the flowmeter tube. The eccentric
orifice is similar to the concentric orifice except that the hole is located off-center and
the outer portion is tangent to the pipe wall. Thus the flowmeter is fully vented and
TRANSDUCERS
-------
fully drained. This type of a flowmeter is less accurate than the concentric type, as
evidenced by the poorer coefficient data in the literature. If the flowmeter can be
calibrated with the fluid to be used, the eccentric orifice is as accurate as the other types
of flowmeters.
The segmented orifice plate is useful for liquids that carry solids in suspension.
The segmented orifice plate covers the upper cross-section of the flowmeter pipe. The
lower section is left completely free so that the solids will not accumulate on the
upstream side of the orifice plate. The chord section is fabricated with a sharp edge;
the rear portion has a radius of curvature whose arc is 98 percent of the nominal pipe
radius, so that the curved surface will not be located below the wall surface of the pipe.
The quadrant type of orifice plate is useful in special cases. With sharp-edged orifices
the flow coefficient increases as the turbulence decreases. In many situations the flow
becomes so highly turbulent that this coefficient change is of no significance. When the
fluid viscosity is above five centipoise, however, and the quantity to be transferred
through the flowmeter is relatively low, the concentric orifice does not operate satis-
factorily because of the large change in flow coefficient with flow rate. In this case the
quadrant type of orifice is useful. This primary element shows little or no change in the
coefficient for low turbulence conditions. With this plate the curvature on the approach
side is the quadrant of a circle and the radius of curvature depends on the throat ratio.
Manufacture of quadrant plates is difficult because of the curvature requirement.
Since the nature of the surface greatly influences the flow coefficient, it is usually necessary
that a field calibration be made. The quadrant plate does exhibit excellent reproducibility.
For flowmeter pipe diameters of less than 2 inches, the published coefficient data become
somewhat questionable. Therefore, it is highly recommended that flowmeters of smaller
pipe diameters be calibrated.
In a venturi tube, the fluid-carrying pipe contracts and expands gradually to form a
smooth convergent-divergent nozzle. According to the basic flow equations, the fluid
accelerates as it passes through the venturi. Although venturi installations are more
expensive and more difficult to fabricate than orifice plates, they are useful when ex-
tremely large quantities of fluids are to be measured. The pressure recovery with the
venturi tube is excellent, and the inner surface of the tube is smooth. The recovery
section generally is designed with a l-to-10 taper; that is, the diameter increases 1 inch
for each 10 inches of length, giving a 20-to-l slope on each side. Steeper or less steep
slopes result in reduced pressure recovery. With small venturi tubes, the Reynolds number
and viscosity affect the measurements. The surface finish and irregularities become more
important as the size decreases.
A flow nozzle is essentially a type of venturi tube without the recovery section. The
converging section is generally short.
AREA FLOWMETERS
In the area flowmeter a float is positioned in a variable area section. The fluid
enters the bottom portion of a tube and passes upward through the metering section
around a float positioned concentrically within the variable area tube. The fluid then
exits at the top of the tube. The metering tube is usually glass, with sides tapered
uniformly so that the cross section at the top is greater than that at the bottom. The
float inside is guided so that it moves up and down concentrically within the tube as the
flow rate changes. This type of flowmeter must Be installed in a vertical plane.
Bellinger 55
-------
When the flow rate is steady, the float assumes a fixed posmon watfcn the tube and
its position can be calibrated in terms of the fluid flow rate The float is m staUc
equilibrium for a constant flow rate because of the equality of the gravity and bouyancy
forces. Some flowmeters of this type incorporate a heavy and a light float and thus
constitute a two-range flowmeter within one tapered-tube housing.
Variable area flowmeters are produced in many styles. Some are armored for use
at high pressures; others include an electrical readout system so that the output signal
can be fed directly into a recorder, an indicator, or a computer.
WEIRS, FLUMES, AND NOZZLES
Head-area flowmeters often are used in open channels to measure liquid flow rates.
These open-channel meters are commonly used in electrical generation stations, water
works, sewage disposal facilities, and water irrigation. One of the most common types
of weirs incorporates a rectangular notch; in other weirs the notches are V-shaped or
trapazoid. In the rectangular-notch weir, the velocity is proportional to the depth at
the weir.
For the calibration of weirs approximate discharge coefficients are available in the
literature. The flow rate through the weir depends on the height of the fluid flow in
the rectangular opening raised to the 3/2 power. The exponent on the elevation term
varies with other shapes of weirs.
Flumes are used in open streams where the flow rates are much greater than those
that can be measured with a weir, up to 70 million gallons of water per day, for example.
The flume restricts the stream and then expands it again in a definite fashion. The head
is measured at a single point about one third of the distance downstream from the inlet
of the flume in the entrance section. Either test data or an empirical formula must be
used to obtain the flow rate.
An open nozzle can be used to measure the flow rate of sewage, sludge, and in-
dustrial waste in pipes and channels that are partially filled. The unique cross-sectional
shape of the nozzle produces a nearly linear relationship between head and flow.
ELECTROMAGNETIC FLOWMETERS
The concept on which the electromagnetic flowmeter is based is that of the electric
generator. "When a conductor moves in a magnetic field, the voltage generated in the
conductor is proportional to the strength of the magnetic field and to the velocity at
which the conductor moves. For the measurement of fluid flow rates by an electro-
magnetic flowmeter, the fluid must be a conductor having a reasonable electrical con-
ductivity. A uniform magnetic field is produced either by a permanent magnet or by an
electromagnet located outside the pipe. The generated voltage is measured by a pair of
insulated electrodes located in opposite sides of the pipe on an axis perpendicular to the
magnetic field.
When the magnetic field is uniform, the voltage developed is proportional to the
velocity of the fluid flow. There are no obstructions in the pipe. The electromagnetic
flowmeter is useful in measuring the flow rates of liquefied metals, particularly those used
in the nuclear industries. For some water-based liquids, use of an AC magnetic field is
desirable, because of polarization difficulties at the electrodes.
Difficulties are encountered when the conductivity of the fluid changes. The response
56 TRANSDUCERS
-------
time is excellent. Since it depends on the frequency of the magnetic field, a high-
frequency field must be used when extremely high response is required. Electromagnetic
flowmeters are suitable for measuring the flow rates of corrosive fluids and slurries.
Results are relatively unaffected by viscosity, density, and turbulence.
DISPLACEMENT FLOWMETERS
The liquid displacement flowmeter calibration method is one of the primary standards
for gas flows. This technique consists of displacing a known or measurable volume of a
liquid with the gas at a known pressure and temperature. The flow rate of the gas must
be constant with this technique unless only the integrated value of the total volume of
gas is needed. The volume of the displaced liquid can be calculated by weight or by
volume over a known interval of time to yield an average volume flow rate of a gas
at the conditions of test.
In some displacement flowmeters the level of liquid remains relatively constant; in
the positive displacement type, some portion of the gas stream energy is required to
move the liquid. Generally, this type of flowmeter measures relatively low flow rates with
a high degree of accuracy. The wet- test meter is a displacement flowmeter which uses
liquid displacement.
TURBINE FLOWMETERS
With turbine flowmeters, a free-running turbine is mounted in the flow stream. By
use of special shapes for the turbine to reduce inertia, friction, and viscosity effects, the
flow rate is determined by measuring the rate of revolutions of the turbine. Ordinarily,
a magnetic-type pickup is employed, and the output pulses are fed to a pulse-counter
system in which the pulse rate is determined electronically. This rate is a function of
the mass flow rate of the fluid flowing through the meter.
The advantages of the turbine flowmeter are that the pressure drop across the flow-
meter section is small, the line element is very compact, and the flowmeter can operate
at temperatures from below zero to above 500°C. They range in size from 1/8 to more
than 8 inches diameter. Response times of the turbine-type flowmeters are extremely
rapid, and accuracy of 0.1 percent can be achieved under special conditions. Flowmeter
units must be interchanged when the change in ranges is more than about 10 to 1.
CHEMICAL COMPOSITION
INTRODUCTION
In many applications the chemical composition of materials must be determined.
Since many techniques are available for determining the composition of fluids or solids,
the choice depends upon such factors as the material, the available instruments, the cost,
and the accuracy required. There is no systematic method by which the best possible
technique or transducer can be selected to determine chemical composition. One can use
transducers based on electromagnetic radiation (including X-rays), chemical affinity or
reactivity, electric or magnetic fields, thermal or mechanical energy, and other principles.
SPECTROSCOPY
Spectroscopy is the measurement of the position of the wave length of interest within
the spectrum and its relative or absolute intensity. Both emission and absorption
spectroscopy are used to determine chemical composition. In emission spectroscopy the
Bellinger 57
-------
material whose composition is to be determined is used to produce a
spectrum. Then, with a suitable optical system which includes a pnsm or d.ftracUon
Sating, and a detector, the characteristic spectrum can be interpreted and the compos.-
tion of the material determined.
The fluids or solids are placed in an arc, spark, gas discharge, or flame and heated
to a point at which they emit their characteristic radiation if they do not radiate naturally.
The recording of the spectra, the measurement of the intensity of the various lines, the
various types of spectrographs, and the interpretation techniques are covered extensively
in the literature.
ULTRAVIOLET TECHNIQUES
The concentration of an ultraviolet-absorbing material in a mixture may be
determined fairly easily. The concentration is related directly to the amount of absorp-
tion from a beam of ultraviolet radiation that can be passed through the mixture. Thus,
with an ultraviolet type of spectrophotometer, a number of absorbing components in a
mixture can be identified simply on the basis of their patterns of absorption as a function
of wave length. In principle, the ultraviolet transducer consists of a source of ultraviolet
radiation, optical filters, a sample cell, a detector, and an output indicator.
The amount of transmission is determined by use of the ratio of the output signal
ascertained with the cell filled to that obtained with the cell empty. The concentration
can be determined from the known absorptivity of the substance by means of the
Lambert-Beer law or by comparison with other samples whose concentration of known
substances is well known. The ultraviolet sources that can be utilized in these trans-
ducers include tungsten lamps and arc discharge lamps that contain mercury, mercury-
cadmuim, hydrogen, xenon, sodium, or other materials. Each type of lamp has its own
characteristics, which must be selected for the particular problem.
REACTION PRODUCT DEVICES
Chemical composition can be determined by the measurement of a reaction product.
First, the desired chemical reaction must be promoted, and then the reaction product must
be measured to determine the presence and quantity of the constituents. A reagent-treated
paper on a fabric tape can be suspended in the stream carrying the fluid. These devices
are useful for monitoring exhausts and smoke, for determining the dusts, aerosols, and
corrosive and toxic gases and vapors in polluted atmospheres, and for continuous
monitoring, by use of moving tape, of the concentrations of specific components of gas
mixtures. Concentrations can be determined from fractions of a part per million up to
several percent.
A small area of impregnated tape is exposed to the gas sample, and a constituent in
the sample then reacts with a reagent in the tape to form a reaction product. The
reagent buried in the tape must be selected so that the reaction product will display a
characteristic that can be detected such as a change in color, in electrical conductance, or
in opacity. Tape can be moved continuously through the sample or individual samples
can be taken. Suitable analytical instruments must be employed to determine the quantity
of the characteristic change such as the change in conductance, etc., and relate it to the
concentration.
With liquids, it is possible to employ a reaction that forms a dilute suspension
Or, the reaction product can cause a change in the color of the liquid. When solids
are formed in the liquid because of the reaction, the amount of solid formed can h
58 TRANSDUCERS
GPO 814-105-3
-------
determined by photometry. From the change in the quantity of suspended particles in
the liquid, the concentration can be measured. By application of the Tyndall effect, this
technique can be employed for the quantitative determination of very small amounts of
any material that is capable of reflecting light.
When color is changed, a differential optical transmission measurement can be
utilized to determine the specific components of gas mixtures. The transmittance through
a reference solution is compared with transmittance through a reagent solution in which
the gas sample has passed under controlled conditions.
Concentrations can be determined also by a change in electrolytic conductance. A
constant-temperature conductance cell is employed for the flowing electrolyte.
pH MEASUREMENTS
To determine the effective concentration of acids and bases in solution, the pH
technique can be utilized. Special electrodes are employed to develop a voltage that is
proportional to the hydrogen-ion concentration in the solution into which the electrodes
are immersed. This technique of measuring pH refers only to the concentration of
hydrogen ions that actually are dissociated in the solution and not to the total acidity
or alkalinity.
Many techniques of pH measurement are available. Ordinarily they involve some
type of an electrode, such as glass, antimony, or hydrogen. Also, a reference electrode is
used, usually calomel or a silver silver chloride unit. A potential measuring instrument,
such as a vacuum tube voltmeter, indicates the output.
MASS SPECTROMETERS
When mass spectrometers are used to determine the chemical composition of a
substance, the material to be analyzed is injected into some type of ionizing device. The
resulting ions are separated according to their mass number by the combination of an
electric acceleration field and a magnetic field. When an electrode is placed at a focal
point, the resulting ion current is a function of the particular mass that is in focus at
that station. By maintaining a constant electric acceleration force and by varying the
magnetic field, one can sweep through a wide range of mass numbers.
If strength of the magnetic field, the charge, and the accelerating potential are known
for a particular instrument the mass that is being received at the electrode may be
determined. Various calibration techniques with known constituents may be used also.
By recording the intensity of the ion current, one can determine the relative magnitude
of the constituents. The most common angles used to bend the ion beam in the magnetic
field are 60, 90, and 180 degrees. When ion source and the ion collector are suitably
located, the ion beam is focused in a line at the collector electrode.
Special mass spectrometers can be designed and constructed for direct isotope-ratio
measurements (or to determine ratios of other constituents), wherein two or possibly three
electrodes are located at the various focus positions and the resulting ion currents are
collected continuously at the various electrodes. The isotope ratio (s) can be measured
directly by taking the ratio (s) of the output signals electrically.
X-RAY TECHNIQUES
X-rays may be used to determine the composition of certain materials and fluids on
Hollinger 59
-------
the basis of fluorescence, emission, absorption, and diffraction. It is possible to deter-
niine qualitatively and quantitatively the basic content of the constituents of complex
mixtures in terms of the elements, and to determine exactly their atomic arrangement
and spacing of the unit crystal. With diffraction techniques, one can determine the
crystal sructure of metals and other materials at high temperatures and observe the
phase changes as the crystal structure is modified by temperature variations.
ELECTRICAL CONDUCTIVITY TECHNIQUES
The ion concentrations in many solutions may be measured simply by electrical
conductivity methods. The concentrations of various materials in simple water solutions
can be determined with relative ease by conductivity techniques. The conductivity-
concentration curve must be known in advance, or the calibration determined experi-
mentally. Polarization effects can be reduced by using an alternating current rather than
a direct current in the conductivity cell. Usually some type of AC Wheatstone bridge is
used to determine the changes of conductivity in the conductivity cell. The cells are
very simple in basic structure; they usually consist of two metal plates or electrodes that
are fixed rigidly within an insulated chamber. Often platinum electrodes are used in
pyrex glass cells.
OXYGEN ANALYZER
The concentration of oxygen in some cases may be measured by using the para-
magnetic property of oxygen. The paramagnetic susceptibility of this component varies
inversely as the square of the temperature of the gas and decreases rapidly as the temper-
ature is increased. Usually the cell must have some type of temperature control so that the
temperature can be held constant while the magnetic susceptibility is measured. The
output signal is a function of the paramagnetic susceptibility of the gas volume and
generally this signal is directly proportional to the oxygen concentration. For these
measurements no paramagnetic substance other than oxygen can be present.
POLAROGRAPHY
Polarography is a method of chemical analysis based on comparative measurements
of current-voltage curves that are obtained during electrolysis of a solution under specified
conditions. Concentration polarization must occur at one electrode and a constant poten-
tial must exist at the other electrode. The various ions and molecules in solution can
be identified and measured by this technique if they are susceptible to oxidation or re-
duction at the polarized indicator electrode by applied potentials in the neighborhood of
a few volts. For many applications this technique is selective and accurate.
NUCLEAR MAGNETIC RESONANCE SPECTROSCOPY
Various isotopes can be identified separately according to their differing nuclear
gyromagnetic constants, the basis for nuclear magnetic resonance spectroscopy. Just as
a certain mass and electric charge are associated with each isotope, so also a spin or
angular momentum is associated with each isotope.
With this method a magnet is employed whose field strength can be changed from
essentially zero up to perhaps 10,000 gauss. A low-frequency radio transmitter supplies
RF energy to a small transmitter coil, which is placed in the magnetic gap. A small
receiver coil is located within the transmitter coil that surrounds the sample material
to be tested. A sensitive radio receiver, which is tuned to the same frequency as that
60 TRANSDUCERS
-------
of the transmitter, is capable of amplifying any signal that might be induced in the
receiver coil. Some type of indicator or recorder measures the presence of these signals,
which can be related to the various individual isotopes.
BIBLIOGRAPHY
1. Baker, H. D., Ryder, E. A., and Baker, N. H., "Temperature Measurement in
Engineering, Vol. 2," Wiley, New York, 1961.
2. Cerni, R. H., and Foster, L. E., "Instrumentation for Engineering Measurement,"
Wiley, New York, 1962.
3. Considine, D. M., "Process Instruments and Controls Handbook," McGraw-Hill,
New York, 1957.
4. Cusick, C. F., "Flow Meter Engineering Handbook," Minneapolis-Honeywell Regu-
lator Co., Philadelphia, 1961.
5. Eckman, D. P., "Industrial Instrumentation," Wiley, New York, 1950.
6. Gray, D. E., "American Institute of Physics Handbook," McGraw-Hill, New York,
1957.
7. Kallen, H. P., "Handbook of Instrumentation and Controls," McGraw-Hill, New
York, 1961.
8. Ladenburg, R. W., Lewis, B., Pease, R. N., and Taylor, H. S., "Physical Measure-
ments in Gas Dynamics and Combustion," Princeton University Press, Princeton,
New Jersey, 1954.
9. Lajoy, M. H., "Industrial Automatic Controls," Prentice-Hall, Englewood Cliffs,
New Jersey, 1954.
10. Lion, K. S., "Instrumentation in Scientific Research," McGraw-Hill, New York, 1959.
11. MacDonald, D. K. C., "Thermoelectricity," Wiley, New York, 1962.
12. Minnar, E. J., "ISA Transducer Compendium," Plenum Press, New York, 1963.
13. Monk, G. S., "Light," Dover, New York, 1963.
14. Stone, J. M., "Radiation and Optics,'' McGraw-Hill, New York, 1963.
15. Tyson, Jr., F. C., "Industrial Instrumentation," Prentice-Hall, Englewood Cliffs,
New Jersey, 1961.
Bellinger
-------
SESSION 3: General
Chairman: Dr. Frank E. Gartrell
Assistant Director of Health
Division of Health and Safety
Tennessee Valley Authority
-------
Dr. Peter K. Stein
Professor of Engineering
Arizona State University, Tempe
SUMMARY
Methods are presented for the classification of transducers and measurement systems.
Transducers are classified by function, by input requirement, and by energy types. Meas-
uring systems, the systems formed by combinations of transducers, are classified as unbal-
ance systems, in which the output quantities are observed directly, and reference systems,
in which output is compared to known quantity. The presentation incorporates general
principles of measurement engineering, on which these classification systems are based.,,,
CLASSIFICATION SYSTEMS FOR TRANSDUCERS AND
MEASURING SYSTEMS
CLASSIFICATION OF TRANSDUCERS
INTRODUCTION
General
The process of measurement consists of transferring information from one com-
ponent in the instrumentation chain to the next, until a final display on the readout
instrument is obtained. This signal transfer from link to link will always correspond
to a transfer of energy from one component to the next.
If energy is drawn from the source of the quantity to be measured, then the very
phenomenon that is to be observed will be altered. When the criterion is that only
a small amount of energy may be drawn from the source system in the process of
measurement, then the word 'small' implies that:
The amount of energy drawn from the source system in the process of measurement
must be small compared to the total amount of energy available in the source system.
Thus some knowledge of the source system must be available to the measurement
engineer. In a temperature measurement in a small, cooling cup of water the available
energy in the observed phenomenon is small and finite. In an atmospheric pressure
measurement the reservoir of available energy is almost infinitely large.
In a good measurement system the necessary transfer of energy from the source
system to the measuring system is minimum.
The measuring process always affects the phenomenon on which the measurement
is made.
This is the first law of measurements.
Every measuring system, no matter what the quantity measured, consists of a chain
of components that transform energy from one form into another. These transformations
of energy may frequently occur in the same discipline. The process by which this energy
transformation occurs is called transduction, and the components performing this
operation are called transducers.
In every measurement chain one must distinguish between different transducer
Stein 65
-------
types depending on their function in the system. Furthermore, in order to identify, under-
stand and express the behavior of a transducer, it is necessary to define certain basic
properties that serve to 'completely' specify the transducer. It finally becomes necessary
to be able to combine transducers into a chain of measuring element links and to pre-
dict the behavior of the resulting measuring system on the basis of the known properties
of the individual transducer elements.
Principles of Transducer Classification
Transducers can be (and have been) classified by a variety of different techniques.
The predominant three are discussed in the following paragraphs.
1. By the function they perform in the measuring system, i.e., whether they are at
the input or output of the measuring chain or whether they act as modifiers of
the information to be transmitted. This manner of classifying transducers results
in the following categories:
a. Input or measuring transducers.
b. Modifying transducers.
c. Output or readout transducers.
2. By the input requirements of transducers. This division results in two basic
classes of transducers:
a. Self-generatinog (active) devices, which produce an energy output for
a single energy input.
b. Non-self-generating (passive or impedance-based) devices, which require
two energy inputs in order to produce a single energy output.
For each of these transducer classes it is possible to relate variables at each
input and at the output in a convenient manner, rendering the system ready for
mathematical operation.
3. By the energy types involved in the transduction process. If one recognizes
eight forms of energy, it is possible to classify all conceivable transducers in a
'Transducer Space' containing 8-cubed, or 512, possible locations.
CLASSIFICATION BY FUNCTION
Measuring Transducers
The measuring transducer is the portion of the measuring system that transforms
the quantity to be measured into another quantity more easily measured. Usually more
than a single process of transduction is involved in this stage of a measuring system.
A thermocouple measuring temperature in a moving gas stream, for example:
a. The gas temperature is transformed into a related temperature of the thermo-
couple junction.
b. The temperature of the thermocouple junction is transformed into an electrical
output in the form of voltage or current, depending on the instrumentation
conditions.
A pressure-measuring device, for example:
a. The pressure is transformed into a force acting on a mechanical structure
(diaphragm, bellows, etc.).
66 TRANSDUCERS AND MEASURING SYSTEM
-------
b. Some consequence of this force is then measured. For example, the displace-
ment of the diaphragm is measured 'with a differential transformer; the strain
in the diaphragm may be measured with a strain gage, etc.
An accelerometer, for example:
a. The acceleration, acting on a mass, is transformed into a force.
b. The force is transformed into an electrical charge by action on a piezo-electric
material, or it may be transformed into an electrical resistance by means of a
piezo-resistive material, etc.
Note that the three examples cited have one thing in common:
All the phenomena listed as items (a) are basic phenomena associated with the
physical quantity to be measured. Thus the relationship between thermocouple tempera-
ture and temperature of the gas stream into which it is inserted is exclusively a heat
transfer problem and really has nothing to do with measurement engineering as such,
although this relationship must be known and understood by the measurement engineer
who measures temperatures with thermocouples.
All the phenomena listed under items (b) are basic phenomena associated with the
transduction process. There are almost countless phenomena in the physical world that
respond in some way to temperature, force, etc. An entire branch of measurement en-
gineering is devoted to the study of the physical laws that can be used as the bases for
transducers. A glimpse into this field will ge given in a later section.
Modifying Transducers
Modifying transducers act on the output from the measuring transducers and may
be divided into two varieties:
a. Intentional modification.
b. Parasitic modification.
Intentional Modifications. Intentional modification implies that the modification
(or computing function) introduced by the transducer is at the desire of and under
the control of the measurement engineer. Perhaps the most universal component in
an instrumentation system which exemplifies the intentional modification approach is
an amplifier (mechanical, hydraulic, pneumatic, electrical, etc.). An amplifier is the
prime example oi an intentional signal modifier: it produces at its output a signal that
is a known and desired modification of its input. Examples of other desired modifica-
tions may be integration, differentiation, adding, filtering, etc.
Parasitic Modification. In the process of signal transmission, signal modifications
may occur that are undesired and therefore parasitic in nature. Although the measure-
ment engineer may be aware of these undesired modifications he may not be able to
exercise full control over their presence and action. Such modifications are often called
noise levels.
Prime examples of such modifying systems are transmission systems such as lead
wires, switches, and slip rings. In the piezo-electric accelerometer example cited
previously, the validity of relating the charge of a piezo-electric transducer to the input
acceleration is entirely dependent on the choice of the cable that will connect the
electrical charge with the portion of the measuring system that will measure the charge.
Stein 67
-------
The most usual form of parasitic modification in transmission systems is caused by the
resistance and capacitance of the lead wires connecting the measuring transducer with
the intentional modifying transducer in the measuring chain.
In general, parasitic modifications may be multiplicative or additive in action, i.e.,
they may multiply or add to the desired signal.
Readout Transducers
The readout transducer transforms the modified signal into an indication that
may be observed with human senses: a visual, audible, smellable, tasteable, touchable
form. Examples of such readout transducers are galvanometers, dial indicators, direct-
writing recorders, the color of a titration mixture, the smell of a chemical.
It is normally assumed that nothing follows a readout transducer. This assumption
is somewhat erroneous and depends entirely on where the denned measuring system is
cut off. If the system is cut off at the cathode ray oscilloscope tube, for example, then
considerations of matching the optical properties of the light emanating from the tube
to the optical characteristics of the human eye or photographic film do not enter into
the picture. Carried to extremes, however, the system could be defined as going on
through the human eye into the system within ourselves that transmits the external light
stimulus to our brain and permits us to observe the phenomenon displayed on the
cathode ray tube face.
CLASSIFICATION BY TRANSDUCER INPUT REQUIREMENTS
Introduction
The definition of a transducer as an energy conversion element immediately includes
such commonplace transducers as:
Thermometers: The heat energy input results in mechanical displacement output
against a controlled force in the glass tube.
Bourdon tubes: The pneumatic energy input results in mechanical rotation output
against the restraining torque to the tube.
Other transducers do not seem to be covered by the energy concept as expounded
so far.
Thermometer: Heat energy input results in change of electrical resistance.
But resistance (any impedance in fact) is not a form of energy nor is it a component
of energy. It has been stated that any of the forms of impedance (resistance, capaci-
tance, inductance) have no existence in themselves. To observe such elements one must
either supply a current and observe a voltage or supply a voltage and observe, a current.
Thus, to obtain an energy output from a resistance thermometer one must supply
it not only with thermal energy input but also with electrical energy, so that the
temperature-induced resistance change can be observed.
Thus a certain class of transducers requires two energy inputs to produce a single
energy output. The additional energy input is often called auxiliary or biasing energy
supply, or the minor or modulating input.
68 TRANSDUCERS AND MEASURING SYSTEM
-------
THERMAL
ENERGY
INPUT
ELECTRICAL
RESISTANCE
TRANSDUCER
^
ELECTRICAL
ENERGY
OUTPUT
I
AUXILIARY
ELECTRICAL
ENERGY
INPUT
Transducers are classified as to whether the energy supplied by the unknown
quantity to be measured (hereafter called the UQ) is sufficient to produce an energy
output, or whether additional energy must be supplied to the transducer.
Self-Generating Transducers
Those transducer types for which the energy supplied by the phenomenon to be
measured directly produces output energy are called self-generating transducers.
Examples:
1. Thermo-electricity: heat electricity (thermocouples)
2. Mechanical levers: mechanical - mechanical energy
3. Piezo-electricity: mechanical force - electrical charge
4. Electrical generator: mechanical motion electricity
Non-Self-Generating Transducers
Those transducers that require one or more auxiliary, minor, or biasing energy
inputs to transform the action of the unknown phenomenon into an energy output
are called non-self-generating, passive, or impedance-based transducers.
Examples:
1. Resistance-thermometer: thermal energy into electrical impedance
2. Resistance strain gage: mechanical energy into electrical impedance
3. Photoelasticity: mechanical energy into optical impedance
4. Inductance microphone: acoustic energy to magnetic impedance
Representation of Transducers
General: Since measurement implies the transfer of information through a transfer
of energy, the definition of transducer has required the concept of an energy conver-
sion device.
Energy normally consists of two co-existing physical quantities that are physically
inseparable. Examples of such quantities, for which the product is energy, are:
Force and displacement
Pressure and volume
Voltage and charge
(mechanical energy)
(pneumatic-hydraulic energy)
(electrical energy)
One cannnot measure a force without permitting this force to go through some
displacement. In so doing, the energy drawn from the system supplying this force and
Stein
69
-------
this displacement, could conceivably alter the force being measured. Neither is it
possible to measure a displacement without force, although optical displacement meas-
uring techniques could render such forces exceedingly small.
The representation of a transducer or of a measuring system consisting of a chain of
transducers must, therefore, be in terms of energy flow a concept requiring two
inputs quantities at each input 'terminal' of the transducer. The representation and
associated nomenclature for self-generating and non-self-generating transducers are
shown below:
MAJOR ENERGY INPUT ENERGY OUTPUT
TRANSDUCER
MINOR ENERGY INPUT
(exists only for non-self-generating transducers)
Qp = primary quantity, i.e., the one to be observed.
Qs = secondary quantity; the one that necessarily co-exists with Q .
Major input is that energy input containing the quantity to be observed.
Minor input, also called auxiliary, bias, and carrier input is that second energy
input required for non-self-generating transducers to 'carry' the major-input-created im-
pedance through the transducer to its output.
The only restriction on the choice of primary and secondary quantities is that
dimensionally: Qp x Q8 = Energy.
It will be shown later that under certain special conditions this product could
also be power.
The minor energy input: It has already been stated that the function of the minor
energy input, especially in impedance-based transducers, is to 'carry' the major-input-
created impedance-change to the output in the form of energy.
The properties, capabilities, and limitations of a measuring system are directly a
function of the minor energy input used. These concepts could be further elaborated;
only an indication of the possibilities involved is given below.
The classification of systems centers on two characteristics of the minor energy input,
and one system design parameter. Any wave form can be used. The following are most
frequently selected in commercial systems:
a. An invarient level (DC).
b. A sine wave.
c. A pulse train (square waves are considered as the special pulse train in which
pulse duration and duration between pulses are equal).
The information being transmitted in the energy transfer 'carrier' process may be carried
on any of the properties of the wave form used.
70 TRANSDUCERS AND MEASURING SYSTEM
-------
For a level input:
a. Amplitude
For a sine-wave input:
b. Amplitude (amplitude modulation, AM)
c. Frequency (frequency modulation, FM)
d. Phase (phase modulation, PM)
For a pulse train:
e. Amplitude (pulse amplitude modulation, PAM)
f. Frequency (pulse frequency modulation, PFM)
g. Position (pulse position modulation, PPM)
h. Duration (pulse duration modulation, PDM)
i. Width (pulse width modulation, PWM)
j. Presence or absence of pulses in a specified number of pulses (pulse code
modulation, PCM)
The system performance will depend on whether the non-self-generating transducer
requiring the minor input is located
a. as an input transducer
b. as a modifying transducer
Each of the 20 different systems possible in this classification alone will present
different performance characteristics and can conceivably give 20 totally different an-
swers in measurements of the same physical phenomenon.
Characterizing a Transducer (Self-Generating)
Mathematically, the behavior of a four-terminal system becomes defined when four
specific coefficients for the system are known. This approach will be elaborated in
Reference 5. The section that follows will approach the same problem from an intuitive
way first.
To predict the behavior of a transducer, i.e., its output for any given input (s), one
must establish at least three sets of relationships for self-generating-transducers.
1. Relations at the component input
2. Relations at the component output
3. Relations between component output and input
Relationships at the component input: There are always two quantities acting on
the input of a transducer. The product of these input quantities will be either energy
or power. Energy is the more fundamental form, but since energy is the time integral
of power, it is usually accepted that measuring systems can be treated in terms of either
the energy or the power transmitted through the system.
One of the two input quantities will be the one to be measured, the other quantity
co-exists by physical necessity.
The primary quantity is the physical quantity to be measured. The secondary
quantity is the physical quantity that co-exists with the primary quantity at the input.
Stein 71
-------
The product of primary and secondary quantities will be the energy or power
absorbed by the measuring system from the source system.
primary quantity
The ratio: acceptance ratio = : A
secondary quantity
identifies the reaction of this system component with the one preceding it. This ratio
permits the determination of how much the measuring system influences the physical
process being measured.
The acceptance ratio is a complex number, mathematically, exhibiting both a
magnitude and a phase angle (or a real and an imaginary component).
Each of the components of the acceptance ratio, i.e., its magnitude and its phase
angle, depends on the frequency and the amplitude of the primary quantity.
Hence at the input alone, four characteristic equations or curves identify the re-
action of the transducer to and on the preceding link:
where the symbol 'f is the mathematical
function symbol, 'a' connotes signal am-
plitude, and 'w' is radian frequency.
Relationships at the component output. There are always two quantities emerging
from the output of a transducer. The product of these quantities will exhibit the
dimensions of either power or energy. The dimensions of the product of the two quan-
tities existing at the transducer input and at its output need not be the same. It is
possible for the product of the input quantities to be energy in dimension and for the
dimension of the product of the output quantities to be power. This condition prevails
in all impedance-based transducers such as linear-motion potentiometers, strain gages,
capacitive and inductive transducers, etc.
One of these quantities at the transducer output will be the one to be measured,
i.e., transferred to the next link in the measurement chain. The other physical quantity
co-exists of physical necessity. Once again:
The primary output quantity is the one to be measured. The secondary output
quantity is the one that of necessity co-exists at the transducer output.
The product of primary and secondary quantities will give the power or energy that
the transducer delivers at its output. The ratio
primary output quantity . r,
emission ratio = ; : &
secondary output quantity
identifies the reaction of this transducer with the one following it. This ratio permits
the determination of how much the source system (transducer output) and the meas-
uring system (input of the following transducer) interact and affect each other.
In general the emission ratio of a transducer will depend on the magnitude and
on the frequency of the input quantities and will be a complex number, as was the case
for the acceptance ratio. Hence, to characterize a link in the measurement chain, the
measurement engineer must know:
|E| = f (a)
/E = f (a)
72 TRANSDUCERS AND MEASURING SYSTEM
-------
]E|=f(w)
Note: The concepts of acceptance and emission ratios are analogous to those of
input and output impedance, but they are not equivalent. The impedance concepts can
be derived from the acceptance and emission ratios, but not vice versa. The energy-based
approach can be shown to be more fundamental and more universally applicable.
Relationships between output and input for the component. The relationships be-
tween transducer output and input may be denned as
, . primary output quantity ,
transfer ratio = - - - - - - - - = T
primary input quantity
This ratio identifies the action of the transducer on the transfer process that the
signal undergoes in passing through that component. The transfer ratio (also called
gain, sensitivity, response ratio) will be a function of the magnitude and frequency of
the transducer input. Thus again the ratio will be a complex number and not necessarily
related in a linear manner to the system input.
Hence, to characterize the link in the measurement chain being considered, the
measurement engineer must know:
General note on the ratios. When the statement is made that a ratio is a function
of the amplitude or magnitude of the system input, the implication is that the relation-
ship between input quantity and the ratio is nonlinear. In general it is possible to define
ranges of input quantity for which the relationship is linear within certain limits of
linearity (say one percent deviation from linearity). This range of inputs is then called
the linear input range for the transducer.
The statement that a ratio is a function of signal frequency implies that the
amplitude and the phase angle of the ratio may be both frequency dependent (i.e., the
amplitude of the ratio or its absolute magnitude is frequency dependent and the phase
angle between the numerator and the denominator or the direction of the vector repre-
senting the ratio may be frequency dependent).
Thus, to display the entire properties of each of these ratios the following must
be known:
a. The magnitude of the ratio:
its dependence on the magnitude of the input signal
its dependence on the frequency of the input signal
b. The phase angle of the ratio:
its dependence on the magnitude of the input signal
its dependence on the frequency of the input signal
Interaction between transducers. In measuring systems, it is generally desired that
a minimum of energy be transferred from the source system to the measuring system.
Stein 73
-------
The source transducer is the transducer that immediately precedes the one being
considered.
The two transducers are said to be isolated when they do not interact, i.e., when
the transfer of energy from the source transducer to the measuring transducer is zero
(or very, very small).
Q.
A measure of how isolated two transducers are is the ratio:
acceptance ratio of measuring transducer (Am)
Isolation ratio =
A + emission ratio of the source (E )
As this ratio approaches unity, the isolation between the transducers approaches
perfection.
This ratio, too, is a complex number since it is a function of complex numbers, and
it, too, will depend on amplitude and frequency of signal:
|I| = f(a)
/I =f(a)
|I| = f(w)
/I = f (w)
The isolation ratio also represents the portion of the primary quantity that is
available at the source output under ideal conditions to the primary quantity available
under the existing isolation conditions:
= I
isolation ratio = primary quantity obtained
maximum available primary quantity
It can thus be said to be an efficiency indicator for the measuring system design.
CONCLUSION
If the basic characteristics of a component in the measurement chain are known,
then the interaction and transfer characteristics of the component are known and it
can be intelligently applied, selected, and used. If these characteristics are not available
from the manufacturer of the transducer element, they must be determined experimentally
or analytically; otherwise it is impossible to obtain valid data on purpose. One
merely obtains data instead of making a valid measurement.
Characteristics of pure sources. A pure source of any physical quantity must have
an emission ratio of zero. Only in that case can the isolation ratio between it and
the elements which the source feeds be unity.
74
TRANSDUCERS AND MEASURING SYSTEM
-------
PURE SOURCE
OF Qp
QP
a
isolation ratio = I =
= 1 only for Es = 0
Examples :
1. Pure Source of Force. A mass in the field of gravity acts as a pure source of
force. It must be considered not as a mass that stores kinetic energy, but as
the extreme example of a spring!
SOURCE OF
FORCE
Q,, Force
On = Displacement
Given no restraints, the mass would be capable of undergoing an infinite dis-
placement in order to apply its force (mg) to an object.
Hence its emission ratio: Qp/Qs = mg/oo = 0
A mass in the field of gravity has an infinite spring constant and is to be considered
as a pure source of force.
2. Pure Voltage Source
SOURCE OF
VOLTAGE
Q,, Voltage
Q. = Charge
By definition, a constant voltage source must be capable of supplying any
charge necessary to maintain a given voltage, and its emission ratio becomes
3. Pure Displacement Source
SOURCE OF
DISPLACEMENT
Q, Displacement
OH = Force
A pure source of displacement must be capable of overcoming any force generated
in the process.
Stein
75
-------
CLASSIFICATION OF TRANSDUCERS BY ENERGY TYPES INVOLVED
The basic definition of a transducer implies that some input energy is converted
into some output energy. That there may be more than a single type of input energy
in order to arrive at an output in the form of energy was discussed for passive or non-
self-generating transducers.
Furthermore, it was shewn that energy is usually not the quantity that is to be
measured; that usually the input and the output quantity to be measured are accompanied
by secondary quantities such that the product of the two quantities at the input or at
the output is the form of energy.
All properties of a transducer that describe its reaction with previous and following
measuring system components, and that describe the action of the transducer on the
input energy, were expressed in terms of the primary and secondary quantities at the
input and output.
Basic Types of Energy Conversion
Energy can conveniently be divided into eight general classes, although the lines
of distinction have grown less and less well defined over the years. For example, electro-
magnetic waves in certain frequencies are called light, in others they are called electro-
magnetic radiation; the point at which pressure fluctuations cease to contain acoustic
energy and become mechanical is just as poorly denned. Examples of this type can
be multiplied to encompass almost all the distinctive boundaries between the classical
concepts of energy types.
For single-input energy single-output energy transducers, i.e., active or self-
generating transducers, the classification to be given may cover all the possibilities in
types of transducing principles. Note that not all the conversions have yet been achieved,
nor do they all form bases for transducers that have been achieved in the past.
For passive, non-self-generating transducers, requiring two or more energy inputs
to produce a single energy output, the combination and permuation of these energies in
threes yields a tremendously large variety of transducer types that can be envisioned.
A course in this field should undertake the study of non-self-generating trans-
ducers, since this field is the more general (and also more complex). The approach
taken should be independent of the transducing principle used so that the principles
presented could be applied to any of the possible transducing mechanisms.
Terminology
Some of the terminology applied to interactions between the energy classes is listed
below:
Type of Energy Adjectival and Combining Forms
Mechanics Mechanical, mechano-, piezo-, -strictive, -elastic,
-dynamic
Sound Accoustic, -sonic (ultrasonic)
Heat Thermal, thermo-
lg l Optical, photo-, spectro-, spectral, (infrared, ultra-
violet) , luminescent, phosphorescent
76
TRANSDUCERS AND MEASURING SYSTEM
-------
Electricity Electro-, electric, electrical, electronic, galvano-,
voltaic
Magnetism Magneto-, magnetic, (paramagnetic, ferromagnetic,
ferrimagnetic)
Chemistry Chemical
Physics of the Nucleus Nuclear, subatomic, nucleonics
A General Classification System: The Transducer Space Concept
For self-generating transducers with only one energy input and one energy output,
the above concepts result in 64 possible transducer combinations when 8 forms of energy
are considered.
In non-self-generating transducers the action of the major input, i.e., that of the
physical quantity Q to be measured, creates a variation in passive property of the
auxiliary energy system. Example: the temperature-induced electrical-resistance change
in a resistance-thermometer. This passive aspect of an energy system has been called
impedance and may be mechanical, electrical, thermal, etc.
To transform an impedance into an energy output it is necessary to apply a minor or
biasing energy input as previously explained. Thus it becomes a simple matter to extend
the two-dimensional "lattice" of self-generating transducer energy-conversion methods
into a three-dimensional array with the auxiliary energy input as the third axis, as
illustrated below, resulting in a transducer space.
This system will then permit the classification of any of the physical effects used
in energy conversion, resulting in 512 transducer possibilities (in terms of energy-types
combinations) when eight types of energy are distinguished, or 343 when only seven
are recognized. (Dr. Lion in Ref. 2 combines acoustical with mechanical energy, for
example.)
All transducers can now be classified by their location in the transducer space
coordinate system:
Major Energy Input Minor Energy Input Energy Output
Examples: Piezoelectric devices 503
Thermocouples 803
Electric resistance strain gages 533
Electric resistance thermometer 833
For all impedance-based transducers it is necessary that the minor energy input
be of the same class as the energy output, so that the last two digits of the transducer
classification will always be either "Ox" or *'xx."
An arbitrary number code is used for the arbitrarily selected classes of energy
in the specific illustration of the general concept.
Utilization of the Transducing Possibilities
To study in detail the many possibilities that could be the bases for transducers
would require a tremendous amount of work, and would resolve basically into a study
of physics. What becomes important to the measurement engineer is to have the basic
knowledge of these phenomena available, and a few good references at hand where
Stein 77
-------
additional information can be found. Some of the most useful information in this field
is contained in References 1-4.
OUTPUT ENERGY
i
,
,
1 /
/
!/ ,
//
/' '
/
* \
/ /
/ ' ELECTRIC
/( STRAIN GAGE ^\/
/ ^~^
x xi~~ ;
x /T
T ~I ~1 f~
8
-r T -
--]
J
ALL SELF-GENERATING
7 TRANSDUCERS FALL IN
' THIS PLANE
6
H
J
THERMOCOUPLE
5 PIEZOELECTRIC DEVICES
DEVICES \
4 \
x" iy
3 /' /
^, X ,' V
- "=^/ x x' 5
f FOR ALL IMPEDANCE / , / / /
\ BASED TRANSDUCERS!/ f/ / /
1 MECHANICAL, 2/ xx \ /
1 ACOUSTIC, ETC.) |/ / \/
1 THE MINOR INPUT 7~ A/
i AND OUTPUT ENERGY.3/ l/i/ x
, IS OF THF SAMF L£ IX I/ --K1
H
\
> X
/
/ x^ x-l
t-x / /
^ / ,
1/6/7
V/l /
\x'
/I
/
8 !
x ... mp
X '/'' '' i'NPUT
\1 x' ENERGY
^-4JX ELECTRIC
/i x^ RESISTANCE
_i/ THERMOMETER
CLASS.
ALL NON-SELF-GENERATING
TRANSDUCERS EXHIBIT A
COMPONENTS IN THIS PLANE /
MINOR INPUT ENERGY
' CLASSES OF ENERGY USED
7 TO BUILD THIS PARTICULAR
. "TRANSDUCER SPACE" :
1Acoustic
2Chemical
3Electrical
4Magnetic
5Mechanical
6Nuclear
7Optical
8Thermal
The Transducer-Space Concept for Transducer Classification
CONCLUSION
The advantages of the foregoing definitions and representations become apparent
in further study of measuring systems.
1. They represent a measuring system for what it is, a system of information
transfer through energy transfer.
2. They permit the application of a well-developed mathematical toolthe four-
and six-terminal network theories.
3. They permit direct and simple classification of transducers by the energy
types involved in the transduction process.
4. They permit ths logical inclusion in the classification system of such informa-
tion transfer methods as amplitude, frequency, pulse-width, pulse-duration,
pulse-position, and pulse-code modulation techniques.
78
TRANSDUCERS AND MEASURING SYSTEM
-------
CLASSIFICATION OF MEASURING SYSTEMS
The methods by which a physical quantity may be measured are basically three,
within two classifications:
1. Unbalance Systems
2. Reference Systems
a. Based on unbalance techniques
b. Based on null-balance techniques.
The capabilities and limitations of a measuring system depend to a very great extent
on which of these three methods of measurement is used. Knowledge of the measurement
method used in any commercial instrument applied to a test is vitaL The most blatant
measurement blunders are usually committed by users of transducers who either are
not familiar with the fundamentals of their instrument or do not care to be. It is
inevitable that the transducer itself, or the instrument, is blamed for the resulting
erroneous data never the user.
UNBALANCE SYSTEMS
In any measuring system, one set of input quantities (primary and secondary)
produces one set of output quantities. In the unbalance measuring system the output.
quantities are observed directly and their magnitude is measured.
UNKNOWN
INPUT
"UQ"
TRANSDUCING
SYSTEM
TO UNBALANCE
READ-OUT
Unbalance System
UNKNOWN
QUANTITY
"UQ"
TRANSDUCER
SYSTEM FOR
UNKNOWN QUANTITY
I
AUXILIARY
ENERGY
SUPPLY
TO UNBALANCE
READ-OUT
Unbalance System For Non-Self-Generating Transducers
Examples of unbalance systems are any meter on which needle deflection is taken
as measure of the input quantity; the common spring scale; the speedometer on an
automobile; the loudspeaker in a radio.
Stein
79
-------
In unbalance systems the transfer characteristics of the measunng instrumen are
of vital importance in the interpretation of the measurement; so are the «-«J^J
and emission characteristics. These ratios and their dependence on signal amplitude and
frequency will govern system performance.
For a non-self-generating transducer the output will depend on both inputs for the
unbalance system, and any variation in the auxiliary supply energy will influence system
behavior.
REFERENCE SYSTEMS
In the reference systems of measurement, the transducer output is not observed,
but it is compared to a known quantity. This known quantity is generated within the
reference portion of the measuring system. The reference system output is varied until
the unknown and known signals are observed to be equal or their difference is zero.
Then the measurement is considered complete and the desired value is read from the
reference system. The comparison may be of two types, and reference systems are
distinguished as follows.
Based om the Unbalance Technique
The output of the UQ (unknown quantity) transducing system is compared with
that of the reference transducing system by alternate switching between the outputs from
the two systems to a common unbalance indicating device. The reference quantity will be
called RQ.
KNOWN
INPUT
"UQ" TRANSDUCER
SYSTEM FOR
UNKNOWN
QUANTITY
REFERENCE
TRANSDUCER SYSTEM
FOR KNOWN
QUANTITY
TO UNBALANCE
READ-OUT
SWITCH
TO UNBALANCE
READ-OUT
UNBALANCE
MEASURING
DEVICE
Reference Unbalance System
In such systems the measurement is independent of the linearity characteristics of the
readout device since the unknown and known (reference) quantities are adjusted to
be of the same amplitude. Thus any amplitude distortion in the system would be com-
mon to both signals and its effect on the reading eliminated.
The measurement would not, however, be independent of the frequency response
of the readout instrument. When the reference system is switched into the readout device,
the instrument "sees" essentially a pulse input. The frequency content of this reference
signal may not be identical with that of the unknown signal so that frequency or phase
distortions in the readout instrument would affect the unknown and known signals
differently.
80
TRANSDUCERS AND MEASURING SYSTEM
-------
The measurement is also dependent on the isolation ratio between the measuring
instrument and each of the sources (of the UQ and RQ). It can be shown that so
long as the isolation ratio is high, even large differences in the emission ratios of the
UQ and RQ sources will not materially affect system performance. The measurement also
depends, of course, on all characteristics of the individual components in the UQ and the
reference channels.
An example of the reference-unbalance measuring system is the Norwood Controls
Pressure Indicator. The instrument is designed primarily for the measurement of dy-
namic steady-state pressures such as in internal combustion reciprocating engines. The
unknown pressure wave may appear as in (a). By alternately switching to a reference
circuit, which emits electrical signals equivalent to known pressures (b), one can
adjust the output from the reference circuit to equal that of the unknown phenomenon.
The required setting (c) on the reference circuit then gives the magnitude of the
unknown signal.
Switch to Reference
System is pressed
momentarily
Unknown steady-state
dynamic pressure
(a)
Reference system
output switched
momentarily on
to output
(b)
Reference system output
adjusted to equal maximum
unknown pressure
(c)
Sample Unbalance Reference System
Special case of unbalance reference system zero-reference. In a special type of
unbalance reference system the basic reference may be zero output from the reference
system, which would also correspond to zero output from the unknown system (since
zero is zero).
In this case, it is merely required that periodically, the input to the readout detector
be made zero; in an electrical system this is equivalent to a short-circuit, which is
simple to achieve. In a typical instrument such as the Ellis Associates BA-13 (or BA-12)
Bridge and Amplifier, the input to the amplifier is periodically short-circuited by means
of an electromagnetically driven switch. In its closed position the switch short-circuits
the amplifier input; in its open position the switch permits the signal to be measured to
pass through the amplifier.
Based on Null-Balance Techniques
The UQ output is added to (or subtracted from) the reference system output. The
reference system is then so adjusted that the combined output is zero. Then the reading
on the reference system is equal to the unknown signal (in subtraction) or minus the
unknown signal (in addition). Example: A mechanical balance for weighing.
Since under all conditions of data-taking the system output is maintained at zero,
the system behavior is independent of the transfer characteristics or the acceptance
ratio of the readout instrument. Thus input loading, linearity, frequency response, etc.,
of the readout instrument do not affect the accuracy of null-balance systems. On the
Stein
81
-------
other hand the rapidity with which the reference system can be adjusted to maintain
zero output limits the frequency response of the system as a measuring system.
"UQ" TRANSDUCER
SYSTEM FOR
UNKNOWN QUANTITY
REFERENCE
TRANSDUCER
SYSTEM FOR
KNOWN QUANTITY
ADDER
OR
SUBTRACTOR
TO NULL DETECTOR
Reference Null Balance System
Manual null balance will not accommodate signal frequencies over about % cps;
mechanical servosystems may go to a few cps; electronic techniques can be used to
extend the frequency response of such systems to higher limit.
Special case of non-self-generating transducers. For non-self-generating transducers,
the unbalance reference system may be operated in one of two ways. The separate
auxiliary energy supplies for the UQ and reference systems imply that any change in
either auxiliary supply will affect the reading. It is then possible to take advantage of
placing the transducer system outputs directly in series.
Where separate auxiliary energy supplies are used, the system is called a separate
reference system. Where a common auxiliary supply is used, the system is called an
integral reference system. The word 'integral' means that the reference system is a
part of the non-self-generating transducer and that both the transducer and the reference
system are fed from a common auxiliary supply.
CONCLUSION
Depending on the measuring method selected, the characteristics of the readout
transducer may or may not influence the measurement. Therefore, it is important to
know both the measurement method and the readout instrument characteristics. In-
herently the reference technique is capable of higher accuracy than the unbalance
technique. Most precision measuring systems are based on a form of reference meas-
urement.
A property of all integral reference systems is that a change in zero (or balance)
results in a change in calibration; in other words, the very act of balancing the system
affects its calibration or transfer ratio.
A property of all separate reference systems is that a change in calibration setting
(transfer ratio) results in a change in zero; in other words, a zero-shift will result when
the transfer ratio of the system is adjusted.
82
TRANSDUCERS AND MEASURING SYSTEM
-------
UNKNOWN
INPUT
KNOWN
QUANTITY
"U
TRANS
SYSTE
UNKNOWN
Q"
DUCER
M FOR
QUANTITY
AUXILIARY
ENERGY
SUPPLY
AUXILIARY
ENERGY
SUPPLY
I
REFEF
TRANS
SYSTE
KNOWN C
?ENCE
DUCER
M FOR
QUANTITY
TO SUBTRACTOR FOR MULL BALANCE
TO SWITCH FOR UNBALANCE
REFERENCE SYSTEMS
Reference System With Separate Auxiliary Energy Sources for Non-Self-Generating Transducers
UNKNOWN
INPUT
KNOWN
QUANTITY
"UQ"
TRANSDUCER
SYSTEM FOR
UNKNOWN QUANTITY
AUXILIARY
ENERGY
SUPPLY
REFERENCE
TRANSDUCER
SYSTEM FOR
KNOWN QUANTITY
O
o
m
Reference System With Common Auxiliary Energy Sources for
Non-Self-Generating Transducers
Stein
83
-------
These two properties are disadvantages of each system only under certain specific
test conditions. There are tests in which only one or the other instrument should be used.
REFERENCES
1. Physical laws and their effects, C. F. Hix, R. P. Alley, John Wiley and Sons, 1958.
Compiles a large number of physical laws, relating the different types of energy
one to the other. A brief description of the law is given, an example of its appli-
cation, an indication of the expected magnitudes, and one or two references. The
effects are cross-indexed both by the proper name of the inventors or discoverers and
by their scientific nomenclature.
2. Instrumentation in scientific research: input transducers, K. S. Lion, McGraw-Hill
Book Company, 1959.
Compiles a large number of physical laws which have actually been used as trans-
ducing principles for a large variety of measurements. Each section gives a brief
description of the law, of actual transducers which have been made and used,
based on this principle, and references in the literature applicable to the trans-
ducer are given. The book is organized for methodical presentation.
3. fnternational critical tables, McGraw-Hill Book Company.
Lists the physical laws and the various numerical coefficients which give the mag-
nitudes of the various effects. This is a collection of much of man's experimental
knowledge in all fields of science.
4. Searching the literature for transducer information: Part 1: A guide to the litera-
ture, J. Pearlstein, Report PB 161-320 from Office of Technical Services, Washing-
ton 25, D.C.
5. Measurement engineering, Peter K. Stein, Stein Engineering Services, Inc., 1962.
A systematic survey and text on measurement engineering fundamentals.
81
TRANSDUCERS AND MEASURING SYSTEM
-------
Gerald C. Gill
Professor of Meteorology
University of Michigan, Ann Arbor
SUMMARY
To obtain valid data it is essential that attention be given to the following sequence
of events: careful selection of the most suitable sensors and recorders for the parameter
to be measured; proper installation; regular maintenance and servicing; and regular
recalibration. An area often overlooked in a measuring system is the dynamic response
of the sensors and of the recorder to fluctuating inputs. Grave errors in the recorded
data may result from this oversight. Some fundamental relationships in this area are
discussed and some useful curves reproduced.
DATA VALIDATION*
INTRODUCTION
We shall discuss the main factors that determine the accuracy and fidelity of
recording a given variable. To specify the degree of accuracy and to maintain a high
level of dependability, the investigator must consider the following factors:
1. A clear understanding of the principle of operation of the basic sensor and
a knowledge of its dynamic response.
2. A general understanding of the principle of operation of the indicating or
recording system.
3. Calibration of the system. Some instruments require only static calibration;
others require dynamic calibration to determine the response of the system
to a rapidly fluctuating variable.
4. Proper installation and use of the instruments.
5. Routine servicing.
6. Periodic maintenance.
7. Periodic calibration checks.
8. Alertness for small clues that may indicate errors developing in the system.
Before discussing these topics I will define some of the terms used in specifying
the performance of instruments.
DEFINITION OF TERMS
The sensitivity of an instrument may be defined as the smallest change in the meas-
ured variable that causes a detectable change in the indication of the instrument.
(Example: For a thermocouple recorder having a range of 100°C on 10-inch-wide
chart paper, the sensitivity of a new and properly adjusted instrument might be
* Publication No. 79, Department of Meteorology and Oceanography, The University of
Michigan. Research conducted under Research Grant #AP00233-01, from the Division
of Air Pollution, Bureau of State Services, U. S. Public Health Service, and the sponsor-
ship of the National Center for Atmospheric Research.
Gill 85
-------
± 0.1 °C, which corresponds to about 0.01 inch of pen movement. But the sensitivity
might be as low as ± 1.0°C if the sliding contact were badly worn or the servo ampli-
fier poorly adjusted.)
The accuracy of an instrument (including application of its calibration curve)
is the precision with which the instrument will measure the variable in terms of inter-
nationally accepted units. (Example: the accuracy of the thermocouple recorder might
be ± 0.5 °C over the complete range when it is new and properly adjusted, but aa
poor as ± 2°C with worn sliding contacts and a weak servo amplifier.)
The term speed of response of an instrument is variously applied. Often it indicates
the time required for the indicator or recorder to follow 90 percent of a sudden full-scale
change in the measured variable; sometimes 99 percent of full scale. Sometimes the term
indicates the time that elapses from the application of a sudden square-wave change
until the recorder reading is steady. The term must always be defined.
For most sensors and recorders having a first-order response* the term time constant
is much better, since it has only one meaning. Suppose the thermojunction of our
thermocouple thermometer is suddenly transferred from an air stream at a constant
temperature 00 to a warmer air stream at constant temperature 0e (Figure 1). The
thermo-junction will not instantly assume the new air temperature 0e but will change
at a rate depending on the instantaneous temperature difference (0e 0), and will
approach the new temperature asymptotically. The time constant is the period that is re-
quired for the temperature sensor (thermojunction) to respond to 63.2 percent (1 1/e)
*See "Dynamic Response of Sensors."
T, -J STEP FUNCTION. OR
/ SQUARE WAVE TEMPERATURE CHANGE
Figure 1 Response of a Thermometer at Temperature 0O and Time Constant T to a Sudden
Change in the Environment (Step Function) to a New Temperature 0
86
DATA VALIDATION
-------
of the stepwise change in temperature. (The significance of this constant is given in a
succeeding section.)
For some sensors the term distance constant is more appropriate than the term
time constant. For instance, when a three-cup anemometer is suddenly transferred from
quiet air to a wind of 10 ft/sec the time constant might be 3.0 seconds, but if the
same instrument were transferred from quiet air to a wind speed of 20 ft/sec the time
constant would be only 1.5 seconds. The same amount of air (3.0 sec x 10 ft/sec = 30
ft; 1.5 sec x 20 ft/sec = 30 ft) will have passed in each case for the sensor to respond
to 63.2 percent of the speed change. Thus the term distance constant is more appropriate
for such a sensor. This is likewise true for propeller anemometers, propellor-type flow
meters, etc. The distance constant of a sensor is the length of air column (or water col-
umn) required to cause it to respond to 63.2 percent of the square-wave change in speed.
In the calibration of an instrument the indications of the instrument are usually
plotted against known values of the parameter for a number of points over the range
of the instrument. Since these points generally do not yield exactly a straight line, in-
strument manufacturers usually draw a "best fit" straight line through the calibration
points and specify the linearity as the maximum deviation of any points from the
straight line. This linearity, often expressed as a percentage, refers to percentage of
full scale deflection rather than percentage of the indication. (Example: For the thermo-
couple recorder, the linearity might be expressed as ± 0.5 percent. This would indicate a
deviation of ± 0.5°C from true value over the complete range from 0° to 100°C).
Some manufacturers specify that the straight line must pass through the zero of the
recorder. In such cases the linearity then specifies the maximum deviation of any point
from this straight line.
Most of you probably have conducted a static calibration of an instrument, and
are familiar with the problems of making a reliable calibration; yet many of you are
probably unfamiliar with the pitfalls of using such an instrument for measuring
fairly fast fluctuations of the variable (ten fluctuations per second, or perhaps only
one fluctuation per minute). Accordingly, it seems appropriate to outline the behavior
of some general types of sensors with stepwise and with sinusoidal fluctuations of the
variable being measured and to supply a set of curves and formulae that will be valuable
in conducting dynamic calibrations of sensors and sensing systems.
DYNAMIC RESPONSE OF MEASURING SYSTEMS
DYNAMIC RESPONSE OF SENSORS
Sensor With First-Order Response (equation of forces being a first-order
differential equation).
Consider a thermometer, initially at a temperature 90, which is suddenly transferred
into a moving air stream whose temperature is 9e (see Figure 1). Experiment shows that
the indicating thermometer will approach the new temperature 0e asymptotically at a
rate depending on the temperature difference Qe 0. This relationship may be expressed
by the equation:
J*2J £\ r\
= (a first-order differential equation) (1)
dt A
where 0 = instantaneous indication of thermal bulb at time t
0e = temperature of new environment (assumed constant)
00 = initial temperature of thermometer
Gill 87
-------
t = elapsed time (sec) after thermometer immersed in new environment
A = constant, depending on shape and composition of thermometer
bulb, and properties of new environment. (Note that \ has the
dimensions of time in the equation.)
Solving this differential equation we get
(ee-e) = (0e-e0)e-t/A (2)
Now when time t = \,
(©e e) = ( ee - e0) e-x/\ = ( ee -e0) e-i
= ?~o =o.368 (ee-e0) (3)
2.718
that is, after the elapse of A sec the instantaneous difference in temperature (0e 0)
has been reduced to 36.8 percent of its original value, or, in time \ the thermometer
will have responded to 63.2 percent of the initial temperature difference.
This constant \ having the dimensions of time is called the time constant T and
corresponds to the elapsed time required after a sudden change in the environment temp-
erature for the indicated temperature difference to be reduced to _L of its initial value.
e
Response of a first-order sensor to square-wave (step function) input. In time T
seconds the sensor will have responded to 63.2 percent of the initial temperature differ-
ence. In the succeeding T seconds the sensor will have responded to 63.2 percent of the
remaining temperature difference 0.368 (@e 00) ; that is, in 2 T seconds it will have
responded to 86.5 percent of the initial temperature difference. Table 1 relates percentage
response of the sensor to other values of elapsed time, measured in terms of the time
constant.
Table 1 Recovery of Sensor With First-Order Response After a Step-Function Input.
Recovery, % 50 63.2 90 95 99 99.5 99.9
Elapsed Time 0.7T LOT 2.3T 3.0T 4.6T 5.3T 6.9T
In determination of the time constant T by the method given above, other points
on the curve beside 0 may be used. For instance, T± = elapsed time after ^ for tem-
perature indication to reach 02, where 02 is determined by equation, (0e 02) 0.368
(0e 0.,). T may also be determined by drawing a tangent line to the curve and noting
elapsed time T3 where it cuts line 0e (Figure 1).
The unique value of the term 'time constant' is shown in Figure 2, which illustrates
what happens when the same thermojunction sensor has been immersed in an air
stream at the same speed as before and at constant temperature 00, when at time
t = o the temperature is raised at a constant rate. The temperature sensor does not
immediately respond to this constant rate of temperature rise but takes about 40 seconds
to reach this rate. The time constant T is the lag time T or T .
For a given air speed (or water speed) and a given temperature sensor the time
constants TI? T2, T3, T4, and TE should be the same within a few percent.
(Note The time constant for a thermometer exposed in an air stream at a certain
speed is about 60 times greater than it is for the same thermometer exposed
in a water stream at the same speed.)
88 OATA VALIDATION
-------
CORRESPONDING THER-
MOMETER INDICATIONS
{SAME THERMOMETER
AS IN FIG II
NOTE: T, - T. 20 SEC
Figure 2 Response of a Thermometer (With Time Constant T) That is Exposed in an Air Stream
That is Suddenly Heated at a Constant Rate
Response of a first-order sensor to a sinusoidally fluctating input. The accuracy of
indications of the amplitude of a sinusoidally fluctuating input is given by the following
equations:
(1) Amplitude ratio for a single-capacity system (e.g., bare resistance wire, or butt-
welded thermojunction) :
1
V
(o>T)
(x0 /x)* -1
(2) Amplitude ratio for a double-capacity system (bulb in well) :
x 1
(4)
(5)
(6)
0 V 1 + UT^)2 V 1 + (WT2)2
where x = indicated amplitude
x0 = actual amplitude
2
^ = angular velocity (radians/sec) = 2-n-f =
f = frequency of fluctuation
P = period of fluctuation =
T = time constant
Tj, and T, are time constants of bulb alone and well alone.
Figure 3 is a graphical representation of Equation 5 relating the time constant
Y
T and the period P of the cycle to the amplitude ratio of the sensor. (Example:
Gill
89
-------
Suppose the time constant of a thermometer in a wind of 10 mph were 100 seconds an
that sinusoidal air temperature fluctuations of ± 5° F were occurring at 5-minute periods
(P = 300 sec). The ratio JL = 15P_ = 0.33.From the graph the amplitude ratio would
P 300 . .
be 0.43. The sensor would show only 43 percent of the true temperature fluctuation..)
Thus by knowing the time constant of the sensor and the period of the fluctuations,
we can quickly specify the fidelity response of the temperature sensor. If we want the
temperature sensor to respond to 90 percent of the temperature fluctuations, the ratio
must be 075 or less; that is, the time constant of the sensor could not be more than
P
7.5 percent of the shortest period of fluctuations the sensor is to follow.
TIME CONSTANT OF SENSOR
~ PERIOD OF FLUCTUATION
Figure 3 Relationship Between the Time Constant T of a Temperature Sensor, the Period P
of a Sinusoidal Temperature Fluctuation of the Environment, and the Fidelity of Recording
this Fluctuation.
K
Figure 4 illustrates the response of several average temperature sensors to a sinu-
soidally fluctuating air temperature having a period of 300 seconds.* The effects of
time constant on amplitude ratio and phase shift are clearly demonstrated.
The following formula (experimentally determined) relates the diameter of cylindri-
cal metal temperature sensors, the air flow rate, and the time constant:
T = 6000 d1-3* v ~o.*o (7)
where T = time constant (sec)
d = diameter of cylinder (inches)
v = air speed (ft/ min).
Note that the time constant is roughly proportional to the square root of the wind speed.
* £°urtesf °f E- W- Jensen and K. C. Kiesling, Eastman Kodak Co., "Response of
15 17mi950 S' A- Instrument Maintenance Clinic, Buffalo, New York, Sept.
90
DATA VALIDATION
GPO 814I OS4
-------
Sensors with first-order response. Essentially all temperature-sensing instruments
have first-order response, e.g., mercury-in-glass thermometers, gas thermometers, resistance
thermometers, etc.
Figure 4 Response of Typical Temperature Sensors to a Cycling Air Temperature of 5°F in
Amplitude and of 300 Sec Period
Many flow-measuring sensors have first-order response, e.g., cup, propeller, and hot
wire anemometers; cup and propellor water-speed-measuring sensors and turbine types
of sensors.
It should be noted that a first-order sensor never indicates a larger change in the
measured variable than the true change, even with a step-function input.
Sensors With Second-Order Response (equation of forces being a second-order
differential equation).
Response of a second-order sensor to a square-wave (step function) input. In the
electric circuit of Figure 5 after switch S has been closed for some time the gal-
vanometer reading G will have become steady at some value, say A degrees. If
at time t = O switch S is opened, the galvanometer coil will quickly start turning
toward its zero position; overshoot this value by maybe 60 percent; reverse direction; and
again overshoot, executing a simple harmonic oscillation of decreasing amplitude, as
shown in Figure 6. (Here the sensor does indicate a larger change than the true change
larger by a factor of 60 percent thus differing markedly from sensors with first-
order response.) The equation of forces is:
dt*
de
dt
K0 =
(8)
Gill
91
-------
where I = moment of inertia of the coil suspension system
C = damping constant (primarily self-induced electromagnetic
damping; secondarily air damping)
K = spring constant
0 = angular deflection (measured from rest position)
./VWNA-
Ru 10,000 n
(DECADE BOX, INITIALLY
SET AT 1000 (1)
R, 2n
0
R0 - 18 n
R, = 100 - 10,000 n
Figure 5 An Electric Circuit to Determine the Critical Damping Resistance of a Galvanometer
or Indicating Meter
\ / A,""' (LOGARITHMIC DECREMENT)
Figure 6 Typical Galvanometer Decay Curve (h = 0.16)
92
DATA VALIDATION
-------
A general solution to this equation (when the decay curve is similar to that of Figure 6)
is:
Q Ae -at cos (yt -)- 8)
where A = initial displacement from zero (9)
_ C
cos8 = V 1 h2
As mentioned previously, this decay curve is a simple harmonic motion (cos yt)
of decreasing amplitude with envelope defined by the dashed curve Ae "a*.
If the electrical resistance Ra were decreased, the damping coefficient would be
increased and the decay curve would show fewer oscillations, each of decreased amplitude.
For a certain value of (Rd + Rj) the instrument deflection returns to zero in a
minimum of time without any overshoot. This condition is known as. critical damping
and is shown in Figure 7 by curve 6, labelled h = 1.0. Other values of the damping
ratio h from 0.0 to 3.0 are shown in Figure 7.
Figure 7 Damped Oscillations Galvanometer Decay Curves for Damping Ratios of h = 0.0
to h = 3.0
In the decay curve of Figure 6, the galvanometer has a damping ratio h = 0.16,
and a damped frequency of 2 cps (damped period ta = 0.5 sec). In the circuit of
Figure 5 with Rd = 1000 ohms, if a sinusoidal voltage of constant amplitude but of
varying frequency were applied across resistance R, the galvanometer would swing
back and forth at the same frequency as the input signal; when the input frequency
approached 2.0 cps, the galvanometer would resonate and show amplitude fluctuations
up to 3.0 times the true amplitude! Thus the dynamic response of the sensor can
greatly distort the true form of the input signal.
Gill
93
-------
Figure 7 shows the decay curves of a galvanometer whose damping ratio h as
been varied in steps from h = 0.0 to h = 3.0. Where h = 0.0 the galvanometer would
execute simple harmonic motion without decreasing amplitude indefinitely. V is
would represent a frictionless galvonometer without electrical or air damping, a theo-
retical case.) Time is given in units of the natural period (tn) of the galvanometer.
Note that as the damping ratio increases the "damped period" ta increases. The relation-
ship between the "damped period" and the damping ratio is given by
tn= V 1 -h2xtd (10)
or, fn = fdH- V 1 _h" (IQi)
where tn = natural period of oscillation of galvanometer
fn = natural frequency of oscillation of galvanometer
td = damped period
h = damping ratio
Note that for h = 0.2 the first overshoot = 52 percent of the initial displacement; the
second overshoot 52 percent of first overshoot, etc.
The damping ratio can be determined from the decay curve of the sensor by the
use of the following equation:
h=
1.862n" + (Iog10 -^- ) 2 (ID
An
in which A0 is the first considered amplitude or displacement measured from the rest
position and An is the amplitude on the nth succeeding swing past the rest position.
As an alternative to solving this equation for each test, Figure 8 relates the
damping ratio h with the first overshoot after release, (that is, in Equation (11) AQ =
initial displacement; A1 = first overshoot; n = 1). (Example: In Figure 7, curve 2,
the first overshoot is approximately 52% percent. Referring to Figure 8, with an
abscissa of 52% percent the damping factor h = 0.20, which agrees with the value of
h given for curve 2.)
Response of a second-order sensor to a sinusoidally fluctuating input. In most appli-
cations we are not concerned primarily with the response of the sensor to a step-
function (square-wave) input, but rather with its response to a sine-wave input.
Figure 9 shows the dynamic response of sensors with damping ratios from h = 0.1 to
h = 1.0. This graph shows that for a galvanometer with a damping ratio of 0.2, if the
impressed frequency fj were 0.5 that of the natural frequency fn of the galvanometer
(that is, f,/fn = 0.5), the galvanometer would indicate sinusoidal fluctuations 1.25 times
that of true; when f,/fn = 0.95, the galvanometer would show fluctuations up to 2.50
times that of true (amplitude ratio = 2.50) ; and at ft/fn = 2.0, the amplitude ratio
would be only 0.32, or % that of true. Thus for a galvanometer having a damping
ratio of 0.2 for good fidelity in indicating, the ratio il/ia i> 0.2, or, the impressed
frequency should never exceed 20 percent of the natural frequency of the galvanometer.
This is a serious limitation on the use of the system because often one cannot limit
the input frequency. If the galvanometer were damped to h = 0.64, the galvanometer
would record the^true input amplitude (within ± 2%) for all input frequencies where
the ratio f,/fp =^0.6; and would record input signals less than true for all higher
input frequencies. This value h = 0.64 is the desirable damping ratio for most
94
DATA VALIDATION
-------
implications, and manufacturers generally specify the input circuit resistance needed
to achieve this ratio. (Note that if the galvanometer were critically damped, h = 1.0,
and high-fidelity recording were desired (within ± 2% of true), the input frequency
should never exceed 15 percent of the natural frequency of the galvanometer, a very
serious limitation on the system.)
40 50 60
FIRST OVERSHOOT, percent
Figure 8 Relationship Between the First Overshoot and the Damping Ratio of a Galvanometer
Most galvanometers have a damping ratio of 0.2 or less in a high-resistance circuit.
This would be true of ammeters, voltmeters, etc., were they not provided with air or
electro-magnetic damping. Such meters on open circuit usually have values of h
0.4 to 0.7, depending on their intended use.
Thus to specify the accuracy of recording fluctuating input frequencies, one must
know the damping ratio of the sensor, its natural frequency, and the range of frequencies
of the input signal.
Methods of increasing the damping of sensors with second-order response. As al-
ready mentioned, for best dynamic response (least distortion), the sensor should have
a damping ratio of approximately 0.64 or higher. For electric meters this damping can
be arranged by decreasing the input circuit resistance, or providing air or oil damping,
or both.
In a wind vane with a particular area of vane, little can be done to change the damp-
ing factor C, but the moment of inertia I of the vane can sometimes be reduced without
reducing the torque constant K in Equation (8). For most commercial wind vanes
h = 0.1 to 0.3. Thus Figure 9 shows that such wind vanes will resonate with gusts of
certain wave length, showing fluctuations up to 2 to 3 times the true angular fluctuations.
Gill
95
-------
By reducing the moment of inertia of the vanes (using very light plastics) damping
ratios as high as 0.6 have now been obtained. In this way sensors have been made
that do not erroneously magnify the angular movements of the wind.
015 02 03 04 0.5 06 08 10
MPRESSEO FREQUENCY / NATURAL FREQUENCY
Figure 9 Relationship Between the Damping Ratio of a Galvanometer (or Voltmeter) and its
Dynamic Response to Sinusoidal Input Voltages of Constant Amplitude but Varying Frequency
Sensors with second-order response. Electric meters generally are in this category.
Fortunately, manufacturers usually provide their meters with damping rates of 0.6 or
higher or indicate the circuit resistance that should be used to attain this value.
Flow meters that incorporate a tapered tube with a float have second-order response.
Force sensors in which the force is balanced against a spring (either longitudinal
extension or angular rotation) usually have second-order response.
In all such cases the dynamic response of the sensor must be known or measured
if the accuracy of recording is to be specified.
Note Sensors with first-order response are essentially special cases of second-
order response, in which h = 1.0, that is, the sensor is critically damped.
DYNAMIC RESPONSE OF INDICATING METERS AND RECORDERS
Aronson* lists ten basic types of recorders. Of these, the galvanometer types and
the null-balance types probably account for over 75 percent of the analogue recorders
in routine use.
f M. H. Aronson, "Basic Types of Recorders," Recorder Manual. 1962 Edition
Instruments Publishing Company, Inc.
96
DATA VALIDATION
-------
Galvanometer Types (second-order response)
For most indicators (voltmeters, ammeters, etc.) some mechanical damping is
incorporated in the meter circuit to bring the damping ratio in the region of 0.4
to 0.7. If the damping ratio is not given in the specifications of such instruments, it
can easily be obtained by the use of a circuit similar to Figure 5.
Most direct-writing galvanometer recorders (such as Esterline-Angus and Texas
Instrument) incorporate some internal damping to permit movement of the recorder
without shorting of the terminals. The manufacturer usually specifies the value of
this damping factor or provides a set of typical response curves of the instrument for
step function input of varying internal resistance. Fast-response recorders (such as
Sanborn, Brush, etc.) are not damped in this way, but again the manufacturer supplies
the dynamic response of the sensor. With all of these recorders the manufacturer's
recommended circuit resistance should be used for best dynamic response.
Null-Balance Potentiometer Recorders
Quoting from Aronson
Null-balance recorders are servo-operated devices that are generally referred
to as potentiometers. The basic advantages of the null-balance potentiometer are
(1) high sensitivity, down to microvolt signals, and (2) independence of lead
length. The sensitivity is realized by the inherent amplification in the servo
system; independence of lead length is realized by cancelling out the input signal
so that no signal flows at balance.
These two basic advantages are gained at the expense of response speed;
potentiometers cannot operate at speeds faster than about % second full-scale pen
travel, limiting the response to signals of less than 1 cps. However, at these
frequencies the potentiometer principle opens up vast areas for recording.
Most null-balance recorders have first-order response, and therefore, shows no
overshoot for a stepwise input. In the specifications of such instruments it is usual
to state the time for the recorder to indicate 90 percent or 99 percent deflection after
application of the step function. If either of these is given, it is a simple matter to
obtain the time constant of the recorder by reference to Table 1.
Null-balance potentiometers are made in many different forms. Some types could
have a second-order response; these provide enough damping that negligible or no
overshoot occurs with the step-function input. Accordingly, first-order response can be
expected from most null-balance recorders.
DYNAMIC RESPONSE OF SENSORS PLUS RECORDERS
First-Order-Response Sensor; First-Order-Response Recorder
With this combination, where the time constants of both the sensor and the recorder
are known, the dynamic response of the system may be obtained simply by use of
Equation (6). Such a system will never over indicate fluctuations.
First-Order-Response Sensor; Second-Order-Response Recorder
If the input sensor is slow in response relative to the recorder, then the recorder
damping ratio will approach 0.64 independently of circuit resistances. But if the
response of the sensor is very fast relative to the recorder, the sensor will follow
sinusoidal fluctuations in the variable without overshoot; to avoid overindication of the
input variable the circuit resistance of the system must be designed so that the recorder
damping ratio is 0.60 or greater. (Example 1: Consider a thermocouple sensor having
Gill 97
-------
a time constant of 20 seconds and a galvanometer recorder having a natural frequency
of 1 0 cps (e.g., Texas 0-1 ma recorder). From Figure 3, for the sensor to respond to
90 percent of the sinusoidal fluctuations T/P = 0.08, P = T/0.08 = 250 seconds, or
the impressed frequency to the recorder = 1/250 cps. Referring to Figure 9,
impressed frequency __ l/2jO^_ _^1_ w{,icn js offscale to the left. Thus the recorder
natural frequency 1 250
would faithfully follow the fluctuations without distortion whether h = 0.1 or h = 1.0
or higher. Example 2: Consider a hot wire anemometer, whose time constant is 0.1 sec-
ond, connected to the same recorder. From Figure 3, for the sensor to respond to 95 percent
of speed fluctuations, T/P = 0.05, P = 0.1/0.05 sec = 2 sec, or, the impressed
frequency = 0.5 cps. In Figure 9 when ^pressed frequency = 0.5 = 0 5 ^ damping
natural frequency 1.0
ratio should be 0.60 or greater if no overindication of wind speeds is to be recorded. If the
circuit resistance were such that h = 0.60 for this recorder, the sensor and recorder
would be almost ideally matched for good dynamic operation.)
Other Combinations
The techniques just discussed will also apply to these combinations: second-order-
response sensor with first-order-response recorder; and second-order-response sensor with
second-order-response recorder.
STATIC CALIBRATION OF MEASURING SYSTEM
The term calibration is used to relate the indications of an instrument to inter-
nationally accepted units of measurement. Some recorders, which are built for a specific
purpose, are equipped with special charts that indicate the desired units directly.
Others operate with universal charts, whose values must be converted. In both cases
one subjects the sensor to a series of known values of the variable, notes the corresponding
deflections, and plots the calibration curve. For recorders with special chart rolls that
read units directly, the calibration curve relates errors in indication to values of the
parameter over the instrument range. For recorders with universal charts, the calibration
sheet usually indicates the true value of the variable versus divisions on the chart roll,
and the set of points is joined by a smoooth curve.
If the instrument system has been in prolonged operation, the system should be
calibrated before it is adjusted or serviced. The calibration then applies to the readings
that were taken during previous operation. For future use of the recording system,
the basic sensor and the recorder should be carefully checked before the second calibra-
tion is made. For null-balance potentiometer recorders, one should check the freedom
of operation of the writing system; the absence of backlash in the writing pen; the
absence of end play in the chart-drive roller; performance of the servo-drive system
(as shown by the pen returning to within ± 1/100 inch when deflected to right or
left) ; condition of the battery, if any; and operation, adjustment, and lubrication of all
other moving parts of the system. Galvanometer recorders require fewer adjustments
but should be oiled and checked for proper adjustment before a calibration run.
Full calibration is usualy done in a laboratory, but sometimes it is desirable to
calibrate the sensors and recorder in the field. As an illustration, if the recorder is
normally mounted on * wall that is 30= F warmer or cooler than normal room tempera-
ture, the calibration should be conducted with the instrument in place. If large diurnal
temperature fluctuations occur at the recorder site, the calibration should include tests
98
DATA VALIDATION
-------
to determine any errors due to these fluctuations. If accuracy of calibration within ±
1% of range is required, one may use four or five calibration points to cover the full
scale of the instrument. If the accuracy of ± 0.3% of full range is desired, the system
should be checked for at least ten values. For such accuracies the error involved in
measuring the value of the parameter must be significantly less than the precision
desired in the calibration. (Example: If a temperature system is to be calibrated
within ± 0.3° C, the actual temperature must be measured with an accuracy of at
least ± 0.1° C.) Both the sensor and the recorder must be allowed to come to an
equilibrium position.
Multi-point recorders must be checked for any internal errors caused by the
switching circuit. Usually a full calibration is not required for each multi-pen position,
but it is well to record a particular sensor on each of the multi-points in succession to
determine whether differential heating of the terminal block or of the switching circuits
causes any error. (Example: In a thermocouple temperature recorder of a supposedly
reputable manufacturer, the terminal block was located near one end of the amplifier
system, causing differential heating of the block. This caused a progressive error in
the circuits, so that temperature indications at the circuit nearest the terminal block
differed by as much as 2° C from those at the junctions at the opposite end. When
the terminal block was moved to a point remote from any heat source, this error was
reduced to 0.1° C.)
If nonidentical sensors are used on a multi-point recorder, each sensor should be
calibrated with the recorder.
Present-day good quality recording systems usually require full calibrations only
once a year, if check calibrations are made periodically. One might check a multi-point
temperature recorder by immersing a temperature sensor in a well-stirred bath of carbon
tetrachloride at bimonthly periods. The temperature of the bath would be measured by
a mercury or alcohol thermometer whose calibration was known. If the instrument were
still within the previous limits of error of the system, full calibration would not be
required.
Calibration checks can be built into some systems. (Example: in a resistance
thermometer system one could use one or more precision resistances that would be
automatically sampled at regular intervals, either with a strip-chart recorder or in a
punch card system.)
MISCELLANEOUS FACTORS AFFECTING THE ACCURACY
OF A RECORDING SYSTEM
Proper installation of the sensors and the recorder is imperative for accurate,
reliable measurements. Many excellent instruments in which the basic sensor was poorly
located have yielded observations that were almost valueless. (Example 1: If a wind-
direction-measuring instrument were placed at the recommended height of 30 feet above
ground but located within 60 feet and in the lea of a building 40 feet tall, the recorded
wind directions would not represent the general area but would only indicate the eddies
around the building. Example 2: If an accurate thermocouple system were installed with
the thermojunctions at selected heights above ground but exposed to direct solar radiation
without radiation shields and without artificial aspiration, the temperature readings could
be several degrees high on calm sunny days and several degrees low on clear calm nights,
even though the calibration was accurate within ± 0.3° C over the complete range.)
Gill 99
-------
For reliable observations over periods of months the recorder should be checked
daily at a specified time to insure proper operation and to place time marks on the
chart roll. Daily maintenance should include a check for proper inking, for proper
indication of the time, and for general system operation. Whenever chart rolls are
changed, the operator should place enough data on the starting end of the roll to
distinguish it positively from any other charts that might be used in the system complex.
For instance, the wind direction chart at one level on a tower might be identified as
foUows: "Wind direction, 256 ft level, Charlevoix, on 0803 EST, Feb. 4/62, John Doe."
A similar entry placed on the end of the roll thus completely identifies the chart records.
Generally such instrument systems should be thoroughly checked at about quarterly
intervals. This check should include routine checks on the basic sensors, oiling and
servicing where appropriate, and full maintenance and servicing of the recording system.
Some inking systems require only very occasional cleaning of the pen points and the
ink wells, say, at quarterly intervals. Other systems will require thorough monthly
flushing of the ink wells and weekly cleaning of the pens for consistent fine-line traces.
A careful maintenance and servicing routine can yield good records 99 percent of the
time, whereas moderately careless servicing may yield less than 50 percent.
An alert technician detects trouble before it becomes serious, takes corrective
action, and thus avoids loss of continuous records. He should report any variation from
normal operation to his superior. Servicing personnel should be encouraged to obtain
continuous, reliable records nearly 100 percent of the time; 95 percent is poor; less than
90 percent may make the record almost unusable. Most researchers are frustrated when
even 1 hour of data is missing in a month; 35 hours (5 percent) of missing data con-
stitutes a very serious loss.
The collection of accurate, reliable data is no accident. It is possible only through
proper selection and installation of the measuring system, adequate maintenance and
servicing, careful calibration at regular intervals (interspersed with routine checks),
and continuous alertness for possible errors or failures in the recording system.
100 DATA VALIDATION
-------
Richard S. Green
Chief, Basic Data Branch
Division of Water Supply and Pollution Control
U. S. Public Health Service, Washington, B.C.
SUMMARY
As pertinent data become more widely needed by groups and agencies involved in
air and water pollution control, greater importance must be placed on uniformity of
sampling and analytical procedures and on the accessibility of reliable data from all
sources. Wherever acquired data are likely to have lasting value, serious thought should
be given to some system of storage and retrieval through which potential users of the
data can obtain the information they need in a usuable form at minimum cost and with
reasonable speed and can be assured that all reliable information is included and that
all extraneous information is excluded. To be workable, the system must be compre-
hensive, flexible, and simple. The Division of Water Supply and Pollution Control of
the Public Health Service has devised such a system for storing and retrieving data for
water quality control.
THE STORAGE AND RETRIEVAL OF DATA FOR
WATER QUALITY CONTROL-A SUMMARY
INTRODUCTION
Collecting data and putting data to use costs a great deal of money, as those in this
room know probably better than most others. A relatively simple chemical analysis of a
water sample, with no unusual determinations, for example, costs even the most efficient
laboratory $30 to $50 to run. We ought to get the most out of every dollar spent for such
work, and we can help to reach this goal if all reliable data are made easily available
to those needing them.
Special studies of all sorts produce large amounts of data, but little thought can
usually be given to possible use of the data by others, for different purposes. Because of
variations in objectives and requirements resulting in different quality parameters,
levels of concentration, period of sampling and the like the data in original reports
cannot easily be presented in a uniform format. Moreover, a large body of valuable
data never appears in print at all, but remains in inaccessible files until discarded.
Whenever data are likely to have lasting value, we should give serious thought to
some system of storage and retrieval wherein potential users of the data:
1. Can obtain the information they need in the form they need it,
2. Will be assured that all reliable data, wherever produced, are included in the
material requested, and that all areas of interest are covered.
3. Will not be bothered with data they do not need.
4. Will get this service at minimum cost and with reasonable speed.
Any such system, to be workable, must possess three important characteristics.
It must be:
Green 101
-------
1. Comprehensive have the ability to handle oil possible quality and related
parameters, both those now in use and those that may be significant in the future.
2. Flexible be able to take into account geographic and environmental differences.
3. Simple be relatively easy to use and within reasonable cost range.
Many problems are involved in the design of a system of this scope. Most have
been encountered and solved in a procedure for storing and retrieving data for water
quality control devised for general use in the operations of the Division of Water Supply
and Pollution Control of the Public Health Service. A description of the elements of that
system, which has been named STORET, will bring out many of the basic principles
involved.
ORIGINS
This system was developed from ideas brought together in a brief informal con-
ference held in the Public Health Service about 2 years ago. The thoughts and suggestions
of a few state officials who had been concerned about this problem were contributed
by PHS personnel familiar with their views. Operating procedures in Indiana, New York
State, and Pennsylvania were especially helpful. I should like to pay special tribute to
the skill, tenacity, and patience of Assistant Sanitary Engineer Clarence Tutwiler, of our
staff, who has been responsible for the electronic computer programming required in
this system. He has adjusted and readjusted the storage and retrieval procedures several
times as the full potential of the original concepts brought out in the 1961 meeting have
become apparent.
SCOPE OF SYSTEM
The size and complexity of this data handling problem dictates the use of electronic
computers with their great storage capacity and ready access to selected items. Two
major concepts are being applied uniformly throughout the country, regardless of the
computer equipment or programming technique utilized. These entail:
1. A single procedure for the identification of point locations pertinent to the data,
whether they are water quality sampling points, points of waste discharge or of
water intake, or any other locations for which data are to be secured.
2. A uniform coding system for the identification of specific parameters of water
quality or other items of interest, such as data on flow, precipitation, and the
like.
In the time available I shall only summarize these two concepts. The text of the
full paper describes the storage and retrieval procedures in detail.1
Location Code. The location code permits the retrieval of data in the hydrologic order
that is desirable for studies of basin problems. Since the system is complete and "open
ended," it is possible to identify any point on any stream by it. Once a data point or
station is given its proper location code, this "label" remains with that data point in-
definitely. The location code is not, at present, completely adapted to points in estuarial
waters where interlocking channels cannot easily be fitted into the concept, nor to large
open water bodies. It was deemed unwise, however, to delay the application of other
features of the plan on this account. Some form of coordinate location system will
probably be used for such points.2
102 STORAGE AND RETRIEVAL OF DATA
-------
Parameter Code. One of the most troublesome features in the handling of data, par-
ticularly water quality data, for wide geographic areas is that we find, first, a large
number of different kinds of data of concern to us. Also, in any given type of measure-
ment there are usually widely varying limits to the values reported from place to place:
Chlorides in New England streams are usually low, whereas those in some streams in the
Southwest may be very high. If we provide a fixed field size on a punch card to
accommodate the maximum values to be reported, we waste columns when less than
that number is used. Furthermore, if we make room on a fixed field card for several
parameters of data, not all of which are reported in each use of that card type, we
waste still more columns. The availability of magnetic tape data storage enables us to
overcome these difficulties. The parameter code adopted will handle up to 100,000
different parameters, Blocks of numbers have been assigned for specific parameter
groups, leaving wide areas unfilled for additional future determinations and related data
of interest.3
In this system, the code for any individual parameter is the same wherever the
data are secured, and any potential user of the stored information can call for specific
kinds of data through use of the proper code numbers. A special comment is necessary
with respect to the handling of biological data. The number of individual entities here,
i.e., species of organisms or fish, that may need to be reported and retrieved from
storage for a given station, the need to record supplementary data about each entity,
and the need to store and retrieve these data in associated groups require a slightly
different method of storage and retrieval than that proposed for all other types of data.
The procedures for this modification of STORET to handle biological data are now
being developed.
The problem of variable number of digits in the value for any given parameter is
handled by limiting the reported value to four significant figures, with the decimal point
coded as the applicable exponent of 10.
Statistical Analyses of Data. Early in the design of this system it was decided that no
attempt would be made to build in any procedures for statistical analysis of the stored
data since this would only complicate the job of storage and retrieval. Since the output
from the system can be in the form of data on tapes in prearranged order as requested
by the user, however, subsequent statistical processing is a simple matter.
Status of the System. The programming of all routines involved in this data storage and
retrieval system has been completed. In preparation for full-scale use of the system,
index coding and stream mileage measurements are being undertaken at the present
time by several of the comprehensive river basin projects within the Division of Water
Supply and Pollution Control.
REFERENCES
1. The Storage and Retrieval of Data for Water Quality Control. Richard S. Green.
PHS Publ. No (in press).
2. Location Coding for the STORET System, Basic Data Branch, Division of Water
Supply and Pollution Control, Public Health Service.
3. Parameter Code List for the STORET System, Basic Data Branch, Division of Water
Supply and Pollution Control, Public Health Service.
Green 103
-------
DISCUSSION
Mr. Ransell expressed concern about duplication of eflort in gathering water
quality data and asked whether the storage and retrieval system described would be a
repository for all valid data. Mr. Green indicated that the Public Health Service needs
a national system for its own operations. The system described is capable of accepting
data from all sources.
Mr. Ransell inquired about the organization required for handling input to the
central system from various groups. Mr. Green explained that within the Public Health
Service each comprehensive project is responsible for handling its own data. Therefore,
most of these projects will probably operate within the system as self-contained units.
It is expected, however, that the data from one part of the country will be useful and
assimilable in other locations. Discussions have been held with State representatives and
others about the broader use of a system (not necessarily this system) for handling
data on water quality and related variable data throughout the country. Several partici-
pants of these discussions feel that a truly national system of handling water data should
be formulated. The Public Health Service is interested in this and would be willing to
contribute towards this goal.
Mr. Ransell cited current duplication of data-collecting effort as a reason for estab-
lishing one national collection system. Mr. Green pointed out that the storage and re-
trieval system described here is designed to accept data from various agencies and that
ready availability of this information should be a stimulus toward reducing duplication.
Dr. Gartrell asked whether this storage and retrieval system, which will incorporate
data from the National Water Quality Network, could be used for handling the vast
amount of water quality data collected by the TVA. Mr. Green explained that although
the built-in mechanical features of the system would permit this, the budgetary and ad-
ministrative problems would have to be worked out. The Conference of State Sanitary
Engineers and others are interested in this whole problem area, and future discussions
are anticipated. Criteria for determining whether specific data will be stored or dis-
carded must be formulated by all participating agencies.
104 STORAGE AND RETRIEVAL OF DATA
-------
SESSION 4: Measurements of Air Environment
Chairman: George J. Taylor
Supervisory Air Sanitation Engineer
California State Department of Public Health
-------
John S. Nader
Chief, Physical Research and Development Section
Division of Air Pollution
U. S. Public Health Service, Cincinnati
SUMMARY
Two major automated data acquisition systems are now being used in the United
States for air quality measurements. These systems, operated by the Los Angeles County
Air Pollution Control District and by the U.S. Public Health Service (Continuous Air
Monitoring Program), are reviewed in detail; plans for automated data handling by
the California State Health Department are discussed briefly. Design and operation of
these systems are reviewed in terms of sampling, detection, recording, data validation,
and data display.
DATA ACQUISITION SYSTEMS IN AIR QUALITY
Automatic data collection and data processing in air quality had an early start in
the Air Monitoring Network of the Los Angeles County Air Pollution Control District
(LACAPCD), which was initiated in 1948. The U.S. Public Health Service began its
Continuous Air Monitoring Program (CAMP) in September 1961. Currently, the
California State Health Department is implementing its Berkeley station with auto-
matic digital recording equipment as a pilot study toward a uniformly automated net-
work of stations throughout the state.
This review is primarily directed to the first two of these air quality data acquisition
systems, with respect to their major operations and their component elements. These
networks are essentially the only air quality data acquisition systems that are fully
automated and encompass the various operations in an environmental measurement
system from the sampling of the ambient air to the display of validated data on pollu-
tant concentrations in an accepted tabulation.
An air quality data acquisition system can be shown (Figure 1) to consist of the fol-
lowing basic operations:
1. Sampling
2. Detection
3. Data Recording
4. Data Validation
5. Data Display
The first two operations usually are performed by an integral and automatic
instrument for sampling and analysis. Often the data recording is partially included
as an analog recorder that produces a strip chart recording, which normally must be
converted either manually or instrumentally to digital data to be compatible with
subsequent data handling operations. Consequently, it is convenient to consider the
first two operations as the components of one major subsystem, which generates the
analog data for various parameters under study. The last three operations may be
viewed as another major subsystem, which acts on the analog data to produce an ac-
ceptable display of information for subsequent operations of data analysis, interpretation,
and drawing of conclusions.
Nader 107
-------
Although the Los Angeles County Network takes historical precedence, the air
quality data acquisition system of the USPHS is discussed first for convenience in
presentation. The USPHS system is the more completely automated, particularly with
respect to the recording of the analog data in digital form; in addition, all stations are
equipped with identical equipment.
AIR QUALITY ANALYSIS
DATA REDUCTION AND DISPLAY
Figure 1 Air Quality Data Acquisition System
U. S. PUBLIC HEALTH SERVICE CONTINUOUS AIR
MONITORING PROGRAM
Objectives of the USPHS Continuous Air Monitoring Program1 may be stated as
follows:
1. To provide information on the concentrations in major American cities of
various gaseous air pollutants, which may be related to auto exhaust.
2. To provide continuous data as basic information for research studies, including
a study of programming data generation to optimize routine monitoring pro-
cedures.
3. To provide basic data for the prediction of dosage levels to which people may be
exposed and to which health effects may be related from epidemiological
findings.
Six cities (Chicago, Cincinnati, Philadelphia, San Francisco, St. Louis, and Wash-
ington, B.C.) were selected by the Public Health Service to provide data directly for
these specified objectives. Corresponding data from Los Angeles and Detroit are avail-
able from measurements by the Los Angeles Air Pollution Control District and from a
study of health effects on animals exposed to urban air by Wayne State University
in Detroit.
SAMPLING
Except for minor differences, the stations in five of the cities are essentially the
same in that the shelters were built specifically to house the instruments. In the re-
108
ACQUISITION SYSTEMS IN AIR QUALITY
-------
maining cities facilities already available are utilized. The constructed buildings provide
approximately 400 square feet of floor space, part of which is used to accommodate a
desk for the technician, who is in daily attendance. Two of the buildings are made of
Armco prefabricated metal units and are rectangular in plan. Three of the stations are
of Pease geodesic dome construction (Figure 2). Air conditioning and heating are
Figure 2 Pease Dome Air Monitoring Station, Philadelphia, Pa.
provided, as well as facilities for water, electricity, and sewage disposal. Ambient air
is sampled within 10 to 15 feet above ground level at each station through an air intake
on top of the building.
These stations are located in downtown areas as close as possible to the center
of each city's business district (Table 1). Some of the considerations in selecting a
Table 1 Sampling Station Sites
City
San Francisco
Chicago
Cincinnati
Philadelphia
St. Louis
Washington, D. C.
Los Angeles
Detroit
HuiWing
Garage
Armco
Armco
Pease
Pease
Pease
Laboratory
Laboratory
Location
Union Square Garage, Inc.
Union Square
445 South Plymouth Court
Ann and Central Avenues
c/o Franklin Institute
2031 Race Street
215 South 12th Street
1027 First Street, N.W.
434 South San Pedro Street (13)
St. Antoine and Gratiot
Nader
109
-------
suitable area were openness of the surroundings, availability of utilities, proximity to
atypical sources, and approval of city building commission or other authorities. The
main criterion was that the air being sampled in these locations is typical and repre-
sentative of the air to which people are exposed in downtown areas.
Sampling probes, made of unbreakable glass pipe, are used to introduce the
ambient air through the center of the roof of the building. Inside the building the
1.5-inch-diameter probe branches into 1-inch-diameter arms, which serve as manifolds
from which individual instruments sample (Figure 3).
Figure 3 Interior of the Cincinnati Air Monitoring Station
DETECTION
To eliminate instrument differences as a varying parameter, identical gas analyzers
were selected for the six cities for each of the seven pollutant gases. To provide for opti-
mum performance of the analyzers in terms of reliability, sensitivity, stability, etc., a com-
plete set of specifications was written for each type of gas analyzer. Wherever possible,
specifications were the same for comparable components of different analyzers to
allow for an interchange of components. This uniformity, together with unitized con-
struction in which subassemblies are replaceable as unit components, provides for
optimum maintenance and servicing procedures.
The pollutant gases under study are nitrogen dioxide, nitric oxide, sulfur dioxide,
total oxidants, carbon monoxide, total hydrocarbons, and ozone. The operation of the
analyzers is based on the methods of detection described in the following paragraphs
110
ACQUISITION SYSTEMS IN AIR QUALITY
-------
Nitrogen Dioxide and Nitric Oxide
The operations of the NO and NO, analyzers are interconnected in the sampling
operation. Air sampled from the manifold is analyzed for N02. The effluent from the
N02 analyzer is serially analyzed for NO after passing through a potassium perman-
ganate solution (2.5%), which oxidizes the nitric oxide to nitrogen dioxide. Thus, the
NO analyzer is essentially an N0? detector operating on a pretreated sample.
In both analyzers the N02 is reacted with Saltzman reagent to form a visible
color.2 A ratio photometer measures the color change with respect to the unreacted
reagent and an electrical analog voltage is generated in a photovoltaic cell. The 90 per-
cent response time is about 15 minutes, the time required for the gas absorption and
color formation in the reagent and for the reacted reagent to pass to the point in the
analyzer at which the colorimetric detection is made. The concentration range is 0 to
1.0 ppm full scale.
Sulfur Dioxide
Sampled air is passed through a dilute aqueous sulfuric acid solution containing
hydrogen peroxide. Absorbed S02 is oxidized to sulfuric acid. Concentration of S02 is
detected as the difference in conductivity in the reagent before and after S02 absorp-
tion, since the change in conductivity of the solution is proportional to the change in
its sulfuric acid content.3 The conductivity is measured with conductivity cells in a
balanced bridge circuit on alternating current. Concentration ranges are 0 to 2 and 0
to 10 ppm full scale. Resonse time for full scale reading is less than 1 minute.
Total Oxidant
Oxidants in the sampled air are absorbed in a buffered 10 percent potassium iodide
solution. The reacted solution is measured colorimetrically with respect to fresh reagent
by a ratio photometer with filtered light (350-370 m^) .* Photovoltaic cells generate the
analog voltage. Concentration range is 0.3 ppm midscale and 0.5 ppm full scale. Ninety
percent response time is equal to or less than 5 minutes.
Carbon Monoxide
Carbon monoxide analysis is based on the principle of selective absorption of
energy by the gas to which the instrument is sensitized.5 Air is passed through a sample
cell, through which infrared energy is transmitted from an ac-powered filament source
to a pair of detector cells in series. The detector cells are sensitized with a mixture of
carbon monoxide and argon. The sample side of the detector has a lower concentration
of CO relative to argon than the reference side. Carbon monoxide in the sampled air
is measured as the difference in infrared absorption in the sample and reference de-
tectors. Each detector has a capacitor diaphragm, which moves in response to gas
volume changes brought on by infrared energy absorption. Analog voltage is generated
in proportion to the difference in energy absorption in the two detector cells, thus
giving a measure of CO concentration in the sampled air. Concentration range is 0 to
100 ppm.
Total Hydrocarbons
The operating principle of this analysis is the hydrogen flame ionization technique.6
Sampled air is mixed with hydrogen and burned in a combustion chamber. Combustion
Nader HI
-------
of the hydrocarbon gases in the hydrogen flame increases the production of ions, which
are collected at a collector ring near the flame as a result of an electric potential
applied between the ring and flame. The migration of ions constitutes an ion current,
which is proportional to the carbon atom content of the hydrocarbon pollutant under-
going combustion. Detection of the analog pico amperes generated is by an electrom-
eter. Concentration ranges are 0 to 100 ppm measured as carbon atoms.
Ozone
Sample air is contacted with a solution of potassium iodide to allow a reaction with
the ozone pollutant with the liberation of free iodine. About 0.24 volt is applied to a
sensor electrode cell, and the polarization current produces a thin layer of hydrogen
gas at the cathode. Removal of the hydrogen by its reaction with the free iodine re-
establishes the polarization current and the reaction cycle. For every ozone molecule
reacting in the sensor, two electrons flow through the external circuit. Thus, electron
generation is directly proportional to the oxidant mass concentration. Detection of the
electron flow as a function of ozone content in the sampled airflow is by a microam-
meter. Full-scale concentration is in the range of 0 to 1 ppm. This method of ozone
analysis suffers some interference from oxidants such as NO and NQ2. Therefore, it is
presently referred to as coulometric oxidants analysis to distinguish it from the color-
imetric oxidants analysis.
Response Time
In the analytical methods that involve scrubbing the sampled air in a chemical re-
agent, there is inherent in the method a minimum amount of integrating or response
time such that rapid or peak concentration changes are not resolved but instead are
averaged out over the response-time interval. These periods can range from 2 to 15
minutes for the wet chemical methods. For some of the physical methods, such as those
for CO and hydrocarbons, the inherent response time is relatively short, about 1 minute.
For the hydrocarbon analyzer a volume container was introduced into the sampling
line of the analyzer to give about a 5-minute response time that would correspond to
the printout interval. Similar plans are underway for the CO and coulometric analyzers.
The response time for the various analyzers as they are operated in the system is
measured from the time the specific pollutant at known concentration is introduced at
the sampling probe of the analyzer to the time the analog recorder shows a response
equal to 95 percent of the final concentration. The time lag introduced by the sampling
manifold is 30 to 45 seconds, which should be added to the following response-time
values for the various analyzers:
Analyzer Response Time, minutes
NO, NO, 15
SO2 10
Hydrocarbons 5a
Oxidants (colorimetric) 5
Oxidants (coulometric) 1
CO 1
"Surge bottle attached to sampling line to give longer response time.
112 ACQUISITION SYSTEMS IN AIR QUALITY
-------
Calibration
Calibration is the procedure by whicb correspondence is established between the
electrical analog output of the pollutant analyzer and the pollutant concentration of
the air sample entering the instrument.
The broad geographical distribution of the sampling stations and variety of
pollutant analyzers within a station necessitated considerable attention to calibration
techniques and procedures to assure the collection of accurate and valid data.
The procedure followed in this network involved: (1) the initial calibration, in
which a calibration curve was established for each instrument; (2) the standard cali-
bration check, in which periodic checks of the standardized initial calibration are made
to allow for drift and variations that may occur in the operation of some of the instru-
ment components over a period of time; and (3) the reference calibration, which is
common to all instruments that measure the same pollutant at different stations.
Static calibration techniques are used on the chemico-physical analyzers such as
the colorimetric, coulometric, and conductometric instruments. This method of calibra-
tion is applied to the detection and recording operations of the data acquisition system,
and the sampling and chemical reaction operations are omitted. In one widely used
method, a standard solution, chemically equivalent to reagents that have absorbed and
reacted with known concentrations of pollutant gases, is substituted in the detection
component of the analyzer. The CAMP staff has developed a refinement of this method
by using colored pieces of cellophane. These serve as optical filters and reproduce
the detected property spectral of optical density. Therefore, these optical niters are
checked to determine their pollutant equivalents and are used for static calibration
checks on the colorimetric analyzers.
Dynamic calibration applies to all the operations involved in sampling and analysis
of the gaseous pollutants. This type of calibration must be made initially in the operation
of any instrument.
Availability of gas mixtures containing the desired pollutant gas of known con-
centration and purity is essential to the dynamic calibration of the CAMP analyzers.
The CO and the hydrocarbons analyzers are calibrated dynamically with gas mixtures
contained in pressurized cylinders. These gas mixtures are prepared by CAMP personnel
or are purchased and analyzed at the Sanitary Engineering Center with an analyzer cali-
brated against prepared bag mixtures.10
Some gas mixtures, such as ozone, cannot be prepared to accurately known con-
centrations having sufficient stability for dynamic calibrations in the field. In such cases
the calibration sample must be analyzed concurrently by an accepted reference method.
A dilution board (Figure 4) is used to prepare calibration samples at low concentration
for S02, NOX, and O3. The board provides two sample streams, one for the sampling
and analysis instrument under test and one for manual sampling and laboratory analysis,
the latter serving as a. reference calibration.
DATA RECORDING
All of the various methods of sampling and analysis discussed above generate an
electrical analog signal (Figure 5). The analog strip-chart recorders for the various
Nader 113
-------
tf CLEAN
JJ DILUENT
CONCENTRATED AIR
GAS
BLEED OFF
MANUAL SAMPLE
Figure 4 Dilution Apparatus for Dynamic Referee Calibration
AMBIENT
AIR
SAMPLING
r PARTIC-~
ULATE PRE-
FILTERING
for gas
analyses
CHEMICAL
TREATMENT
PHYSICAL
DETECTION
ANALOG
SIGNAL
S02
NO, NO2 O
HC
03
CO
PARTICULATE
IONIZED SOLUTION
COLOR DEVELOPMENT
FLAME IONIZATION
OXIDATION-REDUCTION
. CONDUCTIVITY
. SPECTROPHOTOMETRY
. ION CURRENT
. ION CURRENT
. SPECTROPHOTOMETRY
. TRANSMITTANCE
- REFLECTANCE
Figure 5 Physical and Chemico-Physical Analyzer Systems
gas analyzers, however, are standardized to permit interchangeability and to facilitate
servicing and maintenance. The output analog signals of the analyzers are either atten-
uated or amplified to be compatible with the 0 to 1 millivolt input range of the analog
recorders. Thus, in the data handling operations from introduction of the analog signal
into the analog recorder to the final data display operation, all operations are common
to each of the pollutant gases (Figure 6).
114
ACQUISITION SYSTEMS IN AIR QUALITY
-------
Analytical data obtained in monitoring air pollution are usually presented in the
form of continuous strip-chart recordings. This system offers several advantages: (1) a
graphic display that can be scanned visually for immediate interpretation of the data;
Figure 6 Data Reduction and Display, USPHS Continuous Air Monitoring Program
(2) relatively instantaneous values in the form of a continuous record that is easily
checked for anomalous data, which might represent malfunction of the detection sys-
tem; and (3) a fairly reliable and accurate measurement system for a nominal price.
Strip-chart recorders were obtained for these analyzers primarily for the advantage
offered in item (2).
The tremendous quantities of data being acquired in this project prohibit the use of
the normal procedure, which requires manual or semi-automatic reduction from the
strip-chart recording of digital data onto punched cards, punched tape, or magnetic
tape compatible with the input of electronic computers. For the seven gases continuously
monitored in the eight cities every 5 minutes throughout the year, 112,896 items of data
are generated in a single week, or approximately 6 million items in a year. To over-
come the problem of handling strip-chart data, a digital punch-tape recorder has been
incorporated with each strip-chart recorder.
This digital punch-tape recorder is a modification of the Fischer and Porter
analog-to-digital recorder (ADR) designed for rotary shaft input (Figure 1). The
modification (henceforth referred to as the modified ADR) involves the addition of a
servomechanism assembly and related electronic circuitry. A retransmitting slidewire
on the strip-chart recorder drives the modified ADR.
Nader
115
-------
The ADR input is an angular positioning of two digitally encoded wheels geared
together in a 100:1 ratio. Each of the wheels presents two digits in range from 00 to 99
in one revolution. The wheels are marked so that a visual reading is indicated at all
Figure 7 Analog-to-Digital Punch-Tape Recorder
times. An electric timer programs the punch mechanism to punch the digital data di-
rectly on paper tape every 5 minutes.
Paper tape is provided with hourly interval markings and 12 punch spacings to
accommodate 12 items of data programmed within the hour. Synchronization of data
punching with time of day is checked visually, and any malfunction of equipment is
detectable.
The servomechanism modifies the rotary shaft input requirements of the ADR so
that analog information existing as ac voltage on the retransmitting slidewire can be
converted directly to digital data on punched tape.9 In principle this system utilizes the
null-balancing-type circuit, consisting of a voltage amplifier (amplifier of a strip-
chart recorder is automatically switcher! in to serve in this capacity) and a balancing
potentiometer and motor, both coupled to the input shaft of the ADR. The modified
ADR is designed to give 3-digit full-scale output (000 to 999) for full-scale signals of
1.0 millivolts.
For a changing signal level at the input, the servomechanism will continuously
follow to maintain a null balance and consequently cause the ADR to give the correct
116
ACQUISITION SYSTEMS IN AIR QUALITY
-------
instantaneous output at all times. The digital tape punch of the analog input takes
place only when the programmer commands the ADR to punch a reading on the paper
tape, at which time the encoder wheels lock in place and the reading at that instant is
punched.
DATA VALIDATION
The digital punch-tape recordings are forwarded weekly from each station to the
Robert A. Taft Sanitary Engineering Center for evaluation. These data are unconnected
upon arrival and as such, coming directly from the analyzer, are treated as "raw" data.
The technician at each station maintains a daily operator's log on each instrument for
each gas pollutant. This log includes such things as calibration checks, zero-drift cor-
rections, instrument malfunctions, and bad data recording as indicated by the strip-
chart recorder. An operator's log sheet and a strip-chart record accompanies each
corresponding punch-tape record sent to the Center.
Punch-tape data received at the Center are transferred directly onto IBM cards
by means of the Fischer and Porter translator so that "raw" data are in a form
compatible with electronic computers. The translation from tape to cards is achieved by
three component instruments: a programmer and reader, which together comprise the
translator, and an IBM key-punch machine. The programmer provides the time and
date the data were taken and a 7-digit identification code, 5 digits to identify the city
and 2 digits to identify the pollutant. The sequencing of time and date is auto-
matically programmed. The reader transfers the punch-tape data of gas concen-
trations to the key-punch unit, in which 1-hour sets of readings per gas (12 items
of data) are tabulated per card. The resulting deck of IBM cards incorporates the "raw"
data in a form that can be handled directly by a computer. The "raw" data together with
the calibration information and corrections indicated in the operator's log are pro-
grammed through the computer to give raw data on 5-minute concentration values in
ppm on magnetic tape. These concentration values are subsequently screened by a
computer program for invalid data to give corrected data on magnetic tape.
DATA DISPLAY
Computer programs have been prepared to present the corrected data in the form
of monthly summaries and daily summaries on punched cards, and for various sta-
tistical analysis studies such as dosage, exposure time, frequency distribution, etc.
LOS ANGELES COUNTY APCD STATION NETWORK
Since its inception the Los Angeles County Air Pollution Control District has
employed an air monitoring network to ascertain the magnitude and character of the
air contaminants. This network has varied both in number of stations and scope of
sampling.11 Late in 1961 a comprehensive physiological study of the problem of air
pollution from auto exhaust was started by the University of Southern California under
a contract with the U.S. Public Health Service. Four locations in the Los Angeles
Basin were selected to conduct the study, in which experimental animals were to be
exposed to the sampled air. A separate but corollary contract was executed between
LACAPCD and USPHS to provide three of the exposure sites and to provide air moni-
toring services at all locations.12 During 1962 three of the LACAPCD Network Stations
were relocated to meet the requirements of the USC-PHS Study.
Nader 117
-------
LACAPCD currently operates 10 air monitoring stations within the confines of
the County. The objectives of the continuously recording automatic instrument systems
employed in this program may be summarized as follows:
1. To support research efforts such as trend evaluation and meteorological, at-
mospheric chemistry, and animal exposure studies.
2. To implement an emergency regulation pertaining to the buildup of certain
known toxicants.
3. To ascertain the effectiveness of air pollution control regulations.
SAMPLING
Currently, LACAPCD Network provides coverage of the Los Angeles Basin for
monitoring air quality by means of 10 stations (Table 2), which include the four sites
Table 2 Air Monitoring Installations
No.
1
51
60
64
68
69
70
71
72
Location
Downtown Los Angeles
El Segundo
Azusa
Pasadena
Inglewood
Burbank
General Hospital
West Los Angeles
Long Beach
Freeway Site
CONTROL CENTER
Street Address
434 South San Pedro Street
359 Maryland Street
803 Loren
862 East Villa
5037 West Imperial Highway
228 West Palm Avenue
1411 North Eastlake Avenue
2351 Westwood Boulevard
3648 Long Beach Boulevard
608 Heliotrope Drive
434 South San Pedro Street
Telephone
MAdison 9-4711
Ext. 66032
Cumberland 3-5967
MUrray 1-8748
ORegon 8-6362
Victoria 9-3642
225-4085
478-6754
424-5420
666-2672
MAdison 9-4711
Ext. 66011
selected for the physiology study. Each of the four sites was selected to fulfill a re-
quirement of the study: the USC Medical School site to represent high levels of both
primary and photochemical automotive-related air pollution; the Burbank site to rep-
resent a fairly densely populated suburban area subject to relatively high levels of air
contaminants; the Azusa site to represent a pollution receptor area as designated by
previous air quality measurements and further defined as one in which measurements
of photochemical pollutants such as ozone are relatively high while those for CO and
NOX are relatively low; and a freeway to represent a location adjacent to a major
traffic artery.
A pyrex-pipe manifold system is used to introduce sampled air to the analyzers.
Air is sampled both from the outside ambient air and from the purified-air control
rooms in the animal studies. In the latter case, for a relatively continuous check of the
contaminant level, a sampling valve operated by a timing circuit obtains samples al-
ternately from the ambient air and from the control room. Cycling time permits
118
ACQUISITION SYSTEMS IN AIR QUALITY
-------
calculation of hourly averages for each sample. This sampling technique is limited to
analyzers having relatively short response time, such as the hydrocarbon instrument.
The longer response time of the oxides of nitrogen instruments requires grab sampling
in the control room and subsquent laboratory analyses.
DETECTION
For the greater part, the LACAPCD Network analyzers are similar in operating
principles to those discussed for the USPHS CAMP analyzers. The flame ionization
hydrocarbon analyzer, coulometric ozone, colorimetric oxidant, conductometric S02, and
colorimetric oxides of nitrogen instruments are the same type as the CAMP instruments.
The remaining instruments have been described elsewhere in some detail,12'13 and only
essential differences from USPHS CAMP equipment will be mentioned here.
Ozone
Measurement of ozone by the ozone photometer is based on ultra-violet absorption
by the ozone at wavelength of 2537 angstroms. A dual-cell differential detector measures
the ambient air stream against a parallel air stream from which the ozone has been
removed by catalytic decomposition by means of a manganese dioxide coated tube. The
difference in UV absorption is a linear function of the ozone concentration.
Carbon Monoxide
Carbon monoxide is measured in a nondispersive infrared analyzer. This analyzer
incorporates a parallel pair of detector cells in a dual-beam arrangement as contrasted
with the series detector cells and single beam of the analyzer in the CAMP instrument.
Paniculate
An automated participate sampler and analyzer determines filterable black aerosols
by light reflectance and transmittance immediately after the sample is collected. Air
is sampled at 25 cfh through a paper filter medium, which is advanced intermittently.
Since the flow rate changes slightly during the sampling interval, three flow rate
measurements are made during each sampling period of 1 hour and the average flow
rate calculated. Detection is by means of a photovoltaic cell, which generates an analog
voltage signal as a function of the light reflected from or transmitted through the
filtered sample spot of particulates.
Calibration
The chemico-physical systems, i.e., oxidants, ozone, oxides of nitrogen, and sulfur
dioxide are calibrated dynamically by use of a dilution system, and samples are ob-
tained simultaneously for subsequent reference analysis as discussed previously. CO and
hydrocarbon analyzers are calibrated dynamically by use of prepared known gas mix-
tures in pressurized cylinders.
DATA RECORDING AND VALIDATION
Analog data generated in the Los Angeles Network are recorded on strip-chart
recorders. Only the Downtown Station has been equipped with the ADR recorders to
provide digital data directly. For the remaining nine stations the analog data are edited
Nader 119
-------
manually and, with supplemental field reports, digital tabulations are prepared manually
for a key punch to produce corrected and validated data (Figure 8).
TABULATION
FOR
KEY PUNCH
Figure 8 Data Reduction and Display, los Angeles County Air Pollution Control Network
DATA DISPLAY
From the corrected data on punched cards, a summary of daily maxima is obtained
directly after sorting and collating. A computer program applied to the collated punch
cards provides Basic Monthly "Tab A," Statistical Analyses, frequency of daily maxima,
and frequency of hours and episodes.
CALIFORNIA STATE HEALTH DEPARTMENT
A brief review of early plans of the California State Health Department will be
of interest with respect to design of their data handling system. Five air pollutants will
be recorded: carbon monoxide, hydrocarbons, oxides of nitrogen, and oxidant. The
data-generating analyzers are very much the same as those in the Los Angeles Net-
work; these are located in a number of stations (about 15) throughout California.
A station at Oakland is now being equipped with digital punched-tape apparatus as a
pilot study. Tentative plans are that the digital raw data will go through a computer
program in conjunction with operator's information to give pollutant concentrations on
punched cards. These data will be stored by a computer on magnetic tape and will
be available for computer programs to give various summaries (Figure 9).
120
ACQUISITION SYSTEMS IN AIR QUALITY
-------
The data-logging system consists of Coleman Digitizers attached to the analog
strip-chart recorders; a Coleman Data Processor, which samples the Digitizer shaft
positions in sequence; a Coleman Tape-Punch Control Unit, which encodes data from
IN O^tKAMUN
1
ANALOG
SIGNAL
1
STRIP
CHART
CORDER
DIGITAL
PUNCHED
TAPE
RECORDER
EDITING
EDITING
r
i
i
L
r
1
1
1
1
COM- j
PUTER |
1
/ARIOUS 1
SUM- L_
MARIES \
J
1
1
1
1
COM-
PUTER
1
-I
1
1
1
1
1
j
! RAW ]
J DATA
,PUNCHED
j CARDS |
r
1
I
1
1
1
L
COM-
PUTER
1
U-
1
1
IcONCENTRA-
1 TION (ppm)
-1 PUNCHED
| CARDS
,
i !
COMPUTER)
i i
i i
L_ r _l
1 MAGNETIC!
1 TAPE 1
1 STORAGE 1
Figure 9 Data Reduction and Display, California State Health Department
the Data Processor and data from a digital calendar and from a digital clock for entry
into a Friden motorized tape punch.
The eight-channel punch tape is standard and operates with serial entry as con-
trasted with the ADR sixteen-channel punch tape with parallel entry. Five channels are
used for binary coded digits: 0,1,2,4,8. One channel is for parity check and another
for "end of line" to identify the end of the cycle. The remaining channel is unwed.
The tape entry for one cycle has the following sequence (Figure 10) :
1. Station number two digits
2. Date and time eight digits
3. Pollutant identification one digit
4. Mode of operation one digit
5. Pollutant concentration three digits
Items 3, 4, and 5 are repeated for each of five pollutants within the cycle.
6. End of cycle identification one digit
This gives a total of 36 character words per cycle. A complete cycle is punched out
within 6 seconds exclusive of any balancing time delay added by the strip-chart re-
corders. The digital recording of the five pollutant readings is made at 5-minute intervals.
Nader
121
-------
Feed holes
Parity bit check (ODD)
(
[
1
j
1
1
;
*r
: L]
^ r
4 <
i 1 l
V
< * i
(
H l
" t
-------
5. Waters, J. L. and Hartz, N. W. An Improved Luft-Type Infrared Gas and Liquid
Analyzer. Instrument Society of America Meeting, Houston. 1951.
6. Morris, R. A. and Chapman, R. L. Flame lonization Hydrogen Analyzer. JAPCA.
11, 467. 1961.
7. Mast, G. M. A New Ozone Meter. Summer Instrument and Automation Conference
of the Instrument Society of America, San Francisco. 1960.
8. Christman, K. F. and Foster, K. E. Calibration of Automatic Analyzers in a Con-
tinuous Air Monitoring Program. Presented at the Air Pollution Control Association
Annual Meeting, Detroit, Mich. 1963.
9. Nader, J. S. and Coffey, W. L. Direct Digital Recording of Air Pollution Measure-
ments. Presented at Air Pollution Control Association Annual Meeting, New York
City. 1961.
10. Altshuller, A. P., et al. Storage of Vapors and Gases in Plastic Bags. Intern. J.
Air and Water Pollution. 6, 75. 1962.
11. Taylor, J. R. Methods and Procedures Employed in the Recordation and Processing
of Air Quality Data. Analysis Paper No. 35, LACAPCD, Los Angeles, 'California.
August 10, 1960.
12. Bryan, R. J. Instrumentation for an Ambient Air, Animal Exposure Project.
JAPCA. 13, 6:254. 1963.
13. Bryan, R. J. and Romanovsky, J. C. Instrumentation for Air Pollution. Instruments
and Automation. 29, No. 2, December 1956.
Nader 123
-------
Dr. Harrison E. Cramer
Director, Round Hill Field Station
Massachusetts Institute of Technology, Cambridge
SUMMARY
Automatic data collection and data processing techniques have found in the past
decade important application in empirical studies of low-level atmospheric structure and
in diffusion problems associated with the operation of nuclear reactors. A typical
data acquisition system comprises jour major subsytems: sensors, telemetry, central con-
trol, and displays. A review of basic features of existing acquisition systems at several
installations shows u, wide variety of subsystem designs and focuses attention on factors
that must be considered in the .selection of system components: sensor response charac-
teristics, sensor location and density, data sampling rates, parity checks, and time or space
averaging techniques. Acquisition systems designed for use in air pollution studies or
control should be capable of handling three scales of meteorological information:
macroscale, mesoscale, and microscale. To illustrate the application of engineering design
criteria, an idealized system is described in detail.
DATA ACQUISITION SYSTEMS IN METEOROLOGY
INTRODUCTION
In meteorological studies of atmospheric diffusion mechanisms and the structure
of turbulence, automatic data collection and data handling techniques have become
practically indispensable. The relatively large number of observations required for a
statistically significant description of characteristic air properties effectively precludes
the use of manual techniques for data acquisition, reluction, and analysis. This trend
has been facilitated by the vastly increased capability of small computers that have
been introduced in the last 3 or 4 years. The introduction of automatic techniques in
meteorological instrumentation systems has proceeded rather cautiously, in part because
of economic factors. Piecemeal procedures and general lack of over-all planning and
system engineering characterized many data acquisition system developments in the
past. Basic uncertainties as to the specific operational or research requirements to be
met by the measurement system and the time and space variability of meteorological
parameters have contributed to this situation.
A typical data acquisition system comprises four major subsystems: sensors, -which
provide electrical or mechanical analogs of meteorological variables; telemetry, which
provides for the transfer of sensor information to a central collection point; central con-
trol, which provides for the interrogation of sensors, the recording and processing of
sensor outputs, and the routing of processed data to displays or storage; and displays,
which present processed data in a convenient form. Figure 1 illustrates these relationships.
This paper begins with a brief summary of the design and operation of existing systems
at various locations. Next, some of the fundamental problems associated with the
measurement of meteorological variables and the basic operations involved in a data
acquisition system are described. Finally, the application of engineering analysis tech-
niques to system design is illustrated by consideration of a hypothetical system designed
to serve the need of an urban air pollution study.
SURVEY OF EXISTING DATA ACQUISITION SYSTEMS
One of the first automatic data acquisition systems was installed at Dugway Proving
Cramer 125
-------
Ground approximately 10 years ago, primarily for the purpose of collecting information
useful in small-scale climatological studies. In this system, measurement of wind speed,
wind direction, air temperature, vertical temperature gradient, surface pressure, relative
RECORD
STORAGE
Figure 1 Schematic Diagram of Basic Components for Data Acquisition System
humidity, and radiation at a number of widely separated stations were periodically trans-
mitted in a digitally coded form over telephone lines to a central collection point. Here
the data were decoded and printed out as numerical sequences by an electric type-
writer. One of the main difficulties experienced in the use of this system was the lack
of an adequate facility for translation of the acquired data into a record-form that can
be processed by an automatic computer.
During the past 10 years other acquisition systems have been developed at Brook-
haven National Laboratory (Brown, 1959), Round Hill Field Station (Cramer, Record,
Tillman and Vaughan, 1961), Argonne National Laboratory (Moses and Kulhanek, 1962),
Oak Ridge, Tennessee (Meyers, 1956), National Reactor Testing Station, Idaho (Islitzer)
and other places. These systems have all been aimed at producing a punched paper
tape suitable for direct processing by an automatic digital computer. The Argonne
Laboratory system contains an automatic programmer, which sequences through read-
ings of the various meteorological sensors on a preset schedule controlled by a digital
clock. This system punches a paper tape, which is read directly through a teletype tape
reader and printed. The tape is subsequently converted to punched cards for processing
by an automatic computer. Characteristics of some of these systems are illustrated in
Figure 2 and Table 1.
The first serious step toward actual on-line (real-time) computation of data was the
Air Force WIND system, which became operational in 1961. This computer-controlled
system automatically acquires micrometeorological data and provides diffusion-prediction
information for operational use on a continuously updated basis (Haugen, Meyers,
Taylor, 1962). Information from the various meteorological sensors is transmitted in
analog form over wire lines to an analog-digital converter controlled by the computer.
The sensor multiplexing, or switching, is controlled from the computer; all readings
are directly processed in the computer, which performs the necessary diffusion compu-
tations and punches summary data on a teletype tape and a typewriter. Although, from
a modern system engineering point of view many of the components of the WIND sys-
126
ACQUISITION SYSTEMS IN METEOROLOGY
-------
Table 1 Idaho Falls Data Acquisition System (Fast System)
Parameter
Temperature
Temperature Gradient
Solar Radiation
Dew Point
Wind Speed
Wind Direction
Vertical Wind Directio
(TYPE 1) C
SFIMSDRS r '
(TYPE II) C
SENSORS . r
(TYPE III)
SFIMSORfi
(TYPE IV)
Sampling Period,
Integration Period Minutes Accuracy
Instantaneous
Instantaneous
Also 60-Min Avg
60-Min Avg
Instantaneous
( 10-Min Avg
| 60-Min Avg
( 10-Min Avg
\ 60-Min Avg
Continuous
2-Min Avg
10-Min Avg
j 30-Min Avg
60-Min Avg
Manual
'OTENTI-
3METERS
3OTENT|-
DMETERS
SYNCRO 1NTE-
RECEIVER GRATOR
EN-
CODERS
INTE-
GRATORS
10 Min LOT
10 Min O.IT
60 Min
60 Min 3%
10 Min LOT
10 Min 3%
60 Min
10 Min ±3.5%
60 Min
Continuous 1.0°
2 Min
10 Min
30 Min
60 Min
Manual
TELETYPE
'
DATEX
PRO-
GRAMMER
PAP
COUNTER TA
" *-" puf,
DATEX
-DIGITAL
CLOCK
ER
PE
CH
Figure 2 Simplified Block Diagram Argonne Meteorological Data Processing System
tern are far from up to date, it is giving faithful service at both Cape Kennedy and
Vandenberg installations and represents a significant step in philosophy and approach
to the handling of this type of meteorological data. A simplified schema of the data flow
in WIND is shown in Figure 3.
Cramer
127
-------
There are currently underdevelopment at various national test ranges computer-
controlled meteorological data collection systems of substantially greater capacity than
the systems now existing, both in terms of numbers of sensors under control and the
REAL-TIME
DIFFUSION CALCULATION
= A x° (o-(0)) (A T
PRINTED COPY
-S-, A T , o-(6), u, x
Figure 3 Simplified Block Diagram Real Time Data Processing for WIND
rate at which these sensors are read. The current trend is strongly toward on-line
computer control of these functions. The advent of the digital computer as a central
component in electronic systems is a phenomenon of the past 5 to 10 years that is only
now beginning to be fully appreciated in the area of meteorological instrumentation.
Flexibility of the modern digital computer enables it to replace literally hundreds of the
special-purpose devices formerly used for the acquisition, filtering, transmission, re-
cording, and processing of data. The replacement of numerous small components and
the elimination of the interface problems, which proliferate when large quantities of
electronic gear are tied together, will usually more than offset the expense of the
computer. At the same time the increase in automation makes it incumbent upon the user
or the system planner to specify more carefully, in advance of equipment purchases,
the system functions and the engineering philosophy to be followed.
BASIC CONSIDERATIONS IN THE DESIGN AND OPERATION
OF METEOROLOGICAL DATA ACQUISITION SYSTEMS
GEOPHYSICAL CONSIDERATIONS
The basic purpose to be served by a meteorological data acquisition system is
generally to provide a satisfactory description of atmospheric structure within a speci-
fied reference volume. In air pollution problems the horizontal dimensions of the refer-
ence volume are fixed by the areal extent of an urban complex, for example, and the
vertical dimension is set by the maximum height attained by any pollutants that are
transported across the complex by the wind. Relevant properties of atmospheric struc-
ture within the reference volume include the three-dimensional distribution of mean air
temperature, moisture, wind speed, wind direction, and the turbulent fluctuations of
the latter two variables. In mathematical terms, the mean value, M of a meteorological
variable obtained from a time series at a fixed point is expressed as
M =
t0 + T/2
M(t)dt
to T/2
(1)
where T is the length of record and t0 is an arbitrary reference. It also follows from
the argument presented above that the value of M at any arbitrary time t is given by
and
M = M M'
M' = O
(2)
(3)
128
ACQUISITION SYSTEMS IN METEOROLOGY
-------
where M' is the departure of M from the mean. Generally, M is a function of both
space and time variables:
M = f (x,y,z,t) (4)
The choice of appropriate integration limits is dictated both by the scale of the problem
to be investigated and the form of the energy spectrum of the meteorological variables.
Most characteristic air properties exhibit a spectrum of variability that is at least quasi-
.continuous (consists of disconnected, continuous segments) over a very wide range of
time or space frequencies. Since the spectrum is quasi-continuous over a broad range
of frequencies, measurements of its properties is limited at high frequencies by the re-
sponse characteristics of the instrumentation and at low frequencies by the length of
record. Pasquill (1962) has shown that the omitted portions of the spectrum can be ap-
proximated by weighting functions of the form
sin2 TT n t , 1 sin2 77- n T
and =
where t is the response time of the measurement system, T is the length of record, and
n is an arbitrary frequency.
To derive meaningful relationships between meteorological variables and diffusion
patterns, for example, it is usually necessary to choose an averaging time T such that
M is reasonably stable.
The space and time variability of meteorological variables is only partially under-
stood at present, and such choices as we have been discussing are usually not routine.
Determination of upper and lower frequency limits that will include the "significant"
portion of the spectrum for a given variable is still, in the last analysis, a matter for
experienced judgment.
Because of the importance of these scaling considerations, it is usually necessary
in meteorological applications to relate the functional requirements to three scales of
observations: macroscale, mesoscale, and microscale. Macroscale data are those nor-
mally used in describing the general weather conditions prevailing over a large area, per-
haps 100 miles on a side. Mesoscale data pertain to the general environmental condi-
tions within a few miles distance and in particular to the deviations of the local param-
eters from the general macroscale weather. Microscale data deal primarily with the
fine structure of the local atmosphere for distances of 1 mile or less. With each of these
scales of observation are associated different sensor and system input requirements, differ-
ent data rates, and different processing and display requirements. A properly conceived
system will provide for adequate integration of all three data streams.
SYSTEMS ENGINEERING CONSIDERATIONS
The proper development of an automatic data acquisition system for meteorological
use involves both meteorology and data systems engineering techniques. Failure to
recognize this fact at the outset results in a system based on many practial compromises
that may fail to meet the application requirements optimally or even at all. The
engineering design of such a system should reflect both the realities of current data
systems technology and the ultimate application for which the measurement system is
intended. It is well worth the effort to develop a systematic plan for the implementation
of the system in advance of choosing specific pieces of hardware. This effort requires
technical skills in the areas of operations research, communications systems, computer
Cramer 129
-------
systems engineering, and programming as well as meteorology to develop a definition o
the system concept adequate for determination of subsystem specifications and require-
ments.
The first step in such a study involves ascertaining and specifying precisely what
is to be measured, when, where, and why. In particular, this includes descriptions of the
significant range of the spectrum of each variable, the appropriate averaging times, and
the portion of the complete time cycle during which information on each variable is
wanted. On this basis one can determine data rates in the various parts of the informa-
tion-gathering network the fundamental consideration upon which engineering design
considerations must be based. It is, of course, also the basis for specification of the
meteorological sensors that comprise the sensor subsystem of the data system. The next
step involves the specification of the engineering philosophy and logical procedures to
be followed in data transmission, as dictated by the inherent data rates; by requirements
for mobility, expandability, and change in the system; and by cost. For example, it
makes considerable difference whether each meteorological measurement is to be re-
corded on « punched paper tape for later computer analysis and study, or whether these
measurements are connected on-line to a computer for automatic real-time sorting and
processing. The choice should be based upon a broad, scientific examination of the
total information-handling problem and total cost. Frequently it is less expensive to
replace a variety of special devices with a central general-purpose control element, even
though the speed and flexibility requirements of the system do not require this.
The third step in this activity entails analysis of data recording and data processing
requirements, including necessary mathematical modeling and calculations. A fourth
step involves a thorough analysis of requirements for display of the processed informa-
tion to a potential user.
Such information, properly organized to reveal the interdependence of the various
functions, in proper detail, is the necessary foundation for specifying system perform-
ance criteria; it is also essential for an organized approach to the technical problems
of subsystem hardware and for the integration of subsystem interface requirements into
the design approach. It is of the utmost importance for the success of a given meteoro-
logical system that this be done in advance of procuring pieces of equipment for the
system and in the light of the broader uses of the meteorological measurement program
that this system is intended to implement.
The foregoing factors must be investigated against the background of various
constraints on the design and implementation of the system. These constraints reflect
known limitations in physical, engineering, economic, and human factors, which will
significantly affect the feasibility of utilizing specific techniques or equipment in the
system. These factors may be grouped in the following categories:
1. Economics The approximate amount of money available for the development
of the data system is usually the overriding constraint on system design. A thorough
analysis of the problem, such as we have described, is primarily aimed at obtaining
maximum performance from the proposed system within the available funds.
2. Physical Environment The geographical and climatological regime in which
the system must operate is a factor that automatically rules out certain approaches.
Included in this category are considerations of morbidity, or the ease with which the meas-
uring system is available for relocation.
3. Personnel The training, capabilities, and number of personnel required to
130 ACQUISITION SYSTEMS IN METEOROLOGY
-------
operate and install the system can have a significant effect on system design philosophy
and procedures.
4. Interfaces of the Meteorological System with Other Functions A meteorologi-
cal data collection system designed for furnishing information to an air traffic controller
entails significantly different problems from one designed for study of atmospheric
pollution around a city.
With the information developed from analysis of these factors we can establish
system functional objectives consistent with the broad application requirements, tech-
nological state-of-the-art, and so-called "practical" limitations on the system design and
development. These functional objectives properly described with their relationship to
one another, form the basis for more detailed determination of the specific requirements
for each of the subsystems: sensors, communications, data processing, and display.
The final phase of the design study involves specification of requirements for each
subsystem and its important components. For the sensor subsystem characteristics such
as the following must be specified:
1. Form of the sensor outputs (digital or analog).
2. Accuracy, resolution, and range of the sensor readings.
3. Required ruggedness and reliability, including protection from the environment,
and electrical and mechanical functioning.
For the communication subsystem, major considerations will be:
1. Function of the communication system over fixed-wire channels or by radio
telemetry.
2. Necessary bandwidth as determined both by the data rates generated from the
meteorological measuring instruments and by reliability considerations.
3. Coding techniques, particularly whether signals are to be transmitted in
analog or digital form.
4. Power requirements.
5. Physical maintenance requirements.
For data processing subsystem we must specify:
1. Which parts of the processing are to be automatic and which are to be handled
manually.
2. Appropriate processing speeds and memory requirements for the automatic
data handling.
3. Necessary automatic (real-time) computer inputs, if any.
4. Appropriate forms of recording data not directly entered into the computer, such
as punch cards, punch paper tapes, or strip-chart graphs.
5. Signal conversion operations, such as analog-to-digital conversion.
6. Computational requirements, such as objective forecast models, turbulent diffusion
models, and atmospheric statistics.
7. Data recording requirements and appropriate forms, such as printed copy,
magnetic tape, punch cards, etc.
Cramer 131
-------
For the information display subsystem of our data collection system, appropriate
forms and numbers for display devices and materials must be indicated. These include:
1. Necessity for automatic moving displays, such as cathode ray tubes and various
types of projection systems.
2. Printed record requirements.
3. Automatic alarm signal requirements.
4. Physical location and number of displays required.
HYPOTHETICAL METEOROLOGICAL INFORMATION SYSTEM
The system analysis techniques discussed above can be illustrated by a specific
example. Let us consider a rather comprehensive system of measurements over a refer-
ence area of a few hundred square miles; say, a 20-mile by 20-mile square. In addition
to knowing the general regional weather conditions that affect the area in which this
square is located, we desire information on the local meteorology (mesoscale) ; further,
we are interested in analysis of the turbulence structure on certain sub-regions of this
square, say of about a few hundred yards on a side (microscale). To bring out all
facets of the problem, we will assume stringent performance requirements for the data
system. Such requirements are frequently met in missile launch control and chemical
weapons testing, for example. More to the point, we believe that through careful pre-
liminary planning such scope and flexibility can be achieved at no more cost than
that of many current systems having restricted information-gathering and processing
power.
We will suppose that we are interested in obtaining all of the meteorological data,
including that pertaining to the turbulence structure, in digested form essentially in-
stantaneously (in real-time). Hence we adopt the premise that our measurements will
read directly into an on-line general-purpose digital computer, which will then furnish
printed and moving displays. In addition we will assume that all data are to be re-
corded for future research and examination.
Macroscale weather data are to be entered manually into the computer in order
to furnish information on general weather conditions to be expected. Mesoscale data
acquired from our local meteorological instrumentation system are to be used for trac-
ing and predicting the gross trajectory of various air contaminants. This tracing will
be further refined through knowledge of the atmospheric turbulence structure obtained
from our microscale measurements. We will refer to the measurements pertaining to
mesoscale data as the meso-network, or simply mesonet, and those from the microscale
measurements as the micronet.
The micronet must furnish information concerning temperature, relative humidity,
wind speed, and wind direction at selected points over the reference area. These measure-
ments must be made at sampling frequencies and over time intervals consistent with
the form of the power spectrum characteristic of each parameter. In particular, the key
to turbulence structure is in wind-direction fluctuations. Experience shows that frequen-
cies up to 2 or 3 cycles per second in the wind-direction vector frequently contain
significant energy. To measure this portion of the wind energy spectrum, wind direction
must be sampled at a rate of, say, 10 times per second, at least during those intervals of
time when we are seriously concerned with the effects of small-scale atmospheric turbu-
lence. Temperature and relative humidity measurements will be made once per minute
throughout our network.
132
ACQUISITION SYSTEMS IN METEOROLOGY
-------
Our mesonet will be designed to furnish spatial measurements of atmospheric
pressure, temperature, temperature gradient, and net radiation, in addition to the wind
measurements. An effective mesonet might consist of 20 to 25 sounding stations capable
of providing meteorological data up to altitudes of 3,000 feet. Each sounding would
furnish measurements of vertical distribution of ambient temperature, relative humidity,
wind speed, and wind direction at 100-foot intervals. In addition a single net radiation
sensor could be located at each sounding station. Although measurements of tempera-
ture and relative humidity would be obtained by direct measurement from the sound-
ing instrument package, wind data must be computed from successive position informa-
tion. We further envision that these position data are provided by simultaneous auto-
matic tracking of perhaps five sounding packages. The location of the individual
measurement stations is shown in Figure 4; a mirror image of this configuration, aligned
along the opposite direction, is not shown in the figure. The micrometeorological meas-
urement network is contained within the triangular array.
O300-FT TOWER
o 100-FT TOWER
02-m TOWER
AMESONET
STATION
Figure 4 Hypothetical Networks
The instrumentation subsystem associated with these networks must be capable of
supplying reliable data to support two measurement objectives: (1) to obtain opera-
tional data required for routine estimates of pollution levels, and (2) to obtain the
measurements that provide background information necessary for analysis and research.
Often these requirements are significantly different from the operational requirements,
both in the quantity of data required and in the timeliness of the information. The
proper development of an automatic data acquisition involves the fusion of these
differing needs into a well-organized system design.
BASIC MEASUREMENT REQUIREMENTS
The results of a synthesis of these requirements mentioned above for both the
operational needs and for a research program are summarized in Table 2. Ten levels
Cramer
133
-------
of measurement are provided for a 300-foot tower, eight levels of measurement for e
100-foot towers, and a single level for the 2-meter towers. The number of sensors °
each type and the total number of sensors are given for each network. In addition^ to
the number of sensors, the maximum sampling rate for each type of sensor ^is of Vital
concern to the systems engineer. The highest sampling rates are associated with bi-vane
measurements of the horizontal and vertical components of the wind vector. Under
normal operation, these instruments would be sampled at the rate of once per second
(I/sec) with intermittent periods of the high data sampling. Aerovane measurements
of wind speed and direction taken at the 2-meter towers are sampled once per second
(I/sec). Temperature, temperature gradient, and dew point instruments can, how-
ever, be sampled as slowly as once per minute (1/60). This set of measurements gen-
erates our micronet data.
Table 2 Measurement Configuration for Hypothetical System
MICRONET
Parameter
Temperature
Dewpoint
Temp. Gradient
Radiation
Azimuth Wind
Elevation Wind
Wind Speed
Wind Direc.
Total
One 300-Foot Tower Four 100-Foot Towers Twelve 2-Meter Towers
No. of
Sensors
10
10
1
4
4
10
6
45
Max Sampling
Freq, Sec -1
1/60
1/60
1/60
10
10
10
10
No. of
Sensors
28
24
8
8
8
76
MESONET
Parameter
Temperature
Dewpoint
Azimuth Angle
Elevation Angle
Time
Total
No. of
Sensors
15
15
15
15
15
75
Max Sampling
Freq, Sec ~l
1/12
1/12
1/12
1/12
1/12
Max Sampling
Freq, Sec ~1
1/60
1/60
10
10
10
No. of
Sensors
12
12
24
Max Sampling
Freq, Sec -1
1
1
SUMMARY
No. of
Sensors
48
24
75
73
Total 220
Max Sampling
Freq, Sec -1
10
1
1/12
1/60
To produce a reasonably complete description of the mesoscale features of the
atmosphere, rocket- or balloon-launched instrument packages are timed to transmit
temperature and dew point data at the rate of five times per minute. Azimuth and
elevation angles of the package are obtained by ground tracking instruments. Table 2
also shows the measurement program for the mesonet.
A total of 220 separate meteorological measurements are provided by these net-
works. Equally important to the system analysis is the summary of various categories of
sampling rates. Only about one third of the sensors are sampled at rates of once per
second or greater, with the higher rates occurring only intermittently. This information
134
ACQUISITION SYSTEMS IN METEOROLOGY
-------
must be evaluated along with prescribed system timing requirements to develop initial
estimates of the system data loads.
TIMING REQUIREMENTS AND DATA RATES
The combination of operational and research requirements imposes rigid timing
constraints on performance of the measurement program necessary for pollution predic-
tion. The next stage in the development of the system is to investigate the effect of
these constraints on the sequences of measurements and the necessary time relation-
ships between major system functions. These considerations are at the heart of the
real-time operational problems associated with on-line computer systems such as we
are considering. We again consider our hypothetical networks as the basis for illustrat-
ing the development of system specifications.
It is convenient to consider a typical 8-hour period, which we divide into two princi-
pal operational phases. The first phase consists of preparations and includes such ac-
tivities as subsystem activation, system checkout, and preliminary atmospheric sampling.
When a serious pollution problem is foreseen, we move to a second-phase measurement
program, gathering more data, particularly on small-scale turbulence structures.
The timing requirements for such a program are illustrated in Figure 5. During
phase one, 10-minute periods of high-frequency sampling are scheduled at hourly inter-
vals along with mesonet releases 15 minutes in length. At these peak periods, data rates
will be approximately 6600 bits/sec where the estimates of data rates are based on a
binary code with 13 bits/word. This 13-bit word format assumes only data information
and does not include bit requirements for control purposes. The addition of identification
and control data to the message format can be included in a final evaluation of data
rates by proportional scaling of the values shown in the graph. These data rates also
will require modification as a consequence of the method of sensor interrogation em-
ployed and the number of re-transmissions used for checking and validation.
10,000
1,000
ce
<
100
ASSUMES: 13 bits/word
v«
lu.
u
u
I .
4 3 2 1 Z 1 2 3 4
TIME, hours
Figure 5 Meteorological Information Data Rates
Cramer
135
-------
During the second phase, the high-data-rate periods are repeated at half-hour in-
tervals. The frequency release of mesonet sounding package, however, decreases toward
the end of this phase. During low-data-rate sampling, the curves show a steady system
data load of approximately 1000 bits/sec.
Another aspect of the data load picture, particularly from the standpoint of data
recording and storage requirements, is illustrated in Figure 6. This curve shows the
time profile of accumulated meteorological information. If all data from our hypothetical
system are recorded during the single 8-hour experiment, a total of 5.6 million data
words will be accumulated.
~T
4
3
2
1 Z 1 23
TIME, hours
Figure 6 Cumulative Dafa Words Versus Time
This type of information and the procedure followed in its development are
essential steps in the preliminary phases of the system design. The estimates thus de-
rived may be modified at a later date but serve the very important role of providing
a basis upon which initial specifications of the sensor, telemetry, data processing, and
display subsystems may be made.
Let us review here the functional requirements of this system. We are interested
in obtaining all of the meteorological data, including that pertaining to the turbulence
structure, in digested form essentially instantaneously (in real-time) to facilitate short-
term pollution predictions. Our measurements will read directly into an on-line general-
purpose digital computer. The computer will then furnish printed and moving displays.
In addition we have assumed that all data are to be recorded for future research and
examination.
SUBSYSTEM CHARACTERISTICS
Analysis of these functional requirements and data rates subject to the above design
considerations leads to the following description of basic characteristics for the four
major subsystems.
1. Sensors Table 3 presents a summary of the number and required accuracy for
each type of sensor to be installed in the hypothetical system. This information may
serve as a basis for further investigation of appropriate specifications and for surveying
commercially available sensors. Results of this survey may indicate that accuracy re-
quirements must be relaxed if commercially available sensors are to be used.
136
ACQUISITION SYSTEMS IN METEOROLOGY
-------
Table 3 Summary of Sensor Subsystem Specifications
Parameter
Minimum
Type Number Accuracy Digitizing Interval
Temperature (Temperature Thermocouples 34
Gradient)
± .02°C
1.0 Min
Wind Speed
Anemometer 30 5% of Wind Speed 0.1 Sec
or 2 ft/sec
Horizontal Wind Direction
Horizontal i
Wind Vane
y . i >Wind Direction Bi-Vane
Dewpoint
Net Radiation
Dewcell
Radiometer
Total
18
24
38
1
145
1°
1°
1%
2%
1.0 Sec
0.1 Sec
1.0 Min
2. Telemetry To transmit the measurements from remote sensor locations to
the central location for processing, a telemetry system is required. Accuracy and relia-
bility dictate the need for digitizing of all data to be transmitted over any significant
distance. The data rates indicate a peak transmission rate of approximately 6600 bits/
sec. Control and checking requirements may increase the necessary telemetering capac-
ity by a factor of 2 to 4. These factors strongly indicate the use of r-f data links. If we
allow an additional factor of 2 to 4 for adequate modulation and signal-to-noise ratio,
a bandwidth of from 50 to 100 kc will be required for a single r-f communication channel.
Radio transmission problems peculiar to the area over which the system is to oper-
ate may dictate a requirement for the use of relay stations. In the hypothetical system
it is assumed that one relay station is required because of obstructions in the center
of the area. Separate channels are required for transmission and reception and for
communication between the relay station and the central data collection and between
the relay and the remote sensor location. This leads to the requirement for five separate
50 to 100 kc channels. These requirements are listed in Table 4.
Table 4 Summary of Communications Subsystem Specifications
Type of Transmission r-f
Frequency 200-250 Megacycles
Number of Channels 5
Bandwidth 50-100 kc
Number of Relay Stations 1
3. Data Processing The central control element of the meteorological measure-
ment system is a digital computer. The need for this element to be a general-purpose
computer is of major importance, since flexibility is a primary objective of the hypo-
thetical meteorological measurement system. This flexibility cannot be achieved with a
special-purpose computer with wired-in programs. The computer controls the reading
sequence of the meteorological sensors, actuates the digitization and communication
operations, computes necessary control parameters and test decision criteria, and oper-
ates printed and moving display outputs. In addition the computer edits and records
all raw measurements on magnetic tape. With the speeds and asynchronous control
available on small computers today it is reasonable to suppose that diffusion predictions
Cramer
137
-------
and research calculations are time-shared concurrently with these operations. Table 5
presents a listing of basic features of the computer.
Table 5 Summary of Data Processing Subsystem Specifications
DIGITAL COMPUTER CHARACTERISTICS
8,000 Words of Memory (Expandable to 16,000)
20 Microseconds Memory Cycle Time
Magnetic Tape Units For Recording
Auxiliary Drum Memory
Automatic Interrupt
Real-Time Clock
Simultaneous Computing With Input and Output
Line Printer or Flexowriter
Paper Tape Reader-Punch
4. Display The display requirements are derived from the need for real-time
visual presentation of the wind profile over the entire area. To accomplish this we
propose a cathode ray tube display (CRT), which displays the wind profile as vectors
the length of which are measures of the wind speed and the direction of which indicate
wind directions. Such a vector will appear for each tower location at a fixed sensor
height and the height level chosen will be under control of the observer.
In addition to the wind profile it is desirable to observe the trends of the meteoro-
logical parameters. From these trends it would be possible to make short-term predictions
of these parameters. Commercially available trend recorders fulfill this requirement.
In many cases there are some sensors which are of vital importance to the observer.
These important sensors require so-called ''go-no-go'' displays which display a red light
if the measurement being reported is outside tolerable limits, green if it is within
tolerance, and a third color if something appears wrong with the sensor.
Additional printed display of parameters and summaries of computed quantities
can be furnished through an electric typewriter. These requirements are listed in
Table 6.
Table 6 - Summary of Display Subsystem Specifications
Type Number
Cathode Ray Tube Display 1
Trend Recorders ., 50
Automatic Go-No-Go Displays 60
Automatic Alarm
ACKNOWLEDGMENTS
Three staff members of Systems Research Laboratory of Geophysics Corporation of
America made important contributions to the contents of this paper. Mr. David D. Dix
is responsible for much of the material on basic system design. Mr. David Farrell and
Mr. Paul Morgenstern developed the data rates and other details of the hypothetical
system.
138 ACQUISITION SYSTEMS IN METEOROLOGY
-------
REFERENCES
Brown, R. M., 1959: An Automatic Meteorological Data Collecting System, /. Geophys.
Res., 64, 2369-2372.
Cramer, H. E., F. A. Record, J. E. Tillman, and H. C. Vaughan, 1961: Studies of the
Spectra of the Vertical Fluxes of Momentum, Heat, and Moisture in the Atmospheric
Boundary Layer, Annual Report (Contract DA-36-039-SC-80209), Mass. Inst. of Tech.,
130 pp.
Haugen, D. A., R. F. Myers, and J. H. Taylor, 1962: Design and Development of a
Micrometeorological Data Observing and Processing System for Air Pollution Appli-
cations at Cape Canaveral and Vandenberg Air Force Base, Paper Presented at Fourth
Conference on Applied Meteorology, American Meteorological Society, Hampton,
Virginia, 10-14 September 1962, 22 pp.
Moses, H. and F. C. Kulhanek, 1962: Argonne Automatic Meteorological Data Processing
System, /. Appl. Meteor., 1, 69-80.
Myers, R. F., 1956: A Weather Information Telemeter System, Bull Amer. Meteor. Soc.
37, 108-117.
Islitzer: Personal Communication.
Cramer 139
-------
Dr. Oscar J. Balchum
Hastings Associate Professor of Medicine
University of Southern California School of Medicine, Los Angeles
and
Dr. Frank J. Massey
Associate Professor of Biostatistics
University of California at Los Angeles
SUMMARY
// we are to determine the effects of air pollutants on the human respiratory sys-
tem, we must know more about the physical and mechanical properties of the chest and
lungs. Techniques for measuring air pollution effects must be sensitive, accurate, and re-
peatable. Application of computer analysis to these measurements should then yield
useful and reliable information on data acquisition systems for physiological studies.
Investigations now under way are described: the parameters measured, the recording
and coding of data, and the methods of analysis.
DATA ACQUISITION SYSTEMS IN PHYSIOLOGY*
INTRODUCTION
During the past decade researchers have suspected that chronic exposure to low levels
of foreign gases and aerosols is a factor in the etiology of chronic respiratory disease.
Investigations of the physiological reactions of the lungs of animals and man to low
concentrations of particulates, aerosols, and gases, singly and in combination, have
begun only recently. Although measurement systems are being developed for these
studies, the techniques developed thus far are not sufficiently sensitive, accurate, or re-
peatable. Because our knowledge of the properties of the chest wall and lungs is meager,
it has been difficult to measure small degrees of response to low levels of irritants. Both
increased knowledge and improved instrumentation are required for effective investiga-
tions of the physiological effects of pollutants in the ambient air.
In work now under way, various properties of the respiratory system are measured
and recorded in forms suitable for computer processing. This presentation describes
the parameters measured, the recording and coding of data, and the methods of analysis.
MEASUREMENTS
At present the physiological reactions of the lungs are described by data obtained
in a battery of tests.1 These tests may be classified on the basis of the property of the
respiratory system being measured.
VOLUMES OF THE LUNGS (Figure 1)
Vital Capacity (VC)the greatest amount of air that can be exhaled after the
deepest possible inhalation. Vital capacity is measured by means of a 13%-liter Collins
*Supported in part by the grant for the National Institutes of Health (AP207) and
by a contract with the Air Pollution Division of the United States Public Health
Service (PH 86-62-162).
Balchum 141
-------
Spirometer. Values are related to sex, age, and height; prediction nomograms have been
established from these values.
Examples: Men 39-50 years, VC = 3450 ml
Standard Deviation (s) = 980 ml
Coefficient of Variation (CV) * = 28%
Standard Error (SE) = 80 ml
Women 40-67 years, VC = 2880 ml
s = 630 ml
CV = 22%
SE = 60 ml
* Coefficient of variation, or the SD expressed as a percentage of the mean.
Functional Residual Capacity (Resting Level or FRC) the amount of air in the
lungs at the end of an ordinary exhalation.
Examples: Men 2180 ml
s = 690 ml
CV = 32%
Women 1830 ml
s = 420 ml
CV = 23%
Residual Volume (RV) the amount of air still remaining in lungs after the deepest
possible exhalation, measured by the helium dilution method. The apparatus consists of
a closed circuit, with a spirometer, a C02 absorber, and a helium thermoconductivity
cell in series. The circuit is filled with 15 percent helium in air. The patient wears
a nose clip and is attached to the two-way valve of the apparatus by a rubber mouth-
piece. At the end of an ordinary exhalation, he is connected to the circuit and then
breathes the 15 percent helium in air mixture.
Oxygen is supplied to the circuit at the same rate at which it is consumed. After
helium has diffused into the lungs so that equilibrium or a plateau of the concentration
curve has been achieved, the final concentration of helium is read on the galvanometer.
Calculation of FRC: (Initial He Cone) (Vol of Circuit) = (Final He Cone)
(Vol of Circuit + FRC).
The only unknown, the FRC, can then be calculated. Residual Volume (RV) is
obtained by subtracting the Expiratory Capacity (obtained from the spirogram) from
the FRC.
Examples: Men 1140 ml
s = 430ml
CV = 38%
Women 995 ml
s = 280 ml
CV = 28%
Total Capacity = Vital Capacity + Residual Volume.
Examples: Men 4590 ml
s = 1300 ml
CV = 28%
142 ACQUISITION SYSTEMS IN PHYSIOLOGY
-------
Women 3880 ml
s = 800 ml
CV = 21%
RV/TC % = Residual Volume x
Total Capacity
Examples: Men 24.9%
s = 4.7%
CV = 19%
Women 25.7%
8 = 4.7%
CV = 18%
TLC: TOTAL LUNG CAPACITY IR:
VC: VITAL CAPACITY TV:
RV: RESIDUAL VOLUME ERV:
1C: INSPIRATORY CAPACITY REL:
FRC: FUNCTIONAL RESIDUAL
CAPACITY
INSPIRATORY RESERVE
TIDAL VOLUME
EXPIRATORY RESERVE VOLUME
RESTING EXPIRATORY LEVEL
Figure 1 Definition of Lung Volumes.
VENTILATION
Maximum Breathing Capacity (MBC) the greatest amount of air that can be
inhaled and exhaled into and out of the spirometer. The subject is asked to breath
in and out as rapidly and as deeply as necessary in order to ventilate as much air as
possible. This is done at a rate above 80 breaths per minute for 12 seconds. The resulting
volume is multiplied by 5 to obtain MBC in liters per minute.
Examples: Men 103 1/min
s = 32 1/min
CV = 28%
Women 89 1/min
s = 20 1/min
CV = 22%
Forced Expiratory Volume (FEV or Timed Vital Capacity) the amount of air
exhaled (upon command) as rapidly and completely as possible after the deepest possible
breath, with the drum of a spirometer rotating at 600 or 960 mm/min. The FEV can be
measured from the spirogram.
j: the liters of air exhaled during the first second of a forced expiratory volume.
Balchum
143
-------
FEV3 : the liters of air exhaled during the first three seconds of the FEV.
Volumes are corrected to BTPS, or body temperature (37°C), saturated. If a lung
volume was measured at a spirometer temperature of 25 °C, and 750 mm Hg atmospheric
pressure, the correction would be:
., , v, (273 + 37) (750 _ 24)
VolumeBTPS - Vol ^ + ^ (750 _ 47)
24 and 47 mm Hg-are the water vapor tensions at room and body temperatures,
respectively.
The FEV and FEV, are also used as a percentage of the total FEV:
FEV ,% = FEVi X 100
FEV
Average FEV, = 83% (Range 70-90%)
Average FEV3 = 97% (Range 90 - 100%)
Example: FEV1 = 3.93 1, s = 0.67 1
FEVX = 82.0%, SE = 1.12
FEV3 = 97.6% (Range 92-100%)
Maximal Mid-Expiratory Flow Rate (MMF)
MMF 25_T5% = the rate of flow of air during the middle 50 percent of the FEV,
expressed in liters per second.
Normal values not well established.
Example: 4.49 I/sec.
s = 1.3 I/sec
SE = 0.25 I/sec
Peak Flow Rate the maximal or peak rate of air flow measured during a rapid
or blastlike exhalation (after a deep inhalation) into a Wright Peak Flow Meter. The
meter consists of an encased light vane that rotates and stops at the point of peak flow.
The dial of the instrument is calibrated in liters per minute.
Other instruments (Puff meter, Pneumotachy graph) consist of small resistances, the
pressure drop across which is proportional to rate of air flow. The Pufrmeter has as
its resistance a porus cuplike grinding wheel, and the Pneumotachygraph a 400-mesh
stainless steel screen. Recordings are made by use of amplifier and strip-chart recorder
units of suitable frequency response.
LUNG MIXING
Rate of fall of helium concentration during performance of residual volume
measurement.
7-Minute Lung Nitrogen Washout Oxygen is inhaled by the subject, who
breathes in a normal fashion. The nitrogen concentration is recorded during each ex-
halation and falls as the subject continues to breathe oxygen. Normal air distribution
and lung mixing will result in a nitrogen concentration less than 1.5 percent in 7
minutes, with no appreciable increase during the performance of a maximal exhalation
at this time.
144 ACQUISITION SYSTEMS IN PHYSIOLOGY
-------
LUNG DIFFUSION
The rate of passage of a tracer gas (0.05% carbon monoxide) from the air sacs
(alveolae) of the lungs into the blood, as measured by an IR spectrophotometer. The gas
is rapidly taken up but not released by the hemoglobin of the red cells and therefore
exerts little back-pressure.
The concentration of CO breathed in and that exhaled, and the volume of gas mix-
ture breathed are recorded during a period of 4 to 5 minutes. During the latter half of
such a period the rate of passage of CO from the inhaled CO-air mixture becomes
constant, the CO concentration of the exhaled gas mixture forming a plateau on the
record.
Indices Calculated
Uptake of CO, % = (Min.Vol) (C0insp - C0exp)
(Min. Vol) (C0lnsp)
Example: 51.1%, s = 4.66%
Diffusion Capacity, ml/min per mm Hg =
(Min. Vol) (C0insp - C0exp)
(End-Tidal Cone. CO) (Barometric Pressure - 47)
Example: 23.3 ml/min per mm Hg
s = 4.93 ml/min per mm Hg
Mean difference between first and second paired estimates = 0.64, s = 3.12
AIRWAY RESISTANCE AND THORACIC GAS VOLUME
BY BODY PLETHYSMOGRAPHY
Airway Resistance the subject sits in an airtight box or plethysmograph. Box
pressure (Box P) is recorded by means of a strain gage and rate of airflow by means
of a pneumotachygraph screen while the subject pants at a rate of about 120 times
per minute at the FRC level. A vector loop of flow rate (Y axis) is plotted against box
pressure (X axis), and the slope of the long axis of the vector loop measured on a CRO.
A few seconds later, at the instant exhalation reaches the FRC level, a solenoid
completely obstructs the tube between the patient's mouth and the pneumotachygraph
screen while he is still panting.
Mouth pressure (Y axis) is plotted against box pressure (X axis) on the CRO. During
the few seconds of complete obstruction, the pressure in the alveolae (air sacs) of the
lung is considered to be in equilibrium with that in the mouth, and alveolar (ALVP) or
lung pressure = mouth pressure.
Resistance of Airways, cm HO per I/sec =
Flow Rate/Box P
Example: 1.5 cm H20 per I/sec
s = 0.37 cm H20 per I/sec
Thoracic Gas Volume (TGV), liters = 97°
AP A V
Balchum 145
-------
[Note: 970 cm H2O is the atmospheric pressure minus water vapor tension at body
temperature (37°C), the conditions in the lungs.]
Example: 2.97 1
s = 0.22 1
SE = 0.07
THE RECORDING AND CODING OF DATA
After computation with pencil, paper, and desk calculator' the results of these pul-
monary function tests are entered on coded forms suitable for punching on IBM cards.
The analyses of these data have been programmed for the IBM 7094 computer of the
Western Data Processing Center, Los Angeles. Figures 2 through 8 present example forms
for coding the results of vital capacity and timed vital capacity, lung nitrogen washout,
carbon monoxide lung diffusion, thoracic gas volume and airway resistance by body
plethysmography, and residual volume. Forms are given also for recording of air pollu-
tant concentration levels (Figure 7) and of objective signs detected upon examination
of the chest and of symptoms obtained by the questioning of the patient (Figure 8).
These forms are a part of the system now being used for recording data in a study of
patients with chronic respiratory disease. Records are taken while the patients reside for
days in a room supplied with ambient Los Angeles air, and again while this room is
supplied with air filtered through absolute and activated charcoal filters, at the Los
Angeles County General Hospital (USC).
Similar methods are being used in a second study for coding and entering data
for transfer to IBM punch cards according to a planned format for later transfer to
magnetic tape. Here data on occupation, smoking, exposure to lung irritants, respiratory
symptoms, etc., and the results of physical and x-ray examination of the chest and of
lung function tests are being recorded annually, in an effort to depict the course or
natural history of individuals who are "normal" or bronchitic, or who are already
emphysematous. A year-to-year comparison of any data can be programmed from this
longitudinal clinical and physiological investigation into the development of chronic
respiratory disease in man.
ANALYSIS AND INTERPRETATION
The data can be checked by a complete printout (Figure 9), with inspection for
values that appear to deviate more widely than expected, and for missing values. A
complete listing of variables is printed out (Figure 10), with tabulation of the variable
numbers, variable names, the number of non-zero cases, the means, standard deviation,
and high, low, and range of values. These two tabulations aid greatly in the detection
of punching errors even though the card punching already has been verified, and in the
identification of missing values or a wrong order of cards including those cards directing
the programmed sequence of steps in the analysis. These tabulations also have been of
great help in locating technical or measurement errors.
A histogram and cumulative frequency polygon (Figure 11) are used to describe
the distribution of values and to give the percent of cases within the range of limits
of the variable selected. It should be noted that any restrictions can be placed on the
variables selected. These restrictions may include characteristics such as age, sex, or a
limit of pulmonary function test result, etc.
A two-way plot or scatter diagram (Figure 12) with computation of the mean and
146 ACQUISITION SYSTEMS IN PHYSIOLOGY
-------
standard deviation of each variable, the coefficient of correlation, and the equation of
the line of regression is very informative. A two-way table (Figure 13) with or without
restrictions enables obtaining various frequencies according to the ranks selected, and
the estimates of variation (SD, Chi square).
Tables can be obtained with various restrictions on variables, with printout of the
means and ranges of the values for each, and printout of individual values (Figure 14).
Row and column restrictions of any combination desired can be handled (Figure
15). Nested distribution tables (tables within tables) are useful in analyzing studies
involving multiple variables (Figure 16). Analyses of data according to selected restric-
tions with computation of correlation coefficients and regression coefficients are available
(Figure 17).
On-line transmission of instrument output signals directly to a computer or to a
magnetic tape recorder is not yet in use. Exploration of the method is beginning, and
the method has been used in the recording and analysis of vectorcardiograms. A "com-
puter-spirometer'7 is available, having a readout of calculated values derived from the
FEV, such as the VC and FEV1, %. All these systems would be useful in surveying by
spirometric methods large numbers of individuals for chronic respiratory disease.4
Pulmonary function testing is based upon the careful handling and instructing
of patients in the breathing maneuvers desired, adherence to the conditions of each test,
and a careful setting up of the instrument system; hence it demands trained personnel.
On-line instrument-computer systems probably would be too costly for the usual hospital
pulmonary function laboratory. Smaller systems for spirometry are probably feasible, since
this is a commonly performed yet extremely informative measurement.
REFERENCES
1. Balchum, 0. J. Instrumentation and Methods for Measuring the Physiological Effects
of Air Pollution. ISA Biomedical Sciences Instrumentation Symposium, June 14-18,
1963, Los Angeles, California. Symposium Proceedings, Plenum Press, 227 West
17th Street, New York 17, New York.
2. Swann, H. E., Brunol, D., and Balchum, O. J. An Improved Method for Measuring
Pulmonary Resistance in Guinea Pigs To Be Published.
3. Brunol, D., Balchum, 0. J., and Swann, H. E. Mechanics of the Chest and Lungs:
Physical Basis for Pulmonary Resistance Measurement To Be Published.
4. Balchum, 0. J., Felton, J. S., Jamison, J. N., Gaines, R. S., Clarke, D. R., and Owan,
T. A Survey for Chronic Respiratory Disease in an Industrial City. Amer. Rev. of
Resp. Dis. 86:675, 1962.
DISCUSSION
Asked whether any known correlation had been shown as a result of his studies,
Dr. Balchum replied that insufficient data have been collected for practical statistical
evaluation and that at this time his results are inconclusive. This possibly might be due
to the low-level pollutant concentrations in the ambient air now used for evaluation. Dr.
Balchum indicated that better correlation between pollutants and physiological effects
might be expected when higher pollutant levels found during the California smog season
are used for exposures.
Balchum 147
-------
When asked how subjects are obtained for the physiological research studies, Dr.
Balchum indicated that the University hospital maintains a roster of approximately 300
respiratory patients who volunteer as subjects; also, new patients are solicited to serve as
study subjects.
Dr. Zavon asked whether any physiological reactions other than respiratory response
are being measured. The reply indicated that no other measurements, such as blood
or urine analysis, are being attempted. Dr. Zavon pointed out that the chemical reaction
of such air pollutant substances as 3-4 benzpyrene are being investigated, both in this
country and abroad, for response in other portions of the human system.
A participant asked whether the ambient air used as test atmosphere for the pa-
tients is altered when it is passed through the spyrometer during tests. Dr. Balchum
replied that although no tests have been made to determine whether the NOX or oxidant
levels are reduced in passing through the spyrometer, he believes that because of the
relatively short time that the patient breaths through this transducer such losses are not
a significant factor for a 24-hour test period.
DAILY PUmoNARY FUNCTION STUDIES
FrLIEBED ROOM STUDY
I. SPIROMETRY
Middle
Card No.
Reg. No.
Day of Year
Condition:
1. Filtered
2. Ambient
3. Pre-entry
Duration, Hours
(6)(2)
( ) 9
()()() 12
Day
FIRST TEST
Time of Day Performed ()()()() 16
Vital Capacity, predicted, liters ( )< X )( ) 20
Vital Capacity, observed, liter ( M X X > 24
Obnerved VC/Predlcted VC tt) ( ) ( ) ( M ) 28
Timed VC, 1.0 Sec., Liters ( M X X ) 32
7. Obeerrcd VC ( ) ( U ) ( > 36
WOTF (mid SOJL), Hters/oec < X X X ) 40
Ht., Inchee ( )( X ) 43
Wt., pound. ()()() 46
BSA, M2 ( W )( ) 49
Y«r ( ) 80
Card No. (6)(3) 2
Beg. No. ()()() 5
Day of Year ()(>() 8
Condition:
1. Filtered
2. Ambient
3. Pre-entry ( ) 9
SECOND TEST Duratton, Houra ()()() 12
Time of Day Performed <)<)()() 16
Vital Capacity, Predicted, Uteri ( M )( X ) 20
Vital Capacity, ob.erved. Uteri ( M )( )( ) 24
Ob«erved VC/Pradlcted VC ( )( )( M ) 28
lined VC, 1.0 Sec., Uteri > ( M )( )< ) 32
I Obaarvad VC ( )( M >( ) 16
mEF (.Id 501), Ut«r«/iac ()().( X ) 40
Year ( ) go
Figure 2 Data Form: Spirometry Tests
148
ACQUISITION SYSTEMS IN PHYSIOLOGY
-------
Card No. (6)(4) 2
Reg. Ho. ()()() 5
DAILY PULMONARY FUNCTION STUDIES Day of Year ()()() 8
FILTERED ROOM STUDY C?"« ^""'.l ,
1.Filtered 3. Pre-entry
II. LUNG NITROGEN WASHOUT 2.Ambient <)9
Duration, Hours ( )( )( )12
Last First Middle
P.F. No.
Day Ho. Year
Hour test performed ()()()( )16
7 minute nitrogen washout. Z N-, end tidal, observed (MM )19
Observed/Predicted, 1 ()()()( )23
Forced expiratory N., I at 7 minutes (MM )26
Time to reach plateau, minutes .... (MM )29
End tidal N? at plateau, Z ( )( )< )32
Forced expiratory N,, TL at plateau ()()( )3S
Volume air expired In 7 minutes, liters ( M M M M )&0
Volume air expired to plateau, liters ( M M M M )45
Respirations to 7 minutes ( )( )( ).48
Respirations to plateau ( )( )( )^1
Year ( )BO
Figure 3 Data Form: Lung Nitrogen Washout
DAILY PULMONARY FUNCTIOH STUDIES
FILTERED BOOM STUDY
III. CARBON MONOXIDE DIFFUSING CAPACITY, REST
Card No. (')<5) 2
Re(. No. ()()() 5
Name Day of Year ()()() 8
U>t Pint Middle Condition!
1.Filtered 3. Pre-entry
2.Amblent ( ) 9
P.F. Ho. Dur.tloo, Hour. ( )( )( )12
Day Ho.
B.S.A.
Hour teit performed . . . ()()()( )16
I Upt.k. CO ... ( )( )( M >20
1 Predicted Upt.k. ( )( )( M )24
Dlffuilng Opacity (»l/»rfl8/«ln). . . . . ... ( )( X >"
I Predicted Dlff. Capacity ... ( )( )( M )31
Kin. Vol. (Liter/.In) . . . ( )( M )( )35
Kin. Vol./M2-I./.ln/>|2 ( )( M )( )39
Reiplr. Rate/Mia . . . . . . ( >< X<*1
Tidal Volm (L ) .... .... . ( « X )( >«
Conductance (1/ilnAnHi) ... ( )( M )*'
I Predicted Conductance ... .... .... ( )( )( M )52
Oxygen uptake, ail/Bln/H2 ( )( )( ).5S
Ventilation, Reat, I. /aln/H? ... (XV)58
OxyBen Extraction fro. Inaplred Air, t . . . .... ()().()«
Year < )80
Figure 4 Data Form: Carbon Monoxide Diffusing Capacity, Rest
Balchum 149
-------
DAILY PULMONARY FUNCTION STUDIES
FILTERED ROOM STUDY
V. PLETHYSMOGRAPHY
Date
Last First
Card No.
Reg. No.
Day .of Year
Condition;
1. Filtered, 2. Amb:
3. Pre-entry
Duration, Hours
Time of Test
TCV, L. (
ER, L. (
PRV, L. (
VC, L. (
TC, L. ( )(
PRV/TC 7. (
TGV/TC 7. (
Compliance
Thorax <
Lung (
Reaiatance
Airway
Tlaaue
Interrupter 1
Interrupter 2
Interrupter 3
Peak expir .
Eaoph.preaa. + ( }
Minimum
Baoph.preaa. +( )
Day
Teat Mo.
Year
Middle
!
(6) (7) 2
()()() 5
( X X ) 8
Lent,
( ) 9
( )( X )12
()(.)( XH6
).( X X )20
).< X X )24
).( X X )28
).( )( )( )32
).( X X )37< )
)( )( ). ( )41
) ( X ).( M5
).( X )( )49
).( )( X )53
( X X< )56
()().( )59
( X ).( )62
( X ).( -)65
( X ).( )68
l-( )( ).( )72 ±(
'-( )( ).( >76 +(
( )77 ~
( 178
( )79
( )80
2
Day
<6)(7) 2
( X )( ) 5
( X X
(
( X X
()()()(
( ).( X X
( >!( x x
( ).( X X
( ).( X X
( ).( X )(
( X )( ).(
( X X ).(
( ).( X X
( ).( X X
( X ).(
( X ).(
( X ).(
( ) ( ). (
( X ).(
)-( X ).(
)-( X ).(
(
(
(
(
) a
) 9
)12
)16
)20
)24
)26
)32
)37( )
)41
)45
)49
)53
)56
)59
)62
)65
)68
)72 +(
)76 +(
)77
)78
)79
)60
Mo. Year
3.
(6X7) 2
( X X ) 5
( X X ) 8
( ) 9
( X X H2
( )( X X >16
( ).( X X )20
( ).( X )( )24
( ).( X X )28
( ).( X X )32
< ).( X )( )37( )
( X X ).( )41
( )( X ).( )45
( ).( X )( )49
( ).( X X )33
( X ).( )5«
( X ).( )59
( )( ).( >62
( )( ). < )63
( X ).( )68
)-( X ).( )72±(
)-( )( ).( )76 + (
( )77~
( )78
( )79
( )80
P.F. Ho.
4.
(6X7) 2
( X X ) 5
( X X ) 8
( ) 9
( X X )12
( X X X )16
( ).( X )( )20
( ).( X X )24
( ).( X X )28
( ).( )( X )32
( ).< X X )37( )
< )( )( ).( )45
( ).( X X )49
( ).( X X )53
< X ).( )56
( X ).< )59
( X ).( )62
( ) ( ). ( )65
( X ).( )6»
>-( )( ).( )72 +<
)-( X ).( )76 +(
( )77
( )76
( >eo
5-
( X X ) 5
( )( X ) 8
( ) 9
( X X )12
( X X )( )16
( ).( X X >20
( ).( )( )( )24
( )( X X )28
( ).( X X )32
( )( X X )37
( )( X >!< )45
( ).( X )( )49
( ).< X X )53
( X ).( )56
( X ).( )59
( X ).( )62
( X ).( )65
( X ).( )68
)-( )( ).( )72
)-( >( ).< )76
( )77
< )7B
( )79
( )«o
Figure 5 Data Form: Plethysmography
FILTERED ROOM STUDY
HELIUM DILUTION
NAME Card No. (7)(0) 2
Reg No. ()()() 5
P.P. No. Day of Year ()()() 8
Condition
AGE HEIGHT WEIGHT 1. Filtered
2. Ambient
3. Pre-entry ( ) 9
Duration, hours ()()() 12
Time of day . . . ( )( )( )( ) 16
FRC (He dilution) (L) . . ( )( ).( )( )( ) 21
Pred FRC (L). . . . ...()().()()() 26
Obs. FRC, 7. of Pred . . ()()().() 30
Vital capacity, slow, sitting
EVR, slow, sitting
RV ().()()() 34
Pred RV (L) ().()()() 38
Obs. R.V., % of Pred R.V ()()().()*2
TLC (He) (L) ()()-()()() 47
Pred TLC (L) ()()-()()() 52
Obs. TLC, % of Pred TLC ()()()() 56
Obs. RV/obs. TLC X100 ()().() 59
Year ( ) 80
Figure 6 Data Form: Helium Dilution
150 ACQUISITION SYSTEMS IN PHYSIOLOGY
-------
FILTERED ROOM STUDY
AIR POLLUTANTS
Card No.
Room No.
Day of Year
Condition:
1. Filtered
2. Ambient
3. Pre Entry
Interval
(XX)
()()() 12
Time of Day
CO, ppm
NO, ppm
N02, ppm
Oxidants, ppm
Temp. , "C
Rel. Hum. , %
Time of Day
CO,1 ppm
NO, ppm
N02, ppm
Oxidants , ppm
Temp. , °C
Rel. Hum. , %
Time of Day
CO, ppm
NO, ppm
N02, ppm
Oxidants, ppm
Temp. , 'C
Rel. Hum. , %
()()()()
( )( M )
( X X )
(MX)
( X )( )
( )( X )
00
0000
(XX)
( X X )
( X( )( )
( X X )
( )( ).( )
( )( )
( )( )( )( )
( )( X )
(MX)
(MX)
( X X )
( )( X )
( )( )
16
19
22
25
28
31
33
37
40
43
46
49
52
54
58
61
64
67
70
73
75
Card No.
Room No.
Day of Year
Condition
1. Filtered
2. Ambient
3. Pre Entry
Interval
(7)(2) 2
()()() 5
()()() 8
( ) 9
()()() 12
Time of Day
CO, ppm
NO, ppm
NO. , ppm
Oxidants, ppm
Temp. , C
Rel. Hum. , %
Time of Day
CO, ppm
NO, ppm
N02, ppm
Oxidants, ppm
Temp. , "C
Rel. Hum. , %
Time of Day
CO, ppm
NO, ppm
N02, ppm
Oxidants , ppm
Temp. , 'C
Rel. Hum. , %
( X )( )( )
( )( X )
( X( )( )
( X )( )
(XX)
( )( X )
( X )
( X )( )( )
(XX)
(XX)
( X )( )
( X( X )
( )( M )
( )( )
( )( X X )
(XX)
(XX)
( X( )( )
(MX)
(XX)
( X )
16
19
22
25
28
31
33
37
40
43
46
49
52
54
58
61
64
67
70
73
75
Year
( ) 80
( ) 80
Figure 7 Data Form: Air Pollutant Concentrations
Balchum
151
-------
FILTERED ROOM STUDY
DAILY RECORD OF SYMPTOMS AND SIGNS
- ~ Card No. (6) (0) 2
Reg. No. ()()() 5
Day of Year ()()() 8
Condition:
1. Filtered
2. Ambient
3' Pre-entry ( ) 9
Middle Duration, Hours ()()() 12
Hour ()()()() 16
P.F. t
__
Day Month Year
1. Objective Signs
A. General
Vigor; Normal 1, Below Normal 2, Poor 3 ........... ( ) 17
Cough; None 1, Minimal 2, Moderate 3, Marked 4 ...... ( )" 18
Sputum; None 1, Minimal 2, Moderate 3, Marked 4 . ..... ( ) 19
Wheeze; None 1, Minimal 2, Moderate 3, Marked 4 ....... ( ) 20
Breathing difficulty; None 1, Minimal 2, Moderate 3, Marked 4 . ( ) 21
Cyanosis; None 1, Minimal 2, Moderate 3, Marked 4 . ..... ( ) 22
B. Chest and Heart examination:
Re op. rate .................. . . ( ) ( ) 24
Heart rate .......................... ()()() 27
Heart rhythm; regular 1, regular with prematures 2,
irregular - 3, other « 4 ........... C ) 28
Auscultation:
Expiration: (1. Normal, 2. Prolonged)
Right .......................... C ) 29
Left ......................... ( ) 30
Expiration Time, Seconds ......... ..... C ) ( ) 32
Breath Sounds:
Intensity (1. Normal, 2. Increased, 3. Decreased)
Right ........................... ( ) 33
Lett ........................... ( ) 34
Type (1 Vesicular, 2 Broncho -vesicular, 3 Bronchial, 4 Loud and Harsh) ( ) 35
Right ........................... ( ) 36
Left .......................... ( ) 37
Adventitious Sounds:
Rales (1 Absent, 2. Preaent)
Right base ....................... ( ) 38
Upper lung field ................... ( ) 39
Left base ......................... ( ) 40
Upper lung field ................... ( ) 41
Rhonchi (1. Absent, 2. Present)
Right ............................. ( ) 42
Left ............................. ( ) 43
Wheezes (1. Not present, 2. Insplr., 3. Ejcpir., 4 Both)
Right ............ ................ ( ) 44
Left ............................. ( ) 45
Is there a significant change in the Behest findings as compared
to the previous examination. 1. No, 2. Yes ............. ()46
[If yes]: 1. Improved; 2. Worse; 3. Changed, without
overall significant change .............. ( ) 47
C. Oral Temperature, previous 24 hours (highest), "C ........... ( ) ( ).C ) 50
II. Symptoms
Cough (No 1., Yes 2) .......................... ( ) 51
[If /es]: Unchanged 1, Improved 2, Worse 3, compared to yesterday . ( ) 52
Unchanged 1, Improved 2, Worse 3, since first day of study ( ) 53
Most difficult on arising 1, arising and all day 2,
night 3, all 24 hours 4, DK 5 .............. ( ) 54
Sputum (No 1., Yes.) .......................... ( ) 55
[If yes]: Unchanged 1, Improved 2, Worse 3, compared to yesterday , ( ) 56
Unchanged 1, Improved 2, Worse 3, since first day of study ( ) 57
Most difficult on arising 1, arising and all day 2,
night 3, all 24 hours 4, DK 5 ..... ....... ( ) 58
Shortness of breath (No 1., Yes 2.) .................. ( ) 59
[If Yes]i Unchanged 1, Improved 2, Worse 3, compared to yesterday . ( ) 60
Unchanged 1, Improved 2, Worse 3, since first day of study ( ) 61
Most difficult on arising 1, arising and all day 2,
night 3, all 24 hours 4, DK 5 . .......... ( ) 62
152 Figure 8 Data Form:
-------
Appetite (Normal 1, Increased 2, decreased 3) ( ) 63
[If 2 or 3]: Unchanged 1, Improved 2, Werse 3, past 24 hours. ... ( ) 64
Cheat Tightness or Congestion (No 1, Yes 2) ( ) 65
[If Yes]: Unchanged 1, Improved 1, Horse 3. past 24 hour ( ) 66
Chest p«tn (No 1, Yes 2, Don't know 3) ( ) 67
[If Yea]: Changes vlth breathing. No 1, Yes 2, Don't know 3 .... ( ) 68
Changes with cough No 1, Yes 2, Don't know 3 ( ) 69
Worse with exercise No 1, Yes 2, Don't knov 3 .... ( ) 70
[Examiner: Angina? No 1, Yes 2] ( ) 71
T««r ( ) 80
Card No. (6)(1) 2
Keg. Ho. ()()() 5
Day of Year ()()() 8
Condition:
1. Filtered
2. Ambient
3. Pre-entry ( ) 9
Duration, tours ()()() 12
Hour ( )( )( )( ) 16
Previous 24 hours only. No 1, Yes 2, Don't know 3
Sore throat ( ) 17
Nasal congestion ()18
Abdominal discomfort or pain ( ) 19
Eye Irritation ( ) 20
Any effect of smog on breathing ( ) 21
Was smog present In the room ( ) 22
Did temperature effect breathing ( ) 23
[If Yes]: Improved lf Worsened 2, Don't know 3 ( ) 24
Change related to elevated 1, or reduced 2, temperature ( ) "
III. Smoking!
Number cigarettes smoked In past 24 hour ( >( )( ) !«
Number plpefulls tobacco smoked pait 24 hours ( )( ) 30
Number cigars smoked past 24 hour ( )( ) 32
8 A.M. to S P.H.
IV. Sputum
Volume. .1 ()()() 33
Colon Clear 1, White 2, Green 3, Gray 4, Yellow 5,
Yellow-gray 6, Yellow-green 7, Black-brown B. Other 9 ... ( ) 3*
Blood! None 1, Streaks or flecks 2, More than streaks 3 ( ) 37
Purulence: None 1, Huco-purulent 2, Purulent 3 . . ()3fl
Physical character: Wster 1, Vlalc 2f Layered 3,
Clumped 4, Other 1 ( ) 39
Odor: None 1, Minimally unpleasant 2, Foul 3,
local 4, other 5 ( ) *0
8 P.M. to B A.M.
Sputum
Volume, ml <)()()«
Colon Clear 1, White 2, Green 3r Gray 4, Yellow 5,
Yellow-gray 6, Yellow-green 7, Black-brown 6, Other 9 ... ( ) 44
Bloodi None 1, Streaks or flecks 2, More than streaks 3 ( ) 45
Purulenee: Hone 1, Muco-purulent 2, Purulent 3 ( ) *6
Physical character: Water 1, Vlslc 2, Layered 3,
Clumped 4, Other 5 ( ) *'
Odor: None 1, Minimally unpleasant 2, Foul 3
Fecal 4, Other 5 < ) *
V. Medications [No 1, Yes 2]
Positive Pressure Breathing ( ) *9
With Bronchodllator ( ) *>
Number times per 24 hours < )( ) H
Bronchodllator other than with PPB (Nebullser, etc.) ( ) 33
Antibiotics (Specify) ( ) 5*
Oral chest medications (PET,Amesec, Tedral, etc.) ( ) "
Cortlcoaterolds ( ) 56
Year ( ) «°
Record of Symptoms and Signs 153
-------
10" 11621MR 2021983 15065602*291 73291 17 143292386 075 0099340 02390617 06
100111981315055855809002000088888068888488000040*00969600700000000 06
IOC 11 19862 1MR3 150000000000000000000000000000000000000 500000023003703701290 100
;001119eMR 3 175002290010 299000090000909900009999090990000999909006 0100448
...... "" ...... '
10011198MR3
10013342MJ2
J 90000999909006
021840406240831209624101264246293 055 0099440 00000000
055855806001000088888088888488005000000969600000000000
00
0 4000000000000000 00 000000 0000000000000 10000000000034Q3400375 100
^ ~-. = =
27060375004599000090000909900009999090990000999909000
817131706656439017*31113315300363 075
0100158
990000999909000 Q10015
0099390 01390648 (J 1
10013226MJ4
10013226MJ4
10032222131 055855806001000088888088588488000020000969600000000000 01
10032222381MA31700000000000000000000noO00000000000001000000025003503501800101
10032222MA31*219072690011*219037500091682008000095990000999909001 0397178
1 0032222WA3 99000099990=>001 0357178
10051571GG191205225070672*10262391201"0279353 07& 0099*80 01770756 05
100511031225052151103002000088888086flR6*88C00020000969600000080803 05
10051103571 GG22501688*105000000001101010101090U"09901010i035.:290ai2902*i51o5
10051103GG2022370*15 00 5*9 90000900009 Cc;900009?9«0^0990000y'?9909005 10Z3234
10051103GG2 990000^9990*005 102323*
10052662GM153173225063373182191510430*10981*2 000 9622180 02*90884 07
100522942225015955806002010088888038838*88000060000969600000068808 07
10052294662GM22503288340500000000105101010101011201903Q3C3035200033334i300107
10052294GM2168200B00009599000090000909000009999090990000999909007 010000C
10052294GM2 990000999909007 0100000
100615618T1851932?6066541380161301102302'*353 055 0099460 012703*3 05
1006110*1226055855^06002010088888088833*66000000000969600000088608 05
10061104561BT2260173417QOOOOOOO0080310101010103651090505050553000752*02023105
10061 10*sr2131360686*03H31360685102 I 1313oOiS570Zl23679002300*605 0*28*56
10061104BT2 990000999909005 0428*58
100627023Di77H3226063*212601&1380b30*6132221 075 0099410 026907*5 08
10062290222605565560600200906888621100010S007000000969600700758806 08
IOC*22907028D22600000000000000000000C,1030000000000000404050553000752002800 108
10062290BO21682008000095990000°00009C^00009999390990000999909008 0100000
100
IOC
100
1"C
LOO
100'
100
2290BD2 990000999909008 0100000
1602HE19818622306036523*13136053034150201 075 0099260 02850776 06
120022 23055 8 55fl06004009oa8n3308fl8d8*86002000ri0096960040l 188801
1?0060?HE2230000000000000000000000000000000000000500000021003903901825106 32
1200HE214409Q726901110870062500219900009999090990000999909006
... _ 0210101
1200HE2 990000999909006 0210*01
1321Pri35l773li07064l*8SI5324^52i2317*3fc035 0099*60 01150446 00
IOC 811231J14055855806001000068883088888486002020200969600000000000 00
1?081128321DF314018681*0500000000110101010109011409910030J1057000331821988*00
1008 1128PF3059 15055500 1602 2 37(.^Jd00249900009999090-J90000999909oOO 0209146
.008H28Pr3 990000999909000 0209148
10C9H51TW2202063 1506543 1265 16i J 6090 1361 9323 3 055 0099455 01770531 &7
1009119*13l?055e5590AOO2000088988201000118000000000969600000000000 07
10091194661TW3I 5018H8480500000000110101010109011*0990000000000004104100290107
10091194TW3033*90290008299000090000909900009999090990000999909007 1000376
' ">091194Th.> 990000999909007 1000*76
52
12
1*
li
lo
23
25
Figure 9 Printout: Summary Form
JON-
ZERO
CASES MEAN
.COUf»r 3F CASCS PEHCc'ir OF CASES
1.5 1.5
66LOH IJ A60VE BELOrt TO 4BOVE
4 MT IH POUNDS 949 16
01 ASTHMA 392
^.OB64 32.4806 294.0000 83.0000 211.0000
.8444
.0530 9.0000 1.0000 a. 0000
0
0
11
0
88
25
0
490
966
903
0
0
936
6GO
0
851
0
518
459
393
446
949
0
0
0
0
949
1
0
0
0
3
5
5
5
. 0. iLi
U. 10
.16 9B.8<.
0. 10
.27 90, 3
.63 97. 7
0. 10
.31 54. a
.63 48. 7
.64 40.36
.00 47. JO
Figure 10 Printout: Listing of Variables
154
ACQUISITION SYSTEMS IN PHYSIOLOGY
'GPO 8141056
-------
100. 0 «
289.500 100.0 *
269.500 100.0 *
Z49.50Q 99.1 n
6 1.3 .*
229.500 97. B »
34 7t* ::
90.5
IZT 2T.5 .
114 2*. 7 -
149.500 19. T *
77 16.7 .
124.500 3.0 *
12 2.6 .
109.500 0.4 *
2 0.4 -
....
....
'!"!
...
...
PERCENT
i*
»
:';
: :
.'
.
.
.
t
.
.
Figure 11 Histogram and Cumulative Frequency Polygon
GRAPH 1 WITk- 461 POINT!. 01 THE HGHIZOtiAL. IS XI 6) HT. INCHES ON THE VERTICAL IS VI 4) XT IN
5C.51C 56.500 62.500 68,300 74.500 BO.SOu
^i.500 59.500 65. SCO 71.500 77.500
300.50
J96.10
2--l.tr.
pnj.qii
276.50
274.1U
765.30
260.90
756.50
252. 10
243.30
238.90
P34.50
225.70
221.10
216.90
217.50
209. 10
190.50
1B6.10
IB 1.70
172.90
16B.50
159^70
150.90
146.50
133.10
128.90
174.50
111.30
106.90
102.50
89. JU
84.90
*
. ,
'
2 2 *
" 2
2
7
6
\ t
3 5
4 5
6 2
2 2
6
2 3
5 2
. ,
;
2 J
2
2
2
323
2 3
.
3
2 2
BO. SO t
3. .0.5'
e?.9
65.3
56.5
52.1
7.7
4.5
5^7
6.9
12.5
19.)
9*. 9
61.7
72.9
6^.5
59^7
55.3
50.9
42. 1
37.7
24.5
20.1
15.7
02.5
8T.3
B4.9
80. 5
50.500 56.500 62.500 68.500 74.500 S0.«:no
53.500 59.500 b5,5QU 71.500 77.500
Figure 12 Printout: Scatter Diagram
Balchum
155
-------
TWO WAY TABLE
ROW VARIABLE NUN BE PI 2t AGE
COL VARIABLE NUMBER <,, «T IN POUNOS
RESTRICTIONS ON THE DATA ARE
VARIABLE 1 CROUP NF 0 5,
VARIABLE 3 SEX .
IS NOT LESS THAN
IS NOT LESS THAN
0.5000 OR GREATER THAN
0.5000 OR GREATER THAN
L.5000
1.5000
FREQUENCY TABLE
129.5000
949 POSSIBLE INDIVIDUALS SATISFIED THESE RESTRICTIONS
TOTAL
169.5000 209.5000 229.5000
70.
3
60.
2
50.
0
40.
0
11
14
2C
21
10
17
26
32
2
1
4
2
0
0
2
24
26
34
57
12
1 3
17
29
.44
.62
.53
168
169
171
176
.7917
.5882
.5088
25.
25.
24.
0477
9231
0623
9493
16.62 DEGREES OF FREEOC" - l«
CH1-SOUARE/DF
TOTAL 6 7? 96 10 4
PCT J.ll 39.90 49.74 5,18 2.07
(-EANS 60.8333 51.1299 48.635* 40,9000 38.2500
S.O. B.4242 14.5280 13.0796 13.5191 6.0208
CHI-SQUARE -
RCW FERCENTS
59.5000
49.5000
COLUMN PERCENTS
129.5000 169.5000 209.5000 229.5000
16.7 14.3 11.5 10.0 0.0
69.5000
50.0 14.3 10.5 20.0 0.0
59.5000
33.-. 18. £. 17. fl 10.0 0.0
49.SOOO
0.0 26.u 27.1 40.0 50.0
39.5000
u.u 27.j 33.4 20.0 50.0
129.
.2
.6
,S
.0
,0
5000 169.
45,9
42.4
41.4?
3ft. 5
36. *
,5000
45.9
3B.5
50.0
50.0
56.1
209.5000
4.2
7.7
3.0
7.7
3.6
229.5000
0.0
0.0
0.0
3.9
i. n
Figure 13 Printout: Two-Way Table
RESTRICTION
e
BCUP NF Q S
Gt
1TAL CfiPACl
*'£
I
ro
9
H S
70000
30000 I
99000
0.
.43305
.08966
.
3
GH
.00000
.00000
.99000
LO
1
41
9
4
00000
00000
99000
NOT ZERO »
1.70000
70.30000
1.62400
9.99000
AN
0
11
-0
S.D. 5.E.
<-H305 0.15275
OB98S, "..13438
-0.
L.OOO
1. 000
2.000
Z.OUO
?.ooo
2.000
?.ooo
2.000
2.000
41.000
ST. 000
bB.OOP
re. ooo
74.000
2.000
4.000
7.000
8.000
60.000
03.000
01.000
55.000
57,000
16.000
29.000
IB. 000
98.000
.700 0
.460 3
.090 1
.720 2
.720 62
.500 61
.550 60
.460 SB
.010 58
000
000
000
000
000
000
000
000
000
».990
.990
.990
.990
.990
.990
.990
.990
.990
10 CflSFS ARP tISTEn.
Figure 14 Printout Showing Restrictions on Variables
156
ACQUISITION SYSTEMS IN PHYSIOLOGY
-------
THE FOLLOWING TABLES USE 949 CASES UITH 19 VARIABLES EACH
TRANSFORMATIONS ARE HADE AS FOLLOWS
-0 BOOLEAN
-0 TRANS-GENERATIONS
-0 BOOLEAN
-0 TRANSGENERATIONS
COUNTS OF THE CASES MEETING SPECIFIED RESTRICTIONS ARE GIVEN'IN a ROWS AND
ROW AND COLUMN RESTRICTIONS
ROW 1
VARIABLE t, AGE
ROW 2
VARIABLE It AGE
ROW 3
NCNE
COLUMN 1
VARIABLE ^. SEX
COLUMN 2
VARIABLE 3. SEX i IS NOT LESS THAN 1.5000 OR GREATER THAN
TABLE 1 USES VARIABLE 4 HT IN POUNDS IF BETWEEN 99.5000 AND 200.5000
TABLE 2 USES VARIABLE 5 SURFACE AREA IF BETWEEN 1.2500 AND 2.4500
TABLE 3 USES VARIABLE 6 HT. INCHES IF BETWEEN 55.5000 AND 80.5000
TABLE 4 USES VARIABLE 7 VITAL CAP1CI IF BETWEEN 0.7500 AND 1.5000
, IS NOT LESS THAN 29.5000 OR GREATER THAN
, is NOT LESS THAN 49.5000 OR GREATER THAN
, is NGT LESS THAN 0.5000 OR GREATER THAN
49.5000
79.5000
1.5000
2.5000
MEANS OF VARIABLE < WT IN, POUNDS
167.11 142.60
16*1.66 151.69
165.62 146.18
ROW-COLUMN COUNTS AND MEANS OF VARIABLE 4 WT IN POUNDS
1 2
1 167.11 142.60
203 255
2 164.66 151.69
176 187
3 L65.62 146. 18
368 448
Figure 15 Printout Showing Row and Column Restrictions
TABLE i. VAMABLE 4t AGE
IS USED IN 6 SUBTABLES.
VARIABLE
S, SEX
LIMIT
69.50
ft 4. 50
59. 50
32 7.0 81.5
4T 9.9 84.1
10 3.2 79. 3
22 10.4 T9.1
9.4 88.0
49. SO
44.50
54 11.7 65,
39.50
34.50
29.50
TOTALS
PEAKS
ST.OEV.
460
49.9B91
13.6024
477
49.1195
13.1670
193
49.6083
13.6916
267
50.1199
13.9691
211
HO r 5 166
14.2037
266
46.0113
12.1963
Figure 16 Nested Distribution Table
Balchum
157
-------
CORRELATION TABLE 1
RESTRICTIONS ON THE DATA ARE
VARIABLE 7 VITAL CAPACl.
IS NOT LESS THAN
937 OUT OF 949
POSSIBLE INDIVIDUALS SATISFIED THESE RESTRICTIONS
VARIABLE MEAN
2
6
7
AGE
VITAL CAPACI
49.5464
3.4883
ST. OEV.
13.3825
1.0282
5.E.
0.4372
0.0336
HIGH
96.0000
294.0000
78.0000
6.5600
LOU
30.0000
83.0000
56.0000
0.7800
RANGE
66.0000
ZLL.OOOO
22.0000
5. 7800
CORRELATION COEFFICIENTS
2 4
5 -0.03*2 0.930
6 -0.1449 0.427
7 -0.4551 0.284
THE REGRESSION CDEFF
2
2 1,0000
5 -0.0005
6 -0.0399
7 -0.0350
LEAST SQUARE LINES M
COEFFICIENTS. FOR EX
SECOND COLUMN IS GIV
567
1.0000 0.7198 0.5118
0.7198 I. 0000 0.7319
0.5118 0.7319 1.0000
CIENTS OF COLUMN VARIABLES USEO
4567
.0127 -2.3163 -0.5262 -5.9231
.0057 1.0000 0.0306 0.0984
.0485 13.4147 1.0000 2.6234
.0090 2.6614 0.2042 1.0000
Y 8E WRITTEN FROM THE ABOVE MEANS AND
NPLE, THE REGRESSION LINE OF THE FIRST
N BY
REGRESSION
ROW ON THE
0.0127.1X1 41- 162.30521
Figure 17 Data Analysis for Selected Restrictions With Correlation and
Regression Coefficients
158
ACQUISITION SYSTEMS IN PHYSIOLOGY
-------
PANEL MEMBERS
Robert Bryan
Director, Technical Services
Air Pollution Control District, Los Angeles
Dr. Paul B. MacCready, Jr.
President, Meteorology Research, Inc.
Altadena, California
Dr. Benjamin V. Branscomb
Associate Professor of Medicine
Medical College of Alabama, Birmingham
Dr. Ralph I. Larsen
Field Studies Branch, DAP
U. S. Public Health Service, Cincinnati
DISCUSSION: DATA ACQUISITION SYSTEMS
Mr. Bryan indicated that the prime question asked by an administrative group
responsible for a practical data acquistion system is "Why do we measure?" This question
is answered at the local level by the need to establish trends 'or background information.
Air quality monitoring indicates whether proposed or enforced standards are being met
and whether control activity is producing the desired effect.
Dr. Branscomb pointed out the inadequacy of our measuring devices in precision
and accuracy. He noted that measuring devices seemed to fall into three categories:
those designed principally around the chemical aspects of the measurement, those designed
principally for the engineering aspects of sampling and analysis, and those oriented
toward electronic interpretation of the measurement.
Dr. MacCready mentioned several items not yet considered at the meeting. He
stated that the ambient air is a very poor laboratory because it varies in three dimen-
sions and also in time. He pointed out the value of the light airplane in assessing air
pollution problems. The mobile airplane appears highly flexible and relatively inex-
pensive in comparison with the money and manpower required by extensive ground
networks. He also mentioned the use of naturally occurring topographical configurations,
such as craters, which are good for special stable air mass studies. Dr. MacCready
commended the use of tetroons as a means of remaining with a particular air parcel
and noting its change during transport. The atmospheric laboratory can be made more
quantitative by introducing more flexible means of measurement and analysis.
Dr. Branscomb noted that medical disease and its ramifications are just as difficult to
define as air quality. One of the problems in studies of biological variations in subjects is a
lack of definite knowledge that medical effects are due to a specific pollutant. The
engineer appears to be ahead of the physician in defining variables. Diagnosis alone is
unsatisfactory as a goal for measurement. Although approximately 10 percent of the
adult population over 40 have emphysema, the medical body is divided in its opinion
of just what comprises this disease. Most of the information accepted by the medical
profession is inferential; often there are no clear-cut proofs to substantiate medical
knowledge. Therefore, scientists have every right to question the physician when he
submits a medical diagnosis or finding as a goal for measurement.
Discussion 159
-------
Dr. Branscomb pointed out the uselessness of a static tool such as the x-ray for
diagnosis of emphysema. Present instrumentation used for measuring loss of respiratory
function is not sensitive enough to determine small incremental amounts of such loss
due to possible air pollution effects on the lungs. Commonly used instrumentation is
seriously impaired by overshooting and damping effects and does not produce valid data
at the frequency cycle associated with human breathing rates. A spyrometer developed
at the Alabama Medical College was cited as a considerable improvement over existing
instrumentation, but this unit is still barely adequate to meet the investigative needs.
Dr. Branscomb noted that in the Alabama respiratory study the item that correlated best
with reduced pulmonary function was positive response to the query "Does weather
influence your breathing?"
Asked whether airborne particulates are important in health considerations, Dr.
Branscomb replied that there is no evidence that particles alone cause detrimental health
effects. He noted, however, that a recent study with guinea pigs showed that carbon
particles exposed to nitrous oxide produced lesions in the lung. When the animals
were subjected to nitrous oxide and carbon particles individually, they exhibited no
such tissue damage. Although these two substances produce no noticeable health defects
individually, a detrimental effect was caused by their inhalation simultaneously.
Dr. Branscomb further noted that emphysema has now become the second highest
cause of disability in the United States; the magnitude of this problem has become so
great that we must act on the basis of preliminary information. We must continue to
accumulate facts regarding the effects of air pollutants in the production of emphysema
and other disorders, so that industry may act in the public interest rather than being
guided by public imagination. Mr. Nader commented on a paniculate (sulfur dioxide)
study now under way at the Harvard School of Medicine. Present indications are that
particle size is important, since smaller particles seem to yield a greater physiological
effect.
160 DATA ACQUISITION SYSTEMS
-------
SESSION 5: Measurements of Water Environment
Chairman: Leo Weaver
Chief, Water Quality Section, Basic Data Branch
Division of Water Supply and Pollution Control
U. S. Public Health Service
-------
Samuel S. Baxter
Water Commissioner and Chief Engineer
Water Department, City of Philadelphia
and
Joseph V. Radzinl
Chief, Research and Development
Water Department, City of Philadelphia
SUMMARY
The Philadelphia Water Department and the U. S. Geological Survey have established
a water quality monitoring network along the Delaware River estuary. The status of
automation and its application to the water industry are evaluated. If it is assumed that
standard biological waste treatment, low flow augmentation, and treatment of water
supplies by conventional plants and methods will continue to be used to deal with pollu-
tion, the required data acquisition systems are already on the market. These include
equipment for data transmission, recording, storage and retrieval, and the actuation of
secondary devices. The missing elements are certain sensing and detecting devices and
the full knowledge of what parameters or variables reveal the cause and effect relation
within a system. These items are explored in detail, including the economics of an
automatic system. The authors believe that we are at the point of no return that
automation is the key to the water industry today.
DATA ACQUISITION SYSTEMS IN
WATER SUPPLY
Back in 1960, in Cincinnati, and under the same auspices as this meeting, the senior
author presented a paper, "High Quality Water Without High Quality Data Is It
Possible?"1 At that time the City of Philadelphia had just placed in operation on the
Delaware River a modern water treatment plant with complete facilities for automatic
chemical application and with other automatic plant operational features. This Torresdale
Plant has been termed an "automatic" or "push button" plant and is considered to be one
of the most modern facilities of its kind. The principal purpose of the 1960 paper was
to tell about the new Load Control Center in Philadelphia, which gathers intelligence
from various instruments throughout the City through a system of micro-wave stations
and land wires and with this information maintains surveillance over the distribution
system, logs data for record purposes, and maintains supervisory control over pumping
stations. The paper also mentioned the beginning of the cooperative venture between
the Philadelphia Water Department and the U. S. Geological Survey (USGS) in establish-
ing a monitoring network for river quality measurements along the Delaware River estuary.
The 1960 paper posed challenges for water quality treatment and control, ranging
from practical available instrumentation to "blue sky thinking." It was hoped from the
discussions that arose in connection with the paper that the water industry and public
and private research in government and industry appreciated the problems and would
take real action in attempting to solve them.
As we appraise the situation today, very little has been done. The title of the paper
given today is a paraphrase of the one given in 1960. Although this paper will attempt
to view the problem of Data Acquisition Systems in water supply from a broad viewpoint,
Baxter and Radziul 163
-------
it will be natural that many of the illustrations will evolve around the Philadelphia
plants on the Delaware and Schuylkill Rivers. Some of these illustrations may have a
limited value to others, since the Delaware River source is a tidal estuary.
We look at Philadelphia's modern water treatment plants, its new Load Control
Center, and the automatic monitoring stations on our rivers, and wonder where we go
from here. The answer oomes with a real impact to the authors of this paper. It is that
we are at the point of no return, not only for ourselves in Philadelphia, but for everybody
in the water industry. Automation becomes the key word. It now affects a substantial
segment of our entire society, and its accelerated impetus is becoming everyone's interest
and responsibility.
Automation cannot be ignored because in many places it has demonstrated that it
provides more goods and services of superior or higher quality at lower costs. The report
of the Committee on Public Works of the U. S. Senate5 emphasizes the need to provide
maximum service at minimum cost for all public works.
Private industry is making more and more use of automation in process industries
and in manufacturing in general. The water industry and other related public works
operations, generally tied into government, are far behind in developing and using
automatic features.
One reason for this is the fact that each water industry is a utility, whether it is
governmentally owned or privately owned. For the governmentally owned water utilities,
there is the complete absence of the profit motive that provides the stimulus for private
industry to lower costs and to increase quality. In the privately owned utilities, the
regulation by commissions may have a somewhat similar effect.
In the water industry therefore, we should try to find a substitute for the profit
motive. We should have the desire to turn out water of better quality and to reduce
operating and capital costs. It would seem to the authors that in the water industry it is
only through automation, instrumentation, and remote control that operating costs and
reduction in personnel can be achieved. We recognize that new processes may be invented
and developed, but point out certainly that these should be fully automated. The water
industry is not much different from many other process industries that operate every
hour of the day and week, and we should take a lesson from our brothers who, in such
industries as power plants and oil refineries, have used automatic features for many years.
We cannot afford to ignore the effects automation has throughout industry, and
with particular reference to skills of personnel and working hours of personnel. If the
work week in industry in general will be reduced as a result of automation, the water
industry will have to face this problem, including competition for skilled and professional
people.
The 1960 paper suggested that the next step beyond automating treatment plants
and automatic raw water sampling was a digital computer for quality control. The paper
should have properly stressed the more obvious first need, which was the need to make
all operations automatic so that they would eventually lead to automation.
We would then be ready for cybernation, which might be described as the science
that deals with the marriage of automated systems and machines with computerized
analyzing and decision-making machines. Perhaps, we may be able to strike some middle
ground between cybernetics or blue sky thinking and waiting or doing little.
Blue sky thinking will entail the complete evolution of a new system of treatment.
This would call for the invention of new water treatment processes based upon presently
164 ACQUISITION SYSTEMS IN WATER SUPPLY
-------
unknown concepts. The optimum system might be one in which water and sewerage
systems are integrated in a realistic, functional, recycling total water use relationship.
It would be wonderful if we could have all of this. While waiting for the inspiration
to bring it about, what can we do now that is in the realm of practical reality and
achievement? Let us assume that standard biological treatment and low flow augmenta-
tion as we know them today will be the facilities to deal with pollution, and that con-
ventional plants and methods with minor modifications will continue to be utilized for
water treatment and quality control.
If we start from this assumption, it seems that the required Data Acquisition Sys-
tems that could possibly be used in water supply systems are available on the market
today. This includes equipment for data transmission, recording, storage, and retrieval
and the actuation of secondary devices.
What then is missing? The missing elements are certain sensing and detecting
devices, and the full knowledge of what parameters or variables reveal the cause and
effect relation within the system. This paper will explore these items in more detail,
from the viewpoint of the municipal water manufacturer.
SOURCE OF SUPPLY
The working agreement between the Philadelphia Water Department and USGS has
been centered in recent years on the establishment of a system of water quality moni-
toring stations along the Delaware River estuary. The initial objectives of the Water
Department in this work were:
1. To maintain a surveillance network and to warn of spills above and below the
raw water intake so that remedial action can be taken in time at the treatment
plant.
2. To obtain a continuous record of certain water quality parameters for analysis
so that some of the major cause and effect relationships within the estuary
ecosystems can be resolved.
3. To provide essential raw water characteristics as input for the eventual cyberni-
zation of the water treatment plant, or during the transition period to provide
more meaningful data for better plant control.
At the present time, the U. S. Public Health Service is engaged in a water quality
pollution abatement survey of the Delaware River estuary. This work is being done in
cooperation with, among others, the states of Pennsylvania, New Jersey, and Delaware,
and the Philadelphia Water Department. In this survey, greater need and use have been
found for the water quality data obtained from the monitoring stations. Recently Quigley8
cited another threefold purpose for which the data from the monitoring stations may be
used in part:
"A. The determination of the cause and effect relationship between pollution from any
source and the present deteriorated quality of water in the estuary.
B. Development of methods of forecasting variation of water quality due to natural
and man-made causes.
C. Methods of optimal management, including necessary waste removal and flow
regulation to control the quality of water in the estuary for municipal, industrial,
agricultural, fisheries, recreation, and wild life propagation."
Parker" had considered the six parameters of pH, oxygen, conductivity, temperature,
turbidity, and sunlight intensity to be of major significance for the Incodel-sponsored
Baxter and Radziul 165
-------
automatic water quality monitoring stations on the non-tidal Delaware River above
Trenton. Cleary,6 during the development of the Ohio River Valley Water Sanitation
Commission (ORSANCO) Robot Monitor, questioned and explored water quality param-
eters for the purpose of determining the minimum number of significant parameters
that would be most useful in the Ohio River operation. Cleary and Parker were of
accord with the exception of oxidation reduction potential (ORP), chlorine ion, and
turbidity. Thomann's dissertation9 on the use of systems .analysis to describe the time
variation of dissolved oxygen in the tidal stream says that dissolved oxygen alone is
fundamentally a function of six variables.
"A. The velocity field and diffusion.
B. The temperature field.
C. The salinity field.
D. The presence of organic matter capable of utilizing oxygen in its stabilization
(bio-chemical oxygen demand).
E. Photo-synthetic action by aquatic plants.
F. The presence of chemicals which would utilize or produce oxygen in certain
reactions."
Thomann and Sobel10 describe techniques for the forecasting and optimum manage-
ment of water quality in an estuarine environment. These techniques are predicated
upon the understanding of water quality variations. O'Connor13 in his discussion on
the oxygen balance of an estuary gave no consideration to photosynthetic oxygenation
in the Delaware River. On the other hand, a study1* on the same river by Dr. Hull of
Johns Hopkins University, with the cooperation of the Philadelphia Water Department,
produced real evidence that photosynthesis is a major contributor of oxygen. These
items are noted in an attempt to show the nature and complexity of the unknowns and
their entire relationships. As our knowledge of the Delaware estuary has increased arith-
metically, our awareness of our ignorance has increased geometrically.
Reid15 brings out the complex of interrelating factors involved in the study and
use of streams. From his review of estuarine streams, there can be seen the many
disciplines involved: biology, chemistry, physics, geology, hydrology, hydraulics, mathe-
matics, oceanography. These items and others form the present day concept of the
ecology on which the life and use of our streams is based.
Some of the questions for which answers are needed are
1. What significant parameters of water quality should be measured, for an alert
system, for treatment plant control, for a quality forecasting system, for a river
management system?
2. What should be the periodicity or time interval in collecting specific data?
3. What are the cross correlations of these parameters?
4. Are there any synergistic relationships between the parameters?
5. What is being accomplished to develop instrumentation that can gage quantita-
tively those essential parameters, such as BOD, that are not being measured
automatically at the present time?
If in Philadelphia we could have the answers to all these questions, we could make
further advances and progress in a fully automated water quality treatment operation.
It would appear to the authors that all users of inland surface waters will be faced with
answering these questions at some time in the future, in view of the increasing demand
166 ACQUISITION SYSTEMS IN WATER SUPPLY
-------
for water and the increasing pollution abatement problem. We must close the knowledge
gap and develop the missing sensing and detecting devices before we can optimize the
use of data acquisition systems for sources of water supply.
Much is being done in determining causal relations in the study now under way
in the Delaware estuary. Without the continuous-type data that are made available only
through the use of instrumentation from the new monitoring stations, progress on the
study would be much slower. Further use of these stations will be made, when the Public
Health Service installs on each monitoring station a digital recording system that will
handle up to 10 variables. There is some thought on our part that we may eventually need
15 or 20 positions.
THE WATER PURIFICATION PLANT
Although it is apparent that the United States is in the midst of a science and
technology revolution, with many new products and procedures having come into
existence only since World War II, no new major concepts have been recently developed
in the water treatment field. A Robert A. Taft Sanitary Engineering Center report4
notes that "the basic methods of municipal water treatment have not changed sub-
stantially for almost 50 years." Erdei2 points out some minor advances in water treat-
ment, while spot-lighting the urgency that "in the era of scientific hygiene, the specifica-
tions for water, in accordance with the physiologic needs of man, must be more mean-
ingful and exact."
Busch16 reports that Dr. Keilin of Aero Jet General Corporation invented a new
thin plastic membrane that filters salts and bacteria of body wastes, and viruses and
detergents, and that "this filter could save 90% of the water that cities now discard."
Without discounting such new processes unduly, we believe that the conventional plant
and methods will be with us and in use for some time. What is said following is based
on this assumption, since the authors believe that progress in automation and instrumenta-
tion should be made now.
INSTRUMENTATION AND AUTOMATION IN THE
WATER PURIFICATION PLANT
Although the experience of the authors has been primarily with large water treat-
ment plants located on large rivers, we believe that the comments and suggestions that
follow may also apply to smaller operations. No one should write off instrumentation
and automation simply because of size.
Any consideration of instrumentation and automation of water purification processes
quickly encounters the obstacle that in two of the most important areas where control
is needed coagulation, and taste and odor removal no signal is available on which
to hang instrumentation. There is no analytical method for directly indicating what
steps are to be taken to produce coagulation of a specific water or that proper coagula-
tion has taken place. Similarly, no signal is available to indicate that compounds are
present in water that must be processed to remove bad taste and odor or that the final
processed water is free of taste and odor objections.
Instruments are available for continuous determination of turbidity, dissolved
oxygen (DO), color, temperature, radioactivity, pH, ORP, specific conductivity, phenols,
residual chlorine, and chlorine demand. In general, these instruments are reliable and
are reasonably economical in cost of operation; they give greater accuracy and permit
Baxter and Radziul 167
-------
greater frequency of testing than manual means and techniques, and they have ability
to actuate secondary devices.11
Since most purification operating problems are due to fluctuations in raw water
characteristics and since any of the above instruments will indicate successfully the
fluctuations in the parameter it measures (and record the same), it is surprising that
only in the case of chlorine residual and demand instruments is there direct, general
application of these instruments to control of the water purification process.
There may be some use of the turbidity instrument in adjusting the coagulant
dose, but since there appears to be no direct relationship between this variable and the
amount of coagulant, the turbidity analyzer cannot be converted into a control mechanism.
The automatic pH recorder is of some help, but its use as a direct-control instrument
is limited at the present time.
This leaves the two key purification processes of coagulation and taste and odor
control without a method of direct measurement or means of controlling procedures with
instruments.
There is apparently the lack of complete information about coagulation processes
and the chemicals used in these processes.12 Without such knowledge, the plant operator
can do no better than employ trial and error empirical methods. Although a good
operator can obtain good results most of the time, such methods can result in waste or
misuse of chemicals and poor results. To be on the safe side, many operators probably
overdose with coagulants.
To produce an acceptable water from the standpoint of taste and odor, plant
operators must rely on a periodical manual performance of a time-consuming test, both
of the raw water and water in process. At this point, the control chemist is faced with
the fact that sensitivity to taste and odor in water varies greatly with individuals. It is
possible that in many plants some of the duty laboratory men are not capable of per-
forming an acceptable odor control test.
There are several reasons for the limited use of continuous automatic-analysis
instrumentation in water purification plants. Some of these are the absence of equip-
ment to measure some of the most important factors directly; the fairly high cost of
instruments available; the belief that instruments are so complicated that only a
highly trained man can keep them in satisfactory operation; and a lack of full knowledge
of the benefits to be gained by continuous sampling and analysis.
The automatic residual chlorine instrument is an example of these points, since
it is in fairly common use, is reliable, and can be used to control an essential water
purification chemical. The cost of these instruments at the present time probably elimi-
nates them from consideration by managers of small or medium-size plants. On the
other hand, continuous automatic analysis of residual chlorine content at several points
in the purification process would seem to be a valuable tool. If more plants would use
them, the price would probably come down.
Instrumentation has not yet caught up with the basic requirements of water
purification plants, and part of this may be due to the lack of fundamental knowledge
of water purification processes and controls. This is a matter that affects everyone who
uses public water supplies. It points out the need for research on a national basis.
This need for research on a national basis should not, however, prevent individual
research by operators of water purification plants. In the Philadelphia Water Depart-
168 ACQUISITION SYSTEMS IN WATER SUPPLY
-------
ment, its water quality and research divisions conduct studies whose objective is the
development of better methods of control of chemical dosages and greater efficiency in
the use of chemicals. Bean, Campbell, and Anspach17 have made studies of the Zeta
Potential method of coagulation control and efficiency at the Torresdale Water Treat-
ment Plant. Studies have also been made on the use of polyelectrolytes18 in coagulation.
One conclusion that has been reached is that better control of chemical treatment is
possible after the establishment of optimum rates by measurement of turbidity and Zeta
Potential. Our opinion is that the door could be open for complete automatic control
of coagulant application if instrumentation could be developed for measuring Zeta
Potential automatically and continuously.
Here is the point where science and research meet head on with economics. The
Philadelphia Water Department spends about $1,150,000 annually for chemicals used
in this water treatment process, including a large amount of alum. If automatic monitor-
ing and controls could save 5 percent of this, the annual saving of $57,500 would carry
the capital charges on a large amount of instrumentation. This is the carrot we would
like to dangle before the noses of both water treatment operators and instrument manu-
facturers. If the 5 percent seems too high, even 3 percent would do wonderful things.
PLANT DESIGN
In the continuing process of evolution in all of our manufacturing and commercial
operations, there runs the trend of a minimum of attention and physical effort by plant
operators. This is an element the water treatment plant designer and operator should
not ignore, if for no other reason than the difficulty in attracting and holding competent
personnel.
For this reason, serious consideration should be given in new plants for provision of
centralized control and automatic operation. If fully automated operation is not possible
at the present time, the possibility of such operation in the future should not be ignored.
DISTRIBUTION SYSTEM
Considerable progress has been obtained in automating various distribution elements
of a water supply system. The Load Control Center at Philadelphia20 now transmits and
records to a central point full information about an entire system and also controls from
this same central point nearly all of the pumping stations. The aim is for more complete
and economic control of the system, however, since yearly power costs are $1,600,000
and a small saving in this amount would justify additional instrumentation.
Because of the ability to serve any distribution district with two or more pumping
stations with different power factors and pump efficiencies, a fairly complex problem is
posed when the most economical combination of dispatching water to a district is
considered. It seems to the authors that we will only realize minimum power expendi-
tures when load dispatching is regulated by computers with complete automatic equip-
ment. Brock19 has reported experience in the Dallas City water works of developing a
computer program for distribution network operation.
Bean3 stated that there is great need to obtain analytical information on water
at the point of delivery to the customer. In this case, we are confronted again with water
quality criteria for which there may not be a sensor. Of primary interest to the customer
would be clarity, palatability, tastes and odors, and pressure. The treatment operator
would also be interested in these factors, plus basic chemical and bacteriological criteria.
Baxter and Radziul 169
-------
PERSONNEL
There is another and rather odd form of data acquisition system in the presence of
homo sapiens. It has been said in several places that the half-life of the present
engineering graduate is 10 years, unless he updates himself. If the water industry is to
proceed along the path of automation, being followed by other industries, it will have
to update its personnel or its homo sapiens data acquisition system. It seems manda-
tory and imperative that industry and education accept the responsibility of updating
the men in this field. There are many ways of doing this, but it should be thorough
and complete. Without belittling in-training courses or self-teaching, there is also room
for the 6-month or 1-year sabbatical for instruction of personnel who can assimilate the
new techniques. The water industry itself will have to recognize that the cost of such
complete training is as important as the cost of a new building or of power or chemicals.
Rather it should be said that it is more important than these items.
The water industry has talked about the shortage of trained men, ranging from
top professionals to technicians and operating personnel. A fair share of the best men
must be attracted to the industry. They will come if they are given the same challenges
and the same modern operating features they will find in other industries.
Much of water treatment plant operation, including the laboratory work, is routine.
Good men will shy away from work that is all routine. They need the challenge of some
amount of research and application of new ideas. The field of automation in water
purification can provide that.
ECONOMICS
As indicated earlier, economics and cost cannot be ignored in the field of automation
and instrumentation. A water works manager can only justify the cost of instrumentation
if he can prove that it will result in lower operating costs or in better quality, for which
an economic value can be given. This was outlined in detail in the discussion of the
Philadelphia Load Control System.20 Therefore, instrumentation in the treatment plant
must be balanced by a reduction in cost of personnel, in chemicals, and in general
maintenance and operation.
The other factor that cannot be ignored is the cost of instrumentation. If development
costs must be reclaimed through the sale of only a few instruments, the cost of these
instruments will be high. If there is much use and demand, the costs will be lower.
Since instrument cost is a basic factor in this, possibly it behooves all of us in the
water industry to lift ourselves by our boot straps by using new instrumentation where-
ever possible.
REFERENCES
1. "Water Quality Measurement and Instrumentation" Transactions of Seminar, R.
A. Taft Sanitary Engineering Center, Cincinnati, Ohio (August 1960).
2. Erdei, Joseph F. "Advances in Water Treatment" Journal AWWA 55:845 (July 1963).
3. Bean, E. L. "Progress Report on Water Quality Criteria". Journal AWWA, 54:1343
(November 1962).
4. "Biological Problems in Water Pollution" Transactions of Seminar, R. A. Taft
Sanitary Engineering Center, Cincinnati, Ohio (April 1959).
170 ACQUISITION SYSTEMS IN WATER SUPPLY
-------
5. Committee Print No. 3, "Study and Investigations of Use of Materials and New
Designs and Methods in Public Works" Committee on Public Works, U. S. Senate
(1962).
6. deary, E. J., "Development of a Robot System". Journal AWWA 50:1219 (September
1958).
7. Parker, B. W., Freeberg, J. A. and Barber, S. B., "Automatic System for Monitoring
Water Quality". J 1. Sanitary Engineering Division, A.S.C.E. Paper 2554, SA 4. Vol.
86, p. 25, July 1960.
8. Quigley, James M., "Statement on Water Quality Management of Delaware
Estuary". Presented before Natural Resources and Power Subcommittee on Govern-
ment Operations, at Trenton, N. J., August 9, 1963.
9. Thomann, Robert V., "The Use of Systems Analysis to Describe the Time Variations
of Dissolved Oxygen in a Tidal Stream." A dissertation in the Department of
Meteorology and Oceanography submitted to the faculty of the Graduate School
of Arts and Science in partial fulfillment of the requirements for the Degree of
Philosophy at New York University, N. Y. (November 1962).
10. Thomann, R. V. and Sobel, M. J., "Estuarine Water Quality Management and
Forecasting". Presented at ASCE Water Resources Conference, Milwaukee, Wis-
consin. (May 15, 1963).
11. Jones, R. H. and Joyce, R. J., "Instrumentation for Continuous Analysis", Journal
AWWA 53:713 (June 1961).
12. Larson, P. E., "Research, Needs, Priorities, and Information Services", Journal
AWWA 54:657 (June 1962).
13. O'Connor, D. J., "Oxygen Balance of an Estuary", ASCE 1961.
14. Hull, C. H. J., "Photosynthetic Oxygenation of a Polluted Estuary", Report No.
XIII, Low-Flow Augmentation Project, The Johns Hopkins University, January 1962.
15. Reid, W. C., "Ecology of Inland Waters and Estuaries", Reinhold Publishing
Corporation, 1961.
16. Busch, H., "Pollution Problem Gets More Attention", The Ensign, July-August 1963.
17. Bean, Campbell and Anspach, "Some Aspects of Zeta Potential", Presented before
Pennsylvania Section of AWWA, June 5, 1963.
18. Campbell, S. J., "Coagulation Studies Nalco 614, Jaguar W.P.B.0 and Narvon
Activated Clay Z3". Unpublished report, Philadelphia Water Department, July
26, 1963.
19. Brock, D. A., "Closed-Loop Automatic Control of Water System Operations",
Journal AWWA 55:467, April 1963.
20. Baxter, S. S. and Appleyard, V. A. "Centralized Load and Quality Control
Systems at Philadelphia". Journal AWWA 54:1181 October 1962.
DISCUSSION
Mr. Baxter was asked whether prizes have been used as an incentive to manufacturers
to develop instruments needed for environmental measurements. He indicated that
Baxter and Radziul 171
-------
although this has been considered, the main incentive for industry probably is more
business and a wider market. Instrument manufacturers must have a reasonable prospect
for the sale of the instruments they develop. In the meantime we may have to make
greater use of the instruments now available.
Although no automatic instrumentation was in use in the Delaware Kiver project 3
years ago, an instrument that measures six parameters continuously is now in operation.
Manufacturers of this instrument anticipated a sufficient market to warrant the develop-
ment costs. Mr. Baxter challenged water researchers to use more instruments in the
production of water and suggested that if more instruments were purchased, the manu-
facturers would do more to develop cheaper and better instrumentation.
Mr. Mentink asked what use is made of the water quality data collected by the moni-
toring installation above the Philadelphia water intake. Mr. Baxter replied that the
information from the automatic instruments is not being used to change or to regulate
day-to-day operation of the water plant. If continuous valid water quality data were
available, it could be used to adjust the automatic facilities used in operation of the plant.
Mr. Mentink commented that there are apparently cost differences in processing
water with differences in turbidity, pH, dissolved solids, chlorides, etc., and that some
savings in cost should result from knowing these different qualities. Mr. Baxter agreed that
cost differences, especially with turbidity, could be significant. The cost of alum, one of
the largest chemical costs in water processing, could be reduced if water quality were
known in more detail. For example, if a plant operator wants to make sure that the
chemicals are adequate to accomplish coagulation, he will generally use an excess.
Better information on the quantity of chemicals needed would allow savings of some of this
excess. As another example, if continuous information showed the presence of chemi-
cals that cause taste and odor, such as phenol, an operator could initiate the addition
of activated carbon, which would not ordinarily be used.
172 ACQUISITION SYSTEMS IN WATER SUPPLY
-------
Paul De Falco, Jr.
Director, Raritan Bay Studies Project
Division of Water Supply and Pollution Control
U. S. Public Health Service
SUMMARY
In a data acquisition system for water quality control, the "need to know" and the
ability to use the information collected must be carefully examined. Past data should
be reviewed to establish the frequency and location of representative stations. The data
handling and analysis .system must be within the program's limitations of need, interpre-
tation, personnel, equipment, time, and money. Data should be collected in accordance
with a plan that best meets all these requirements. At regular intervals during the
program of studies, data collected to date should be reviewed so that it can be determined
whether needs are being met and whether alterations in the system are necessary. Any
changes should be fed back in such a way that an operating system will continue to
meet the program's requirements.
DATA ACQUISITION SYSTEMS IN
WATER QUALITY CONTROL
An impressive assortment of data collection systems was described at the "Symposium
on Water Quality Measurement and Instrumentation" held here in August of 1960.
I left the meeting with the feeling that collecting data was the easiest part of the problem
in our business. What I had not heard discussed at any great length was just why we
were collecting these data and what we were supposed to do with them once collected.
Since that time, a lot of water I don't know what the quality was has passed under
the bridge. We are still collecting data.
Collecting data is usually a simple procedure. What complicates the picture is
answering the questions:
Why are the data needed?
Where should they be collected?
When should they be obtained?
What will be done with the information?
In the development of measurement systems in water quality control we have reached
the point where we must critically answer these questions before embarking upon a
course of studies requiring collection of data. Data collection is expensive. It costs
between $25 and $50 to analyze a sample and almost again that much to collect it.
When the costs of processing and storing data are also considered, it becomes obvious
that no data should be accumulated without a positive justification.
Recently, I was asked by a representative of a state water pollution control program
to assist in the development of a water quality surveillance system for his state. My
answer was to ask him a series of questions:
What were his agency's needs for information?
What objectives would this system be required to meet?
What parameters were of value?
De Falco 173
-------
What past records of water quality were available to indicate frequency of
sampling and the location of stations to best describe changes in quality?
What data utilization system was available in the department to handle, analyze,
and interpret the data collected?
Let us face the facts. The data acquisition system we usually adopt is a compromise
between the system we need and the system we would like. We sometimes collect data
because it is convenient ... it is nice to know . . it may be important some day . . .
and because everybody else collects it. We should collect it because we have a use for
it, and more importantly, because when interpreted it will answer the needs of our
organization. The limits of an/ single element in the chain from collection to utilization
should set the limits for the individual steps. The extreme shortage of skilled engineering
and scientific personnel in our field does not allow us the luxury of collecting interesting
or unduly refined data unnecessary to the needs of our organization.
ESTABLISH PROGRAM NEEDS
The design of a water quality control data acquisition system must evolve from the
needs for the information in the operation of specific programs. These needs can generally
be classified as follows:
... To determine compliance with a given standard.
... To determine or forecast the effect of a water resource project.
... To determine treatment needs in the use of the water.
... To provide water quality control.
To determine compliance with a given standard, or criterion, several levels of
sophistication are available. These range from the simple go no-go decision in a measure-
ment that indicates whether a level is being exceeded or not to the more elaborate
models with built-in warning systems to indicate in advance that remedial action may
be required at a future time.
Water resource projects must be evaluated for the effect they may have on the
quality as well as the quantity regimen of the stream. This requires the measurement of
changes by "before and after" studies of the project. Forecasting changes without
knowledge of the ''after" conditions is possible, although difficult. It requires careful
correlation to changes encountered in similar projects elsewhere.
The measurements required in the utilization of water, such as in a water treatment
plant, usually are well identified as those elements that may be controlled by treatment
or that may affect the safe usage of the water.
The measurements needed to provide quality control, either through stream or waste
flow regulation, are usually those that are indicative of the problem being controlled.
These measurements generally are required only for the period during which the flow
is available for regulation.
DEFINE OBJECTIVES
The objective of a measurement and data acquisition system is to temporally and
spacially characterize the quality of the stream or body of water with respect to the
parameter chosen. Definition of changes that occur in quality between periods of time
or given locations satisfies the needs of most water quality management agencies; how-
174 ACQUISITION SYSTEMS IN WATER QUALITY CONTROL
-------
ever, the degree of sensitivity to change that is required by each agency depends upon
the ultimate use that the agency makes of its collected data. This in turn determines
the degree of sensitivity needed in the measurement system.
Adequate time must be given to the planning of a data acquisition system. Such
planning includes programs for data analysis and interpretation prior to and con-
current with data collection, as well as after collection. One must continually examine
the data being collected to ensure that the objectives of characterization of base quality
and the changes in quality are being met.
SELECT PARAMETERS
The decisions on the parameters to be measured are most important. If the agency's
need for water quality data has been carefully denned, the parameters that are a direct
or indirect measurement of the water quality need to be fulfilled are normally obvious.
The problem lies in the addition of parameters that are not necessary to the needs of
the agency, or to the neglect of parameters that are interrelated with the parameter desired.
The addition of parameters to a study should be carefully weighed with respect to
the cost of collection and analysis as well as interpretation. Any extra cost might better
be used for expanding the temporal or spacial network for sampling the parameters of
direct interest. For example, an agency conducting a program to determine compliance
with a bacteriological standard should weigh carefully the productivity of additional tests
for chemical quality, as opposed to productivity of increased bacteriological examinations
in the fulfillment of its mission.
Conversely, the interrelationship of some parameters requires the measurement of
additional parameters so that the phenomena being observed can be described more
adequately. An example of this is dissolved oxygen, which is interrelated with tem-
perature, conductivity, BOD, turbidity, algae, solar radiation, wind, and hydraulic
characteristics.
The parameters themselves often dictate limitations on the sampling system. Many
of the parameters we are most concerned with cannot presently be measured auto-
matically. Others require such extensive laboratory work that the temporal or spacial
grid desired for proper interpretation is limited. Modification of test methods is sometimes
warranted to establish screening procedures. When the presence of the parameter has
been qualitatively established, a more intricate quantitative analysis can be set up.
Tests for phenols are a good example.
SEARCH HISTORICAL DATA
Prior to the establishment of a data acquisition system, existing data should be care-
fully reviewed. Many agencies and organizations collect water quality data for special
purposes. Much of this information, although not necessarily oriented to the need of
the proposed data acquisition system, provides background knowledge of the type of
variability existent in waters to be monitored. Information on the type of temporal or
spacial variability is a prerequisite to the development of a good measurement system.
Sources for information of this type include past studies by the agency concerned;
by other public agencies in the water quality management field; by public, private, and
industrial water suppliers; and by sewage and industrial waste treatment plants. The
sampling and analytical procedures used by others must be carefully evaluated to qualify
the value of the data collected. On the Delaware River, we found that more than 40
De Falco 175
-------
groups were collecting water quality data. The analysis of these data permitted a signifi-
cant reduction in the field work required to meet the program's objectives.
DEVELOP DATA UTILIZATION
At this point in the development of a data acquisition system, it is necessary to
develop more completely the data utilization program to be followed. Included in this
program, in order of consideration, are data interpretation, data anlysis, and data
handling.
Data interpretation is simply the meaning that is to be placed upon the possible
values that may occur in the selected parameters. By this time, the specific questions to
be answered by the study should have been formulated. These questions formulated in
terms of hypotheses to be tested by measurement enable the statistical consultant to
design an adequate program of sampling, data handling, and data analysis.
Data analysis may be limited by the resources of the agency. The collection scheme
must, however, meet the limitations of the analysis system available to the agency. And,
in turn, the data handling system, whether it be simple forms, or punch cards, or tape,
must mesh with the data analysis system available.
It is valuable to test the system chosen with historical data to determine whether
it responds with the required degree of refinement. Additional tests of the system should
be made at regular intervals during the study to determine whether the collection system
is meeting the needs of the program. The feedback from this series of checks should
augment or correct the collection program as needs are determined.
DETERMINE FREQUENCY OF MEASUREMENT
Another area of decision confronting the engineer planning a data acquisition system
is the frequency of measurement This, again, is determined by the "need to know."
At times a continuous measurement of quality is required such as an alarm or alert
system in which a parameter goes out of control and immediately requires remedial
action. The water treatment plant is an ideal example of this area of decision. In other
cases, a much longer time interval between measurements meets the needs of the agency.
In most cases, the changing quality of a body of water is being characterized and for
this a knowledge of the time behavior of the chosen parameter is required. Most
parameters vary with time, i.e., with natural changes in water flow and temperature.
Superimposed upon these are the transient effects of waste discharges. The design of
a measurement system requires a knowledge of the types of changes that occur and the
periodicity or trend of these changes. The data collection program must be designed
statistically to develop meaningful information with respect to these changes and
their causes.
LOCATE SAMPLING STATIONS
The geographic or spacial distribution of sampling stations requires the same careful
consideration that has been given to the other elements of the system. Where possible,
continuation of existing locations should be considered to give continuity with past
studies; however, the choice of stations must first meet the needs of the program.
Location must be representative of the water body being sampled and indicative of the
changes that are occurring in the parameters being monitored.
ACQUISITION SYSTEMS IN WATER QUALITY CONTROL
-------
Consideration should be given to obvious factors such as horizontal and vertical
stratification due to temperature or specific gravity. Knowledge of the location of dis-
charges should be considered before selection of stations. A series of dispersion studies
to determine proper sampling locations may be required before a final decision is made.
DISCUSSION
Mr. Fry commented that selection of parameters should be greatly emphasized.
It may be a waste of time to collect data on a given parameter, such as BOD, simply
because our profession has accepted it as an essential parameter in the measurement of
water quality. Rather than obtaining simple measurements of BOD in a stream, we might
attempt to measure the rate of change of oxygen consumption or perhaps the total carbon
content of a water. We should look for direct measures of water quality rather than
trying to apply formulas and empirical approaches in our attempts to understand
streams by indirect measures.
Mr. Radziul observed that the first two papers presented had definitely carried the
undertone that we do not know enough about water quality and that we must determine
the cause and effect relationships that exist. He agreed that the yet-to-be-discovered
parameters may be the controlling ones and that old ones may have to be discarded.
Mr. Weaver commented that this discussion underscores the point that regardless of
the black boxes and transducers that may be developed, manpresumably with some
professional background and judgmentwill still be very much in the picture.
Dr. Williams pointed out the need for better communications between scientists
and engineers'to make effective use of the available information. For example, although
the development of chemical and physical instrumentation is urgent and desirable, plank-
tonic organisms can be used now to provide valuable information on water quality. The
species diversity of the planktonic organisms in raw water is a new parameter that has
been worked out and is available for use.
Mr. Stern mentioned that an instrument system has been developed that will measure
total carbon in a given water or waste water sample. Experimental units are in operation
now, and further development is under way. This system will be available soon, and
more such instruments should become available as time goes on.
Mr. Cohen noted that the present data acquisition systems are mainly based on
probe-type devices the DO probe, pH probe, and probes for temperature, specific con-
ductivity, and chlorides. He pointed out that investigators in air pollution are now using
four or five different wet-chemical constant-feed devices that yield continuous data, and
that water researchers should give greater consideration to this type of device. While
the work these probes are doing for us is fine, we should be looking toward other types
of devices as well.
De Falco 177
-------
W. L. Isherwood
Hydraulic Engineer
Water Resources Division
U. S. Geological Survey, Washington, D.C.
SUMMARY
The U. S. Geological Survey has put u great deal of effort into the development of
equipment and techniques for recording river gage heights in the field in such a way that
the data can be efficiently processed into river flow data by use of a digital computer.
Some false starts were made, but they now have a workable system, described here, by
which records for nearly a thousand river gaging stations are being routinely processed
through a computer. They plan to expand the system at the rate of about a thousand
stations a year until nearly full conversion of their stream gaging network is reached.
In addition, they are beginning to apply similar techniques with slightly modified equip-
ment to the recording and processing of other types of hydrologic data, such as precipita-
tion, temperature, chemical quality of surface water, and depths to ground water in wells.
DATA ACQUISITION SYSTEMS IN HYDROLOGY
BACKGROUND
For many years the Geological Survey has collected records of river gage heights and
other types of hydrologic data on strip charts. Therefore, it was natural that we should
first think of automation in terms of beginning our processing by an automatic reading
of the data directly from the strip charts. We spent considerable time and money trying
to develop a photoelectric chart scanner, and at times it seemed that we almost had the
problem licked. But we were never quite able to overcome the basic fact that we could
not satisfactorily control the quality of inked lines drawn automatically on strip charts
at thousands of isolated field installations. Some lines would always be too watery,
other would soak into the paper or smear, and there would often be some dirt spots and
occasionally paper flaws indistinguishable from the real data line. Somewhat reluctantly,
we finally came to the conclusion that automatic chart reading was impractical and that
we must have a really unambiguous record if we were to be able to process large masses
of data automatically with any real efficiency. Punched paper tape seemed to be the
best recording medium to fit our requirements, because each bit of data is represented
by either a hole or no hole in the tape, with no possible intermediate condition. We then
investigated several different kinds of paper tape punching devices to try to find one
capable of battery operation over extended periods at isolated field installations and
still simple enough so that it would not be unduly expensive. A device manufactured
by the Fischer and Porter Company seemed to have possibilities for adaptation to our
needs. The unique feature of this device was a system of recording on punch tape by
positioning a disc containing ridges and valleys in such a combination that a complete
reading in parallel mode could be punched on a wide tape with a single stroke. That is,
the input could be continuously positioning the code disc until a reading was called for
and then a single throw of a punching device would punch all the digits in that reading
at once in a single row of holes.
DEVELOPMENT OF THE DIGITAL RECORDER
Of course, many modifications were necessary to make this existing device fill our
particular needs. For instance, the original device was operated by a-c power but we
Isherwood 179
-------
needed battery operation. This took only a simple modification. The original device
used a single large code disc divided into 1000 code divisions so it could record only a
three-digit number, but we needed to record four-digit numbers. This required a more
fundamental modification. To make it punch four digits, it was necessary to change
from one large code disc to two smaller discs, each divided int« 100 code divisions. The
two discs were connected by a 100 to 1 worm gear so that one revolution of the low-order
disc turned the high-order disc one division. Then it was necessary to devise a mechanical
non-ambiguity system to prevent trying to punch somewhere between two divisions on the
high-order disc. A cam and lever system was devised to adjust the high-order disc
exactly to the proper discrete digit position just prior to the moment a reading is
punched out. Figure 1 shows the digital recorder as finally developed.
Figure 1 Digital Recorder Developed to Record River Gage Heights.
TAPE CODING AND TRANSLATION
The format of the punched output was carefully considered. Parallel punching
provided information with the least power consumption and with the simplest field in-
strumentation. But it was recognized that such a format would require translation before
entry into any computer because computers require serial input. The necessity for
180
ACQUISITION SYSTEMS IN HYDROLOGY
-------
intermediate translation was something of a blessing in disguise, however, because it
made our field instrument entirely independent of computer requirements. We expect
eventually to have several thousands of these field instruments and we expect them to
last many years before replacement. But we were already in the midst of changing
computers while this development was taking place and we knew that computers were
changing so rapidly that we would have no way of knowing what kind of computer we
might have 10 years from then. The answer seemed to be to make a translator that
would have a fixed input for field tapes but would have a completely flexible output
that could be easily changed to fit the input requirements of any computer we might
have. This is the system we adopted. The translators we now use can punch out
serial-coded paper tape in any sequence or grouping of characters required by a computer.
FIELD TESTING
After the successful production of the basic recorder we conducted large-scale
field tests covering widely varying climatic conditions and a large range of river regime
types. The initial field tests involved installation at 80 gaging stations for a full year,
20 in each of 4 areas, the New England states, Alabama, Kansas, and California. After
working out the few bugs that showed up in these field tests, we were really ready to
start using the digital recorders in large numbers.
PROCESSING PROCEDURES
The processing done on these river flow records involves first the paper tape transla-
tion using off-line equipment, then a primary computation performed immediately on
receipt of the record from the field, and at the end of each water year an updating
process and a final print for publication. The primary computation computes figures of
daily mean discharge plus several other useful items for each day and prints these pre-
liminary results on a sheet with one line for each day (Figure 2). At the same time a
summary of the daily data is stored on magnetic tape. The updating process at the end
of the year allows insertion of data for periods not available on the original record,
substitution of gage height or discharge figures for periods when unusual hydraulic
conditions prevailed, and recomputation of figures of discharge on the basis of more
up-to-date information on the stage-discharge relation. The final printout sheet of daily
discharges with monthly and yearly summaries is in a form usable as offset manuscript
for publication (Figure 3).
ALTERNATE METHODS OF COMPUTING DISCHARGE
Of course, discharge at all gaging stations cannot be computed by exactly the same
method. We started out by programming the computations for the most frequent situa-
tion where simple stage-discharge relations can be developed. This type of computation
can be used for at least three-quarters of all our gaging stations. Later we added alterna-
tive programs for some of the more difficult hydraulic conditions, and we will continue
to add other alternative computation methods for other situations from time to time.
One of the alternative computations now available is for the so-called "slope stations"
where simple stage-discharge relations do not apply. For this type of station, gage
heights are recorded by two separate instruments at both ends of a suitable reach of
channel, and interrelationships between stage, fall, and discharge are used to compute
figures of discharge. This method works well where steady flow or nearly steady flow
conditions prevail a major portion of the time. For sites where unsteady flow conditions
prevail generally, such as in reaches affected by tide, a much more complicated computa-
Isherwood 181
-------
c/>
3
O
cc
on
HH
«
«!
o
P9
o
o
o
7-182?.1C NEOSriO RIVEP AT PURLINGTON.
82
32
81
80
79
6.78
6
8
9
10
10
9
9
8
8
9
10
9
8
8
8
7
7
7
7
7
7
7
13 7
e
,
.
^
a
*
*
9
*
*
^
»
.
78
07
69
63
30
68
99
35
40
45
71
45
57
22
02
90
80
71
66
61
52
46
44
°F9IOD 11.61 6.50
WATER YEAR ENDING StPT. 30. 1963
BI-HOu^LY GAGE HEIGHTS ISrMBOL TEST DIFF 0.1)
itAN Q 0200 0405 0600 Obju 1000 12'.0 1400 1600 loOO 2000 2200 2400 TMAX
314 06b2 OoB2 0681 Coal 0661 0681 134=
312 w681 0631 0681 0681 0681 uc,bl 06dl 06bl 0681 06bl 0681 0681 2400
309 0681 0681 0681 06B1 0661 u6t>v 0630 0680 0630 0630 0680 0680 1000
3J6 06bO 0630 Oo80 ^680 0680 0679 0679 0679 0679 0679 0679 0679 10l5
301 C679 0679 C67B 0678 0673 0676 0678 0678 06/8 0678 0678 0678 0530
30C ,678 0678 0678 C678 0678 067d 0678 06/8 U676 0678 0678 0678 24QO
300 06/8 0678 0678 C678 0678 0676 C677 0677 0673 0678 0678 0678 1830
1020 C678 0678 0680 07U9 0732 0754 C801 0372 0908 0921 0932 0960 2400
2190 0993 1018 1032 1033 1019 0996 0967 0938 0915 0898 0892 0900 0800
2B30 C9^6 0966 1011 1056 1097 112d 1151 1160 1159 1153 1135 1114 1730
2540 1039 1067 1048 1031 1022 1015 1013 1009 1008 1004 1000 0996 00l5
218J 0936 0980 0969 0958 0944 0932 0919 0916 0947 0987 1025 1068 2400
2360 1094 1087 1064 1040 1019 099b OS7d 0958 0936 0918 0900 0887 0315
1240 0673 0663 0351 Oo43 0634 0829 0^23 0819 OB16 0811 0808 080* Ool5
1280 0812 0814 0816 OB19 0821 0625 0331 0342 0654 0870 088s 0902 2400
2050 0911 0914 0917 0918 0921 0928 0939 0951 0*6B 0962 1002 1022 2400
2780 1045 1066 1080 1088 1094 10*3 1090 1082 1072 1057 1044 1028 124=
2050 1013 0994 0979 0964 0950 0936 0925 0916 0905 0699 0391 0884 0015
1430 03/9 0373 0666 Ud63 0860 0854 0352 0347 0844 0840 0835 0834 0015
1130 -S33 0829 0827 0324 0821 0321 0818 Odl6 Ool5 0813 0811 Ob09 0045
979 0808 0306 0805 ^302 0802 0802 0799 0799 0797 0797 0796 0796 0100
894- 0794 0793 0793 0789 078B 073d 07Sb 07dB 0736 0766 0786 0785 0115
825 C734 0783 0783 C7S1 0780 07/0 0778 0776 0776 0774 0774 0774 0300
766 07/2 0772 0771 0771 0771 077U 0770 0770 U769 0767 0767 0767 0545
736 0767 0767 0767 0767 0766 0764 0754 0764 0764 0764 0764 0764 0930
706 C7o3 0763 0763 07&3 0760 076iJ 0760 0760 0759 0758 0758 0756 011 =
654 j755 0755 0754 0753 0751 0751 0751 0751 0749 0749 0749 0743 011=
619 0748 074B 0747 0746 0746 U746 0746 0744 0744 0744 0744 0744 044=
6C6 0743 0743 0743 U743 0744 0744 1300
Figure 2 Preliminary Printout Showing Bi-Hourly and Daily Gage Heights and Discharges.
-------
tion has been programmed to obtain figures of discharge. Again, two records of stage
are needed, one at each end of the reach. But for each 15-minute time interval, an
analysis of the unsteady flow condition is made. This analysis involves an approximate
numerical solution of two first-order quasi-linear hyperbolic partial differential equations
of two dependent and two independent variables. This type of computation is roughly
10 times as expensive as that for the regular gaging station, but it is presently the only
successful method we have for computing flow in tidal reaches of large estuaries.
Another program has been developed for "deflection-meter"' stations. At these stations
a movable vane mounted in a fixed position in a channel gives an index of velocity and
direction of flow. Two recorders at the same site are used, one for the stage record and
the other for the deflection-meter record. For each 15-minute time period, discharge is
computed as the product of an area and a velocity. Since direction of flow is taken into
consideration, this method is usable for canals or small streams in tidal reaches as well
as for small channels whose very low velocities make it impossible to establish a definite
stage-discharge relation and for stations on small channels where conditions other than
tides cause changes in flow direction.
STATION NO. 3-
-------
fied slightly by the Geological Survey for use as precipitation gages. In this simple
digital precipitation gage, a small float attached to a pulley on the digital recorder input
shaft measures accumulated depth of water in the rain gage reservoir. Incidentally, the
Weather Bureau has recently sponsored development of a much more elaborate weighing
rain gage in which the same basic recording techniques and the same type of punched
output tape are used.
The standard recorder with very minor modification is being used for measuring
depths to ground water at about 40 points in a project in Arizona where a detailed
coordination with simultaneous records of soil moisture and nearby stream flow records
is required.
The only major modification of the standard recorder made by the Geological
Survey has been the adaptation of the basic recorder to accommodate electrical rather
than mechanical inputs. This is important because sensing devices for chemical quality
items generally have electrical outputs. Also, with electrical input, it only takes a small
additional modification to allow for multiple inputs. The modification for electrical input
involves using a positioning motor on the input shaft in place of the direct mechanical
input. If multiple inputs are desired, a stepping switch can be inserted to connect the
recorder to each electrical input sensor in turn at the time that each set of readings is
desired. The Geological Survey has added such modifications to a few standard recorders.
For instance, one recorder has been adapted by our Quality of Water office in Florida
to record four items consisting of top and bottom temperature and conductivity in a
particular stream. The manufacturer can supply units already adapted for single or
multiple electrical inputs.
In addition to the modifications for different kinds of input, a telemetering device
can be attached to the standard recorder. In fact, the basic recorder was specifically
designed with this option in mind. Space was provided just behind the punch block for
a set of contacts that are operated by the punch pins at the moment of punching. A
wiring cable brings this information to an external box containing circuitry to store
the last punched information until the next reading and to decode the information from
binary-decimal to straight-decimal and to transmit it as a sequence of recognizable tones
to telephone or radio transmitting equipment. A number of the telemetering attachments
with telephone transmitting equipment are being added to our river gages in a coopera-
tive project with the Weather Bureau. Certain of our river gages now equipped with
digital recorders can be used by the Weather Bureau for flood forecasting.
CONCLUSION
No doubt, there are other, as yet unconsidered, possibilities for use of this same
basic field recorder in hydrologic investigations. But our uses of this recorder and the
computer processing techniques for streamflow data alone save us significant amounts of
manpower at regular stations and enable us to obtain flow data in places where we could
not obtain the information in any other manner. The release of some of our technical
manpower from the drudgery of routine data processing allows us to do more interpre-
tive work that will lead more directly to the solution of specific water problems and to
the expansion of our general knowledge of the behavior of water in nature.
DISCUSSION
Mr. Isherwood indicated that three translator systems are being used in the
system. The first two cost about $6,000 each and the third, which operates four times
184 ACQUISITION SYSTEMS IN HYDROLOGY
-------
faster, costs about $12,000. It is expected that a magnetic tape translator that operates
10 times faster than the last will be needed soon and will cost about $15,000.
He indicated that the group has used three different computers. In each case they
started out at a low use rate but within 3 to 4 years were operating 24 hours per day; they
expect to need a larger unit in the near future. The paper tapes are being read by a
standard photoreader that reads about 1000 characters per second; it rents for about
$14,000 per month. When thousands of stations are in operation, however, this reader
will be too slow. They plan then to use magnetic tape that can be read at 25,000 to
50,000 characters per second.
In response to a question about the use of paper tape under high humidity, Mr.
Isherwood indicated that the papers have held up remarkably well. The National Bureau
of Standards indicated that so-called waterproof papers get just as wet as standard pa-
pers, but not as quickly.
Recently, a foil-backed paper with more mechanical strength has been introduced.
Mr. Isherwood further indicated that the electrical components are more sensitive to
moisture than the mechanical parts of the system.
Isherwood 185
SPO a 14-105-7
-------
John J. Gannon
Associate Professor of Public Health Engineering
School of Public Health
University of Michigan, Ann Arbor
SUMMARY
Several graphical and statistical procedures are presented for the interpretation and
analysis of hydrologic data from the standpoint of influence on water quality. These in-
clude the hydrograph; development of seasonal patterns on normal and log-normal
probability papers; analysis of drought flows with examples of procedures used in a
Michigan study; determination of time of passage by displacement calculations and tracer
methodology; and comparison of potential regulated flows and natural flows. These
procedures used with good judgment have proved their usefulness in many water
pollution investigations.
THE INTERPRETATION AND ANALYSIS OF
HYDROLOGIC DATA
Several graphical and statistical procedures are available for the interpretation and
analysis of hydrologic data from the standpoint of influence on water quality. It is the
intent of this paper to present and discuss the use of the hydrograph both continuous
and daily average; normal and log-normal probability papers and their use in developing
seasonal patterns; log-extremal probability paper and its use in analysis of drought
flows, including certain adjusting and summary procedures; systematic studies of the
influence of flow regulation; and the importance of knowledge of the physical character-
istics of the river channel and the use of this information in determination of river
time of passage. Certainly, this is not an exhaustive list of all hydrologic considerations,
but these graphical and statistical procedures have proved useful in many water pollution
investigations.
THE HYDROGRAPH
The hydrograph consists of a graph of time versus river flow at a particular location.
The time scale in some cases is presented on a continuous basis, resulting in an instan-
taneous hydrograph frequently expressed in terms of river stage rather than runoff. In
other cases, the time scale is presented on a daily basis, with the flow averaged over the
day resulting in a daily hydrograph. In still others, it is presented on a monthly basis,
with the flow averaged over the month resulting in a monthly hydrograph. Each has
its advantages and disadvantages. The continuous and daily-average hydrographs are
presented and discussed below; the monthly average hydrograph is considered later in
connection with the development of seasonal patterns of runoff.
THE CONTINUOUS HYDROGRAPH
Most of the important stream-gaging stations maintained by the U. S. Geological
Survey in the United States operate on a continuous basis, generally making a continuous
recording of river stage with time. This river stage must, of course, be converted to
discharge, by means of an appropriate rating curve. In river situations where there are
diurnal fluctuations in flow frequently induced by activities of man it is important
to know of these fluctuations and to minimize their effect in the design of any stream
sampling program. Here is where a continuous-gage chart can be extremely valuable.
Gannon 187
-------
Unfortunately, these charts are not routinely published and must be obtained from
the files of the appropriate district engineer's office of the Geological Survey.
In two recent intensive stream surveys conducted by the writer, hour-to-hour fluctua-
tions in stream flow through the critical reach of river were observed during the dry
warm-weather period of August. This is the ideal period for evaluation of a water
quality problem related to organic wastes.
The first survey involved the Clinton River in Michigan and covered the section of
the river from below the Pontiac waste treatment plant outfall to the village of
Rochester, a distance of 11.41 river miles. Fortunately, a Geological Survey continuous-
recording stream-gaging station is located at Auburn Heights in the critical stretch
3.02 river miles below the Pontiac waste treatment plant outfall. Figure 1 shows the
5,2
intensive
sampling period
Wed
AUGUST 17
Thurs
18
Fri
19
Sat
20
Sun
21
1960
Mon
22
Tues
23
Wed
24
Thurs
25
Figure 1 Gage Chart for Clinton River at Auburn Heights, August 17 through 25, 1960.
Auburn Heights continuous-gage chart for the period August 17 through 25, 1960.
It can be seen that a definite hour-to-hour flow fluctuation exists and is produced pri-
marily by flow variation from the Pontiac waste treatment plant. Average flow during the
intensive 48-hour sampling period, August 23, 24, and 25, 1960, was 33 cfs at Auburn
Heights, while the average flow from the waste treatment plant was 16.9 cfs; thus, the
treatment plant effluent made up more than 50 percent of the total river flow during
this period. Also, it can be seen that while the average flow was 33 cfs, the actual
flow ranged from 26 to 46 cfs. Although it was not possible to alter the flow pattern,
it was possible to minimize this influence by collecting river samples every 4 hours
around the clock for a 48-hour period through the critical section.
The second survey involved the Tittabawassee River in Michigan and included the
section from below Midland and the Dow Chemical Company waste treatment plant
outfall to Saginaw, a distance of 19.25 river miles. Fortunately, as in the previously
mentioned cases, a Geological Survey continuous-recording gage is located on the
Tittabawassee River at Midland opposite the grounds of the Dow Chemical Company.
Figure 2 shows the Midland continuous-gage chart for the period August 17 through
25, 1961. Diurnal fluctuations in flow are induced during weekdays by a hydropower
installation upstream from Midland, and these fluctuations are illustrated by the usual
variations on August 17 and 18. For purposes of conducting an intensive stream
sampling program under steady-flow conditions below Midland, arrangements were made
with the hydropower company to lower their reservoir on August 21 and 22, and then
to stop operations and hold back the river flow for 48 hours August 23, 24, and 25,
188
INTERPRETATION AND ANALYSIS OF HYDROLOGIC DATA
-------
1961, thus, creating an artificial drought condition. This proved to be an extremely
successful operation, resulting in the accumulation of a considerable amount of useful
data in a short time, under favorable flow conditions.
Thurs
AUGUST 17
1961
Figure 2 Gage Chart for Tittabawassee River at Midland, August 17 through 25, 1961.
It should be reported that a certain amount of the Tittabawassee River water is
diverted around the Midland gage by the Dow Chemical Company, and although this
adds to the total river flow below Dow, it is of a steady nature and does not induce
further pulsations in flow. The diversion will be discussed later.
THE DAILY HYDROGRAPH
For many purposes a daily hydrograph of the daily average flow versus time is
useful. Generally, the daily average flow is the shortest period of flow regularly reported
in the Water Supply Papers1 of the Geological Survey and is, therefore, readily available
for all published stations. The daily hydrograph is useful in characterizing a river as
"flashy" or ''stable," i.e., rapid change in flow from day to day or, gradual change from
day to day. Also, it can be useful in relating water quality data to the flow conditions
that prevailed during the sampling period, including such things as high or low runoff,
rising or falling hydrograph, and stable or unstable flow conditions.
Figure 3 is the daily hydrograph for 1961 for the Clinton River at Auburn Heights,
Michigan. It might be characterized as a "flashy"' hydrograph resulting from several
drainage area characteristics, including its small size of 123 square miles. Routine
sampling days are indicated across the top of the hydrograph, allowing an immediate
visual comparison of runoff conditions during and preceding these sampling periods.
In contrast, Figure 4 is the daily hydrograph for 1951 for the Savannah River near
Clyo, Georgia, which might be characterized as a "stable1' hydrograph. Probably, one
of the most important factors contributing to this stability is the large drainage area of
9850 square miles. A period of intensive water quality sampling is indicated in August
1951 when the runoff during and preceding the sampling period was relatively stable.
Just as the daily hydrograph can be useful in relating previous runoff and water
quality conditions, it can also be helpful in planning stream surveys, especially if they
are to be the intensive type conducted over short periods of time under steady, low
runoff conditions. Plotting and studying daily hydrographs for a particular river loca-
tion for several years preceding a planned stream survey period tend to identify the
time of the year most likely to have a steady, low-flow condition. This then serves as
a guide for the assembly of the necessary sampling personnel and equipment, together
Gannon
189
-------
with the supporting laboratory facilities. As the planned survey period approaches,
the maintenance of a current daily hydrograph, together with knowledge of the weather
forecasts for the survey period, enables the investigator to know whether runoff conditions
are approaching an acceptable level, and also, whether there is a reasonable chance
of a dry period that will result in a steady flow condition. This approach has been useful
in planning several stream surveys, and generally, results in the accumulation of con-
siderable data under desirable runoff conditions.
300
260
JAN
FEB MAR APR
MAY JUNE JULY AUG SEPT OCT NOV
1961
DEC
Figure 3 Daily Hydrograph for Clinton River at Auburn Heights, Michigan, a 123-square-mile
Drainage Area.
JAN FEB MAR APR MAY
JUNE JULY
1951
AUG SEPT OCT NOV DEC
Figure 4 Daily Hydrograph for Savannah River near Clyo, Georgia, a 9850-square-mile
Drainage Area.
PROBABILITY PAPERS AND SEASONAL PATTERNS
Probability paper facilitates the application of statistical theory in summarizing data,
and several types including normal, logarithmic normal, linear extremal, and logarithmic
190
INTERPRETATION AND ANALYSIS OF HYDROLOGIC DATA
-------
extremal, have been used in summarizing hydro-logical and meteorological observations.
The early work of Hazen2 and the more recent work of Velz3 are to be particularly
noted. This section of the paper will deal with normal and log-normal probability
papers, and their application in the development of seasonal patterns of selected
hydrologic phenomena.
NORMAL PROBABILITY PAPER
Graphical methods describing the relationships expressed by the normal distribution
are available in terms of normal probability paper. Here is a quick and easy procedure
that makes available most of the advantages of the statistical method.
Briefly, normal probability paper is constructed by summing the area under the
normal probability curve from left to right, thereby obtaining an expression for the *
or horizontal axis of percent equal to or less than. Such a grid is illustrated in Figure 5,
STANDARD DEVIATION,
579
576
0.01 0.05 0.2 0.5 1 2
10 20 30 40 50 60 70 80 90 95
PERCENT EQUAL TO OR LESS THAN
98 99
99.99
lay 1860
Figure 5 Monthly Mean Elevation for the Lake-Michigan-Huron System for Mi
through 1957.
where it is noted that a clustering of percentage occurs around the 50 percent or center-
ing value, with a considerable spread toward the upper and lower end of the scale.
Further, a definite relation is observed between the standard deviation (Q-) scale
across the top and the percent equal to or less than scale across the bottom following
the normal distribution. The vertical, or y scale, is linear and is assigned the units of
measurement of the observations involved.
Gannon
191
-------
Data that follow the normal probability curve plot as a straight line on this grid;
thus, there is available a quick method of testing the normality of a series of observa-
tions. Furthermore, the slope of the line is a measure of variation, the steeper the slope
the more variation, the flatter the slope the less variation. There is a definite relation.
ship between the slope of the line and the standard deviation, making it possible to
determine standard deviation graphically.
A more complete discussion of normal probabilit, paper has been presented else-
where by Gannon* and Velz,3 including the mechanics of plotting on the grid, and
will not be repeated here.
Figure 5 is an illustration of the application of normal probability paper in defining
the variation of the monthly mean lake level for May for the Lake Michigan-Huron
system for the period of record 1860 through 1957. This illustration is taken from a
recent publication of Velz and Gannon.5
In Figure 5 it can be seen that the points describe a straight line, thus, indicating
that the data are normally distributed. Furthermore, from a statistical standpoint it is
possible to graphically determine the mean (X), which in this case has an elevation
of 580.8 feet. In addition to the mean, it is possible to define variation around the mean,
such as the 90 percent confidence range, i.e., 90 percent of the individual values fall
within this range around the mean, while 5 percent are less than the lower limit, and 5
percent are greater than the higher limit. The lower limit indicated at point (a) of
the distribution opposite the 5 percent equal to or less than line is seen to have a value of
578.4 feet, while the upper limit indicated at point (b) of the distribution opposite the
95 percent equal to or less than line is seen to have a value of 583.1 feet. Thus, normal
probability paper has been useful in defining the mean monthly lake level during May
for the period of record, together with the 90 percent confidence limit of these monthly
values.
SEASONAL PATTERN OF LAKE LEVELS
By an analysis similar to that indicated in Figure 5 for each month of the year,
it is possible to develop a seasonal pattern of lake levels such as that illustrated in
Figure 6 for the Lake Michigan-Huron system. Curve A is seen to be the most probable
monthly average lake level, while the 90 percent confidence range around this most
probable value is indicated by a dashed line. It is interesting that the low monthly
average level usually occurs in the winter months of January and February, whereas
the high monthly average level generally occurs in the summer months of July or August.
In addition to the 90 percent confidence range, the highest and lowest observed monthly
average level for the period of record is indicated for each month. Thus, there is
available in a single chart most of the important summary data for the lake levels of
the Michigan-Huron system.
LOGARITHMIC NORMAL PROBABILITY PAPER
Certain types of data do not plot as a straight line on normal probability paper;
however, in some cases, they straighten out on a logarithmic vertical scale. In some
instances, it is not possible to anticipate whether the data will follow a normal or a
logarithmic normal distribution, and the only practical solution is to try both. A typical
logarithmic normal probability grid is illustrated in Figure 7, where it can be seen that
the probability scale is the same as it would be on normal paper, whereas the vertical
scale is logarithmic instead of linear. The mechanics of plotting on logarithmic proba-
192 INTERPRETATION AND ANALYSIS OF HYDROLOGIC DATA
-------
bilty paper are the same as for normal probability paper, and generally the same type
of information is obtained.
Curve A most probable monthly average lake level
based on record for 1860 through 1957
range within which monthly average lake levels
can be expected for 90 percent of the years
x highest monthly average level tor period or record
o lowest monthly average level for period of record
E.QA
3OQ
583
582
OJ
O)
- 581
zf
o
<
§ 580
UJ
579
578
^77
1 '
X *
X ^ -
V, ^
o 0 o
1 1
I 1
X
X
X ^* "**
^
^"
^*l*
^^^cur\
^^
/
S
/
0 °
0
1 1
X v
. x
^*
>^
X
*«««a%1^^
IQ A "^S
^*v
o
0 0
1 1
X
>^ x
^. x
-
^*s^^
-
X
\
^^
-^
0
o o -
1 1
J FMAMJJASOND
MONTH
Figure 6 Seasonal Pattern of Levels for Lake Michigan-Huron System.
Gannon
193
-------
Figure 7 is an illustration of the application of logarithmic normal probability paper
in defining the variation of the monthly average now for May for the Kalamazoo River
at Comstock, Michigan, for the period of record, October 1935 to September 1960.
Generally, experience has indicated that monthly average runoff figures are best described
by a logarithmic normal distribution, but there is no fundamental explanation why.
0.5 1 2 5 10 20 30 40 50 60 70 80 90 95 98 99 99.5 99.9 99.99
PERCENT EQUAL TO OR LESS THAN
Figure 7 May Monthly Average Flow of Kalamazoo River at Comstock.
In Figure 7 it can be seen that the points describe a straight line, thus, indicating
that the data are logarithmically normally distributed.^ As with normal probability paper,
it is possible to determine graphically the mean (X), which in this case has a value
of 500 cfs, whereas the 90 percent confidence range around the mean is indicated at
points (a) and (b), which have values of 280 and 900 cfs, respectively.
SEASONAL PATTERN OF RUNOFF
Just as it is possible to develop a seasonal pattern of lake levels, so is it possible
to develop a seasonal pattern of runoff. Figure 8 is an illustration of such a chart for
the Kalamazoo River at Comstock, developed from an analysis of the variation in monthly
average flow for each month of the year on logarithmic normal probability paper, as
illustrated in Figure 7. Such a figure is in effect a type of monthly hydrograph.
It is seen that Curve B in Figure 8 is the most probable monthly average flow
and Curve A the mean for the period of record. In addition, the dashed lines C and
D indicate the 90 percent confidence limits of individual monthly values around the most
probable. For the Kalamazoo River, the high runoff period occurs in the spring months
of March and April, whereas the low runoff period occurs in the late summer months of
August and September.
In contrast to the Kalamazoo River in Michigan, Figure 9 illustrates the seasonal
pattern of runoff for the Platte River at Sinclair, Wyoming, for the period 1940 through
1961. The high flow generally occurs in June whereas the low flow occurs in September
and again in January. Also, the variation from month to month is greater than for the
Kalamazoo River.
194
INTERPRETATION AND ANALYSIS OF HYDROLOGIC DATA
-------
The graph showing seasonal pattern of runoff is helpful in depicting the most prob-
able flow available each month of the year, together with its variation, rather than the
lowest flows only. If one subscribes to the concept of using the total river flow for waste
assimilation purposes, either by means of storing the high river flow in reservoirs and
releasing it during the low flow periods or of storing the waste by means of storage
lagoons and releasing this waste in accordance with river flow, then flow information
of the type presented in the seasonal pattern of runoff is essential.
DROUGHT FLOW ANALYSIS
To meet the need for knowledge concerning the probability of occurrence of drought
flows, particularly as they relate to water quality considerations, special graphical pro-
CJJ
cc
I
o
CtL
UJ
X
I
z
o
A mean for period of record
B most probable monthly average
C to D 90 percent confidence range
100
Figure 8 Seasonal Pattern of Runoff for Kalamazoo River at Comstock, Monthly Average
Discharge, October 1935 through September 1960.
Cannon
195
-------
cedures have been developed and adapted employing the theory of extreme values as
proposed by Gumbel.6-8 These procedures have been successfully used in a state-wide
analysis of the drought flows of the streams of the State of Michigan to compile a
comprehensive report on the subject by Velz and Gannon.5 This section of the paper will
discuss the analysis of drought flows on logarithmic extremal probability paper, including
certain adjusting and summary procedures together with a consideration of natural and
artificial influences. Where it is necessary to establish river water quality standards,
these standards should be related to drought flow levels, and information on the prob-
ability of these flow levels is essential.
10000
8
6
5
4
1000
Q
Ld
DC 6
LU
£ 5
i 4
I 3
2-f
100 H
6
4-1
I-I
A mean for period
B most probable monthly average
C to D 80 percent confidence range
^s^^-^-^
MONTH
Figure 9 Seasonal Pattern of Runoff for Platte River at Sinclair, Wyoming,
Monthly Average Discharge.
196
INTERPRETATION AND ANALYSIS OF HYDROLOGIC DATA
-------
LOGARITHMIC EXTREMAL PROBABILITY PAPER
The extremes of hydrologic observations such as floods and droughts do not follow
a normal symmetrical distribution but rather are skewed (the more severe values deviate
beyond the mean to a much greater extent than the less severe values deviate below it).
Gumbel0-8 has proposed three asymptotic probabilities of extremes suggesting that the
third asymptotic distribution is suitable for analyzing droughts. In the Michigan drought
study, logarithmic extremal probability paper was used and the third asymptotic distri-
bution of smallest values followed as suggested by Gumbel.
Such a grid is illustrated in Figure 10, which was developed in a manner similar
to those for normal and logarithmic normal probability papers previously discussed.
The probability equal to or less than scale is unbalanced to the left, with the more
severe values to the right having the greatest spread in accordance with the skewed
nature of the distribution. Also, an additional scale has been added across the top,
called the return period (T), which is related to the probability scale across the bottom;
this scale is particularly useful in dealing, with hydrologic data. For example, if the
base unit of time from which low flow data are selected is a year, then the return period
of 10 would indicate a l-in-10-year drought.
RETURN PERIOD. T
345 10 20 30 4050
1946
\
00010010 01000.200 0.400 0600 0.800 0.900' 0.9500.9700.980 0.990 0.995
PROBABILITY EQUAL TO OR LESS THAN
0.998 0.999
Figure 10 Minimum 30-day Flow during May through October in Grand River at Jackson,
Michigan (Gumbel's Logarithmic Extremal Probability Paper).
The vertical, or y scale, is logarithmic and is assigned the units of measurement
involved such as cubic feet per second (cfs). For purposes of use, data are arranged
and plotted in order of severity, which in the case of low flows means ordering from
the higher to the lower absolute values. In addition, it is necessary to calculate a
plotting position for the probability scale using Gumbel's refinement as illustrated by
Velz,3 or Velz and Gannon.9
Gannon
197
-------
In Figure 10 it is seen that the minimum consecutive 30-day flows for the Grand
River at Jackson, Michigan, do approximate a straight line, with the straight line fitted
to the data by eye rather than by use of a more rigid mathematical method. Since a
good straight line fit results, there is an indication that these data do follow the third
asymptotic distribution of smallest values. From the fitted line and the return period
scale, the most probable minimum consecutive 30-day drought is read as 44 cfs (more
properly called the characteristic drought indicated as a dashed vertical line), the 1-
in-5-year drought as 26 cfs, the l-in-10-year drought as 20.5 cfs, and the l-in-20-year
drought as 16 cfs.
Not all drought flow data necessarily plot as a straight line on logarithmic extremal
probability paper, especially if storage, either artificial or natural, is involved or if
flow augmentation is involved. An adjusting procedure to handle these cases will be dis-
cussed subsequently in this paper.
BASIC INFORMATION
Three time elements are involved in the definition of drought flow: (1) the base
unit of time from which a low flow is selected from the record, (2) the length of time
over which a low flow is averaged, and (3) the season in which the selection is made.
Ideally, from a statistical standpoint, extreme values selected from consecutive
time units should be independent of each other. With low flows, there is a possibility of
a carry-over influence from one year to the next, but notwithstanding this possible in-
fluence, the base unit of time of the year was used in the Michigan study, primarily
because of the relatively short records available in this State.
The second time element, the length of time over which a flow is averaged, may
vary depending on the particular application intended for the information. To meet
as many time needs as possible, the Michigan study has reported and analyzed five flow
periods: the minimum day; the minimum consecutive 7-day, 15-day, and 30-day averages;
and the minimum calendar monthly average.
The third time period, the season, is important in differentiating between warm-
weather and cold-weather droughts. Generally speaking, in Michigan, cold-weather
droughts are different from warm-weather droughts. Because main interest in this study
was in connection with warm-weather applications such as water pollution control,
irrigation, and recreation uses, low flows were selected from the summer-fall period May
1 through October 31. The base unit of time of the year was retained, but low-flow
selections were made only from this summer period.
DROUGHT DURATION VERSUS SEVERITY
As a summary device and as an interpolating aid, a chart similar to Figure 11
has been prepared for each gage to show the relationship between drought duration and
severity. Information for the construction of this chart was obtained from four separate
logarithmic extremal probability plots similar to Figure 10, covering the minimum daily
flows and the minimum consecutive 7-day, 15-day, and 30-day averages. From each
plot, the most probable, the l-in-5-year, the l-in-10-year, and the l-in-20-year figures
were obtained; these served as a basis for the development of the most probable, the
l-in-5-year, l-in-10-year, and l-in-20-year curves in Figure 11. These curves then serve
as a framework from which a drought of any duration from 1 to 30 consecutive days
can be determined for the indicated return periods. Because of the influence of regula-
198 INTERPRETATION AND ANALYSIS OF HYDROLOGIC DATA
-------
tion, the minimum daily flows are in many instances out of line with the rest of the
data. To caution the user of this fact and to urge care in the use of data in this short
duration range for interpolation purposes, the 1-day and 7-day duration points have
been connected by a dashed line.
2.0
70
60
50
o 40
u.
O
30
20
10
1 IN 20 YEARS
-a''
I
I
-- 1.5
o
0)
w
E
1.0
- 0.5
05 10 15 20 25 30
NUMBER OF DAYS OVER WHICH DROUGHT IS AVERAGED
Figure 11 Chart of Drought Duration Versus Severity for Grand River at Jackson, Michigan.
ADJUSTING PROCEDURES
Not all of the logarithmic extremal probability plots developed as straight lines,
especially where regulation was involved, either artificial or natural. In several instances,
because of the presence of a base flow below which the river flow had not fallen, a
curve developed when the original data were plotted on probability paper. Such a case
is illustrated by the solid points in Figure 12, for the minimum daily flows for the
Kalamazoo River near Battle Creek. Gumbel6 has proposed an elaborate computational
procedure, involving the third moment of the distribution, for the evaluation of the lower
limit and for fitting a curve through the data.
A much simpler technique involves estimation of the base flow by eye, subtracting this
figure from each flow, and replotting the remainders as illustrated in Figure 12. If a
straight line does not develop, a second and a third estimate of the base flow is made
until a straight line results. Thus, in a relatively few trials it is possible to estimate the
base flow, and also, to fit a straight line to the remainder. From this line, it is possible
to determine the most probable, the l-in-5-year, l-in-10-year, and l-in-20-year flows to
which must be added the previously subtracted base flows to bring the flow figures back
to their original levels.
For the illustration of Figure 12, it is seen that the base flow was estimated as
135 cfs and that when this flow was subtracted from the original flows the remainders
formed a reasonably straight line.
Gannon
199
-------
RETURN PERIOD, T
1.0011.011.1 1.5 2 345 10 20304050 100200 500 1000
,000
100
n
J
i
5
10
9
7
6
5
4
3
2
1
« j
o-<
\K
aJ 4.
\> *
^
V
s
J ^
» ur
o AC
.1940 Vqco
«r«-l953.
1939
) 1
i
\
^
\
°\
c
\
JAC
JU
JUSTE
STED (
,1941
D (c
cfs
:ts)
13
5) *
\
\
\
v
\
,25
:20
-15
:10
-8.0
-6.0
i-4.0
-1.5
rl.O
-0.8
rO.6
i-0.4
.
-0.1
1-0.08
0.06
0.04
-CO-OK) .IOO .ZOO .300 .700.800 .900 .930.970.3*0 .990 ,990 .996 .999
PROBABILITY EQUAL TO OR LESS THAN
Figure 12 Minimum Daily Flow during May through October in Kalamazoo River near
Battle Creek (Gumbel's Logarithmic Extremal Probability Paper).
DROUGHT FLOW INDICES
Two important drought flow indices have been developed for all of the gages studied
in Michigan, namely, the yield and the variability ratio. These summary figures allow
comparison of gages within a basin, and in addition, allow comparison of the flow
characteristics of one basin with another.
The yield, defined as the discharge per unit drainage area (cfs/miz), is useful in
reducing discharge figures at gages with varying drainage area sizes to a common base.
Considerable variation was observed in the yield characteristics of the several basins.
For example, the Manistee River near Sherman, with a drainage area of 900 square
miles, shows a l-in-10-year drought as a 7-day average of very high yrfeld, about 0.8 cfs
per square mile. In contrast, the Raisin River at Monroe, with a drainage area of
1034 square miles, shows a very poor yield, about 0.03 cfs per square mile.
The secondary summary index, the variability ratio, defined as the ratio of the
l-in-10-year drought to the most probable drought, is helpful in defining the variation
that can be expected in drought flows from year to year. Because of the nature of
logarithmic extremal probability paper, and also, because of the adjusting procedures
used in some cases, the conventional measures of variation, such as the standard devia-
tion, are not applicable, and it became necessary to develop a new measure. To meet
INTERPRETATION AND ANALYSIS OF HYDROLOGIC DATA
-------
this need, Velz and Gannon5 proposed the variability ratio, which is easy to determine
and which serves as a basis for comparison among gages within a basin, and also,
between basins.
The usefulness of the variability ratio is illustrated in tbe comparison of the
Manistee River and the Raisin River. The variability ratio for the Manistee River is
about 0.9, which is to say that the l-in-10-year drought flow is 90 percent of the most
probable, indicating an unusually stable stream. In contrast, the Raisin River record
develops a variability ratio of about 0.3, which is to say that the l-in-10-year drought is
only 30 per cent of that normally expected, indicating a river of high variability from
year to year and subject to occasional drought flows of considerable severity.
BASIN SUMMARY
In many of the basins of the State where three or more representative gages existed,
it has been possible to establish a linear relationship between the logarithm of the
drainage area size and the logarithm of the minimum consecutive 30-day average most
probable and l-in-10-year droughts. Such a relationship for the Kalamazoo River is
illustrated in Figure 13 on a log-log scale, with Curve A the most probable drought,
and Curve B, the l-in-10-year drought.
DRAINAGE AREA, Km*
3 4 56769
DRAINAGE AREA, mi*
Figure 13 Summer-Fall Drought Flow as Minimum in Kalamazoo Basin, Consecutive
30-Day Average versus Tributary Drainage Area.
In addition to serving as a summary for the key gages in the basin, this chart is
useful in estimating drought flows along the river at points that do not have a stream
gage, but where the tributary drainage area is known. For example, on the Kalamazoo
River at a point having a drainage area of 700 square miles, the most probable 30-day
average drought would be estimated from Curve A as 280 cfs and l-in-10-year average
drought from Curve B as 150 cfs.
Gannon
201
-------
NATURAL AND ARTIFICIAL INFLUENCES
In dealing with drought flows, the investigator must be continually alert to the
possibility of either natural or artificial influences. Natural influences might be re-
flected in a drainage basin with widely varying yield characteristics, whereas artificial in-
fluences might include many of man's activities such as hydroelectric and steam power
production, diversion for irrigation or municipal or industrial use, or possibly naviga-
tion or even flood protection facilities.
A good example of a river with widely varying yield characteristics is the Willamette
River in Oregon, especially in the section of the river from Salem to Portland. Tribu-
taries on the eastern side fed by the melting snows of the Cascades produce high yields,
whereas those from the Coastal Range on the western side produce low yields. Figure
14 illustrates an attempt to estimate the once-in-5-year minimum weekly average flow
at Portland by considering the yield at Salem together with the yields from the individual
tributaries. It will be noted that the Yamhill has a yield of 0.052 cfs per square mile,
whereas the Clackamas draining the Mt. Hood area has a yield of 0.763 cfs per square
mile. The figures used in Figure 14 represent flow conditions prevailing in the Willamette
MT. HOOD
/
CLACKAMAS
Drainage area, 930 mi2
Runoff, 0.763 cfs /mi2
OREGON
MOLALLA-PU DOING
Drainage area. 890 mi2
Runoff, 0.135 cfs/mi2
Drainage area, 7280 mi2
Runoff, 0.391 cfs/mi2
SALEM
TUALATIN
Drainage area, 710 mi2
Runoff, 0.070 cfs/mi2
2850 cfs
YAMHILL
4 fy Drainage area, 770 mi2""
« £
Runoff, 0.052 cfs/mi2
Figure 14 Stream Flow Available Along Willamette River at Once-in-5-Year Minimum
Weekly Average Drought Severity.
202
INTERPRETATION AND ANALYSIS OF HYDROLOGIC DATA
-------
prior to 1950 and are not illustrative of present low flows in the main river, which are
influenced by low-flow augmentation resulting from upstream storage. Notwithstanding
these changes, the illustration does indicate the dramatic differences in yield from the
tributaries on the eastern and western side of this section of the river.
Under the category of an artificial influence might be considered the diversion of
river water around the Geological Survey stream gage on the Tittabawassee River at
Midland by the Dow Chemical Company for industrial use. This diverted flow is re-
turned to the river below the gage, together with a small amount of imported Lake
Huron water, resulting in an augmentation of the natural river drought flows. The
influence of the diversion was illustrated during a special time-of-passage study conducted
by the writer and his associates under controlled river flow conditions on August 15,
1962. Table 1 tabulates the Geological Survey stream gage flow, together with the flows
not reflected by this gage but returned to the river downstream from the gage.
Table 1 River Flows in Tittabawassee River Below Midland, Michigan, August 15, 1962.
Geological Survey stream gage 220 cfs
Dow treatment plant effluent 76.2 cfs
Drain A 13.4 cfs
Drain B 5.3 cfs
Total river flow below Dow 314.9 cfs
It is interesting that a discharge measurement of the Tittabawassee River taken
independently by the writer at the first convenient sampling station downstream from the
diversion amounted to 314.7 cfs, indicating excellent agreement with the sum of the
individual upstream measurements. If reliance were placed only on the official Geological
Survey flow measurement as an indication of downstream flow in the Tittabawassee
River during this survey period, the estimates would be in serious error.
FLOW REGULATION
In many river basins, low-flow regulation can be accomplished by storing high
flows and releasing them during the dry-weather period of the year, thereby eliminating
the most severe drought conditions. As is generally known, one of the most important
elements governing the waste assimilation capacity of a stream is the flow level; the higher
the flow, the greater the capacity; the lower the flow, the less the capacity. Thus, flow
regulation eliminates the need for controlling waste discharges to meet water quality
needs under the most severe drought conditions, and in many cases, makes it possible
to work with guaranteed minimum flows substantially greater than the natural dry-
weather flows. Where storage is in the headwater of a stream, the benefits accrue not
only to the section of the river involving waste assimilation, but to all other sections of
the river below the impoundment, including water for municipal and industrial water
supply, power production, and recreation.
One of the important considerations, of course, is the availability of a suitable site
or sites for reservoir development. A systematic study of the headwater and downstream
tributaries may yield several locations that could be used as reservoir sites, and there-
fore, would merit further analysis. Several years ago such a study was conducted in
the Kalamazoo River basin by Velz and Gannon,10 resulting in the location of a favorable
reservoir site on one of the upstream tributaries.
Gannon 203
-------
Figure 15 shows a comparison of the potential regulated flow and the natural 10-year
drought flow (weekly and daily averages) for the Kalamazoo River at Kalamazoo, at
Battle Creek, and at Marshall for the key dry-weather months of July through October.
It is apparent that at Kalamazoo a regulated flow of approximately 700 cfs could be
maintained in comparison to the natural drought flows in the vicinity of 200 cfs or less.
In addition, it is seen that not only does the river in the vicinity of Kalamazoo benefit,
but also, there is a substantial increase over natural drought flows at Battle Creek and
at Marshall.
700
600
500
400
300
200
100
n
_
-
-
-
-
Atfc
f?
* V
1 .'V,
REGULATED
NATURAL
(Weekly average)
L NATURAL
r (Daily
average)
, '-
'- ^ ~\^
:~* '.
.-, ''»
tx\S
'""I"1-
;
i,1 ,,,
-
fl
,
\
-
~
-
nn "
AT KALAMAZOO
AT BATTLE CREEK
AT MARSHALL
Figure 15 Comparison of Potential Regulated Flow and Natural 10-Year Drought Flow
in Kalamazoo River.
A major pollution problem exists on the Kalamazoo River below the City of
Kalamazoo, and one of the main benefits of increased low flows would be the improvement
of water quality. Unfortunately, an economic study made subsequently by another group
indicated that it would be less costly to improve water quality through this section by
additional waste treatment rather than by means of flow augmentation. As a result, the
proposal for low-flow augmentation by reservoir development was not considered further
and major reliance for water quality improvement is being placed on additional waste
treatment. It may be that as the demand for water increases in the future the economic
balance will change and this proposal will receive further consideration.
Standard procedures for the determination of storage needs by mass curve analysis,
etc., is covered adequately in such text books as that of Fair and Geyer,11 and will not
be considered here.
TIME OF PASSAGE
One of the important elements necessary for an accurate evaluation of the self-
purification capacity of a river where one or more different types of wastes, such as
organic, bacteriological, or chemical contaminants, may exist is the time of flow or
passage along the stream. This information may be obtained or estimated in several
ways. From a knowledge of the channel characteristics and prevailing runoff and use
of an internal or external tracer such as a dye, it can be calculated on a displacement
basis, or from a knowledge of certain generalized data, it can be estimated as proposed
by O'Connor.12
204
INTERPRETATION AND ANALYSIS OF HYDROLOGIC DATA
-------
DISPLACEMENT CALCULATIONS
Where the river channel is of a fairly uniform character, time of passage can be
calculated for a given runoff on a displacement 'basis. This presumes that information
is available on the river channel characteristics so that accurate volumes can be calcu-
lated for a given runoff level. Sometimes this information is available from sources such
as the files of Corps of Engineers units covering flood protection or making navigation
studies. It may be necessary to collect this information in the field; if this is the case,
adequate definition of channel characteristics can generally be obtained by cross-sectioning
the relevant river stretches at about 500-foot intervals. This need not be done with a
high degree of accuracy, but rather emphasis should be placed on more frequent
soundings wherever possible. It may be accomplished by means of a tape and sounding
rod or weighted line, together with a good map for location and orientation in the field,
or if considerable cross-section work is anticipated, it might be desirable to obtain a
portable recording fathometer that gives a continuous record of channel depth.
Volumes can be calculated on an average end-area basis with adjustment to various
runoff levels made by means of an appropriate rating curve. This approach can be
programmed for high-speed digital computers and incorporated as a part of a more
extensive program such as that described by Gannon and Downs13 for programming
river dissolved oxygen calculations.
Figure 16 contains a series of time of passage curves calculated on the displacement
basis for various runoff levels for the Willamette River, for the section of the river ex-
tending from Salem to Oregon City Falls. Fortunately, in this case, detailed charts with
frequent depth figures were available from the Corps of Engineers14 and these data
served as the basis for volume calculations.
0.2 0.4 0.6 0.8 velocity, fps
potential sludge
deposit areas
70 60 50
MILES ABOVE THE MOUTH
Figure 16 Time of Passage Curves Calculated on Displacement Basis, Willamette River.
Gannon
205
-------
In Figure 16 the slope of the time curve is in effect a measure of the average
velocity in that section of the river. This type of plot, therefore, serves as an excellent
guide in identifying those sections of the river that would serve as potential sludge
deposit areas. Velz15, 16 has indicated that at velocities of 0.6 fps or less settleable solids
deposit and tend to accumulate to an equilibrium level. Thus, the nest of velocity curves
in Figure 16 show that a potential sludge deposit area exists in the pool section of the
river extending from approximately mile point 52 to 26, and also, in short sections in
the stretch from mile point 87 to 52. If, therefore, any wastes containing settleable solids
were discharged into the river above these potential deposit areas, it is almost certain
that sludge deposits would develop, resulting in oxygen deficient conditions.
Knowledge of the channel characteristics is necessary for other purposes, such as
the calculation of reaeration for oxygen balance needs where an organic waste problem
exists. Here, it is important to know both mean depths and volumes for the critical
section of the river.
INTERNAL AND EXTERNAL TRACERS
In addition to the displacement approach, time of passage can also be determined
by means of either internal or external tracers. An internal tracer may be classified
as some waste constituent that can be varied in concentration and easily measured in
the stream, e.g., chlorides. An external tracer may be classified as anything that can be
added to the stream and then easily followed and measured, e.g., salt, dye, and radio-
active material. Several investigators have reported on tracer methodology, including
Carpenter17 on Chesapeake Bay, Selleck and Pearson18 on San Francisco Bay, and
Hull19 on the American River in California.
As part of recent investigations on the Tittabawassee River, an opportunity de-
veloped for making comparisons of times of passage (1) calculated on the basis of dis-
placement, (2) use of waste chloride concentration as an internal tracer, and (3) use
of Rhodamine B fluorescent dye as an external tracer, together with a sensitive
fluorometer for detection purposes.
TIME
Figure 17 Typical Fluorometer Tracing, Tittabawassee River at Freeland, Michigan,
August 15, 1962.
Figure 17 is a typical fluorometer tracing of Rhodamine B dye detected at Freeland,
on the Tittabawassee River on August 15, 1962, during a controlled river flow condition
for special time of passage studies. The dye was introduced as a point discharge at the
next upstream station; as a result of longitudinal mixing or dispersion, it took the dye
approximately 3.5 hours to pass the station at Freeland. For a more complete discussion
206
INTERPRETATION AND ANALYSIS OF HYDROLOGIC DATA
-------
of the mixing and diffusion of wastes in streams, the reader is referred to the work
of Thomas.20
Figure 18 is a comparison of calculated and observed times of passage on the
Tittabawassee River below Midland for a total river flow in the range of 315 to 350 cfs.
The calculated times were determined on the displacement basis; however, the channel
was cross-sectioned under higher runoff levels and it was necessary to adjust the volume
down to the indicated runoff levels. The observed flow time was reported by the Dow
Chemical Company on the basis of chloride concentration studies conducted many years
ago. The dye tracer studies were directed by the writer under controlled river conditions
on August 15, 16, 1962.
2.0
calculated: (346 cfs)
observed ( 1st detection: 1 _,_ ,
dye I mean of period: 4 ) 315 cfs
observed chloride flow: « 350 cfs
16 14 12
MILES ABOVE THE MOUTH
Figure 18 Comparison of Calculated and Observed Times of Passage, Tittabawassee River.
The results of the Rhodamine B study are plotted in two ways: first, as the time of
first detection, which might be important if toxic wastes were involved, and second, as
the mean of the period, which should be compared with the displacement calculation.
Reasonable agreement exists to Freeland, but differences are greater farther downstream.
The dye studies were conducted at a runoff level of 315 cfs; the calculated time corre-
sponding to a runoff of 346 may partially account for the differences. Furthermore, a
backwater influence at station 11 from Lake Huron no doubt contributes to the differences
at this station. The river channel is fairly uniform and shallow through this section,
suggesting minimum amounts of short circuiting.
The availability of dyes such as Rhodamine B and Pontacyl Brilliant Pink B,
together with sensitive fluorometer detection instruments, makes an external tracer study
of time of passage a necessary part of any well-planned stream survey.
ACKNOWLEDGMENTS
The assistance of Mr. Jackson R. Pelton in the time of passage studies, and
Mrs. Josephine Toney in the statistical compilations and computations is gratefully
acknowledged.
Cannon
207
-------
Financial support was provided by the Water Resources Commission of the State
of Michigan for the analysis of drought flows in Michigan, while the U. S. Public
Health Service supported studies on the Clinton and Tittabawassee Rivers as research
grant RG-6905 later redesignated WP-187.
The cooperation of Mr. Arlington Ash, District Engineer, U. S. Geological Survey,
in supplying runoff information, and also, the cooperation of Messrs. John Robertson
and Charles Sercu of the Waste Control and Utilization Department, Dow Chemical
Company, Midland, Michigan, in supplying information and facilitating several phases
of the Tittabawassee study is gratefully acknowledged.
Finally, particular recognition is due to Professor C. J. Velz, Chairman of the
Environmental Health Department, The University of Michigan, who has encouraged the
use of the statistical tool in the analysis of hydrologic data, particularly as it relates to
water quality considerations.
REFERENCES
1. "Water Supply Papers." Published annually by U. S. Geological Survey.
2. Hazen, Allen, "Storage to be Provided in Impounding Reservoirs for Municipal
Water Supply." Transactions of the American Society of Civil Engineers, 77, 1539
(1914).
3. Velz, C. J., "Graphical Approach to Statistics." Water and Sewage Works, 99, 4,
R106 (1952).
4. Gannon, John J., "Statistical Basis for Interpretation of Data." Proceedings of
Michigan Sewage and Industrial Wastes Association 1959 Annual Meeting, 34 pp.
(1959).
5. Velz, C. J. and Gannon, John J., "Drought Flow of Michigan Streams." Michigan
Water Resources Commission, Lansing, 771 pp. (1960).
6. Gumbel, E. J., "Statistical Theory of Droughts." Proceedings American Society of
Civil Engineers, 80, separate No. 439 (May, 1954).
7. Gumbel, E. J., "Statistical Theory of Floods and Droughts." Journal of the Institution
of Water Engineers (British), 12, 3, 157 (May, 1958).
8. Gumbel, E. J., "Statistics of Extremes.'' Columbia University Press, New York
(1958).
9. Velz, C. J. and Gannon, John J., "Low Flow Characteristics of Streams." Proceedings
of the Second Annual Ohio Water Clinic, Ohio State University Studies Engineering
Series, 22, 4, 138 (1953).
10. Velz, C. J. and Gannon, John J., "Reservoir Site Study in the Kalamazoo Basin."
Unpublished material.
11. Fair, Gordon M. and Geyer, John C., "Water Supply and Waste-Water Disposal."
John Wiley and Sons, Inc., New York (1954).
12. O'Connor, Donald J., "The Effect of Stream Flow on Waste Assimilation Capacity."
Paper presented at the 17th Purdue Industrial Waste Conference (May, 1962).
13. Gannon, John J. and Downs, Thomas D., "Programming River D. 0. Calculations."
Water and Sewage Works, Part I, 110, 3, 114 (March, 1963) ; Part II, 110, 4, 157
(April, 1963).
208 INTERPRETATION AND ANALYSIS OF HYDROLOGIC DATA
-------
14. U. S. Corps of Engineers, Willamette River, Oregon. Portland, Oregon Office
(Revised to November, 1938).
15. Velz, C. J., "Factors Influencing Self-Purification and Their Relation to Pollution
Abatement Part II Sludge Deposits and Drought Probabilities."' Sewage and
Industrial Wastes, 21, 2, 309 (1949).
16. Velz, C. J., "Significance of Organic Sludge Deposits." Oxygen Relationships in
Streams, Technical Report W-58-2, Robert A. Taft Sanitary Engineering Center
(1958).
17. Carpenter, James H., "Tracer for Circulation and Mixing in Natural Waters.''
Public Works, p. 110 (June, 1960).
18. Selleck, Robert E. and Pearson, Erman A., "Tracer Studies and Pollutional Analyses
of Estuaries." Publication No. 23, State of California Water Pollution Control
Board, Sacramento (1961).
19. Hull, D. E., "Dispersion and Persistence of Tracer in River Flow Measurements."
International Journal of Applied Radiation and Isotopes, Vol. 12, p. 63 (1961).
20. Thomas, Jr., Harold A., "Mixing and Diffusion of Wastes in Streams.'' Oxygen
Relationships in Streams, Technical Report W-58-2, Robert A. Taft Sanitary
Engineering Center (1958).
DISCUSSION
Mr. Gray asked whether observations of dye concentration had been carried on for
any period of time after the low point apparently was reached. He indicated that a
second peak had been observed in a test in which grab sampling had necessitated an
extended sampling period to assure that the dye had passed.
Mr. Gannon observed that this is probably the result of pools in the stream that
hold some of the dye and then feed it back to the river. The dye could not be followed
for any great length of time in the Titabawassee River, since the flow was being reduced
by storage in a reservoir of limited capacity. The investigators observe the time of first
detection and the time of maximum concentration, and estimate the remainder of the
curve. He believes that dye provides a relatively simple, inexpensive means of securing
fairly accurate estimates of time of passage.
Mr. O'Connell noted that the time of first appearance is somewhat short of the cal-
culated displacement time, while the mean time is a little longer than the calculated
time. He asked whether the time of passage as indicated by the peak concentration had
been considered, since it might be closer to the displacement time. The mean might give
a distorted measure of time of passage because of the diffusion that takes place while
the dye passes the measuring station. A synoptic observation would prevent distortion
from this source.
Mr. Gannon agreed that the peak is sometimes used rather than the mean to deter-
mine time of passage. He indicated that diffusion prevents the following of a slug of dye
very far downstream. Observing dye passage from one station to the next is the most prac-
tical method, and problems of background concentrations are avoided by starting with
the downstream station and working up river with a new slug of dye each time.
Gannon 209
-------
SESSION 6: General
Chairman: Bernard B. Berger
Assistant Chief for Research
Division of Water Supply and Pollution Control
U. S. Public Health Service
-------
Dr. John C. Bellamy
Director, Natural Resources Research Institute
University of Wyoming, Laramie
SUMMARY
The purpose of informatic data research is to find better ways of using new equip-
ment that make it possible to acquire and analyze vast amounts of quantitative data.
The kinds of data of primary interest are those that will better inform us about the nature
of the ground, water, air, and near-space environment of our geosphere. New forms of
numerals can advantageously be utilized to paint half-tone pictures that will not only
provide the qualitative information needed to gain understanding of our environment, but
can also serve as a concise and complete data store for whatever arithmetic processing
anyone might subsequently wish to have performed by machine. This is one of a series
of papers describing the progress of Informatic Data Research at the University of
Wyoming.
DATA DISPLAY FOR ANALYSIS
This is one of a series of papers describing the progress of a program of Informatic
Data Research at the University of Wyoming. The purpose of this program is to establish
the principles and practices of utilizing newly possible
m/ormatic ways of representing large sequences of numbers as concise complete
"pictures" or portrayals of
information which can be acquired, processed, recorded, and reprocessed in numerical
detail only with appropriate
automatic equipment if scientific and engineering operations are to become more
economically effective.
In brief review of a previous discussion,1 the goal of informatic data research is to
find ways of utilizing newly possible equipment for better acquiring, analyzing, and
utilizing vast amounts of quantitative data. The kinds of data of primary interest are
those that will better inform us about the nature of our ground, water, air, and near-
spacs environment, or in short, about the nature of the lithosphere, hydrosphere,
atmosphere, and pyrosphere that make up the geosphere.2
Progress to date has shown that new forms of numerals can advantageously be
utilized now that man no longer has to write them. Briefly, these new numerals can be
likened to the variably sized dots that make up half-tone reproductions of photographs or
paintings. Consequently, the goal of informatic data research can be thought of as
being to find ways of utilizing machines "to paint half-tone pictures with numerical data."
Even partial realization of this goal would evidently be very worthwhile. Not only
would the resultant "pictures" provide men with the largely qualitative kind of informa-
tion they need to gain understanding of the nature of their environment, these "pictures"
would also serve as a concise and complete data store for whatever arithmetic processing
anyone might subsequently wish a machine to perform. The sizes of the individual
"half-tone numerals'7 would need only to be sensed by appropriate reading equipment to
re-establish whatever numerical values might be needed for any desired quantitative
analysis.
Bellamy 213
-------
INCREMENTAL DATA
To illustrate, the data-block on the left in Figure 1 is an "incremental" portrayal of
the vertical distributions of values of temperature measured throughout a month above
a particular radiosonde station.3
INCREMENTAL TIME CROSS SECTION
T vs Zp
T-3, Fletcher's Ice Island
1-31 JULY 1952, 0300Z & 1500Z
PENTIADIC TIME CROSS SECTION
T vs Zp
T-3, Fletcher's Ice Island
1-31 July 1952, 0300Z & 1500Z
Units of Resolution
Zp = 100ft
T = 0.30°C
+1 Unit of T per Unit of Zp
0 No Change
. 1 Unit of T per Unit of Zp
^ Missing Sounding Preceding Repeated
Units of Resolution
Pressure Altitude, Zp
Biadic: 100, 1000, and 10,000 ft
Temperature, T:
Penfiadic: 5°C
Time:
Biadic: 12 hr and 5 days
Figure 1 Incremental and ladic Notations.
214
DATA DISPLAY FOR ANALYSIS
-------
To understand the pictorial character of this example, it can be thought of as
though it had been produced as a half-tone reproduction of a photograph of a plaster-of-
paris model of that measured temperature distribution. The height of each point of the
model would have been proportional to the value of temperature measured at the corre-
sponding values of altitude (or pressure) and time, and it would have been photo-
graphed with a point source of illumination above and to the high-altitude side of the
model. The light regions would then have occurred where the illumination was normally
incident upon the surface of the model, or where the temperature increased rapidly
downward; gray regions would have occurred where the illumination was obliquely
incident, or where the temperature was nearly constant vertically; and dark regions would
have occurred in regions of grazing illumination, or where the temperature increased
rapidly upwards. Or, in meteorological terms, the troposphere is light, the stratosphere
is gray, the low-level and tropopause inversions are black, and the various shades of gray
indicate various degrees of lapse rate.
Actually, of course, this incremental data block was produced without going to the
trouble of constructing and photographing a plaster-of-paris model. Rather, it was formed
in accordance with a particularly simple arithmetic formula based upon the character-
istics of continuous data. That is, in order to explicitly represent the value of temperature
measured at each and every altitude, it is necessary that no significant changes of
temperature be omitted and hence that the data be continuous in the sense that no two
successive values differ by more than an appropriately selected and significant unit of
numerical resolution. But then, since the numerical differences, or increments, between
successive values can only be +1, 0, or 1, it is only necessary to record one of these
three possible values of increments between each successive value of continuous data to
designate all but its initial values.
Specifically, units of resolution of 1/3°C, TOO feet, and 12 hours are used for
temperature, (pressure) altitude, and time in this example. Incremental numerals con-
sisting of short, medium, and long horizontal dashes have been used (instead of the
Arabic numerals 1, 0, and +1) to represent, respectively, a unit of 1/3°C decrease,
no decrease or increase, and a unit of -1/3°C increase of temperature over a unit increase
of altitude. Initial or ground level values of temperature for each sounding have been
tallied with similar numerals at the bottom of the data block.
In other words, more than 36,000 measurements of temperature obtained at about
600 increments of altitude in each of 62 radiosonde soundings are contained in this
example. Any or all such values of temperature could readily be re-established with an
optical sensor that need only identify three widths of marks while scanning any one
of the sounding-data lines from bottom to top. Counting the marks without regard to
their size would provide values of altitudes, and counting the marks with regard to their
size or algebraic sign would provide corresponding values of temperature.
IADIC DATA
Although such an incremental form of data provides most of the desired character-
istics of conciseness, qualitative portrayal, and quantitative exactness, it falls short of the
ideal in two respects. First, it is virtually impossible to discern particular quantitative
values manually. Second, it is too nonredundant; any error in sensing an incremental
value would produce a continuing and undectectable error in all succeeding counts of
"whole" values.
An early attempt to eliminate these shortcomings is illustrated on the right in
Bellamy 215
-------
Figure 1. These data are recorded with an "iadic" or "incrementally alternating^
incrementally continuous" notation. In effect, numerals in the iadic notation consist of
marks with variable transverse widths, each width standing for the value (such as zero,
one, two, three, or four in this example) of some particular digit in a digital representa-
tion of a "whole" number. The name of the notation is derived from the fact that a
long longitudinal "dash" is formed by the continual repetition of a particular width of
numeral throughout a region in which the value of the digit remains constant, and that
a one-to-one correspondence exists between a change of value of the digit and a change
of width of the longitudinal "dashes."
In this particular example, a "pentiadic" notation has been used to record the number
of 5°C units contained in the "whole'7 values of temperature represented incrementally
on the left of the page. That is: a "zero width" of iadic marks designates that the
temperature lies between 0° and +5°, 25° and 20°, 50° and 45°; a "one
width" of mark designates a temperature between 5° and 10°, 20° and 15°, 45°
an(J 40°; and so on until a "four width" of mark designates a temperature between
+20° and +25°, 5° and 0°, 30° and 25°, and between 55° and 50°. Conse-
quently, the positions of particular isotherms at 5° intervals are readily apparent as
the positions at which the width of the dashes change. The positions of isotherms at
25° intervals (or at 0°, 25°, and 50°) are especially apparent as being the positions
at which the width changes between the "zero width" and the "four width."
The data display characteristics of the iadic form of notation can thus be summarized
as follows. It provides a readily apparent "picture" of the large scale distributions of
particular quantitative values in much the same way as isotherm or contour maps do.
It would also provide for error-checking in automatic playback of associated incremental
records. In that case, corresponding iadic and incremental data lines (or soundings)
would be scanned simultaneously. They would then be checked to see that each change
in width of the iadic numeral corresponds with a change of value of that digit in a
continuing count of the incremental changes.
TALLIC DATA
The two examples in Figure 1 suffer the disadvantage that although they should
be used together it is very difficult to use them together. For example, it would be
virtually impossible to maintain the degree of mechanical registry required for simul-
taneous error-checking scanning of the two data-blocks as they appear in Figure 1. On
the other hand, if corresponding incremental and iadic sounding lines were to have been
interspersed into juxtiposition, most of the highly desirable shades and shadows pictorial-
ization would have been lost.
To overcome this disadvantage, several attempts have recently been made to
utilize a "tallic'7 or "transversely and Zongitudinally Zabelled, incrementally continuous,"
or "tally-like," form of numerals. That is, it is evidently possible to vary both the trans-
verse width and longitudinal thickness of rectagular "half-tone dots" in order to simul-
taneously tally two kinds of interrelated numbers.
The first and most readily accomplished trial4, 5 resulted in the formulation of the
tallic notation illustrated in Figure 2. The goal was to represent as concisely and
clearly as possible those years for which observational data were obtained at some
particular station with instruments such as stream gages or precipitation gages. This
goal was realized by using four transverse widths of tally marks to designate groups of
5 years each in a repetitive 20-year pattern. Also the availability of full, partial, or no
216 DATA DISPLAY FOR ANALYSIS
-------
record of measurement for a particular year is indicated, respectively, by a longitudinally
thick, medium, or thin tally mark for that year. The utility of this kind of notation is
indicated by the combination of over-all view and copius quantitative detail portrayed by
the enclosed map (Figure 3) of the periods of records available from all stream gage
stations that have ever existed in Wyoming since 1890.
Thick Tally; Full Record-
Medium Tally; Partial Record
Thin Tally; No Record-
Medium Tally; Partial Record i-
Thin Tally; No Record
Station Number
1950
c
1940 5;
1920
1900 ra
1895
1890
2280
Figure 2 Sample Tallic Representation of Periods of Data Records for the Period from
1890 through 1961.
The results of a more ambitious attempt5 to develop and use a tallic notation is
illustrated in Figure 4. This particular example portrays (1) each of the hours through-
out 4 years in which at least 0.01 inch of precipitation fell on the precipitation gage at
Laramie, and (2) the running accumulation of precipitation throughout each of those
years.
Briefly, the periods and rates of precipitation are indicated in the following way.
A row of 365 (or 366) lines, each consisting of 24 side-by-side tally marks, identify each
hour of each day of each year. The occurrence of at least 0.01 inch of precipitation
during any particular hour is indicated by a thick tally mark for that hour. A medium-
thick tally mark is used to indicate that it did not precipitate during that hour, but that
a precipitation amount in excess of 0.01 inch had occured (and had not yet been ac-
counted for) in some closely preceding hour. A thin hourly tally mark indicates (1)
that no precipitation fell during that hour and (2) that the total number of preceding
thick and medium-thick tally marks equals the total number of hundreths of inches of
precipitation that had previously fallen that year.
In addition, accumulated amounts of precipitation are indicated by five different
transverse widths of the hourly tally marks. The narrowest width of mark is used to
indicate accumulations between 0.0 and 0.2 inch, 1.0 and 1.2 inches, 2.0 and 2.2 inches,
etc.; the next wider width of mark indicates accumulations between 0.2 and 0.4 inch,
Bellamy
217
-------
LEGEND
I960L
I940JL
1920 |
Figure 3 Periods of Records of
Stream Gages in Wyoming.
1900}
FULL PARTIAL NO
RECORD RECORD RECORD
218
DATA DISPLAY FOR ANALYSIS
GPO 8141O58
-------
105*00'
Bellamy
219
-------
1.2 and 1.4 inches, 2.2 and 2.4 inches, etc.; and the widest of the five widths of mark
indicates accumulations between 0.8 and 1.0 inch, 1.8 and 2.0 inches, 2.8 and 3.0 inches,
etc. The gray shade appearance of this particular example is determined primarily by
these transverse widths of the tally marks. Hence, it is relatively easy to determine the
full inches of accumulation by counting the number of times they have changed from
their widest and darkest appearing width to their narrowest and lightest appearing width.
The results of a similar attempt6 to develop and use a tallic notation for portraying
rates of flow of three streams throughout 10 years are reproduced in Figure 5. In
each of these examples, a line of 365 (or 366) tally marks is used to identify each day
of each year. Three different longitudinal thicknesses of tally marks are used to indicate
(1) that the flow increased to at least one appropriately selected unit of measurement
of flow more than on the previous day with a thick tally mark, (2) that the flow de-
creased to at least one unit of measurement less than on the previous day with a thin
tally mark, or (3) that the flow remained constant within these limits with a medium-
thick tally mark. Five transverse widths of tally marks are used to indicate that the rate
of flow on any particular day was (for any positive integer, i) between 20i and 20(i+l),
20U+1) and 20U+2), 20U+2) and 20(i+3), 20(i+3 and 20(i+4), and between
20(i+4) and 20U+5) units of flow measurement, respectively.
These latter two examples demonstrate that tallic numerals can provide extremely
concise compilations of complete observational data in an error-checkable way. They
leave much to be desired, however, with respect both to their pictorial characteristics
and to the ease with which particular quantitative values can be discerned manually.
Evidentally these attempts to obtain a "double exposure" of both shades and shadows
pictures and quantitative contours surfer too much from excessive mutual interference.
SIPLIC DATA
A more recent and evidently more successful trial7 that eliminated these disadvantages
resulted in the portrayal of hourly precipitation data reproduced in Figure 6. This
particular form of tallic data is designated as being "siplic" data since it utilizes a
"scaled incremental, periodically labelled, incrementally continuous" notation. It utilizes
a relatively simple kind of tallic numerals to produce readability of quantitative values
without destroying the pictorial "shades and shadows" character of incremental data.
As in the example in Figure 4 of the same precipitation gage data, 365 (or 366)
lines, each consisting of 24 side-by-side tally marks, identify each hour of each day of
each year. In this case, however, three transverse widths (instead of thicknesses) of
tally marks are used to indicate (1) the occurrence of at least 0.01 inch of precipitation
during the hour with a wide mark, (2) the occurrence of an as yet unaccounted for pre-
cipitation in excess of 0.01 inch in a closely preceding hour with a medium width of
mark, or (3) no precipitation during the hour and no previously unaccounted precipita-
tion with a narrow mark.
This incremental data is then "scaled" by inserting a medium-width scaling mark
in the space behind those increments at which the accumulation reaches i(O.l) inches
(for any integer, i), and a wide scaling mark in the space behind those increments at
which the accumulation reaches i(l.O) inches. This technique of indicating values of
higher order digits, rather than interfering with the "shades and shadows effect,"
actually enhances it. In regions of especially heavy rates of precipitation the several
"extra" scaling marks tend to make the portrayal appear even darker, and the variations
in shade of the portrayal are thus determined almost entirely by the rates and durations
of precipitation.
220 DATA DISPLAY FOR ANALYSIS
-------
* O Gt CO
-------
GREEN RIVER UNIT OF RESOLUTION: 50 CUBIC FEET PER SECOND
ENCAMPMENT RIVER UNIT OF RESOLUTION: 5 CUBIC FEET PER SECOND
SYBILLE CREEK UNIT OF RESOLUTION: 0.5 CUBIC FEET PER SECOND
MONTHLY AND DAILY TIME SCALE
YEARLY TIME SCALE
Figure 5 Tallic Notation of Daily Values of Stream Flows.
In addition, "periodic labelling" of the values of higher order digits greatly en-
hances the case of discerning quantitative values of accumulation throughout each year.
It is accomplished with the pentiadic labelling lines alongside each year of incremental
record. The zero, one, two, three, and four widths of numerals represent values between
5i+0 and 5i + l inches, 5i+l and 5i+2 inches, 5i+2 and 5i+3 inches, 5i+3 and 5i+4
inches, and between 5i+4 and 5i+5 inches of accumulation, respectively. The particular
hour of the day in which any full inch of precipitation has accumulated is indicated
directly by the position of a wide scaling mark in the incremental portion of the portrayal
222
DATA DISPLAY FOR ANALYSIS
-------
I I !
UNITS OF RESOLUTION:
Time Horizontally 1 hour
Vertically 1 day
Precipitation Horizontally 001 inch
Scaling 0.1 inch
Vertically 1.0 inch
Figure 6 Scaled Incremental, Periodically Labeled Notation of Hourly Precipitation Amounts
in Laramie, Wyoming, 1958 through 1961.
Bellamy
223
-------
for that day. The total accumulation for the year can easily be ascertained by first
counting the numbers of 5's of inches corresponding to each major change from four-
width to zero-width numerals.
CONCLUSION
It is concluded from these examples that the development of equipments appropriate
for acquiring and utilizing siplic forms of tallic data should be very worthwhile. Evi-
dently, they would be adaptable to portraying measurements of most if not all kinds of
environmental conditions much more completely, concisely, and usefully than has been
possible heretofore. Or, as previously discussed in more detail,2 the development of
such "informatic" kinds of equipments should make the ultimate goal of acquiring and
utilizing "portrayals of everything geospheric everywhere always" much more approach-
able.
It is important in this respect to notice that although the incremental ''pictures"
in Figures 1 and 6 are at least as pictorial as many graphical "analyses" of contemporary
meteorological and hydrological conditions they are in reality more nearly records of
"raw measurements" than a result of "analysis." At least, they would be if appropriate
observational instruments were available to acquire these kinds of continuous data at
environmental measurement stations and if correspondingly appropriate equipments were
available for collecting, compiling, and utilizing the observations in this form as outlined
in Figure 7. Clearly, the key to opening this door to better understanding and utiliza-
tion of our environment is the availability of a wide variety of continuously recording
informatic observing instruments. Their development is now being emphasized in the
University of Wyoming's program of Informatic Data Research.
REFERENCES
1. "Informatic Forms of Data, 1961," John C. Bellamy, Natural Resources Research
Institute, University of Wyoming, November 1961, 11 pp.
2. "Geospheric Data Systems," John C. Bellamy, Natural Resources Research Institute,
University of Wyoming, November 1961, 11 pp.
3. "Study of Usefulness of Unitary Differential Notation for Storing and Utilizing
Meteorological Data," Cook Research Laboratories, Report No. 62-1, Contract No.
AF 19 (604)-1108, June 1955.
4. "Periods of Records of Stream Gages in Wyoming, 1890-1961," Philip M. Hoyt and
John C. Bellamy, Natural Resources Research Institute, University of Wyoming,
August 1962, 8 pp.
5. "Informatic Precipitation Gage Data," Merlin C. Williams and Leonard B. Baldwin,
Jr., Natural Resources Research Institute, University of Wyoming, Septmeber 1962,
14pp.
6. "Informatic Stream Gage Data," Verne E. Smith, Natural Resources Research
Institute, University of Wyoming. (To be published).
7. "Siplic Form of Precipitation Gage Data," Anton C. Munari and Merlin C. Williams,
Natural Resources Research Institute, University of Wyoming. (To be published).
224 DATA DISPLAY FOR ANALYSIS
-------
by SENSING
to get I n p ul J ig no Is
for TRANSFORMING
into o u
c
,4
t p u t Signals
\
for RECORDING
.5
as output Data
?rfl
F
which Represent Occurrences of the Earth s
0g[?D3g^3£
by OBSERVING
get Obse rva tions
(or COLLECTING
of the distri bution of
Everything Geospheric Everywhere Always
lor DISTRIBUTING
for ACTING to utilize
Figvre 7 Geospheric Data Systems.
Bellamy
225
-------
DISCUSSION
Dr. Larsen asked whether Dr. Bellamy knows of machines that produce good half-
tone pictures of the data displays described. Dr. Bellamy replied that some are in the
making at the University of Wyoming in connection with Masters and Doctoral work.
Very little machinery on the market is adaptable directly. Some of the very complex
and expensive machines completely invalidate the simplicity of this display technique.
Mr. Gelmont asked whether Dr. Bellamy finds difficulty in going from an analog
signal to the actual printout without getting into some involved programming or compu-
tational procedures. Dr. Bellamy indicated that the main requirement is a good analog-
to-digital converter and a little memory. The basic thing you are doing is keeping track
of the previous value so you know what the step has been. This can be incorporated into
the analog-to-digital mechanism by means of a stepping servo. Mr. Gelmont then asked
why the data should not be stored on magnetic tape to provide a computer printout, which
presents the topology of the situation. Dr. Bellamy pointed out the difficulty of identifying
what is on a magnetic tape without running it through something. He stated that the
density of storage in his system is compatible with or even better than most magnetic
tapes. Potentially it can be played back at least 10 times as fast as magnetic tapes.
Mr. Linsky commented that a major advantage of this system is inexpensive repro-
duction in large quantities.
226 DATA DISPLAY FOR ANALYSIS
-------
Glenn W. Brier
Chief, Meteorological Statistics Research Project
U. S. Weather Bureau, Washington, D.C.
SUMMARY
In the analysis of experimental data, the problem is to separate chance effects from
true regularities. By the use of the probability theory, certain mathematical models are
constructed that seem to bear at least some resemblance to the real world. This has led
to many useful techniques such as Least Squares Curve Fitting, Analysis of Variance,
Regression and Correlation Analysis, X2 Goodness of Fit, etc. Examples are given that
illustrate some of these techniques of data analysis; some aspects of extreme values are
considered in the examples. Several morals are drawn from this discussion: a knowledge
of the physics of the situation is' necessary before a meaningful variable or parameter
can be chosen for statistical analysis; there is value in knowing something about the
observations how they are taken, the peculiarities of the instrument or the observer,
etc.; and, so that the real effects are not confused with statistical artifacts that could
arise from data that are essentially random, the method to be used for processing the
data should be understood so that a valid interpretation of the results can be made.
TECHNIQUES FOR DATA ANALYSIS
This is a very broad topic and time does not permit a thorough discussion of even
a small fraction of the techniques available. Some of them are discussed in other papers
at this symposium and numerous textbooks are available.1-3,6 The emphasis here will be
on general principles, and good texts along this line are also available. For example,
"An Introduction to Scientific Research" by E. B. Wilson, Jr.,7 is an attempt to explain
simply a number of general principles, techniques, and guides for procedure.
Generally speaking, we are concerned with the analysis of experimental data. The
problem is one of separating chance effects from true regularities and is treated as a
branch of the theory of sampling. By the use of probability theory, certain mathematical
models are constructed that seem to bear at least some resemblance to the real world.
This leads to many useful techniques such as Least Squares Curve Fitting, Analysis of
Variance, Regression and Correlation Analysis, X2 Goodness of Fit, etc. It is doubtful
whether a further enumeration of such statistical techniques or even a brief description
of them is what we want here. A particular rule or a formula can be given, but there
is no assurance that it will be applied correctly or chosen wisely. I think what we really
want are "trained brains, and not a knowledge of facts and processes crammed into a
wider range of untrained minds,'7 as expressed by Karl Pearson. Or as Francis Bacon
would have it, "minds . . versatile enough to catch the resemblance of things (which
is the chief point) and at the same time steady enough to fix and distinguish their
subtler differences."
With these thoughts in mind, I have chosen some examples from the experience
of myself and my colleagues to illustrate some techniques of data analysis that, we hope,
can lead us to some general principles or conclusions. The printed program mentions
''extreme values,'' so perhaps I won't be departing too far from the spirit or intent of
the program to discuss some aspects of extreme values in these examples. The context
m which I discuss extreme values is, however, very different from the one commonly en-
countered in statistical practice. The statistics of extreme values has become a specialized
topic with at least one book5 devoted to this topic alone. A typical application of the
theory is directed toward the problem of estimating the probability that some natural
phenomenon such as an extreme flood will occur within a specified period of time or that
Brier 227
-------
. piece of machinery will fail or break down. Also, studies have been made about rules
relating to the rejection of observations of extreme values that do not appear to nt in
with the rest of the sample.
The first example here refers to a series of measurements of solar intensity by
means of an instrument called the pyrheliometer. This instrument measures the intensity
of the direct solar beam at the surface of the earth and therefore is affected by^the
atmosphere that contains dust and cloud particles, smoke, water vapor, etc. Instructions
to the observers at U. S. Weather Bureau stations are that observations should be taken
at specified solar zenith distances only when there are no clouds obscuring the sun.
This is somewhat subjective, since one observer may take more observations in a period
of time (such as 1 month) than another observer. When monthly averages are taken, they
will tend to run higher for the observer with fewer observations, since he has selected
only the "clearest" skies. Although both observers have chosen the extreme values, in a
sense, one of them has included observations closer to the mean or median of the fre-
quency distribution. One effect of this is shown by the graph for Blue Hill in Figure 1.
A new observer came on duty near the beginning of 1952. He has chosen only the
"clearest" skies, so that the "average" appropriate to his observations is approximately
11 percent higher than the "normal" for the period 1934 through 1951, shown as the
heavy "0" line. This bias might be avoided by using only the highest value each month,
and this might be very desirable if long-term trends in the data were being studied and
more than one observer was involved. Figure 1 came from investigation of the question
whether a volcanic eruption in Alaska in July 1953 produced extensive pollution in
widespread regions of the earth's atmosphere.* The data for both Blue Hill and Table
Mountain shown in Figure 1 give some support to the suggestion that such an influence
existed during the last part of 1953 and the first part of 1954.
1952
j s
1953
M M J S N
1954
M M J S
TTT
TABLE MOUNTAIN
Elevation 7500 feet
i i i i i i i r
i
34° 22'N. 117° 41'W.
Zenith Distance 60°
TTTF
TT
BLUE HILL 42° 13'N. 71° 07'W.
Elevation 672 feet Zenith Distance 70.7° (P.M.)
10
5
0
-5
-10
40
35
30
25
20
(5
10
5
0
-5
-10
~15 K
-20 <
Q.
UJ
Q
1952
1953
1954
Figure 1 Mean Monthly Solar Radiation Intensity in Terms of Departure from the
Monthly Averages for Two Locations.
Long-term
228
TECHNIQUES FOR DATA ANALYSIS
-------
It was possible to investigate this same question by a different type of data. Sky
photometer readings were available from Climax, Colorado, and Sacramento Peak, New
Mexico, for the period 1950 through 1955. This instrument measures the intensity of the
light from the sky at angles near the sun and compares this intensity with the intensity
of the direct solar beam. Dust or other scattering particles in the atmosphere tend to
increase the sky readings, whereas, very low readings indicate the absence of particles
due to dust, smoke, clouds, etc. Table 1 shows a sample of the original data used.
Table 1 Sky Photometer Readings at Climax, Colorado, and Sacramento Peak, New Mexico,
for January 1953.
Climax
Day Greenwich
Time
1
2
3
4
5
8
9 1600
1651
1717
1745
1833
2019
Sky
readings
5
5
5
5
5
5
Sacramento
Greenwich
Time
1718
1655
1759
1809
1844
1902
1550
1607
1644
1702
1515
1531
1729
1745
2105
2120
1740
1758
1524
1542
1545
1654
1705
1712
Peak
Sky
readings
62
26
35
35
23
23
35
35
23
23
90
90
16
16
42
42
15
15
35
31
31
21
31
21
2048 5
2159 15
10 1510 38
1533 38
1542 38
1705 14
1719 13
2129 26
11 1540 31
1633 14
1650 14
1853 35
12 1630 29
1645 30
1716 >500
13 1530 52
1550 50
1555 50
1926 29
Brier 229
-------
Considerable variability is indicated, and the occurrence of values of 500 or greater
would have a large influence on a daily or monthly mean. For the purpose of this study,
it was reasoned that selecting the lowest value each month would make considerable
physical sense. If the eruption of the volcano in 1953 produced an extensive pollution
of the upper atmosphere over widespread areas, then extremely low values of sky bright-
ness should no longer be observed because of the ever present common background of
extra particles producing scattering in the atmosphere. Figure 2 suggests that this is
actually what happened. There is a seasonal factor, with a deficiency of low values during
the summer months, but the winter of 1953-1954 shows an absence of low values for
both Climax and Sacramento Peak. The Sacramento Peak data suggest a return toward
normal seasonal conditions by the end of 1954, whereas the Climax data for 1954 and
1955 suggest the possibility of a "drift" in the instrument. On the basis of the results of
these charts, technicians examined the Climax photometer and found that it needed
to be recalibrated!
55
50
45
40
35
30
25
20
n
.
.
. . .
a "
f
.
*
1950
1951
1952
1953
1954
1955
CLIMAX
60
55
50
45
40
35
30
25
20
15
1 o
lu
5
0
'."'"
.
.-; --J-- -;-.-.-
.
" ' ' .
0 "
. . .
a
*
(
.
1950
1951
1952 1953
SACRAMENTO PEAK
1954
1955
Figure 2 Lowest Values of Sky Photometer Recordings Each Month for Climax
and Sacramento Peak.
Although these studies show the value of using only the most extreme observation
taken during an interval of time, the objection might be made that the extreme value
may depend too much on an instrumental or observational error. For example, if errors
due to mis-reading the scale of the instrument or transcribing the data are frequent and
large, the extreme values may be extreme only because of "goofs." The researcher must
know something about the magnitude and relative frequency of such errors before he
can make rational decisions involving the treatment of extremes. In some cases, it
230
TECHNIQUES FOR DATA ANALYSIS
-------
might be better to ignore the most extreme value and take the second highest (or
lowest), for example. The optimum procedure will depend upon the physical and other
factors involved. Figure 3 shows for Sacramento Peak the lowest daily readings of the
sky photometer for January 1953 and December 1953. In this case, it would not make
much difference what statistic is used to compare the 2 months. January would be
lower than December whether the highest or lowest values were chosen, the second
highest or second lowest, etc.
R rr
Hm HH P
Mlii m n
DEC. 1953
i H H rm m n n
JAN. 1953
m n n n
0 10 20 30 40 50 60 70 80 90 100 110
Figure 3 Lowest Daily Readings of Sky Photometer for Sacramento Peak
During January 1953 and December 1953.
In another type of investigation, it is common to use extremes for correlation or
regression studies. This can have important effects on the interpretation of the results,
especially when it is not recognized that extremes are being used or the investigator
is not aware of some of the subtleties or pitfalls in using extreme or rare events. For
example, in weather forecasting, we might wish to study how the frequency of extremely
large daily rainfalls is related to surface dewpoint. For a medical or health study, we
might wish to study how the frequency of patients with blood pressure over 200 is related
to the amount of salt in the diet, for example. The relationship between a dependent
variable of and a possible causal or independent variable x can be represented by a
scatter diagram like that shown in Figure 4. In this figure, all the data are shown for
x and y, not just the extremes. If only the highest values of y are considered, however,
such as those above the line y = k, it can be seen that the relative frequency of these
events can change very rapidly as x changes from negative values to positive values.
If we assume that x and y are distributed in a normal bivariate distribution with means
x = y = 0 and standard deviations o~x = ov == 1> it is easy to construct Table 2 to
show quantitatively how the relative frequency of an extreme event y changes or x
varies. This depends upon the average frequency (py) of the event y and upon the
correlation r between x and y. Even with a correlation as low as r = 0.01, there is
considerable change in the relative frequency of an event according to the value of x.
For example, if we take the case where an event happens only 13 times out of 10,000
(y = 35, py = 0.00135), we find that it is 100 x (118 103) = 14.6 percent more
103
likely, when x is 2 rr above the mean than when x is 2 u below the mean. For a
correlation as high as r 0.10, the percentage variations are tremendous, even though
a correlation of r = 0.10 means that 1 percent of the variance of y is accounted for by
the regression of y on x. Thus, extreme caution must be used in interpreting the results
of investigations where the frequency of unusual or extreme events (the pathological
Brier 231
-------
30-
2
-------
REFERENCES
1. Bowker, A. H., and Lieberman, G. J., "Engineering Statistics." Prentice-Hall, Inc.
(1959)
2. Brooks, C. E. P., and N. Carruthers, Handbook of Statistical Methods in Meteorology.
New York, British Information Services, 845 Third Ave., (1953).
3. Fisher, R. A., "Statistical Methods for Research Workers." Oliver and Boyd, Edinburgh
(1941).
4. Fritz, S. "Opacity of the Atmosphere after July 1953," Meteorological Magazine, vol.
85, No. 1006, April 1956, pp. 311-312.
5. Gumbel, E. J. "Statistics of Extremes." Columbia University Press, New York (1958).
6. Panofsky, Hans, and Glenn W. Brier, Some Applications of Statistics to Meteorology.
University Park, Pa., Penna. State University, (1958).
7. Wilson, E. B. Jr. "An Introduction to Scientific Research." McGraw-Hill, New York
(1952).
Brier 233
-------
Dr. Donald W. Pritchard
Director, Chesapeake Bay Institute
Johns Hopkins University, Baltimore
SUMMARY
The dangers of manipulating data to conform to an established hypothesis are illus-
trated, in particular the use of data both to formulate hypotheses and to verify them.
Data should be collected to provide answers to clearly defined, specific questions that we
ask about the environment, questions that will prove or disprove a, given hypothesis.
Methods of data collection must be compatible with the techniques of interpretation that
we intend to apply.
INTERPRETATIONS AND CONCLUSIONS
A reading of the program of this symposium indicates that others are scheduled to
speak specifically about data interpretation and conclusions for the separate environments
of air and water. I am here apparently scheduled to present some "basic" statements
applicable to the general subject of interpretation and conclusions. I have interpreted
this situation as giving me the license to speak rather broadly on the philosophy of
interpretation of environmental data.
In what follows, I am indebted to my colleague Dr. Blair Kinsman, who has written
very eloquently on this general subject in his paper "Proper and Improper Use of
Statistics in Geophysics'' (Kinsman, 1957). The first portion of my presentation here is
essentially a paraphrase of a part of Dr. Kinsman's paper, since his thoughts on this
subject so nearly coincide with my own. Where I have found that no rephrasing on
my part adds to the clarity (to me), I have simply, and perhaps lazily, quoted directly
from Kinsman.
The environmental scientist shares with all other scientists the task of telling "likely
stories'" with the intent that the story as told will satisfy the observations that the
scientist has made. The concept that this is a business of scientists is an old one,
dating back at least to Plato. The materials, that is, the set of data or observations,
with which a scientist works are "appearances", that is, sense impressions, and, as ex-
pressed by the old Greek phrase, he tries to "save the appearances." Given a finite set
of observations, this is a fairly straightforward task. "An excellent example is offered
by Ptolemy, who takes astronomical observations back to the reign of Nabonasser and
shows that he can 'save' them, that is fit them into a coherent pattern, by telling either
a geocentric story or a heliocentric story. With two equally satisfactory stories Ptolemy
weighs their likelihood and, on the basis of the physical information available to him,
develops the geocentric story as the more likely. The basis for this choice is hardly so
simple and straightforward. Today the general consensus, except possibly among navi-
gators, is that the heliocentric story is the 'true' one.
"The advent of the word 'true' together with the word 'real' in scientific discussion
has done much to cloud the nature of scientific activity. The request for a 'true' story
instead of for a 'likely' story tacitly postulates the existence of a 'real' world underlying
and giving rise to appearances and asks for information about that 'real' world. The
scientist qua scientist cannot answer such a question since the material on which he
operates consists entirely of appearances. St. Thomas Aquinas would probably have
said that no one could answer, since 'Nothing was ever in the mind that was not first
Pritchard 235
-------
begot upon the senses.' The attempts to bridge the gap between the postulated 'real'
world and the world of appearances which we perceive has a long and uniformity un-
satisfactory history, covering the spectrum from Descartes' assertion that God would not
fool us to Berkeley's retreat into solipsism. If we restrict ourselves to 'appearances,'
'hypotheses,' and 'likelihood,' metaphysical speculation about 'truth' and 'reality' can be
left to the metaphysician with a considerable gain in clarity.
"The point of departure is the appearances. These range from the casual impressions
of any sentient being through the systematic observation of essentially uncontrollable
phenomena characteristic of ... ," for example, environmental science, "to the precise
measurement of the result? of highly controlled experiments characteristic of the
laboratory sciences. The habit of attentive observation, coupled with an overwhelming
urge to fit the observations to a pattern, embryonic in most of the human race, is de-
veloped in the scientist to a high degree. All appearances, however, are not suitable for
scientific activity. Aristotle said that the subject matter of science is that which happens
always or for the most part. The unique event is a subject for history. Poincare (1905)
puts it this way:
Carlyle has written something after this fashion. 'Nothing but facts are
of importance. John Lackland passed by here. Here is something that is
admirable. Here is a reality for which I would give all the theories in the world.'
. . . The physicist would most likely have said: 'John Lackland passed by here.
It is all the same'to me, for he will not pass this way again.'
"Having, then, a set of observations of a recurring phenomenon the next step is to
construct an intelligible hypothesis into which the observations can be made to fit. One
fertile method is the use of analogy. Some other set of phenomena and their pattern being
known, if we can see a similarity, we can transfer the properties of the known system
to the unknown. Since analogies are seldom isomorphisms, the correspondences being only
partial, the dangers of argument by analogy are obvious. For example, the complex
numbers are analogous to the real numbers in the sense that operations of addition, sub-
traction, multiplication, and division can be defined for each. We might then argue
by analogy that, since division by zero is not permitted for the reals, division by zero
is not permitted for the complex numbers. We thus reach a correct result. If we argue
by analogy that since the reals are ordered the complex numbers must also be ordered,
our conclusion is false. Fertile as the argument by analogy is as a source of ideas, it is
almost worthless in support of an hypothesis unless it is shown that the analogous systems
are similar in every essential feature and that no dissimilar features can affect the proper-
ties that we wish to establish. Another method is to search the data for regularities.
If the sample is small . . .", as is generally the case in the environmental sciences,
"this is quite easy to do. However, since even samples drawn from a random number
table will exhibit regularities, results from such a procedure are suspect and once
a regularity is found most scientists feel impelled to provide some rationalization for it,
often based on an ad hoc selection of arguments. . . . Perhaps the most acceptable
method of forming hypotheses is by rational argument from established elementary
principles. It is worth noting that scientists in general seem to feel uneasy about any
hypothesis until it has been presented in this form no matter how it was first conceived.
An argument in Kepler's 'Epitome' is a case in point. Kepler's hypothesis that the
planetary distances are governed by the porportions of the five regular or Platonic solids
seems a little wayward today, but the urge to order the welter of appearances is easily
understandable. Kepler apparently felt the need to deduce his hypothesis from the first
principles of geometry and astronomy since he devotes considerable space to the effort.
236 INTERPRETATIONS AND CONCLUSIONS
-------
His logic is impeccable except for one short section. There the line of the argument has
been blurred, whether consciously or unconsciously cannot be known, so that the ostensible
deduction he was at such pains to make is, in fact, not established. The expenditure of
so much effort in such a cause by a man with first-rate discoveries to his credit shows
the importance attached to this method.
"But the telling of tales is only half the job and the easiest half at that. We still
face the problem of deciding how likely the story is or, if confronted with two different
but adequate stories, which is the more likely of the two. Statistics has been increasingly
concerned with understanding the structure of such decisions and with finding a clear
and objective method of making them. The general problem is far from solved, but
many valuable results have already been obtained. The judgments of the likelihood
of an hypothesis have had so many different bases that even a simple enumeration
would be too long. They include decisions made on entirely extraneous grounds, e.g.
the selection of a flat earth over a round earth on arguments derived from the 'second
coming.' They include decisions made on what I should call aesthetic grounds, e.g. the
selection of uniform circular motion as basic by the Greeks in contrast to uniform
rectilinear motion by Newton. They include decisions in which maximum simplicity is
equated with maximum likelihood. Occam's Razor is still a widely used scientific tool,
although the simplicity of nature is more an article of faith than a proven fact. The
more we refine our studies of nature the more complex things become until it almost
seems as though simplicity were an attribute of the infancy of a science. Fortunately,
there are a few threads to guide us in the labyrinth. Occam's Razor has been mentioned.
If you are willing to commit yourself to the proposition that the relations among
phenomena are fundamentally simple, then you will always choose the simplest story
that explains all the facts. However, it must always be borne in mind that one man's
simplicity may be another man's utter confusion. A heliocentric hypothesis simplifies the
astronomer's calculations but it makes those of the navigator intolerably complex. If one
is willing to forego questions of 'truth' and 'reality' one can escape the dilemma by
accepting both the geocentric and the heliocentric hypotheses (so long as they are not in
logical conflict) and use whichever one is simpler for the immediate purpose. A most
important criterion is that of compatibility with already existing structures. If the
hypothesis under consideration would require extensive revision of major parts of a
successful existing theory with all the labor that entails, clearly one would hesitate to
accept it unless a general improvement throughout the whole theory could be anticipated.
Another equally important basis for decision is the continued agreement of observations
with the hypothesis. Here statistics enters, since no set of experimental measures, if
sufficiently refined, ever agrees exactly with an hypothesis or with other sets. The
tincture of statistics that most of us retain from our formal training seldom goes beyond
the memory of where to find a few computing formulae. This isn't enough. While a
sprinkling of Pearson's xz's, Student's t-tests, and Fisher F-tests do lend an air of
objectivity to any paper, it must be remembered that statistical tests of significance
derive from mathematical models, which in turn are based on different views of the
nature of phenomena. The calculation of parameters is routine, and their use often
obscures a lack or precise thought about the fundamentals of a problem.
"In general, the job of the scientist is to invent a story which accounts for a set
of observations and then to decide how likely the story is.
"The . . ." environmental "sciences share the same theoretical base with laboratory
sciences. The Navier-Stokes equations hold equally well for a beaker of water and
for ..." a river, a lake, or "the oceans. With this common base it is not surprising that
Pritchard 237
-------
the methods for treating . . " environmental "data are often selected by analogy with
those used on laboratory data. Unfortunately, the materials on which these methods are
used in . . ." environmental science "are sufficiently different from those of the laboratory
sciences to require justification of the method, which is seldom explicitly given. In the
first place, the equations which give a full description of any situation are usually too
complex to be handled in their complete form. It is almost always necessary to simplify
them by considering some terms as negligible in order to get an approximate solution.
The laboratory scientist, by controlling the conditions of an experiment, can insure that
terms considered negligible are so in fact and, as a result, he can expect good agreement
between hypothesis and experiment even with relatively small samples. Further, he
can repeat his experiment at will.
"The . . ." environmental scientist "who, in the main can only observe, cannot repeat
his observations in the sense of a repeated experiment. This is a grave difficulty, since
most statistical tests of significance are fundamentally rooted in the idea of repetition.
While such repetition is conceptually possible in . . ." environmental science "it is
seldom realized and, in general, when such tests are used, their use should be supported
by argument. Further, his inability to insure that factors considered neglible are so
should logically force the . . ." environmental scientist "to use the equations for a process
in their complete form or first to show that the neglected factors can be neglected.
Neither of these courses is usually taken. There is nothing wrong with the making of
simplified models so long as they are not offered as 'reality.' When the observations
happen to agree with such models it is cause for gratification and suggests that the
neglected terms were small. If the observations continue to agree with the model we
can feel that we have a satisfactory story. In . . ." environmental studies "the agreement
usually need not be very close before verification is claimed. . . . This attitude stems
from lack of control. In comparing data with a simplified model large dispersions are
to be expected. This means that large samples are necessary if relations are to be
established with any certainty. It is unfortunate when the investigator, having been
forced to accept a simplified model, then feels impelled to insist that the natural
phenomena are themselves simple.
"Another difference arises from the answer that can be given to the question: 'Do
the data describe the phenomenon under consideration?' In the labaratory science
methods can usually be devised either to measure a property directly or to measure some
closely linked property. With the method in hand, the experimenter can then accumulate
enough data for a statistically reliable estimate of the property which interests him.
The . . ." environmental scientist "here labors under two kinds of handicap. First is
the matter of scale. Both the space and time scales are usually unwieldy. If one asks
for the monthly average temperature of the Chesapeake Bay is it enough to dip a
thermometer in once a day at some convenient place? One can hardly say without
knowing a great deal about the structure of the Bay. If it were enough for, say, January
1949, would anything be known about January 1950? To answer such questions an
inordinately expensive observation network would have to be established and maintained
for many years. Salinity records taken daily at Solomon's Island, Maryland, are a case in
point. This set of data extending back to 1938 is the longest unbroken record of salinity
taken anywhere on the Bay. Using monthly means and computing the power spectrum
it was found that there was evidence of a yearly cycle, which was to be expected. How-
ever, the great bulk of the power in the signal occurred at periods greater than two
years. It was calculated that to separate cycles having periods of three years and four
years at the 5% level, the record would have to extend over 285 years. Such a sample
from . " an environmental science "point of view is huge and it is rather sobering to
238 INTERPRETATIONS AND CONCLUSIONS
-------
see how little information it gives. Second, the . . ." environmental scientist "must
frequently work with data which were taken for other purposes and which do not
directly measure the properties that interest him. For example, an oceanographer inter-
ested in the factors influencing the size of fish populations might not have any measure-
ments directly made for that purpose but instead measurements of salinity made at
some point in the region and records of commercial fish catches. It isn't what he wants.
It's what he's stuck with. If he persists, he would have to argue something like this.
Fish population controls fish catch. Fish need plankton for food. Plankton need dis-
solved nutrients. Nutrients are brought to the surface layer by upwelling. Upwelling
influences salinity. The salinity of the region can be determined from the salinities
measured at a point which I know. Therefore, I will look over fish catches and salinities
for possible correspondences and, if I can construct one, I will know the connection
between fish population and environment. Laying aside the questions of whether the
salinity measured at a point represents the salinity over a large area and of whether
fish catch is an adequate measure of fish population, it seems unlikely that definite
clear-cut relation between the ends of such a long and tenuous chain would emerge
from a small sample.
"Another hazard inherent in using existing data taken for other purposes arises
from the temptation to fill gaps in it. If properties A and B are to be related and it is
found that A was measured at some point for a number of years but that the measure-
ment of B was neglected for a part of the time, then the urge to use measurements of
B made somewhere else to fill the gap may be almost irresistible. This procedure en-
larges the sample with an apparent increase in statistical reliability but it introduces
tacitly the very difficult additional problem of showing that the measures introduced to
fill the gap are the equivalent of what would have been secured had B been measured
at the point. This is usually impossible. In using existing data for purposes for which
they were not taken, great care must be exercised to see that wishful thinking does
not govern the make-up of the sample."
Environmental studies generally need large samples. Usually only small samples are
available. "In contrast with the laboratory sciences these small samples are often im-
precise, having been painfully secured in the field over many years, sometimes two or
three generations. To get another sample for testing involves the same long process. Thus,
if the entire initial sample is used in the formulation of an hypothesis, we are forced to
leave its verification or rejection to our grandsons. It is clear that progress of a science
which must either proceed on untested hypotheses or wait for generations to test them
will be either insecure or very slow.
"Verification is to be had only from data not used in formulating an hypothesis.
One possible method of securing data from testing an hypothesis formulated by search,
without delay, is to split the data on hand into two groups, one to be used in formulating
the hypothesis and the other reserved for testing. This may be done in a number of
ways. In some fields dealing with time series every other time unit is grouped to form
the two sets, or the data may simply be split in the middle. Separation by means of
some randomizing device could be used so that the bias, conscious or unconscious, of
the investigator would not invalidate such statistical tests of significance as might be
appropriate. The real difficulty here is 'keeping the game honest.' If the hypothesis is
formed before the test data are taken no question of influence arises. With both sets
of data in existence at the beginning of an investigation there is always the question
of the extent to which the investigator is influenced in his selection of hypotheses by
the test data. A glimpse of it, however fleeting, could bias him toward hypotheses
Prkchard 239
-------
likely to fit both sets. The difficulty could be met if the separation were made before
the investigator saw the data and he inspected only one set until he was ready to test
his hypothesis. Any alterations in the hypothesis after testing would, of course, be
highly suspect. The advantage of this device is that verification can be carried out at
once and an estimate of the value of the hypothesis made. The disadvantage is that the
already small sample size is further reduced, but it may be worth accepting this reduction
in exchange for immediate evaluation. It is well to remember that the information
contained by any finite sample is limited. Manipulating it in this way cannot increase
the amount of information contained. It can only sacrifice information of one kind
to gain information of another."
One basic difficulty in treating environmental data arises from the fact that we
seldom have two unique sets of values of specified parameters that may be paired in a
clearly objective manner for comparison. Normally we have one finite set of observations,
the variation in which we wish to "explain" in terms of the variations in a "controlling"
environmental parameter. Observations of this "controlling" parameter make up a
second set, which, after suitable manipulation provide a series of numbers that are
paired with and compared to the first set. Putting aside the questions associated with
generally circuitous and usually unprovable story we invent to show that the particular
parameter chosen is actually a "controlling" parameter, we are faced with the fact that
frequently all the statistical significance of the final results is destroyed by our use of
the data themselves to determine what manipulations are suitable.
A simple example will illustrate this situation. Suppose we have a set of observations
of the annual harvest of young "seed" oysters from a productive oyster bed for a con-
tinuous period of, say, 20 years. As is the nature of such data, we will find considerable
year-to-year variation in the harvest. As is usually done, we now equate the annual
harvest to the actual production and survival of the seed oysters on the bed in question.
We now want to explain the year-to-year variations. Coincidentally we find that there
have been, over this same 20-year period, daily observations of the salt concentration at
the condenser cooling intake of an industrial plant located not too far (?) from the
oyster bed. We conclude that daily observations are too variable and anyway provide too
many numbers to work with, so we compute the monthly mean salinity of the environment
at a location near our oyster bar.
We now have a series of 20 numbers representing the annual harvest of seed oysters,
and a series of 12 x 20 numbers representing the monthly average salinity. It takes
only a moderate amount of imagination, which we usually in such cases call reasoning,
to invent a story that convinces us that the mean monthly salinity should "control" the
production and survival of the seed oysters. In fact, we would probably reason that the
salinity during one part of the year would influence the condition of the brood stock,
and hence the number of eggs and sperm produced; while the salinity during another
part of the year would influence the fraction of young oysters that survived to the time
of harvest. Unfortunately our story is usually not complete. We are not sure which of
the 12 monthly values of salinity in each year is most important from the standpoint of
production of larvae and which is most important from the standpoint of survival. We
therefore proceed to compare the observed oyster harvest to a computed harvest for
each year based on a multiple regression of all combinations of pairs of monthly mean
salinities from the 12 months just previous to the harvest. Hurrah! We find that if
we use the monthly mean salinities for, say, the previous May and for December in our
regression relationship the computed oyster harvest is highly correlated (on the order
of 0.95) with the observed harvest for the 20 years of record. We have now "explained"
240 INTERPRETATIONS AND CONCLUSIONS
-------
the year-to-year variation in oyster harvest in terms of variations in an environmental
parameter!
Unfortunately we have used the observed data to search for the best relationship. In
point of fact almost any set of numbers showing some type of cyclic variation, such as
the mean monthly value of some environmental parameter, can be made, through suitable
manipulation, to show a high correlation to the annual variation in some other set of
observations. The "explanation" of the variations in oyster production arrived at in the
previous paragraph has, in fact, no statistical validity!
The number of published papers in which essentially the approach described above
has been used to explain the variation of some property of the environment is considerable.
Kinsman, in the work cited previously, analyzed a paper in which the author attempted
to show that the number of icebergs counted by the Ice Patrol in the North Atlantic
in any given year was related to the monthly mean sea-surface temperature anomalously
obtained from measurements at the end of a pier at Key West, Florida. In order to show
the relatively high probability of obtaining apparently significant correlations between
finite series when in fact any physical connection is nonsense, as long as some choice
for manipulation of one set is allowed, Kinsman counted the number of commas per
page in the issue of the journal in which the original paper on icebergs was published.
Kinsman correlated the number of icebergs in a given year to the number of commas per
page in the subject journal, but left himself the option of proceeding either forward
or backward in the page count, and of selecting which page he would start his com-
parison with. He found that when he computed the number of icebergs per year based
on the number of commas per page in the journal, starting with the last page of the
article he was analyzing and proceeding in page sequence toward the front of the paper,
he obtained a correlation of 0.95 with the observed iceberg count for the years 1942
through 1951. The comparison is shown graphically in Figure 1, taken from Kinsman's
paper. He then proceeded to use the relationship thus obtained, together with the
number of commas per page, running backwards, in the article just preceding the one
he had analyzed, to "predict" the iceberg count for succeeding years. As shown in
Figure 1, the prediction for the three years 1952, 1953, and 1954 is quite good. There-
after, as would be expected, the prediction failed completely.
Evidently Kinsman's selection of data was fortuitous; however, this example does
serve as a vivid warning about the way environmental data are often used.
It is my experience that most environmental data have been collected under programs
developed without adequate consideration of how the results will be used. The time
has come when we should severely limit the amount of effort being expended on the
general collection of environmental data, for which we have only vaguely or partially
conceived the use.
What, then, should our course of action be? First, we must recognize that from a
practical standpoint it is impossible to develop a single over-all observational program,
involving even a limited number of environmental parameters and a restricted natural
environment, that will provide data suitable for use in answering all, or even a con-
siderable fraction, of the questions that need to be answered regarding the subject
environment. The methods and timing of data collection suitable for the treatment of
one question about the environment will seldom be satisfactory for dealing with other
questions. Data collection programs designed without regard to specific, completely
stated questions that we want to ask of the environment will generally be not quite
adequate to definitely answer any question.
Pritchard 241
-------
Basically, then, the subject of interpretations and conclusions, which appears near
the end of this symposium program, should in fact be an integral part of the initial
development of an environmental observational program. The first step is to state clearly
UJ
m
1200
1000
800
600
400
200
[ce Patrol counts of Icebergs
o -o Iceberg counts estimated from
commas in Tellus
42 43
44
Figure 1 Correlation of Number of Icebergs in a Given Year to Iceberg Counts
Estimated from Commas in Tellus.
the problem or problems of concern. The next step is to use whatever general informa-
tion on the environment is available (yes, our past efforts at environmental measurement
have some use) to develop alternative hypotheses giving possible solutions to the problem
as stated. Each hypothesis then provides a set of questions that we must ask of the
environment in order to prove or disprove the subject hypothesis. The data-collecting
program should then be designed to answer each of the individual questions required
to prove or disprove the formulated hypothesis.
Admittedly, there may be some areas of environmental study for which so little
general knowledge is available that no reasonable hypothesis may be formulated. I feel
that this situation would be exceptional, for if we know enough to clearly state the
problems that need solution, we must know something of the environment, if only by
analogy to similar, better-studied situations.
242
INTERPRETATIONS AND CONCLUSIONS
-------
In essence, our starling point should be a set of conclusions, and our purpose should
be to find which of these conclusions are most nearly correct and which are clearly not
correct. We must know what techniques of data interpretation are available, and which
would most clearly serve our specific purpose. We then can design an observational
program that will supply data compatible with the interpretation techniques we have
selected.
In my presentation here, I have departed from the proposed content of my paper as
stated by the organizers of this meeting in the printed symposium program. I have not
given any gems of knowledge about trends and cycles, cause and effect relationships,
statistical inferences, or direct and indirect conclusions that will greatly help any of
you interpret the mass of generally inadequate existing environmental data. What I
have tried to do is to present some concepts that I hope might be employed in the
development of the new extensive and expensive environmental studies now being
planned or contemplated.
The modern statistical methods of treating environmental data, such as power
spectrum analysis, have been adequately discussed in the literature, and I only hope
that other speakers at this symposium will have given some general information on what
interpretations and conclusions can be drawn from their use. In the time allotted I must
be satisfied (even if my listeners are not) with this broad statement of the philososphy
that should be pursued in future environmental studies.
Prilchard 243
-------
Dr. Leslie A. Chambers
Professor of Biology and
Director, Allan Hancock Foundation
University of Southern California, Los Angeles
SUMMATION
Some of us can recall occasions, only a few decades back, of conferences on technical
and scientific subjects wherein the objective was brief communication of new ideas and
findings, informal discussion of their signicance, and a comforting absence of preprints.
Very gradually, and in parallel with the ready assumption by scientists and engineers
of a new order of economic and social respectability, group communication among us
has become stylized to an astonishing degree. Now there are indices of status built into
every conference indices which stratify the convening agency and the conferees them-
selves much more certainly than the informational content of the session itself.
Instead of contributed papers we now have symposia of invited speakers on pre-
scribed subjects; instead of concise introduction of contributors by name and title of
his paper, we now have lengthy accolades listing past honors, achievements, and other
biographical notes; instead of a prompt entry into the subject area of the symposium,
we now invariably have an hour or two or more of successive introductions and welcomes
culminating in the expected words from the highest-ranking individual the conveners
have been able to woo away from his normal duties. To cap the procedure, to endow
it with the formal attributes of stature, some near-pensioner, formerly but not now
active in the general area of conference purview, is customarily enlisted to say a blessing
over the whole thing in the guise of a "summation." I am honored by this role this
morning, but have never before sensed so fully the non-essentiality of a symposium
summary.
Those of you who have sat with me through the general sessions and a selection of
the separate subsessions dedicated to water and air will easily recognize the dilemma.
How can one possibly abstract a set of abstracts, epitomize an encyclopedia, minimize a
minimum. The enormous breadth of our subject area environmental measurements
coupled with the extraordinarily successful efforts of the several speakers to compress
their assigned facets of the whole into a few minutes, has given birth to what, when
published, will be a kind of pocket reference manual in the philosophy, technology,
and symbolism of communications theory, experimental design, statistical operations,
machine analysis computer programming, and a variety of other more or less related
concepts. It would be an injustice to some of the excellent papers to squeeze them
further or to take items from them; certainly no purpose can be served by offering orally
an annotated index of speakers and titles. You have the flavor of the conference sessions,
your personal estimates of the several contributions, and you have the papers themselves
to read and re-read if you are intrigued.
All of this leads to the simple fact that I have no intention of attempting any
summarization paper by paper or session by session. With apologies to the individuals
who have contributed, but without specific acknowledgment of their respective contri-
butions, I shall instead, use the next five minutes to summarize my own reaction to the
conference as a whole, and to add a comment or two in the philosophical vein so ably
mined by Dr. Anderson on Wednesday.
Chambers 245
-------
The papers we have heard fall into three general categories: (1) those that dealt
with the philosophy of measurement, information transfer, and data interpretation, (2)
those that offered in didactic but often delightful fashion certain elementary principles of
statistical theory, experimental design, and computer programming, and finally, (3) a
considerable number that reflected their author's preoccupation with the application of
measurement techniques and analyses to specific problem objectives. The third category
has tended to exemplify the inadequacies of the existing theories, techniques, sensing
equipment, and concepts, or more probably has exemplified a crying need for more skill
and understanding in their application to concrete problems of environmental measure-
ment.
Some of the questions that arise in any study of the environment have been asked
here, and some have been answered in part.
Why do we measure? The importance of a clear understanding of the objective has
been emphasized; in the lingo of experimental science, a clear and understood state-
ment of the problem to be solved is the prime requisite. Even at this point, a group of
people such as this will certainly formulate different starting points. Those with an
end-point perspective the engineers, and physicians for'example will measure an
environmental parameter for the purpose of future interpretation in terms of some
possible effect on man or other object. Those who find their pleasure in sheer under-
standing of the properties of a given system may be content with information about the
system for its own sake. In either event, they all qiuckly find that the classes, sub-
stances, or events they have chosen to measure, even if completely mensurable, cannot
by themselves give any final satisfaction. The man whose objective is to hold the en-
vironment in compatibility with human tolerance must include himself and other men
as reactants in the system he considers. If, for example, the concern is with lead in
the atmosphere there can be no useful result from measurement of airborne lead alone;
there must also be data on the ranges of human tolerance to lead as functions of age,
rate of intake, physical and chemical forms in which the lead occurs, and especially there
must be data on lead intake in water and food and the relative importance of intake
by different routes.
If, on the other hand, the concern is with airborne lead as part of the normally
dynamic atmospheric system per se, its role cannot be interpreted from static measure-
ments of lead concentrations alone. Interactions are the norm and their products may
not even be precisely definable as lead in a proper sense.
Such considerations lead promptly to a series of additional questions.
What should we measure in order to attain the defined objective? Certainly airborne
elemental lead values will prove insufficient for almost any purpose.
How should we measure the several parameters essential to attainment of our
special objective?
How much measurement is necessary? In other words, what is the minimum
effort necessary to achieve some reasonable level of significance taking into account the
accumulative errors of the several types of measurement involved?
When should the measurements be made? Is the effect of the man-environment
interaction expected to be a long or short function of time?
How do we process the data to produce a display of interpretable evidence bearing
on the pre-set problem?
246 SUMMATION
-------
And finally, how do we communicate the evidence and findings to create maximal
momentum toward a control objective?
We came here with these questions before us; we leave with the certainty that there
is no pre-mixed formula that will permit transfer of our central functions to the best of
present or future sensing, data processing, and analytical labor-saving systems. It is
important to bear in mind that the thinking necessary to the programming of the finest
systems may be the weakest link in the chains of events we set in motion. On the other
hand we leave impressed by the rapidity with which cybernetic extensions of our in-
herent capabilities are enabling some reasonable approaches to environmental problems
involving multiple parameters.
At the conclusion of his paper, Gaylord Anderson drew from the Homeric version
of the Straits of Messina a classic allegory in which the cooperation of Scylla and
Charybdis absorbed an input of fragmentary evidence and imprecise data and spewed
forth false conclusions. In an earlier portion of the same paper attention was called to
the fact that even with attention focussed strongly on physical and chemical factors
in the environment, one must not forget that man and his behavior have produced the
environmental alterations which we fear. And elsewhere pointed out that the basic
reason for measurement is to determine the magnitude of environmental forces and the
effects they have on man.
Now, the minor logical conflict I read in these two statements (taken out of context)
I find it possible to resolve quite readily. There is no a priori necessity for regarding
either man or his environment as either cause or effect. In a very real sense man is
simply another reactant in the nonhomogenous, many-parametered system with which
we are concerned. It is a single, temporally continuing, constantly interacting system
with which we deal. It is with variations in rates and quantities that we are concerned
since the system, with or without rather transient intermediate steady-state assemblages
such as man, has been here a very long time, and will be here much longer. The time
has almost come, in terms of operational capability, when we can begin to think of the
simultaneous analysis of entire segments of the system and not be restricted to adaptive
actions based on measurements of single parameters plus intuition, hope, and a grave-
yard rabbit's foot.
If we are to tackle analysis of the system of which we are a part, a level of tech-
nical skill, mathematical and logical sophistication, and philosophical detachment not
now generally incorporated into the training of environmental scientists will have to
be attained. This symposium has made its contribution; as realization of the total
requirements to cope with the total problem is attained, succeeding conferences will
undoubtedly be more comprehensive in scope and at the same time, deal in greater
depth with the technologoy of planning, sensing, transmitting, translating, analyzing,
and displaying essential information.
Chambers 247
-------
SESSION 7: Measurements of Air Environment
Chairman: Jean J. Schueneman
Chief, Technical Assistance Branch
Division of Air Pollution
U. S. Public Health Service
GPO 614-1059
-------
Dr. Ralph I. Larsen
Field Studies Branch
Division of Air Pollution
U. S. Public Health Service, Cincinnati
SUMMARY
Interrelations among variables may be determined by the following steps. Plot
the data. Study the variables that show good interrelationships. Determine if the inter-
relation is arithmetic, semi-logarithmic, logarithmic, cyclic, or probabilistic. Plot the data
on a type of graph paper that will give a straight line. Determine the equation of the
line, thus tersely expressing the relationship between variables. Correlate and regress the
data. Construct and test a mathematical model that agrees with the results and makes
good physical sense. Try to understand and explain why the relationship exists. Use the
new knowledge gained to better manage the environment, whether it be air, water, land,
radiation, milk, food, or something else.
DETERMINING BASIC RELATIONSHIPS BETWEEN
VARIABLES
"For all the Athenians, and strangers which were there, spent their time in nothing
else but either to tell or to hear some new thing."1
Thus began the response of Dr. Joel H. Hildebrand upon receiving the 1962 William
Procter Prize from the Society of the Sigma Xi.2 He continued with, "All true scientists
like to spend most of their time in nothing else than either to discover, to tell, or to
hear, some new thing.
"The urge begins with a peculiar combination of genes which produces an insatiable
curiosity. This leads in childhood to endless questions and continual experiments with
things and persons. The behavior is not that of the model 'good child,' who, when told
to run along and not to ask so many questions, obediently 'runs along,' never asks
questions his elders cannot answer."
He continues by noting, "A physical scientist does not merely 'learn' the laws of
thermodynamics; he must try to understand them; he must gain an intuitive feeling for
the concepts of enthalpy, energy, free energy and entropy . . Even so delightful a sub-
ject as calculus can be taught mainly as formulas for differentiating and integrating,
whereas what is really needed is that a person shall understand the various expressions
and operations so well that one can formulate a physical problem in mathematical terms,
translating freely back and forth between English and calculus."
This curiosity and desire for understanding noted by Dr. Hildebrand are two im-
portant motivating forces needed for exploring basic relationships between variables.
THE PROBLEM
Today's environmental studies produce thousands and sometimes millions of num-
bers. The desire to understand their meaning forces one to try to determine their inter-
relationships.3 Ideally, the results should distill into a few cogent formulae, just as
Newton distilled his observations into three laws of motion, his second one being
force = (mass) (acceleration)
Larsen 251
-------
Similarly, Einstein conducted almost no experiments of his own, but used the results
of others to formulate his theory of relativity, and his world-shaking
energy = (mass) (speed of light) 2
APPROACH
The thoughts that follow are from my own limited experience. Others might em-
phasize different points.
Insight into possible new relations between variables or new analytic approaches to
a study seldom comes when I am working hard directly on the study. Instead it comes
when I may be ragging some thoughts over in my mind rather loosely and relating things
from different fields. It is in this atmosphere that new approaches may come to mind.
Also, for myself it seems to work best to do the hardest or most demanding or creative
work in isolation before noon, and maybe communicate and do more routine work
after noon.
So much for attitudes and philosophy. What methods can be used to explore and
determine basic interrelations between variables?
METHODS
The following sequence of operations works best for me:
1. Plot tens or hundreds of plots of one variable or group of variables against the
others. Use simple cartesian coordinate paper. Plot with pencil, punched card tabulator,4
or electronic computer.5
2. Study in detail the plots that indicate good relationships, i.e., without widely scat-
tered points.
3. Determine if the data are cyclic. If not, find a graph paper on which the data
will plot as a straight line.
4. Determine the equation of the line, thus tersely expressing the relationship be-
tween variables.
5. If you want to find out how good the relationship is, correlate and regress the
values. If you have quite a few values, let an electronic coputer do this.4' 6
6. Construct a mathematical model that agrees with the results and, preferably,
makes good physical sense as well. Test the model.
7. Try to understand and explain why the relationship exists.
8. Use the new knowlege gained to better manage the environment, whether it be
air, water, land, radiation, milk, food, or whatever.
EXAMPLES
Let us consider examples of various sets of data and various types of interrelations.
ARITHMETIC
Baulch has recently related sulfur dioxide concentration to wind direction gustiness.0
He classified gustiness into five types and showed distributions of the time percent of
each type for various sulfur dioxide concentrations. Gustiness types Bj and D seemed
to be especially related to concentration. They are defined as follows:
252 DETERMINING BASIC RELATIONSHIPS
-------
Bji Wind fluctuations from 15 to 45 degrees.
D : Short-term fluctuations not exceeding 15 degrees. The trace approximates a
straight line.
The ratio D/I^ looks as if it should relate to sulfur dioxide concentration. Three of the
points plot as a straight line on cartesian coordinate paper (Figure 1). The equation
of a straight line is
y = mx + b
Where m is the slope of the line and b is the value of y at x = 0.
Thus the equation for Figure 1 is
sulfur dioxide cone. = 0.033 + 0.02
(Actually, if all four points were considered, a semi-logarithmic plot would fit best.
An arithmetic plot is used here for an example.)
0.10
2 0.08 -
CC
I-
z
u
o
o
o
0.06 -
0.04 -
£ 0.02 -
cc.
O
I
w 0 1
01234
GUSTINESS CLASS FREQUENCY RATIO, D/B,
Figure 1 Two-Hour Mean Sulfur Dioxide Concentration in Nashville Versus Wind Gustiness
Ratio, October 1958 - March 1959. (Data Source: Ref. 6).
Gustiness type D indicates stable meteorologic conditions; type 'B1 indicates more
turbulent conditions. Thus the equation makes sense in that stability is associated
with high concentrations and turbulence is associated with low concentrations.
SEMI-LOGARITHMIC
Tice has presented steel corrosion data as a function of time exposed (Figure 2) .7
These data look is if they would fit a straight line if the lower years were expanded to
the left. A logarithmic horizontal scale will accomplish this (Figure 3). The equation
for this line may be determined as follows. Again, the equation of a straight line is
Larsen
253
-------
y = mx + b
v will be on a logarithmic scale
y = m (log x) + y at log x = o
log x2 log x±
= 84-36 =JS =
log 10 log 1 1-0
y = m (log x) + yx = 1
y = 48 log x + 36
Weight loss, in g = 48 log (years) + 36
Similar methods can be used to determine equations for the other two lines.
Reducing the data to equations allows describing any line by only two parameters, slope
and intercept. These two parameters may then be used to compare plots.
90
12345678
EXPOSURE TIME, years
Figure 2 Effect of Time of Start of Tests on Corrosion of Steel at New York City.
(Source: Ref. 7).
This semi-logarithmic plot indicates that the same weight loss occurs with every
doubling of time. Thus the same weight loss occurs between 1 and 2 years as between
2 and 4 and 4 and 8. As corrosion occurs, fewer open sites are left to corrode, and
oxidation below the oxide layer is probably slower. This might be one possible explana-
tion for a decreased corrosion rate as a function of time.
This is an example of a logarithmic horizontal plot. Let us now consider a logarithmic
vertical plot, and one with three variables instead of two. You can disregard the
parameters and think only of the mathematics. The parameters happen to be salary versus
Government Service grade and step (Figure 4). It appears that a straight line might
result if the lower grades could be expanded downward. A logarithmic vertical scale
254
DETERMINING BASIC RELATIONSHIPS
-------
would accomplish this (Figure 5). A definite change in slope occurs at GS 11. Thus
one line could describe GS 1 to 11 and another GS 11 to 15.
100
1 2 3 456789 10
EXPOSURE TIME, years
Figure 3 Weight Loss of Steel Versus Years of Exposure. (Data Source: Ref. 7).
I I I I I I I
024 6 8 10 12 14 16
GS GRADE
Figure 4 Annual Salary Versus Government Service Grade, January 1, 1964.
Larsen
255
-------
30
ANNUAL SALARY = $2900 (10°. GRADE)
(1+0.033 STEP) FOR GS 1-11
J L I I I I I
6 8 10
GS GRADE
12 14
16
Figure 5 Annual Salary Versus Government Service Grade, Logarithmic Vertical Scale.
Again, for a straight line
y = mx + b
In this case, y is on a logarithmic scale
log y = mx + log y at x = 0
m log y2 log yt
= log 7,650 log 3,000
10 0
= 3.884 3.478 = 0.406 = 0.0406
10 10
log y = 0.0406 (GS grade) + log $3,000
Take antilogs
Annual salary = $3000 (10°-0406
Annual salary = $1500 (10°-0682
grade)) for GS 1 11
grade)) for (jg 11 _ 15
Step 10 rates plot parallel to step 1 rates (Figure 5). Thus the vertical distance
from step 1 to step 10 is a constant. For this logarithmic scale, step 10 is thus always
30 percent greater than step 1, regardless of the grade.
256
DETERMINING BASIC RELATIONSHIPS
-------
A plot of salary as a function of step is linear (Figure 6), indicating that for a
given grade each step is a constant number of dollars greater than the previous step.
Combining the effects of grade and step, salary may be expressed as follows:
Annual salary = $2900 (10°-°«6 ede) Q + 0.033 step) for GS 1-11
Annual salary = $1450 (10°-°6S2 grade) (i + Q.Q33 step) for GS 11-15
Since three variables are involved, the data may be plotted in three dimensions
using isometric paper. The equations describe the top surface of Figure 7.
20
18 h
list-
's
12
3 8
SS
_i 6
GS - 12-
STEP
Figure 6 Annual Salary Versus Step for Various GS Grades, January 1, 1964.
Possibly an easier way to think of the data is that each grade, from 1 to 11, pays
10 percent more than the previous one; and each grade from 11 to 15 pays 17
percent more than the previous one.
The salaries were probably not determined in this manner, since they do vary
above and below the trend, but the equations seem to give a good estimate of the
interrelations between grade, step, and salary.
LOGARITHMIC
On cartesian graph paper, if one end of a plot tends to become horizontal and the
other end vertical, a logarithmic relation may exist. A plot of the percent oxy-hemoglobin
in the blood as a function of the partial pressure of oxygen seems to satisfy this re-
quirement (Figure 8).8 To fit a logarithmic plot, however, the graph has to be turned
upside down and percent arterial hemoglobin unsaturation used rather than saturation.
A straight line then results (Figure 9).
Larsen
257
-------
Figure 7 Annual Salary as a Function of Government Service Grade and Step,
January 1, 1964.
100
0246 8 10 12 14
OXYGEN PRESSURE, cm of mercury
Figure 8 Blood Oxy-Hemoglobin Concentration Versus Oxygen Partial Pressure.
(Source: Ref. 8).
258
DETERMINING BASIC RELATIONSHIPS
-------
100
II I I I I III
O
IE
D
m
o
_i
o
o
III
I
IE
LI
I-
10
I I I I I I II-
0.1
10
OXYGEN PRESSURE, cm of mercury
Figure 9 Percent Oxy-Hemoglobin Unsaturation Versus Oxygen Partial Pressure.
(Data Source: Ref. 8).
Again, for a straight line
y = mx + b
In this case both axes are logarithmic.
log y = m(log x) + log y at log x = 0
m = log y2 log y±
log x2 log x±
m may also be determined by merely measuring the slope on the graph with a scale or
ruler.
Take antilogs
y = b xm
b is the value of y at log x = 0 (i.e., x = 1).
Thus for Figure 9, since y is 5,000 when x = 1,
% hemoglobin unsat. = 5,000 (oxygen pres. in cm of mercury) ~3'7
This indicates that hemoglobin unsaturation is inversely proportional to oxygen
pressure to almost the 4th power.
PROBABILITY
The distribution plot of many entities in the world is bell-shaped, fitting an
arithmetic-probability or Gaussian distribution. Air pollutant concentration data usually
fit this bell shape, if concentration is plotted on a logarithmic scale,9'10 giving a
Larsen
259
-------
logarithmic-probability plot (Figure 10).* To get a straight line distributions of data
may be plotted on cumulative distribution paper, either arithmetic-probability or
logarithmic-probability, whichever fits a straight line best.
0.01
0.1
30 50 70 90
PERCENT OF DAYS
99.9
Figure 10 Frequency of Various Levels of Total Oxidant Peak Hourly Concentration
at Los Angeles Station 1, 1956-57.
A plot of sulfation (an index of sulfur dioxide concentration) as a function of the
distance from the center of Nashville showed this typical bell-shaped distribution (Figure
II).11 Plots of the data on arithmetic cumulative distribution paper gave straight lines
(Figure 12). Therefore it was possible to use a Gaussian-type equation to express
sulfation in Nashville as a function of the distance from the center of town.
S = S -I- S f **
o Ojj \ oce
where S is sulfation,
Sj, is the background sulfation,
Sc is sulfation at the center of Nashville (minus Sb),
e = 2.718, the base of natural logarithms,
r is the radial distance from the center of Nashville, and
sr is the standard radial deviation, which is analogous to standard deviation.
CYCLES
Some variables vary cyclically. Oxidant concentration in Los Angeles is a function
of sunlight, and thus tends to peak about noon and be low1 at night. Concentrations of
pollutants from motor vehicles tend to peak during the morning and evening traffic
rushes and be lower at other times. Many variables are a function of time of day, day
of week, season, year, or maybe even sun spot intensity (11-year cycle). These variables
may be expressed as sine or cosine waves with none or several harmonics. In fact any
260
DETERMINING BASIC RELATIONSHIPS
-------
continuous curve may be approximated by a sufficient number of harmonics, by means
of Fourier analysis.12 More intricate time series techniques may also be used for auto
correlation and power spectrum analyses.
WINTER
FALL
ANNUAL
" SPRING
* SUMMER
2345 678
RADIAL DISTANCE, miles
Figure 11 Geometric Mean Sulfation by Season Versus Radial Distance from
Center of Nashville.
a
I
u"
o
z
i
a
_i
<
a
1
50 60 70 80 90 95 98 99
AREA UNDER CURVE, percent
Figure 12 Area Under Figure 11 Sulfation Curve.
99.9
Larsen
261
-------
A time plot of minimum sun spot activity and air pollution disasters is interesting
(Figure 13). Four air pollution disasters have occurred during the past three peace-time
periods of minimum sun spot activity, including the period we are in presently. The
Donora disaster is the exception. Whether a real interrelation exists or whether this is
merely happenstance, I do not know, but it is interesting to contemplate, and possibly
to predict, "Look out for the winters of 1962-64."
§ 8
£ *» S
I il I i
z to5 D z <"
^_ 8_SE!J!a 3 i
a «"ui
w .. , rju3m-
Q rf O
111
<2 «SsSs8 a i
S
« :x w a: UJ tr O
5 z ~ UJ o UJ OJ
3 O _1 CQ xm
> Q d 5 UJ5 -
CO
UJ UJ
do o co o J" o SJ X J" UJ fr
mo <* TfoPo>2 O SO
cno 01 oiouJoO^ uj _Q-
rHtD rH rHtDQ^^CO Q r-tO
1930 1940 1950 1960 1970 1980
Figure 13 Air Pollution Disasters and Minimum Sun Spot Activity, 1930-1980.
MATHEMATICAL MODELS
Mathematical models for explaining data can be proposed and then checked for
validity.115 We have just seen in the Nashville example how sulfation can be expressed.
Sulfur dioxide emission data can be expressed in a similar manner. Sulfation can then
be related to emission. A simple mathematical model was proposed to do this.11
sulfation = k (emission strength)
where k is a constant,
x is the distance between source and receptor, and
n is an exponent.
The problem was programmed on a computer and tested for several combinations of
parameters. The best fit to actual data occurred with n = 2, indicating that sulfation
is inversely proportional to the square of distance between source and receptor. This
makes sense, for it indicates that the long-term average diffusion of pollutants from
a point is similar to the diffusion of light or the diffusion of nuclear radiation from a
point. Or maybe it would be better to say that the radiation of air pollutants is similar
to the radiation of other mass or energy.
The climax of many studies is building, testing, and validating mathematical models
to explain the data. Ideally, the model should be a mathematical expression of the actual
physical process involved, or a simplified representation of the process. The model might
describe the interactions between mass and energy in air, water, man, and bacteria.
Algebra, calculus, and statistics could all be used in constructing the model.
262 DETERMINING BASIC RELATIONSHIPS
-------
Model construction and validation may be the most vital and challenging part of
a study. Unfortunately, interest in a study may flag by this time, or preparation for
the next study may be demanding. Thus this vital, key operation may be seriously
neglected. It is important, however, to spend lots of time in thought, cogitation, and
testing at this stage, in order to produce a finished product.
REFERENCES
1. Luke, "Acts of the Apostles," Bible, Acts 17:21.
2. J. H. Hildebrand, "To Tell or to Hear Some New Thing," American Scientist,
51:2-11 (March 1963).
3. R. I. Larsen, "Parameters of Aerometric Measurements for Air Pollution Research,"
American Industrial Hygiene Association Journal, 22:97-101 (April 1961).
4. R. I. Larsen, "A Method for Determining Source Reduction Required to Meet
Air Quality Standards," /. Air Poll. Control Assoc., 11:71-76 (February 1961).
5. R. I. Larsen, "Choosing an Aerometric Data System," /. Air Poll. Control Assoc.,
12:423-430 (September 1962).
6. D. M. Baulch, "Relation of Gustiness to Sulfur Dioxide Concentration," /. Air Poll.
Control Assoc., 12:539-542 (November 1962).
7. E. A. Tice, "Effects of Air Pollution on the Atmospheric Corrosion Behavior of
Some Metals and Alloys," /. Air Poll. Control Assoc., 12:553-559 (December 1962).
8. R. A. McFarland, F. J. W. Roughton, M. H. Halperin, and J. I. Niven, "The Effects
of Carbon Monoxide and Altitude on Visual Thresholds," /. of Aviat. Med., 15:381-
394 (1944).
9. C. E. Zimmer, E. C. Tabor, and A. C. Stern, "Paniculate Pollutants in the Air
of the United States," /. Air Poll. Control Assoc., 9:136 (November 1959).
10. Air Pollution Measurements of the National Air Sampling Network, 1957-1961,
Public Health Service Publication 978, U. S. Government Printing Office, Washington,
D. C. (1962).
11. R. I. Larsen, W. W. Stalker, and C. R. Claydon, "The Radial Distribution of Sulfur
Dioxide Source Strength and Concentration in Nashville," J. Air Poll. Control Assoc.,
11:529-534 (November 1961).
12. H. A. Panofsky and G. W. Brier, Some Applications of Statistics to Meteorology,
The Pennsylvania State University Press, University Park, Pennsylvania (1958).
13. E. K. Harris, D. S. Licking, and J. B. Crounse, "Mathematical Models of Radio-
nuclides in Milk," Public Health Reports, 76:681-690, (August 1961).
Larsen 263
-------
Glenn W. Brier
Chief, Meteorological Statistics Research Project
U. S. Weather Bureau, Washington, B.C.
SUMMARY
Because values in a time series may not be statistically independent, the reliability
of various statistics generated from time-series data may be questioned. The tendency
of each value to be correlated with chronologically adjacent values is known as
persistence, a problem that requires the application of special methods. A procedure for
spectral estimates and a filtering or smoothing function are applied to the analysis of
meteorological data. The significance of high-speed computer technology is emphasized.
INTERPRETATION OF TRENDS AND CYCLES
During the past few days you have heard of a number of statistical concepts or
principles and have been introduced to a few techniques of data analysis. A set of n
observations
has been treated or analyzed as a sample from a "population" by some appropriate
theory that makes use of a mathematical or probabilistic model. The usual assumption
is that the n observations are independent that one actually has a sample of n
observations. In much meteorological or geophysical data, however, the value of a
particular Xj is not statistically independent of the other values in the sample and may
be related to Xj+j or Xj-f. for example, because of proximity in space or time. This
interdependence of values tends to invalidate the standard formulas used to assess the re-
liability of the various statistics estimated from the data, such as means, standard devia-
tions, correlation coefficients, etc. In a time series, as a rule, the successive values of
the series are not independent of one another, and the tendency of each value to be
correlated with chronologically adjacent values is known as persistence. Special methods
are needed to treat this problem of persistence in data; today we will consider a few of
the things that might be done. Some aspects of the problem have been reviewed and
discussed recently by Mitchell.6
One of the oldest questions in meteorology is whether there are any cycles in
weather data, other than the well-known daily and annual cycles. This question has
been investigated by hundreds, who have used the classical methods of harmonic
analysis known to mathematicians for centuries. This technique is a proper one for
investigating the harmonics of a fixed identifiable frequency under the assumption that
the time series is genuinely periodic, i.e., repeats itself exactly every n observation. Its
misuse "when these assumptions do not hold has been responsible for the acceptance of
probably more spurious hypotheses than any other statistical or applied mathematical
tool ... [It] breaks down completly when applied to a statistical fluctuation."5
Now if harmonic analysis is not an appropriate tool for use in investigating non-
randomness and apparent quasi-periodic fluctuations in data, what can we use? One
answer is that given by Tukey,8 who suggested a sound and practical computational
procedure for obtaining "spectral" estimates based on the results of the pioneer work
by Wiener10' " on generalized harmonic analysis.
Brier 265
-------
The recommended procedure provides spectral estimates (Uh) showing how the
variance of the time series is distributed as a function of frequency. Ward8 has recently
described the method in some detail in connection with an application to the geo-
magnetic disturbance indices. Panofsky and Brier' discuss some other applications,
and Blackman and Tukey treat the subject more completely in a recent monograph.1
The spectral estimates are obtained by first computing the sample autocorrelation
function
n-k
Z X' Xi + *
n-k
i = 1
where n is the number of observations used and the Xj are data points expressed in
terms of the deviation from the mean of the series. From the Fourier-transform of the
Rk function, the apparent "line powers" Lj, are determined by
Lvk
M-l
Rk
K = I
M-l
2 V** r> v h
K=l
M-l
L = -J-(R + I11MR ) I l Y (-1) KR
m 2M o M -T M L.
K=l
M is the number of lags for which the autocorrelation function is computed and is
usually about 5 or 10 percent of the number of observations n.
The values of Lh are smoothed to obtain the spectral estimates
U0 = 0.54 L0 + 0.46 Lj
Uh = 0.54 Lh + 0.23 (Lh_ 1 + Lh+1)
UM = 0.54 LM +0.46 LM_1
Tests have been given that enable one to determine whether a spectral peak departs
significantly from some specified base line, such as that expected from a flat or
'white" noise spectrum. The white noise spectrum is one in which all frequencies con-
tribute equally to the total variance of the series, and would be expected if a set of
random numbers were analyzed, for example. Figure 1 shows the spectral estimates
obtained from the analysis of 110 years of annual precipitation values for Copenhagen,
Denmark. Although these data were analyzed in connection with an interest in a possible
11-year sunspot cycle, there is no statistically significant peak near 11 years nor anywhere
in this spectrum.
Although power spectrum analysis has been found to be a valuable tool in the
study of time series, it is limited in application and often should be supplemented by
266 INTERPRETATION OF TRENDS AND CYCLES
-------
other types of analysis. Spectrum analysis discards phase information as well as the
details of any amplitude variation. Sometimes it may be desirable to recover this in-
formation to gain a little insight into what is going on in the original series. One way
of accomplishing this is by smoothing or filtering. In these techniques the original
COPENHAGEN PRECIPITATION SPECTRA
0.080
OL 0.060
HJ
0.040
UJ
n:
0.020
22 11 5.5 4 2
PERIOD LENGTH, years
Figure 1 Spectral Estimated Relative Power (Uh) of the Annual Precipitation Totals for
Copenhagen, Denmark.
series At is operated on by a "filtering function," or perhaps by several such functions.
These methods have been discussed by Holloway,4 Panofsky and Brier,3 and othefs. The
simplest commonly used method is to eliminate or reduce the amplitude of the short-
period fluctuations or "noise'7 by the use of running averages. This is a special case of
the general procedure of treating the observations xt in the time series by the following
linear equation
M
K n
. x t +K
where WK is a particular weight in the filtering function. The weight W0 is known as
the principal weight or the central weight when the filter is symmetrical with n = M.
In the process of filtering the time series, successive observations are cumulatively
multiplied by these weights, producing a new series beginning Ft, Ft ,
\-r,rr l-~ *L 1 _ ~T"
ing in this succession.
Ft I and continu-
A filter that reduces the amplitude of both the high- and low-frequency fluctuations,
leaving a middle range of frequencies relatively unaffected, is called a band-pass filter.
Such a filter is useful in studying the fluctuations of a particular time scale. For example,
Table 1 gives the weights used for a filter having the maximum sensitivity to fluctuations
Brier
267
-------
Table 1 Set Of Weights Used For Band-pass Filter
W-27
W-26
W-25
W-24
W-23
W-22
W-21
W-20
W-19
W-18
W-17
W-16
W-15
W-14
W-13
W-12
w-n
W-10
W-9
W-8
W-7
W-6
W-5
W-4
W-3
W-2
W-l
0.0100
0.0121
0.0140
0.0156
0.0165
0.0163
0.0147
0.0115
0.0065
0.0000
0.0079
0.0165
0.0253
0.0334
0.0400
0.0442
0.0454
0.0431
0.0372
0.0279
0.0158
0.0017
0.0133
0.0279
0.0410
0.0512
0.0577
W-0
W-l
W-2
W-3
W-4
W-5
W-6
W-7
W-8
W-9
W-10
W-ll
W-12
W-13
W-14
W-15
W-16
W-17
W-18
W-19
W-20
W-21
W-22
W-23
W-24
W-25
W-26
W-27
0.0602
0.0577
0.0512
0.0410
0.0279
0.0133
0.0017
0.0158
0.0279
0.0372
0.0431
0.0454
0.0442
0.0400
0.0334
0.0253
0.0165
0.0079
0.0000
0.0065
0.0115
0.0147
0.0163
0.0165
0.0156
0.0140
0.0121
0.0100
of about 25 data points. The weights were chosen in such a way that periods longer than
about 50 units of time, or shorter than about 13 units, would be eliminated. The actual
frequency response of this filter is shown in Figure 2. The ordinate of this curve gives
1.00
0.80 -
0.60 -
§. 0.40 -
in
0.20
0.0
_L
_L
_L
48 24 12 6
PERIOD
Figure 2 Frequency Response of a Band-pass Filter.
268
INTERPRETATION OF TRENDS AND CYCLES
-------
the ratio of the amplitude of a wave of a given frequency f in the time series after
filtering to the original amplitude of the wave before filtering. The frequency response
Rj of a filter is a function of frequency and is given by the formula
n
Rj = Wo + 2 Y WK cos 2 TT fK
K= 1
where f is expressed in terms of cycles per data interval and ranges from 0 to 1/2.
If, on the other hand, Rj is specified, the weights WK can be determined by the
formula
n/2n
WK=R(0)+2 Y R(f) cos2Trf K,
f = %n
f = l/2n2/2n, 3/2n, ..., 1/2
Further details of these procedures can be found in Brier.3
The filter shown in Table 1 has been applied to over 200 years of monthly precipita-
tion data for England. A sample of the data is plotted in Figure 3; the corresponding
filtered output Ft is shown, on an amplified scale, in Figure 4. The purpose of this
study was to investigate whether there was any period of around 24 to 27 months that
maintained a constant amplitude or phase over the 200 years. The results were negative
or inconclusive. Another study was made to learn whether the peaks in Figure 4 (and
the data for the remaining years) tended to be more frequent or to have greater ampli-
tude during some calendar months than others. This was done by tabulating for each
peak the amount of the deviation above the zero line and the month of occurrence.
These were plotted in the polar diagram of Figure 5, which shows an essentially random
distribution. Thus there is no strong evidence of any period of around 24 or 26 months
that is phase-locked with the annual cycle.
220% -
160% -
100%
40% -
1727 1728 1729 1730 1731 1732 1733 1734 1735 1736
Figure 3 Sample Plot of Mean Monthly Precipitation Data for Group of English Stations.
Brier
269
-------
In the discussion on spectrum analysis it could have been pointed out that this
method of analysis is appropriate when the contributions to the total variance result
from a continuous spectrum of frequencies. If there are lines in the spectrum corres-
ponding to genuine periodic terms, then it is usually considered desirable to remove
their effects. The difficult problem may be to determine whether there is a line. If the
amplitude of a true periodic component is small, the line may be hidden in the noise
1727 1728 1729 1730 1731 1732 1733 1734 1735 1736
Figure 4 Output of Filter Used on Monthly Precipitation Data for Group of English
Stations.
12
10
Figure 5 Distribution of Amplitude and Phase of Peaks in Filtered Series of English
Precipitation Data.
270
INTERPRETATION OF TRENDS AND CYCLES
-------
and is not likely to be detected by spectral analysis. This is not the place to discuss
this problem in detail, but one suggestion can be made. If a true period of length p
exists in the data, then it should persist through the entire time series without any
significant change in phase. If the entire record is broken into two equal parts, a Buys-
Ballot table can be constructed for the first half and second half of the record independ-
ently. If, for example, one is interested in examining the time series for a period of
27 days, the data are arranged in 27 columns with day 1, 28, 55, etc. placed in the
first column. Days 2, 29, 56 ... etc. are placed in the second column, and this procedure
is followed until the data are all used and the averages for each column determined.
Figure 6 shows the results of using this procedure for 50 years of precipitation data
o
25 -
20 ~
0.20 0.40 0.60 0.80
DECIMAL FRACTION OF 27 DAYS
1.00
Figure 6 Column Means from Buys-Ballot Table for 27-Day Trial Period. Curve A,
U. S. Precipitation Data 1900-1924; Curve B, U. S. Precipitation Data 1925-1949.
for 1544 weather stations in the United States. Use of a high-speed electronic computer
for computation makes it more convenient to plot the data in terms of the decimal fraction
of the period being examined. In this diagram, there is little or no resemblance between
the curves for the two independent periods, the correlation being rAB = 0.05. With
the computer it was practical to examine periods from 27.000 days to 31.000 days by
intervals of 0.005 day. The highest correlation between A and B was found for a period
of 29.530 days (rAB = 0.71), which corresponds to the lunar synodic period. Figure
7 shows a partial plot of these results, which confirm the findings of Bradley et al.2
Brier
271
-------
Although many additional operations can be applied to the analysis of time series,
the main point I would like to make is that the modern high-speed computer enables one
to do a great many things economically and in much greater detail than would have been
considered possible (or even desirable) as little as 5 years ago.
1.0
29.250 29.300
29.700 29.750
PERIOD, days
Figure 7 Correlation Coefficients Between Column Means for Two Independent Time
Periods for Trial Periods Extending from 29.250 Days to 29.750 Days.
REFERENCES
1. Blackman, R. B. and J. W. Tukey. 1958. The measurement of power spectra from
the point of view of communications engineering. Bell System Tech. J. 37: 185-282,
485-569.
2. Bradley, D. A., Woodbury, M. A., Brier, G. W. 1962. "Lunar Synodical Period and
Widespread Precipitation." Science 137: 748-749.
3. Brier, G. W. 1961. "Some Statistical Aspects of Long-Term Fluctuations in Solar
and Atmospheric Phenomena." Annals of the New York Academy of Sciences. 95:
173-187.
4. Holloway, J. L., Jr. 1958. Smoothing and filtering of time series and space fields.
Advances in Geophysics. IV: 351-389. Academic Press, New York, N. Y.
5. Jenkins, G. M. 1961. "General Considerations in the Analysis of Spectra." Techno-
metrics. 3: 133-190.
6. Mitchell, J. M., Jr. 1963. "Some Practical Considerations in the Analysis of Geo-
physical Time Series." (to be published).
7. Panofsky, H. A. and G. W. Brier. 1958. Some Application of Statistics to Meteorology.
Pa. State Univ., University Park, Pa.
8. Tukey, J. W. 1949. The sampling theory of power spectrum estimates. Symposium
on Applications of Autocorrelation Analysis to Physical Problems. "Woods Hole,
Mass. pp. 47-68.
9. Ward, F. W., Jr. 1960. The variance (power) spectra of Q, K , and A . J. Geophys.
Research. 65: 2359-2373. P
10. Wiener, N. 1930. Generalized harmonic analysis. Acta Math. 55:117-258.
11. Wiener, N. 1949. Extrapolation, Interpolation, and Smoothing of Stationary Time
Series. Technology Press of M.I.T., Cambridge, Mass.
272
INTERPRETATION OF TRENDS AND CYCLES
-------
Dr. L. D. Zeidberg
and
Emanuel Landau
School of Medicine
Department of Preventive Medicine
and Public Health
Vanderbilt University, Nashville, Tennessee
SUMMARY
The hypotheses on which research is based must be limited in scope so that a
measurable aspect of a problem can be defined with precision. Ultimately a chain of
proved subsidiary hypotheses may serve to validate the major program objective, which
is usually based on a major hypothesis.
In the Nashville Air Pollution Study the broad objective was to determine whether
health is adversely affected by air pollution. It was postulated that health is affected
and that the effects of air pollution are measurable. Four studies were designed; two
of these are described in detail to show how various hypotheses were developed and
tested, what conclusions were drawn, and what further avenues of research were opened
as a result of these studies.
DATA INTERPRETATION DRAWING CONCLUSIONS
INTRODUCTION
Research generally is based on the development and the testing of hypotheses. This
is not only the scientific approach, but is also part of the accepted epidemiologic method.
Hypotheses usually have some basis in already established facts. The assembling of
such facts is a necessary prelude to the development of hypotheses. Once developed,
they must be subjected to searching tests made with scientific objectivity. Testing
generally involves the collection and interpretation of data, from which conclusions
may be drawn that will either validate the hypotheses or negate them.
In order to define a measurable aspect of a. problem with precision, hypotheses must
of necessity be limited in scope. It may be necessary to develop and test a whole chain
of subsidiary but related hypotheses in order lo marshall the data required to validate
the major program objective. In all of this, data interpretation is a key step that leads
ultimately to conclusions. It is well to recall the often-quoted words of Frost:
"Epidemiology at any given time is something more than the total of its established
facts. It includes their orderly arrangement into chains of inference which extend more
or less beyond the bounds of direct observation."1
Two of the air pollution studies2' 3 conducted in Nashville, Tennessee, under a
contract with the Air Pollution Division of the Public Health Service* will be examined
to illustrate how the results of the analyzed data were related to initial hypotheses,
what conclusions were drawn, and what new hypotheses were formulated.
PULMONARY ANTHRACOSIS AS AN INDEX OF
AIR POLLUTION
Pulmonary anthracosis is a condition of the lungs in which black pigment is
Zeidberg
-------
deposited as a result of the inhalation of particles of coal dust, and perhaps of other
dusts. Pathologists in Nashville had gained an impression, not tested, however, by
definitive studies, that Nashville residents had more such pigment in their lungs than
did non-residents. If this were true, it should be possible to use the pigmentation of
the lung as an index of air pollution due to combustion of coal in a community.
Several hypotheses were advanced, and a plan was devised to test them. The hypotheses
were:
1. Anthracosis in the lungs of Nashville residents is directly related to air
pollution in Nashville.
2. The degree of anthrocosis will vary among Nashville city and out-of-city
residents.
3. Among Nashville residents the degree of anthracosis will vary depending on
the length of residence in the city.
4. Occupational exposure to coal dust may be a factor that affects the degree
of anthracosis, but not to the exclusion of other exposure.
5. Anthracosis is a cause of ill health.
(a) Anthracosis is associated with specific symptoms; and
(b) Anthracosis is related to the occurrence of cardiorespiratory disease.
To test these hypotheses a series of consecutive autopsies (except those in subjects
under 5 years of age) done at Vanderbilt University Hospital between 1953 and 1956
was studied. The degree of anthracosis in the lungs of 641 subjects was evaluated ac-
cording to standards established by Dr. John Shapiro, professor of Pathology at
Vanderbilt University School of Medicine. The lungs were classified as showing no
pigment, or showing minimal, moderate, or severe anthracosis. The residence of each
subject at the time of death was the determining factor in designating him as a Nash-
ville or out-of-city resident. For persons in the first category, city directories were
searched at 5-year intervals to establish how long they had lived in the city, and in
what part of it. The residential data were then converted into a classification of exposure
to low, moderate, or high pollution, on the basis of aerometric data collected in the
engineering phase of the study.5 For those who were out-of-city residents at the time
of death, it was assumed that they had never lived in Nashville. For city residents
occupational data also were sought by reference to city directories. The data con-
cerning symptomatology and pathology were obtained from the hospital records and
the autopsy protocols, respectively.
In development of the plan of the study it was necessary to make certain assumptions,
the validity of which may be questioned. It was assumed, for example, that out-of-city
residents had always lived out of the city, as noted above. In comparison of Nashville
and non-Nashville residents it was assumed that out-of-city dwellers did not live in a
city and were not exposed to as much coal smoke as residents of Nashville. This assump-
tion would tend to minimize differences in anthracosis in the two groups, and therefore
would make real differences even more significant.
Another major assumption was that the level of air pollution in any specific area
of the city in the period from 1958 to 1959 was indicative of that prevailing as long as
20 years ago. Although abundant evidence indicated that the pollution situation in
Nashville had improved considerably with the years, it was assumed that the improve-
ment had been proportionately comparable in different parts of the city.
274 DATA INTERPRETATION (AIR)
-------
The bias inherent in autopsy material would ordinarily make it extremely hazardous
to extrapolate or apply such data to a general population. It was assumed, however,
that anthracosis would not necessarily bring people to Vanderbilt University Hospital,
or prove fatal, or predispose them to post-mortem examination.
In testing of the hypotheses that had been formulated at the outset, it was important
to exercise extreme care to avoid the introduction of bias. Consequently this was con-
ducted as a "blind" study, with each phase entrusted to a different investigator, who
worked independently, so that the one responsible for classifying the degree of anthra-
cosis in the lungs of subjects, for example, identified them only by number and had no
knowledge of their age, sex, race, residence, or occupation. The importance of such
safeguards cannot be emphasized enough. Where absolute objectivity is essential, as it
must be in any scientific investigation, even the slightest bias could destroy the validity
of interpretations and conclusions.
In testing of the first hypothesis, that anthracosis is directly related to the level
of air pollution in Nashville, multiple regression techniques6 were used. The relationship
between residential exposure to different levels of air pollution and the amount of
anthracosis in the lungs was measured, with age considered as a variable. Analyses
were done separately for males and females. For males, the multiple correlation co-
efficient was 0.489, indicating a highly significant relationship. While the individual
regression coefficients for residence alone, not considering age, were not significantly
different from zero, there was a direct relationship between degree of anthracosis and
degree of exposure to air pollutants. For females, the multiple correlation coefficient and
the individual regression coefficients for residence were highly significant. One can speculate
that the female, who is apt to be more closely related to the residential environment
than is the male, reflects a reaction to the home environment, uninfluenced by an
occupational exposure. It may be concluded, therefore, that the first hypothesis may
have reasonable validity.
The second hypothesis, that the degree of anthracosis varies in Nashville and out-
of-city residents, was tested by comparing the degree of anthracosis for both groups, by
age. These comparisons are shown in Figure 1. Above the age of 25, Nashville residents
showed a consistently higher level of severe anthracosis. Thus, the second hypothesis
appears to be valid also.
For the third hypothesis, that in Nashville residents the degree of anthracosis varies
with the length of residence in the city, the analysis was limited to the 466 subjects
who were over 45 years of age, because age could be a limiting factor. The subjects
were divided into two groups: those with residence less than 20 years, and those with
residence of 20 years or more. Figure 2 shows this comparison and indicates that the
longer-term residents had more anthracosis. Very few of those who had lived in the
city for more than 20 years showed minimal anthracosis. Figure 2 also shows the
comparative degree of anthracosis among out-of-city residents, but this merely another
way of illustrating the data in Figure 1. The third hypothesis may be considered
valid also.
The fourth hypothesis suggested that occupational exposure to coal dust affecting the
degree of pulmonary anthracosis would not exclude the effect of other factors, such as
the residential environment. It was not possible to put this hypothesis to the test because
of the insufficiency of occupational data. The city directories that were searched for
occupational listing of the 329 Nashville residents in the autopsy group were unexpectedly
deficient in this regard. For 129 females, many of them housewives, data were totally
Zeidberg 275
-------
lacking. An additional 17 were of the younger age group, and unlikely to have been
employed. The information obtained was believed to be too meager for interpretation.
Here is an illustration of the limitations of retrospective studies, which are periorce
limited qualitatively and quantitatively to already recorded data in the absence of
additional followup activity.
100-
80-
,_ 60-
2
Ul
o
(£
UJ
CL
40-
20-
0 -
RESIDENCE
AGE
I
~
C
TO'
?
#
%
w
K
0
PAL
y
c
5-
%
0
24
\
C
25
1
0
-44
|
C
45
O
-64
'
=
I~
C
65
\
//
//
*<
V.
o
+
ANTHRACOSIS
NONE TO
MINIMAL
Y/\ MODERATE
C = NASHVILLE 0 = OUT-OF-CITY
SEVERE
Figure 1 Age and Residence Differences in the Degree of Anthracosis Found at
Autopsy in 641 Individuals, Vanderbilt University Hospital, 1953-56.
80
60 -
40-
20-
ANTHRACOSIS
0-1 NONE TO MINIMAL
2 MODERATE
3 SEVERE
0-1 2 3
OUT-OF-CITY
0-1 23 0-123
NASHVILLE<20 YR NASHVILLE>20 YR
Figure 2 Residence Differences in the Degree of Anfhracosis Found at Autopsy in 466
Individuals 45 Years of Age and Over, Vanderbilt University Hospital, 1953-56.
276
DATA INTERPRETATION (AIR)
-------
The final hypothesis, related to anthrocosis as a cause of ill health, was formulated
in two parts. The first of these was that anthracosis is associated with specific symptoms.
To test this subhypothesis, the hospital records of each subject were studied carefully
for symptomatology relating specifically to the cardiorespiratory system. Only the
560 white subjects in the autopsy group were included in this analysis. It was con-
cluded that anthracosis was not characterized by specific cardiorespiratory symptoms.
The second part of the hypothesis suggested that anthracosis is related to the occurrence
of cardiorespiratory disease. The data to test this hypothesis were obtained from hospital
records and autopsy protocols. No specific disease could be related to anthracosis. One
of two possible conclusions may be reached: either there is in fact no related symptom-
atology or pathology, or the data obtained are not sufficently accurate. The first of these
conclusions has support among some pathologists7' 8 but not among others.9'10 It is
possible that our data are faulty and do not reveal pulmonary disease that was actually
present. British investigators report the presence of a focal emphysema, or dilatation of
air spaces in the lungs, associated with anthracosis.11-13 In their studies they inflated the
collapsed lungs to their normal size at autopsy. Our studies are based on small sections
of collapsed lung, in which it would not be possible to observe the changes described
by the British. A recent study in this country, however, refutes the British work and
claims that focal emphysema is the cause, rather than the result, of deposits of anthra-
cotic pigment in the lungs.1* There the matter stands, and we may conclude that we
have not been able to put this last hypothesis to an adequate test. To obtain an answer
it will be necessary to set up new hypotheses and develop studies to test them. For
example, it might be postulated that anthracosis follows rather than precedes the
development of centrilobular emphysema. Retrospective studies could be planned if
inflated lung specimens were available. If not, prospective studies would have to be
done, and these might require a considerable period of time.
From our study of pulmonary anthracosis we may conclude that anthracosis in
the lungs of Nashville residents is a fairly good index of the degree of air pollution to
which they have been exposed during their residence, but we are unable to show that
such deposits were necessarily injurious to their health.
MORBIDITY IN RELATION TO AIR POLLUTION
As further illustration of how research data may be interpreted and conclusions
drawn, some of the features of a morbidity survey conducted in Nashville will be dis-
cussed. The survey was part of a general study of the health effects of air pollution.
The following hypotheses were formulated:
1. The morbidity experience of Nashville residents is directly related to the
level of air pollution in their environment.
2. Illness due to specific causes, particularly respiratory and cardiovascular,
will vary according to the levels of air pollution in different areas.
3. Specific age groups will be affected differently.
4. Occupational exposure will affect the occurrence of illness, but not to the
exclusion of other exposures.
To test these hypotheses a survey of a representative sample of the population was
planned, by means of direct interview of a responsible adult in each of the selected
households. Morbidity in the middle socio-economic class was analyzed because members
Zeidberg 277
-------
of this class were found in all levels of pollution. Further, they comprised the largest
group in the surveyed population.
The testing of the hypotheses did not produce general validation. For the first
hypothesis, that morbidity is directly related to air pollution exposure, no regular pattern
could be shown for any of the four pollutants studied. For white residents over 55
years of age, however, in whom the effects of prolonged exposure to air pollution might
be expectedly most pronounced, a consistent pattern of increasing morbidity with in-
creasing exposure to air pollutants was observed when the soiling index and 24-hour
SO concentration were used as indexes (Figure 3). Because too few of the non-whites
lived in areas of low pollution, comparisons were made between residents of high and
moderate pollution areas only. These comparisons showed the same patterns for non-
white females as observed for the white residents, but for the non-white males only the
soiling index showed a significant correlation (Figure 4).
SULFATION
SOILING
INDEX
HI-VOL
PARTICULATE
24-HOUR SO,
SULFATION
SOILING
INDEX
HI-VOL
PARTICULATE
24-HOUR SO,
WHITE MALE
^
WHITE FEMALE
0 20 40 60 80 100 120 140 160 180 200
PERCENT OF ILLNESS
n 1 1 1 1 1 1
J
HIGH
POLLUTION
MODERATE
POLLUTION
r-j LOW
1' POLLUTION
Figure 3 Percent of Illness for All Causes During the Year Prior to the Survey Among White
Middle Class Individuals 55 Years of Age and Over, by Sex and by Degree of Exposure to
Atmospheric Pollutants. Nashville Air Pollution Study.
For the second hypothesis, that morbidity for specific causes (such as respiratory
and cardiovascular disease) is directly related to air pollution exposure, partial valida-
tion could be shown. No correlation could be shown for respiratory illness, cancer, or
gastrointestinal disease; cardiovascular morbidity increased with exposure to particu-
lates measured by the soiling index and to SO2 for white males, while for white females
direct associations with all four pollutants were observed (Figure 5). Among non-whites
no consistent pattern could be shown.
For the third hypothesis, that specific age groups will be affected differently by
exposure to air pollutants, strong support was evidenced. Only for those over 55 years
of age could any pattern of relationship between morbidity and air pollution be shown.
At this point a new hypothesis might be advanced, that the effects of usual exposure
to air pollution become manifest only after prolonged experience.
278
DATA INTERPRETATION (AIR)
-------
SULFATION
SOILING
INDEX
24-HR SO,
SULFATION
SOILING
INDEX
24-HR SO,
NON-WHITE MALE
NON-WHITE FEMALE
i
i
i
i
20 40 60 80 100
PERCENT OF ILLNESS
120
140
HIGH POLLUTION
MODERATE AND
LOW POLLUTION
Figure 4 Percent of llfness for all Causes During the Year Prior to the Survey Among
Non-White Middle Class Individuals 55 Years of Age and Over, by Sex and by Degree of
Exposure to Atmospheric Pollutants. Nashville Air Pollution Study.
SULFATION
SOILING
INDEX
HI-VOL
PARTICULATE
24-HOUR SO.
SULFATION
SOILING
INDEX
HI-VOL
PARTICULATE
24-HOUR SO-
WHITE MALE
a
WHITE FEMALE
iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiHimmmmiiiiiiiiiMMiiiiiiiiiiiiiiiii
i
I
i
I
10 20 30 40 50
PERCENT OF ILLNESS
60
70
I HIGH
POLLUTION
MODERATE
POLLUTION
LOW
POLLUTION
Figure 5 Percent of Cardiovascular Illness During the Year Prior to the Survey Among
White Middle Class Individuals 55 Years of Age and Over, by Sex and by Degree of Exposure
to Atmospheric Pollutants. Nashville Air Pollution Study.
Zeidberg
279
-------
The fourth hypothesis, that occupational exposure will afiect the occurrence of
illness, but not to the exclusion of other causes, may be advanced to explain the lack
of correlation noted above, particularly for non-whites. In comparisons based on pollu-
tion levels in the residential environment, no account is taken of the influence of occu-
pational environments. For workers, almost one-third of their exposure experience occurs
away from home. This may account for the lack of correlation between morbidity and
air pollution exposure. In order to refine the data, an analysis was made of the morbidity
experience of females 15 to 64 years of age, classified into two groups, working and
housekeeping. The latter would be expected to reflect more accurately the influence ol
the residential environment alone. Because subdivision of the data by socio-economic class
and by specific cause produced cells too small for analysis, only total morbidity was
considered, and rates were adjusted for age. Morbidity rates were higher in general
for the white housekeeping females than for the working females. The former groups
showed a direct relationship between morbidity and level of pollution for all pollutants
except soiling index. For this pollution index the morbidity rates were highest in the high
pollution areas, but no difference could be shown in morbidity rates for moderate and
low exposure (Figure 6) . For the white working females, none of the pollutants showed
SULFATION
SOILING
INDEX
WHITE WORKING FEMALES
iiiiiiiiiimiiiiiiiiimiiiiimmiiiiiiiimiiimimiiiiiiiiiiiiiiiiiiiiiiiiiiiii
iiminmiiiiiiiiiimimiiimiiiiiiiimmiiiimiiiiimimi
HI-VOL
PARTICULATE
24-HR SO.
SULFATION
SOILING
INDEX
HI-VOL
PARTICULATE
24-HR SO,
\\\\\\x\\\\\\\\x\\\\\\\i
WHITE HOUSEKEEPING FEMALES
I
I
I
20 40 60 80 100 120
PERCENT OF ILLNESS
140
160
HIGH
POLLUTION
MODERATE
POLLUTION
f ' 1 LOW
«' '"' POLLUTION
Figure 6 Age-Adjusted Morbidity Rate for All Causes for the Year Prior to the Survey
Among White Working and Housekeeping Females 15-64 Years of Age, by Exposure to
Atmospheric Pollutants. Nashville Air Pollution Survey.
any correlation with morbidity experience. For the non-whites, those few who lived in
areas of low pollution were combined with the moderate pollution group for com-
parison with those in high pollution areas. Both the working and housekeeping non-
white females showed similar patterns wherever numbers were large enough to allow age
280
DATA INTERPRETATION (AIR)
-------
adjustment (Figure 7). Since the non-white female is often employed as a domestic, her
occupational environment is not likely to be in the commercial and industrial areas of
the city where pollution is greatest. Since her occupational environment may not in-
fluence her experience significantly, there may be no difference between her morbidity
experience and that of the housekeeping non-white.
NON-WHITE WORKING FEMALES
SULFATION
SOILING INDEX
24-HR SO,
SULFATION
SOILING INDEX
24-HR SO,
NON-WHITE HOUSEKEEPING FEMALES
lllHlllllllllilllllllllllllllimillllllllllllllllllllllllllll
20 40 60
PERCENT OF ILLNESS
80
100
HIGH
POLLUTION
MODERATE AND
LOW POLLUTION
Figure 7 Age-Adjusted Morbidity Rate for all Causes for the Year Prior to the Survey
Among Non-White Working and Housekeeping Females 15-64 Years of Age, by Exposure to
Atmospheric Pollutants. Nashville Air Pollution Survey.
In interpreting the results of an investigation great care must be taken to discover
hidden bias or error that may seriously influence the data. For example, it was well
known from other studies15-18 and from our own experience in a pre-test of the ques-
tionnaire used in these studies that the interview method does not provide accurate quali-
tative or quantitative measures of the prevalence of illness. Since we were primarily
concerned with the relative frequency of illness in areas with differing levels of air
pollution however, it was assumed that biases or errors of response would not vary
Zeidberg
281
-------
with differences in exposure to air pollutants. Analysis of the data showed that this
assumption was grossly invalid. For, although there is nothing about exposure to high or
love levels of pollution that should produce differences in awareness or reporting of
illness, there was a marked relationship between socio-economic status and exposure
to pollutants in the residential environment. In general, the lower socio-economic classes
lived in the most polluted areas. The socio-economic influence on such factors as educa-
tion, occupation, and utilization of medical services is too well known to require elabo-
ration. These factors could seriously affect the reporting of illness, and may explain
the unexpected finding of more illness reported in upper-class children under 15
years of age than in middle-class children, and even more than in the lower-class
children. To minimize or eliminate possible bias, comparisons of morbidity in relation
to air pollution should be made by holding the socio-economic factor constant.
The final conclusion reached by the interpretation of the morbidity study data is
that the hypotheses advanced have been, at best, only partially validated. This is
rather inconclusive and may mean either that there is no true relationship between
exposure to air pollutants and illness or that the morbidity data obtained are not
sufficiently reliable. It is difficult to accept the first of the two possibilities in the
light of what is known about the effects of exposure to high concentrations of pollutants
such as has been reported in London10 and in Donora, Pennsylvania.20 To the authors,
the second possibility seems to be more plausible. This means that more accurate
methods for measuring morbidity in a community must be employed in studies of this
kind. It should be observed that no note of suspicion concerning the reliability of the
aerometric measurements has been sounded.
The authors have drawn on two of the four phases of the Nashville Air Pollution
Study to illustrate how hypotheses are established and tested, and how new hypotheses
are then promulgated for further testing.
REFERENCES
1. Frost, "Wade Hampton. In the Introduction to Snow on Cholera, New York, The
Commonwealth Fund, 1936.
2. Zeidberg, L. D., and Prindle, R. A. The Nashville Air Pollution Study: II. Pulmonary
Anthracosis as an Index of Air Pollution. Am. J. Public Health, 53:185-199,
Feb. 1963.
3. Zeidberg, L. D., Prindle, R. A., and Landau, E. The Nashville Air Pollution Study:
III. Morbidity in Relation to Air Pollution. Presented before the Epidemiology Sec-
tion, at the 90th Annual Meeting of the American Public Health Association in
Miami Beach, Florida, October 17, 1962. Accepted for publication in the American
Journal of Public Health and scheduled to appear in the December, 1963, issue.
4. U. S. Department of Health, Education, and "Welfare, Public Health Service, Air
Pollution Medical Program Contracts No. SAph 68348 and 69628.
5. Zeidberg, L. D., Schueneman, J. J., Humphrey, P. A., and Prindle, R. A. Air
Pollution and Health: General Description of a Study in Nashville, Tennessee. J.
Air Poll. Control Assoc., 77:289-297, June, 1961.
6. Snedecor, G. W. Statistical Methods. Iowa State College Press, 1956.
DATA INTERPRETATION (AIR)
814-105-10
-------
7. Anderson, W. A. D. Pathology (3rd edition). St. Louis, Missouri, Mosby, 1957.
p. 650.
8. Boyd, W. Textbook of Pathology (6th edition). Philadelphia, Pa., Lea and Febiger,
1953. p. 393.
9. Saphir, 0. (Editor). A Textbook of Systemic Pathology. New York and London,
Grune and Stratton, 1958, Vol. I, pp. 323-4.
10. Moore, R. A. A Textbook of Pathology (2nd edition). Philadelphia, Pa., Saunders,
1951, pp. 512-3.
11. Heppelston, A. G. Essential Lesion of Pneumokoniosis in Welsh Coal Miners. J.
Path. & Bact., 59:453-460, July, 1947.
12. Gough, J., James, W. R. L., and Wentworth, J. E. Comparison of Radiological and
Pathological Changes in Coalworkers Pneumoconiosis. J. Fac. Radiologists. 1,
1:28-29, July, 1949.
13. Oderr, C. P. Emphysema, Soot and Pulmonary Circulation-Macroscopic Studies of
Aging Lungs. JAMA, 772:1991-1998, April 30, 1960.
14. Pratt, P. C., Jutabha, P., and Klugh, G. A. The Relationship Between Pigment
Deposits and Lesions in Normal and Centrilobular Emphysematous Lungs. Am. Rev.
Resp. Dis., 87:245-256. February, 1963.
15. Cobb, S., Thompson, D. J., Rosenbaum, J., Warren, J. E., and Merchant, W. R. On
the Measurement of Prevalence of Arthritis and Rheumatism from Interview Data.
J. Dis. of Children, 3:134-139, Feb., 1956.
16. Trussell, R. E., Elinson, J., and Levin, M. Comparison of Various Methods of
Estimating the Prevalence of Chronic Disease in a Community the Hunterdon
County Study. Am. J. Pub. Health, 46:173-182, Feb., 1956.
17. Krueger, D. E. Measurement of Prevalence of Chronic Disease by Household Inter-
views and Clinical Evaluations. Am. J. Pub. Health, 47:953-960, Aug., 1957.
18. Lilienfeld, A. M., and Graham, S. Validity of Determining Circumcision Status by
Questionnaire as Related to Epidemiological Studies of Cancer of the Cervix. J.
Nat. Cancer List., 27:713-720, Oct., 1958.
19. Committee on Air Pollution: Interim Report, Her Majesty's Printing Office, Lon-
don, 1956.
20. Schrenk, H. H., Heimann, H., Clayton, G. D., Gafafer, W. M., and Wexler, H.
Air Pollution in Donora, Epidemiology of the Unusual Smog Episode of October,
1948: Preliminary Report. Public Health Bulletin No. 306, 1949.
Zeidberg 283
-------
DISCUSSION: INTERPRETATIONS AND CONCLUSIONS
PANEL MEMBERS
G. J. Taylor
Assistant Chief, Bureau of Air Sanitation
California Dept. of Health, Berkeley
Robert A. McCormick
Chief, Meteorological Section, DAP
U. S. Public Health Service, Cincinnati
Dr. R. O. McCaldin
Deputy Chief, Field Studies Branch, DAP
U. S. Public Health Service, Cincinnati
Dr. Harold J. Paulus
Associate Professor of Public Health
University of Minnesota, Minneapolis
Mr. McCormick noted the necessity for extreme caution in research, the conscious
thinking through of any problem before beginning work, and the importance of controls
and auxiliary tests to reduce the number of possible interpretations of experimental
results. He quoted the statement, attributed to Darwin, that "nature will tell you a
direct lie if she can."
The group was asked for a prediction as to when, if ever, air quality measurements
would be as complete and representative as meteorological measurements are presently.
Mr. McCormick replied that meteorological data are not as representative as commonly
thought and that the CAMP stations provide data as representative as meteorological
data, although at a cost tremendously higher than wind and temperature instrument
installations. He noted that the Weather Bureaus receive a large number of inquiries
from the public, and predicted better air quality data when public interest in air pollution
becomes comparable.
Mr. Taylor commented that the amount of air monitoring usually is proportional to
the magnitude of the problem and the level of public recognition. At present, needs
for air pollution data are not similar throughout the nation, as are the requirements for
meteorological measurement; nor are air pollution situations usually the same over vast
areas, as are meteorological phenomena. Air quality measurements in the near future
will still be confined to problem areas, although offering much better coverage of these
areas. Mr. Ljnsky disagreed with Mr. Taylor and observed that the requirements for
meteorological measurements are as broad and varied as are those for air pollution data.
He expressed concern over the seeming lack of interest in interferences that prevent
the measurement of what is intended to be measured, particualrly in air quality measure-
ments, and cautioned against blind acceptance of routine data.
Mr. Bryan asked how long air quality measurements must be made to permit time
series analyses as discussed by Mr. Brier. Mr. McCormick replied that formulas are
available that describe precisely the data required to yield a given precision in series
analysis. Dr. McCaldin noted that economics usually dictate in the air pollution field:
the decision generally is based on how much can we afford, rather than how much
we need.
Discussion 285
-------
Mr. Gruber asked what natural geophysical cycles influence air quality, and how
long we must sample to include the effect of the most important of these climatological
cycles. Mr. Bellamy pointed out that on the power spectra displayed by Mr. Brier the
longest cycles were the largest, and that this would seem to indicate the necessity for
continuous sampling on a permanent basis. He noted that continuous detailed data
will become more readily available when air pollution control agencies use them in their
operations in the same manner that airports presently use meteorological data. Mr.
Schueneman commented that probability techniques similiar to those used by hydrologists
might be useful when records of air quality over long periods become available.
The participants were asked how far one can deviate from an ideal sampling situa-
tion in the interest of practicality without compromising the data obtained. Mr. Taylor
observed that water, sewerage, electric power, land, and economic requirements severely
limit sampling sites, particularly for large installations such as CAMP stations. He
pointed out, however, that present aid quality monitoring is generally not aimed at
research, but at assessing the air pollution situation and getting information to form
research hypotheses, and hence concern should be directed not only at deviations from
ideal sampling but also at deviations from program objectives. Mr. Gruber pointed out
that different contaminants demand different criteria in selection of sampling locations;
for example, dustfall is more location-dependent than soiling index because of the larger,
rapidly settling particles involved. Mr. Schueneman suggested screening various loca-
tions with simple portable samplers to determine location-dependence before selecting
sites. He also pointed out that one must consider limitations in site selection, sampling,
and analysis when determining the degree of confidence in the results. Mr. Linsky added
that sampler location can often be guided by program objectives, i.e., characterization
of over-all air quality, evaluation of large single sources, or consideration of many small
sources. He reminded the group that although many air pollution "sensors" measure
only an effect of air pollution rather than physical or chemical quantities, these measure-
ments are valid and should be encouraged.
Dr. Zeidberg mentioned that in the PHS Nashville study an attempt was made to
determine the minimum sampling necessary to characterize the air, but the results may
not be applicable beyond the Nashville area. He warned against injudicious drawing on
the experience of others in'such matters.
286 INTERPRETATIONS AND CONCLUSIONS
-------
SESSION 8: Measurements of Water Environment
Chairman: Maurice LeBosquet
Office of the Chief
Division of Water Supply and Pollution Control
U. S. Public Health Service
-------
Dr. H. B. N. Hynes*
Department of Zoology
University of Liverpool, England
SUMMARY
The complexities of biological reactions to water conditions are reviewed, as well as
the problems in presentation of biological results in a form that can be readily understood
by workers in other disciplines. The problems of biological sampling are also considered.
It is concluded that no system of presentation of results so far proposed is really
satisfactory, and any system that does not include tabulated raw data is concealing
information that should be recorded.
Interpretation of biological data requires considerable training and must be left
to biologists, who also must design the sampling program for each particular situation.
Biology, like medicine, is too complex a subject to codify, but it is an essential tool
in the full evaluation of water quality.
THE INTERPRETATION OF BIOLOGICAL DATA
WITH REFERENCE TO WATER QUALITY
Sanitary engineers like to have data presented to them in a readily assimilable form
and some of them seem a little impatient with biologists who appear unable to provide
definite quantitative criteria applicable to all kinds of water conditions. I think the
feeling tends to be that this is the fault of biologists, and if they would only pull
themselves out of the scientific stone-age all would be well. I will try to explain here
why I believe that biological data can never be absolute nor interpretable without a
certain amount of expertise. In this respect biologists resemble medical men who make
their diagnoses against a complex background of detailed knowledge. Anyone can
diagnose an open wound but it takes a doctor to identify an obscure disease; and al-
though he can explain how he does it he cannot pass on his knowledge in that one
explanation. Similarly, one does not need an expert to recognize gross organic pollu-
tion, but only a biologist can interpret more subtle biological conditions in a water
body; and here again he can explain how he does it, but that does not make his hearer a
biologist. Beck (1957) said something similar at a previous symposium in this city in 1956.
THE COMPLEXITY OF BIOLOGICAL REACTIONS TO
WATER CONDITIONS
The aquatic habitat is complex and consists not only of water but of the substrata
beneath it, which may be only indirectly influenced by the quality of the water. More-
over, in biological terms, water quality includes such features as rate of flow and temper-
ature regime, which are not considered of direct importance by the chemist. To many
animals and plants maximum summer temperature or maximum rate of flow is just as
important as minimum oxygen tension. The result is that inland waters provide an
enormous array of different combinations of conditions, each of which has its own
community of plants and animals; and the variety of species involved is very great.
Thus, for example, Germany has about 6000 species of aquatic animals (lilies 1961a)
* Now Chairman, Department of Biology, University of Waterloo, Ontario, Canada.
Hynes 289
-------
and probably at least as many species of plants. Yet Europe has a rather restricted
fauna because of the Pleistocene ice age; in most other parts of the world the flora and
fauna are even richer.
We know something about the way in which species are distributed in the various
habitats, especially in the relatively much studied continent of Europe, but we have,
as yet. little idea as to what factors or combination of factors actually control the indi-
vidual species. Thus, it is possible to list the groups of organisms that occur in swift
stony upland rivers (rhithron in the sense of lilies, 1961b) and to contrast them with
those of the lower sluggish reaches (potamon). Similarly we know, more or less, the
different floras and faunas we can expect in infertile (oligotrophic) and fertile
(eutrophic) lakes. We are, however, much less informed as to just what ecological
factors cause these differences. We know they include temperature and its yearly
amplitude; oxygen, particularly at minimal levels; plant nutrients, such as nitrate,
phosphate, silica, and bicarbonate; other ions in solution, including calcium, chloride,
and possibly hydrogen; dissolved organic matter, which is necessary for some bacteria
and fungi and probably for some algae; the nature of the substratum; and current.
We also know these factors can interact in a complex manner and that their action on
any particular organism can be indirect through other members of the biota. Thus, for
example, heavy growths of encrusting algae induced by large amounts of plant nutrients,
or of bacteria induced by ample supplies of organic matter, can eliminate or decimate
populations of lithophile insects by simple mechanical interference. But the change
does not stop there: the growths themselves provide habitats for the animals, such as
Chironomidae and Naidid worms, which could not otherwise live on the stones. Similarly,
if oxygen conditions over a muddy bottom reach levels just low enough to be intolerable
to leeches, Tubificid worms, which the leeches normally hold in check, are able to
build up to enormous numbers especially as some of their competitors (e.g. Chironomus)
are also eliminated. One then finds the typical outburst of sludge worms, so often cited
as indicators of pollution. This does not happen if the same oxygen tension occurs over
sand or rock, however, as these are not suitable substrata for the worms. Many such
examples could be given, but they would only be ones we understand; there must be
a far greater number about which we know nothing. One must conclude, therefore,
that quite simple chemical changes can produce far-reaching biological effects; that
we only understand a small proportion of them; and that they are not always the same.
This seems like a note of despair, however, if water quality deviates too far from
normal, the effects are immediately apparent. Thus, poisonous substances eliminate many
species and may leave no animals (Hynes 1960) ; excessive quantities of salt remove
all leeches, amphipods, and most insects and leave a fauna consisting largely of
Chironomidae, caddis worms, and oligochaetes (Albrecht 1954) ; and excessive amounts
of dissolved organic matter give rise to carpets of sewage fungus, which never occur
naturally. Here no great biological expertise is needed, and there is little difficulty in
the communication of results. It is when effects are slighter and more subtle that
biological findings become difficult to transmit intelligibly to other disciplines.
THE PROBLEMS IN PRESENTATION OF
BIOLOGICAL RESULTS
Because of these difficulties various attempts have been made to simplify the pres-
entation of biological findings, but to my mind none of them is very successful because
of the complexity of the subject. Early attempts at systematization developed almost
independently on the two sides of the Atlantic, although they had some similarities.
290 INTERPRETATION OF BIOLOGICAL DATA
-------
In America, there was a simple division into zones of pollution, e.g. degradation,
septic, and recovery, which were characterized in broad general terms. This simple, text-
book approach is summarized by Whipple et al. (1947), and serves fairly well for
categorizing gross organic pollution such as has been mentioned above. It was, however,
soon found by Richardson (1929) during his classical studies on the Illinois River that
typical "indicators" of foul conditions, such as Tubificidae and Chironomus, were not
always present where they would be expected to occur. This was an early indication
that it is not the water quality itself that provides suitable conditions for "pollution
faunas," but other, usually associated, conditions in this instance deposits of rich
organic mud. Such conditions may, in fact, be present in places where water quality
in no way resembles pollution, e.g., upstream of weirs in trout streams where autumn
leaves accumulate and decay and cause the development of biota typical of organically
polluted water. Samples must therefore be judged against a background of biological
knowledge. Richardson was fully aware of this and was in no doubt about the condition
of the Illinois River even in places where his samples showed few or no pollution indicators.
In Europe, the initial stress was primarily on microorganisms and results were
first codified in the early years of the century by Kolkwitz and Marsson. In this
"Saprobiensystem," zones of organic pollution similiar to those described by the American
workers were defined and organisms were listed as characteristic of one or more zones;
a recent exposition of this list is given by Kolkwitz (1950). It was then claimed that
with a list of the species occurring at a particular point it was possible to allocate it to
a saprobic zone. This system early met with criticism for several reasons. First, all the
organisms listed occurred in natural habitats they were not evolved in polluted water
and there was much doubt as to the placing of many of the species in the lists. The
system, however, did serve to codify ecological knowledge about a long list of species
along an extended trophic scale. Its weaknesses appeared to be merely due to lack of
knowledge; such a rigid system took far too little account of the complexity of the
reaction of organisms to their habitats. For instance, many organisms can be found,
albeit rarely, in a wide range of conditions and others may occur in restricted zones
for reasons that have nothing to do with water quality. We often do not know if
organisms confined to clean headwaters are kept there by high oxygen content, low
summer temperatures, or inability to compete with other species under other conditions.
In the swift waters of Switzerland the system broke down in that some organisms
appeared in more polluted zones than their position in the lists would indicate. Pre-
sumably here the controlling factor was oxygen, which was relatively plentiful in
turbulent cold water. In a recent series of experiments, Zimmerman (1962) has proved
that current alone has a great influence on the biota, and identically polluted water
flowing at different speeds produces biotic communities characteristic of different saprobic
levels. He finds this surprising, but to me it seems an expected result, for the reasons
given above.
Perhaps Zimmerman's surprise reflects the deeply rooted entrenchment of the
Saprobiensystem in Central Europe. Despite its obvious shortcomings it has been revised
and extended. Liebmann (1951) introduced the concept of considering number as well
as occurrence and very rightly pointed out that the community of organisms is what
matters rather than mere species lists. But he did not stress the importance of extrinsic
factors, such as current, nor that the system can only apply to organic pollution and
that different types of organic pollution differ in their effects; e.g., carbohydrate solu-
tions from paper works produce different results from those of sewage, as they contain
little nitrogen and very different suspended solids. Other workers (Sladecek 1961 and
references therein) have subdivided the more polluted zones, which now, instead of
Hyiies 291
-------
being merely descriptive, are considered to represent definite ranges of oxygen content,
BOD, sulphide, and even E. coli populations. Every water chemist knows that BOD and
oxygen content are nol directly related and to assume that either should be more than
vaguely related to the complexities of biological reactions seems to me to indicate a
fundamental lack of ecological understanding. I also think it is damaging to the hope of
mutual understanding between the various disciplines concerned with water quality to
give the impression thaL one can expect to find a close and rigid relationship between
water quality measurements as assessed by different sets of parameters. Inevitably these
relationships vary with local conditions; what applies in a sluggish river in summer will
certainly not apply to a mountain stream or even to the same river in the winter.
Correlation of data, even within one discipline, needs understanding, knowledge, and
judgment. Gaspers and Schulz (1960) showed that the failure of the system to distinguish
between waters that are naturally productive and those artifically enriched can lead to
absurd results. They studied a canal in Hamburg, which because of its urban situation
can only be regarded as grossly polluted. Yet it develops a rich plankton the composition
of which, according to the system, shows it to be virtually clean.
Once the Saprobiensystem was accepted it was logical to attempt to reduce its
findings to simple figures or graphs for presentation of results. Several such methods
were developed, which are described by Tiimpling (I960), who also gives the original
references. In all these methods, the abundance of each species is recorded on some
sort of logarithmic scale (e.g. 1 for present, 3 for frequent, 5 for common, etc.). The
sums of these abundances in each saprobic level are plotted on graphs, the two most
polluted zones showing as negative and others as positive. Or, the various saprobic levels
are given numerical values (1 for oligosaprobic [clean], 2 for fi-mesosaprobic, etc.)
and the rating for each species is multiplied by its abundance number. The sum of all
these products divided by the sum of all the frequencies gives a "saprobic index'' for
the locality. Clearly the higher this number, the worse the water quality in terms of
organic pollution. In a similar way the so-called "relative Belastung" (relative load)
is calculated by expressing the sums of all the abundances of organisms characteristic
of the two most-polluted zones as a percentage of the sum of all abundances. Then
100 percent is completely polluted water, and clean localities will give a low number.
There are various elaborations of these methods, such as sharing of species between
zones and taking account of changes in base-line as one passes downstream. None of them,
however, eliminates the basic weaknesses of the system nor the fact that, as Gaspers
and Schulz (1960) point out, there is little agreement between the various authors in the
assignment of species to the different levels. Therefore, one gains a number or a figure
that looks precise and is easily understood, but it is based on very dubious foundations.
Similar systems are indigenous to North America, but were independently evolved.
Wurtz (1955) and Wurtz and Dolan (1960) describe a system whereby animals are
divided into sensitive-to-pollution and non-sensitive (others are ignored), and also into
burrowing, sessile, and foraging species (six classes). Numbers of these species
represented are plotted for each station as six histograms on the basis of percentage of
total number of species. If the constitution of the fauna from control stations or from
similar localities is known, it is possible to express numerically "biological depression"
(i.e., percentage reduction in total number of species), "biological distortion" (i.e.,
change in proportions of tolerant and non-tolerant species), and "biological skewness"
(changes in the ratios of the three habitat classes). Such results must, of course, be
evaluated, and the definition of tolerance is quite subjective; but the method has the
advantages of simplicity and dependence on control data. Like the Saprobiensystem,
however, it can have no universal validity. It also suffers from the fact that it takes
292 INTERPRETATION OF BIOLOGICAL DATA
-------
no account of numbers; a single specimen, which may be there by accident, carries as
much weight as a dense population.
Patrick (1949) developed a similar system in which several clean stations on the
water body being investigated are chosen, and the average number of species is determined
occurring in each of seven groups of taxa chosen because of their supposed reaction to
pollution. These are then plotted as seven columns of equal height, and data from other
stations are plotted on the same scale; it is assumed that stations differing markedly
from the controls will show biological imbalance in that the columns will be of very
unequal heights. Number is indicated by double width in any column containing species
with an unusual number of individuals. I have already questioned the usefulness of this
method of presentation (Hynes 1960), and doubt whether it gives any more readily
assimilable data than simple tabulation; it does however, introduce the concept of
ecological imbalance.
It has long been known that ecologically severe habitats contain fewer species than
normal habitats and that the few species that can survive the severe conditions are often
very abundant as they lack competitors. Examples of this are the countless millions of
Artemia and Ephydra in saline lakes and the Tubifex tubifex in foul mud. This idea
has often been expressed in terms of diversity, which is some measure of numbers of
species divided by number of specimens collected. Clearly, such a parameter is larger
the greater the diversity, and hence the normality of the habitat. Unfortunately, though,
as the number of species in any habitat is fixed, it also decreases as sample size
increases so no index of diversity has any absolute value (Hairston 1959). If a
definite sample size is fixed, however, in respect to numbers of organisms identified, it
is possible to arrive at a constant index.
Patrick et al. (1954) in effect used this concept in a study of diatom species
growing on slides suspended in water for fixed periods. They identified 8000 specimens
per sample and plotted the results as number of species per interval against number of
specimens per species on a logarithmic scale. This method of plotting gives a truncated
normal curve for a wide variety of biotic communities. In an ordinarily diverse habitat
the mode is high and the curve short; i.e., many species occur in small numbers and
none is very abundant. In a severe habitat the mode is low and the curve long; i.e.
there are few rare species and a few with large numbers. This, again, seems to me to
be an elaborate way of presenting data and to involve a lot of unnecessary arithmetic.
Allanson (1961) has applied this method to the invertebrate faunas of streams in
South Africa and has shown, as has Patrick for diatoms, that the log normal curve is
flatter and longer for polluted stations; the difference, however, is not so apparent that
it does not need exposition. Here, again, I would suggest that tabulated data are just
as informative. Indeed I would go further and say that tabulated data are essential
in the present state of our knowledge. We are learning as we go along and if the
details of the basic findings are concealed by some sort of arithmetical manipulation
they cannot be re-interpreted in the light of later knowledge, nor are they preserved
in the store of human knowledge. This point becomes particularly clear when one
examines some of the early studies that include tables. Butcher (1946) requotes a
considerable amount of data he collected from studies of various English rivers during
the thirties; they are not only clear and easy to follow, but they are also informative
about the generalities of pollution in a way that data quoted only within the confines of
some particular system are not.
Simple tabulation of biological data in relation to water quality, either in terms
of number of organisms, percentage composition of the biota, some arbitrary abundance
Hriies 293
-------
scale, or as histograms, has been effectively practiced in many parts of the world: in
America (Gaufin and Tarzwell 1952, Gaufin 1958), Africa (Harrison 1958 and 1960, Hynes
and Williams 1962), Europe (Albrecht 1954, Kaiser 1951, Hynes 1961, Hynes and
Roberts 1962), and New Zealand (Hirsch 1958) to cite a few. These tabulated data are
easy to follow, are informative to the expert reader, and conceal no facts. Although
the non-biologist may find them tedious, he need only read the explanatory paragraphs.
It is a delusion to think that it is possible to reduce biological data to simple numerical
levels. At best, these can only be produced for limited situations and even then they need
verbal exposition; at worst, they give a spurious impression of having absolute validity.
My final point in this section concerns comparisons. It is claimed that the German
system, in effect, measures an absolute state, a definite level of water quality. We have
seen that this is not a tenable claim. In the other systems, by and large, the need to
establish local control stations at which to measure the normal or "natural" biotic
conditions is accepted, and then other areas are compared with this supposed norm.
This is, of course not always possible as there may remain no unaffected area, or no
unaffected area that is, with respect to such factors as current, nature of substratum
etc., sufficiently similar to act as a base-line for data. Nevertheless, basically, these systems
can be used to compare stations and thus to assess changes in water quality. In doing
this, they can all be used more or less successfully, but I maintain that a table is just
as useful as an elaborate analysis, and I believe that the table should be included with
whatever is done. For a particular situation, however, it is often possible to distill the
data into a single figure as a measure of similarity between stations.
Burlington (1962 and Dean and Burlington 1963) has recently proposed an entirely
objective means of doing this, which involves simple arithmetical manipulation. In
his system a "prominence value" is calculated for each species at each station. This is
a product of its density and some function of its frequency in samples, but the details
of this calculation can be altered to suit any particular situation. Then a coefficient
of similarity between each pair of stations can be calculated by dividing twice the sum
of the lower prominence values of taxa that the two stations have in common by the
sum of all the prominence values of both stations. Identical stations will then have a
coefficient of similiarity of 1.00; this coefficient will be lower the more different the
stations are from one another. This is an easy way to compare stations in an entirely
unbiased way and as such may satisfy the need for numerical exposition; however, it
tells one nothing about why the localities are different and like all the other more or
less numerical methods of presenting data has no absolute value. Moreover, it still leaves
unanswered the fundamental question of how different is "different?"
THE PROBLEMS OF SAMPLING
The systems outlined above are all based on the assumption that it is possible to
sample an aquatic habitat with some degree of accuracy; this is a dubious assumption,
however, when applied to biological data. From what has been said about the com-
plexity of biological reactions to the various factors in the environment, and from the
obvious fact that rivers especially are a mosaic of microhabitats, it is clear that to
achieve numerical accuracy or even some limits of confidence considerable numbers of
samples need to be taken. Indeed, even in so apparently unvaried a habitat as a single
riffle, Needham and Usinger (1956) showed that a very large number of samples would
be necessary to give significant numerical data.
it is
There is a limit to the number of samples that can reasonably be taken and, anyway,
desirable to sample many different types of habitat so as to get as broad as possible
294 INTERPRETATION OF BIOLOGICAL DATA
-------
an estimate of the biota. This is the more recent approach of most of the workers in
Central Europe, who have been content to cite abundances on a simple relative but
arbitrary scale and to convert this to figures on some sort of logarithmic scale for use
in calculations. An alternative is to express the catch in terms of percentage composition,
but this has the disadvantage that micro- and macro-organisms cannot be expressed on the
same scale as they are obtained by different collecting techniques. Also, of course,
implicit in this approach is the assumption that the sampling is reasonably representa-
tive. Here again we run into the need for knowledge and expertise. In collection as
well as in interpretation, the expert is essential. Biological sampling, unlike the simple,
or fairly simple, filling "of bottles for chemical analysis or the monitoring of measuring
equipment, is a highly skilled job and not one to be handed over to a couple of vaca-
tioning undergraduates who are sent out with a Surber sampler and told to get on with
it. This point has also been made by other biologists, e.g., Patrick (1961) who stresses
the need for skilled and thorough collecting even for the determination of a species list.
Alternatively we can use the less expert man when concentrating on only part of
the habitat, using, say, microscopical slides suspended in the water to study algal
growth. This method was extensively used by Butcher (1946), and Patrick et al.
(1954) who studied diatoms in this way. This gives only a partial biological picture,
but is useful as a means of monitoring a stretch of water where it is possible that
changes might occur. It is a useful short-hand method, and as such is perhaps comparable
to studying the oxygen absorbed from potassium permanganate instead of carrying out
all the usual chemical analyses on water. A short method of this kind may serve very
well most of the time, but, for instance, would not be likely to detect an insecticide in
concentrations that could entirely eliminate arthropods and hence fishes by starvation.
It is possible to work out biological monitoring systems for any specific purpose.
The simplest of these is the cage of fish, which, like a single type of chemical analysis,
can be expected to monitor only one thing the ability of fish to live in the water
with no information on whether they can breed or whether there is anything for them to
eat. Beak et al. (1959) describe a neat way in which the common constituents of the
bottom fauna of Lake Ontario can be used to monitor the effluents from an industrial
site. Obviously there is much room for such ingenuity in devising biological systems for
particular conditions, but this is perhaps outside the scope of this meeting.
CONCLUSIONS
It may appear from the previous sections that my attitude to this problem is en-
tirely obstructionist. This is far from being so. Water quality is as much biological
phenomenon as it is a chemical or physical one; often what we want to know about
water is almost exclusively biological will it smell nasty, is it fit to drink, can one
bathe in it, etc.? I suggest, therefore, that it is desirable to organize water monitoring
programs that will tell one what one wants to know. There is no point in measuring
everything biological, just as there is no point in performing every possible chemical
analysis; what is measured should be related to local conditions. It would be a waste of
time to measure oxygen content in a clean mountain stream; we know it to be high, and
it becomes worth measuring only if we suspect that it may have been lowered by
pollution. Similarly, there is little point in studying the plankton in such a stream; we
know it only reflects the benthic flora. In a lake or in a slow river, on the other hand,
if our interest in the water lies in its potability, records of the plankton are of consider-
able importance as changes in plankton are, in fact, changes in the usability of the water.
For long-term studies, especially for the recording of trends or changes induced by
Hynes 295
-------
pollution, altered drainage, agricultural poisons, and other havoc wrought by man, one
can expect informative results from two principal techniques: First, we can study micro-
scopic plant and animal growth with glass slides placed in the water for fixed periods;
second, we can obtain random samples of the benthic fauna. The algae and associated
microfauna tell one a good deal about the nutrient condition of the water and the changes
that occur in it, and the larger benthic fauna reveal changes in the trophic status,
siltation due to soil erosion, effects of insecticides and other poisons, etc.
The study of growths on glass slides is reasonably skilled work, but can easily be
taught to technicians; like chemical monitoring, such study needs to be done fairly often.
Sampling the benthos is more difficult and, as explained above, needs expert handling;
unlike most other monitoring programs, however, it need be done only infrequently,
say, once or twice a year. Inevitably sampling methods will vary with type of habitat;
in each case, the question will arise as to whether it is worth looking at the fish also.
It is here that the biologist must exercise judgment in devising and carrying out the
sampling program.
Judgment is also needed in the interpretation of the data. It is for this reason
I maintain that it should all be tabulated so that it remains available for reassessment
or comparison with later surveys. If need be, some sort of numerical format can be pre-
pared from the data for ad hoc uses, but it should never become a substitute for tabula-
tions. Only in this way can we go on building up our knowledge. Perhaps some day we
shall be able to pass all this information into a computer, which will then be able to
exercise better judgment than the biologist. I hope this will happen, as computers are
better able to remember and to cope with complexity than men. It will not, however,
pension off the biologist. He will still be needed to collect and identify the samples.
I cannot imagine any computer wading about on rocky riffles nor persuading outboard
motors and mechanical graps to operate from the unstable confines of small boats. We
shall still need flesh and blood biologists long after the advent of the hardware water
chemist, even though, with reference to my earlier analogy, a Tokyo University computer
recently outpointed 10 veteran medicals in diagnosing brain tumors and heart disease.
It should be pointed out, however, that the computer still had to be fed with information,
so we are still a long way from the hardware general practitioner. I believe though that
he is likely to evolve before the hardware biologist; after all, he studies only one animal.
REFERENCES
Albrecht, M.-L. (1954). Die Wirkung der Kaliabwasser auf die Fauna der Werra und
Wipper. Z. Fisch. N. F. 3:401-26.
Allanson, B. R. (1961). Investigations into the ecology of polluted inland waters in the
Transvaal. Part I. Hydrobiologia 75:1-94.
Beak, T. W., de Courval, C. and Cooke, N. E. (1959). Pollution monitoring and pre-
vention by use of bivariate control charts. Sew. Industr. Wastes 31:1383-94.
Beck, Wm. M., Jr. (1957). The use and abuse of indicator organisms. Transactions of
a Seminar on Biological Problems in Water Pollution. Cincinnati.
Burlington, R. F. (1962). Quantitative biological assessment of pollution. /. Wat. Poll.
Contr.Fed. 34:179-83.
Butcher, R. W. (1946). The biological detection of pollution. J. Inst. Sew. Purif. 2:92-7.
Gaspers, H. and Schulz, H. (1960) Studien zur Wertung der Saprobiensysteme. Int.
Rev. ges. Hydrobiol. 45:535-65.
296 INTERPRETATION OF BIOLOGICAL DATA
-------
Dean, J. M. and Burlington, R. F. (1963). A quantitative evaluation of pollution effects
on stream communities. Hydrobiologia 27:193-9.
Caufin, A. R. (1958). The effects of pollution on a midwestern stream. Ohio J. Sci.
58:197-208.
Gaufin, A. R. and Tarzwell, C. M. (1952). Aquatic invertebrates as indicators of stream
pollution. Pub, Hlth. Rep. 67:57-64.
Hairston, N. G. (1959). Species abundance and community organization. Ecology
40.404-15.
Harrison, A. D. (1958). The effects of sulphuric acid pollution on the biology of streams
in the Transvaal, South Africa. Verh. int. Ver. Limnol. 73:603-10.
Harrison, A. D. (1960). The role of river fauna in the assessment of pollution. Cons.
sci. A/r. Sud Sahara Pub. 64/199-212.
Hirsch, A. (1958). Biological evaluation of organic pollution of New Zealand streams.
N.Z.J. Sci. 1:500-53.
Hynes, H. B. N. (1960). The biology of polluted waters. Liverpool.
Hynes, H. B. N. (1961). The effect of sheep-dip containing the insecticide BHC on the
fauna of a small stream. Ann. trap. Med. Parasit. 55:192-6.
Hynes, H. B. N. and Roberts, F. W. (1962). The biological effects of detergents in the
River Lee, Hertfordshire. Ann. appl. Biol. 50:779-90.
Hynes, H. B. N. and Williams, T. R. (1962). The effect of DDT on the fauna of a
Central African stream. Ann. trap. Med. Parasit. 56:78-91.
lilies, J. (1961a). Die Lebensgemeinschaft des Bergbaches. Wittenberg-Lutherstadt.
lilies, J. (1961b). Versuch einer allgemeiner biozonotischen Gliederung der Fliessgewasser.
Int. Rev. ges. Hydrobiol. 46:205-13.
Kaiser, E. W. (1951). Biologiske, biokemiske, bacteriologiske samt hydrometriske under-
sogelser af Poleaen 1946 og 1947. Dansk. Ingenforen. Skr. 3:15-33.
Kolkwitz, R. (1950). Oekologie der Saprobien. Uber die Beziehungen der Wasser-
organismen zur Ummelt. Schr. Reihe ver Wasserhyg. 4:64 pp.
Liebmann, H. (1951). Handbuch der Frischwasser imd Abtuasserbiologie. Munich.
Needham, P. R. and TJsinger, R. L. (1956). Variability in the macrofauna of a single
riffle in Prosser Creek, California, as indicated by the Surber sampler. Hilgardia
24:383-409.
Patrick, R. (1949). A proposed biological measure of stream conditions, based on a
survey of the Conestoga Basin, Lancaster County, Pennsylvania. Proc. Acad. Nat. Sci.
Phila. 101:277-341.
Patrick, R. (1961). A study of the numbers and kinds of species found in rivers in
Eastern United States. Proc. Acad. Nat. Sci. Phila. 113:215-58.
Patrick, R., Hohn, M. H. and Wallace, J. H. (1954). A new method for determining
the pattern of the diatom flora. Not. Nat. Phila. Acad. Sci. 259:12 pp.
Richardson, R. E. (1929). The bottom fauna of the middle Illinois River, 1913-1925;
its distribution, abundance, valuation and index value in the study of stream pollution.
Bull. III. not. Hist. Surv. 77:387-475.
Hvnes 297
-------
Sladecek, V. (1961). Zur biologischen Gliederung der hoheren Saprobitatsstufen. Arch.
Hydrobiol. 58:103-21.
Tiimpling, W. v. (1960). Probleme, Methoden und Ergenbnisse biologischer Giiteunter-
suchungen an Vorflutern, dargestellt am Beispiel der Werra. Int. Rev. ges. Hydrobiol.
45:513-34.
Whipple, G. C., Fair, G. M. and Whipple, M. C. (1947). The microscopy of drinking
water. New York.
Wurtz, C. B. (1955). Stream biota and stream pollution. Sew. industr. Wastes 27:1270-8.
Wurtz, C. B. and Dolan, T. (1960). A biological method used in the evaluation of
effects of thermal discharge in the Schuylkill River. Proc. Ind. Waste Conf.
Purdue, 461-72.
Zimmerman, P. (1962). Der Einfluss auf die Zusammensetzung der Lebensgemeinschaften
in Experiment. Schweiz. Z. Hydrol. 24:408-11.
298 INTERPRETATION OF BIOLOGICAL DATA
-------
Dr. Werner Stumm
Associate Professor of Applied Chemistry
Harvard University
Cambridge, Massachusetts
SUMMARY
This paper considers some of the chemical reactions that may, at least partially,
determine the composition of fresh water. Examples are given that demonstrate how
elementary principles of physical chemistry can aid in the identification of various
interrelated variables that determine the mineral relations in natural water systems.
In a hypothetical experiment, a unit volume of "natural" fresh water was prepared
by sequentially mixing with distilled water some of the pertinent constituents, starting
with more abundant ones. After each addition, the equilibrium composition was calculated
and compared with the composition of that in a real natural water system. Throughout
the experiment, standard reference tables on the energies or relative stabilities of the
various compounds were used. The stability relationships are shown in simple graphs.
CHEMISTRY OF NATURAL WATERS IN
RELATION TO WATER QUALITY
Natural waters acquire their chemical characteristics through direct solution and
chemical reactions with solids, liquids, and gases with which they have come in contact
during the various parts of the hydrological cycle. The final composition of a natural
water is the result of a great variety of chemical, physical, and biological reactions.
This paper considers some of the chemical reactions that may, at least partially,
determine the composition of fresh water. Obviously, this discussion of the physical
chemistry of natural waters cannot be comprehensive. The author has concentrated on
some examples that are in his opinion best suited methodologically and didactically to
demonstrate how elementary principles of physical chemistry can aid in identifying the
various interrelated variables that determine the mineral relations in natural water
systems. In writing this discussion, the author could not avoid being strongly influenced
by Sillen's excellent paper on the "Physical Chemistry of Sea "Water."1
THE MODEL
Since it is not possible to evaluate all the various chemical process combinations and
the various environmental factors, e.g., mineralogical and geological environment, rate of
water circulation, biological activity, temperature and pressure, etc., a simplified model
will be chosen. In a hypothetical experiment, we shall prepare a unit volume of "natural"
fresh water by sequentially mixing with distilled water some of the pertinent constituents,
starting with the more abundant ones. After each addition the equilibrium composition
will be calculated. For this calculation we will use free energy data (equilibrium con-
stants, redox potentials) found in standard references.2.3 The composition of the water
in a model at equilibrium will be compared with the composition of that in a real natural
water system.
LIMITATIONS OF THE MODEL
This hypothetical experiment is didactical. The sequence of addition of chemicals
to the pure water is not an attempt to follow the geological history and is thus rather
Stumm 299
-------
arbitrary. The comparison between the equilibrium model and the real system must take
into consideration that a true equilibrium is not necessarily attained in all respects in
the real system. In a natural body of water only the upper layers are in contact with the
atmosphere and only the deepest layers are in contact with the uppermost layers of the
sediments. The mixing in the real system is further impaired by density stratification
due to vertical temperature differences. On the other hand, the real systems are sub-
ject to periodic overturns; geological time spans have elapsed, and therefore, reactions
that reach equilibrium very slowly in the laboratory may have come nearer equilibium
in real systems. We must also be aware that biologically mediated reactions, e.g., photo*-
synthesis and respiration, can lead to significant localized disturbances of the equilibrium
composition.
The results obtained for the equilibrium model, of course, contain only that informa-
tion (free energy data for the species considered) that has been used for their computa-
tion. The available free energy values or equilibrium constants are frequently not known
with sufficient precision, some data are lacking, and occasionally we may overlook a
pertinent species. In view of these inadequacies, not much attention has been paid to
activity corrections and all calculations are based on 25°C. Consequently, in most in-
stances the results obtained represent an oversimplified picture. Nevertheless, it is
gratifying to see that the predictions are frequently in reasonable accord with observed
behavior in real systems.
The comparison between natural systems and their idealized counterparts is an
essential prerequisite to isolation of the variables responsible for observed mineral
relations. The equilibrium calculations and the comparison of the results with those of
the real systems will permit us to make some speculation on the type of solid phases and
dissolved species one may expect in fresh water systems. The value of the model thus
lies primarily in providing an aid for the interpretation of observed facts. Discrepancies
between equilibrium predictions and chemical data of the real systems can give us
valuable insight into those circumstances where chemical reactions are not sufficiently
understood, where non-equilibrium conditions prevail in the real systems, or where the
analytical data available are not sufficiently accurate or specific.
MAJOR COMPONENTS
SILICON AND ALUMINUM
At first sight, it might appear somewhat puzzling that we start our imaginary experi-
ment with these two elements as major constituents. It has frequently been assumed that
both silicon and aluminum do not hold an important position in mineral water quality
relations. This is only true if we consider waters in isolation from their natural en-
vironment. But it is so frequently forgotten by those who deal with water resources that
every lake and every body of natural water has a bottom (igneous and sedimentary
rocks). Dissolved mineral matter originates in the crustal materials of the earth; water
disintegrates mineral rocks by erosion and weathering and acts as a solvent on almost
all of them. Goldschmidt* has estimated that for each kilogram of ocean water some
600 grams of primary igneous rock must have been decomposed. Similar estimates cannot
be made in such a general way for fresh waters, but it might be safe to assume that
practically every ground and surface water has been in extensive and intimate contact
with sedimentary rocks. A qualitative illustration of such rock mineral and water inter-
action is given by records of the U. S. Geological Survey,6 which show that from 70 to
86 tons of soluble matter were carried, on the average in 1950, from each square mile
of drainage area of the James River above Richmond, Virginia; the Iowa River above
300 CHEMISTRY IN RELATION TO WATER QUALITY
-------
Iowa City; and the Colorado River above Grand Canyon, Arizona. Higher rates were
observed for streams draining limestone terranes. Silicon and aluminum are, besides
oxygen, the most abundant elements in igneous and sedimentary rocks. Although rela-
tively small amounts of these elements become homogeneously dissolved in water, the
various equilibria for heterogeneous chemical reactions between the solid and solution
phases are probably of utmost significance in the chemistry of natural waters. Un-
fortunately, much of the mineral and solution chemistry of these elements is not yet
well understood.
In making our artificial body of water, we add solid Si02 to pure water. Since not
very much of this Si02 will become dissolved, it is not critical how much SiO, we add
as long as we maintain an excess of solid SiO . It might appear reasonable to add
about 2 mole of SiO, per liter of pure water. The various reactions that can occur are
listed with their respective equilibrium constants in Table 1.* Reactions 1 and 2 describe
Table 1 SiO - Equilibria (Reference 6)
Reaction No.
1
2
3
4
5
Reaction
SiO, (quartz) + H,0 = Si(OH)4
Si02 (amorph) + H20 = Si(OH)4
Si(OH)4 = [SiO(OH)3] - + H+
[SiO(OH)3] -= [Si02(OH)2]-2 + H+
4Si(OH)4= [Si406(OH)6]-o + 4H20 + r:
logK
3.7
2.7
9.5
12.6
[+ 12.6
the solubility equilibrium of SiO,. It is obvious that quartz is the thermodynamically
stable form of Si02, whereas amorphous Si02 is metastable and about 10 times more
soluble than quartz. (Ortho-) silicic acid, Si(OH)4, is * very weak acid (reactions 3 and
4); its conjugate monoprotic and diprotic bases, the silicates SiO(OH)~ and
Si02(OH)2~2, are not important constituents in the common pH range of natural
waters (pH 6 to 9). Thus, for the dissolution of Si02, reactions 1 and 2 primarily have
to be considered. In our mixture, we will find about 2 x 10~*M dissolved Si(OH)4 if
we use quartz (or sandstones), and about 10 times more if we use amorphous silica as
the source of Si02. Natural waters can thus contain up to approximately 56 milligrams
of dissolved silica per liter, if we assume that the amorphous forms of SiO are the
major source of silica in natural waters. In most natural waters concentrations range
from 0.5 to about 15 milligrams per liter, although concentrations up to 50 milligrams
per liter are not uncommon. The solubility of Si02 as Si(OH)4 increases with increasing
temperature; thus, hot springs frequently have higher dissolved silica than cold waters.
On the basis of the data given in Table 1, it must be concluded, contrary to earlier
beliefs, that silica in water does not occur as a colloid. Most natural waters (below
pH 9) do not contain silicate anions in appreciable concentrations. Therefore, dissolved
silica, under these conditions, is not a part of the operationally determined alkalinity oi
a water, nor does the dissolved Si(OH)4 have any marked influence on the buffer
capacity of fresh waters.7
* For most of the reactions listed in the tables in this paper, different authors may have
determined different equilibrium constants. The constants given in these Tables have
generally been selected frorn tabulated values given in references 2 and 3. Only in
those instances where other sources have been used will special reference be given.
All constants given apply to 25°C and do not always refer to the proper ionic strength
of natural waters (5 x 10~* to 5 x 10~3).
Stumm 301
-------
Figure 1 gives a solubility diagram for Si02 (total soluble silica as a function of
pH). The dissolution of SiO? becomes significant at very high pH values (water glass)
(reactions 3, 4, and 5). The polymerization of Si(OH)4 to tetrameric silicate (reaction
5) occurs only under alkaline conditions (pH > 10). If alkaline concentrated solutions
that contain polymeric silicates are acidified to lower pH values, the solubility of silica
is exceeded and Si02 precipitates. Within neutral and slightly alkaline pH ranges, rela-
tively stable negatively charged collodial dispersions of SiO (activated silica) are formed.
0
no
° -2
-4
amorphous SiO2
quartz
9
PH
11
Figure 1 Solubility of Quartz and
Colloidal SiO2.
Besides Si02, various silicate minerals, metal silicates, and clays are important com-
ponents of mineral rocks that represent sources of dissolved material in water.
We now add about 1 mole of Al(OH) per 1 liter to our water. As a first approach,
we might ask ourselves how much of the Al(OH) would become soluble. The chemistry
of Al(III) has not been elucidated in great detail, but some of the more recent
equilibrium information is summarized in Table 2.
Table 2 Hydrolysis and Solubility Equilibria of Aluminum
Reaction No,
6
7a
7b
8
9
10
Reaction
Al+s + H20 = [A10H]+2 + H+
Al+3 + 3 H20 = Al(OH)3(s) + 3 H+
Al(OH)3(s) = Al+s + 3 OH-
Al(OH)3(s) +-H20= [A1(OH)J- + H+
6A1+3 + 15H20= [Al6(OH)16]+8 + 15 H+
8Al+3+20HO I A~\ fOH} 1+4 4- 9n TT+
logK
5.0
9.1
32.9
12.7
47
302
CHEMISTRY IN RELATION TO WATER QUALITY
-------
On the basis of this information, a solubility diagram has been constructed (Figure 2).
From this, it is evident that Al+3 is very easily hydrolyzed to various hydroxide com-
plexes. Aquo-aluminum ion is an acid that exhibits acidity similar to that of acetic acid.
There is some uncertainty about the various hydrolysis products that might occur in
the slightly acid to neutral pH range. Although the behavior of dilute Al (III) solutions
can be reasonably well interpreted on the basis of reaction 6, the potentiometric investi-
gations of Brosset and co-workers8 on the reaction of A1+3 ion with water in the presence
of various concentrations of OH ions have revealed that the monomeric hydrolysis
species A1(OH)+2 is not the main hydrolysis product if it exists at all. Brosset was
able to interpret his data by postulating a soluble polymeric aluminum hydroxo complex
with a stoichiometric ratio of OH- to Al(III) of 2.5. He suggested [A16(OH)15] +3
as the most likely structure. On the basis of colloid chemical investigations, Matijevic
and co-workers9 postulate [Alg(OH) 20] +4 to be the most likely formula. From the data
given in Table 2 and Figure 2, it is evident that Al(OH)3(s) exhibits amphoteric
properties. The solubility of A1(OH)3 increases in the acid and alkaline range. With
increasing pH, more hydroxide ions are coordinated to the aluminum, and soluble
aluminate, [Al(OH) J~ or [A12(OH) J"2, is formed.
Figure 2 indicates that in the common pH range of natural waters the predominant
soluble aluminum species appears to be aluminate, (A1(OH)4]~. Total soluble Al(III)
in equilibrium with Al(OH)8(s) amounts to approximately 10~6M at pH 7 and
approximately 10~5M at pH 8. At these pH values, A1+3 would account for only
about 10~12M and 10-16M, respectively. Total soluble Al(III) in most natural waters
should vary between about 30 micrograms per liter (pH 7) and 300 micrograms per liter
(pH 8). Little reliable analytical data on the Al(III) content of natural waters are
available for comparison with these calculated equilibrium results.
Figure 2 Solubility of Aluminum Hydroxide.
All the Al(OH)g that has been added to our mixture will, however, ultimately react
with Si02 to form aluminum silicate minerals such as kaolinite, Al2Si,,O5 (OH) 4(s). Since
SiO, is in excess of A1(OH), the silicic acid content of the solution will not change.
Stumm
303
-------
Aluminum silicates like kaolinite can rearrange their structure in such a way that
Mg+2 (or Ca+2) may substitute for aluminum in its octahedral coordination arrange-
ment. In a similar way aluminum may replace silica in its tetrahedral structure. Nega-
tively charged aluminum silicate frameworks with layer structure (clays) are built up
in such a manner. Because of their electronegative nature, these clays exhibit cation-
exchange phenomena. Although it has not been established which solid clay phases
represent true equilibrium states, we must take into account that a great variety of
clays are encountered as metastable solid phases in aquifers, in sediments of surface
waters, and in suspension. Ion-exchange equilibria between dissolved constituents of
natural waters and clays and minerals with which these waters come into contact in-
fluence the concentration of H+ and other cations. Sillen1 has identified hetereogeneous
silicate equilibria as comprising the principal pH buffer system in oceanic waters. A
plausible reaction scheme for strongly pH dependent silicate equilibria has been given
by Sillen1:
3Al,Si,0B(OH) ,(s) +4SiO,(s) + 2K+ + 2Ca+- + 9H,0 =
2KCaAl3Si501"6(H,0)6(s) + 6H+ " (11)
,B
Although equilibrium relationships of such reactions are not yet well understood, it. is
obvious that exchange reactions such as given in equation 11 must exert considerable
influence upon the hydrogen ion concentration of natural waters.
CALCIUM CARBONATE
We will now add to our mixture CaC03 in the proportion of about 0.5 mole per
liter, thus introducing Ca+2 and carbon simultaneously.
Since the previous additions of A1(OH)3 and Si02 did not have any appreciable
influence upon the pH of the solution (there has been a very slight reduction in pH in
this unbuffered system) and since the already dissolved species will have no influence
upon the CaC03 solubility equilibrium, our problem of equilibrium calculation can
essentially be reduced to that of pure water being in contact and in equilibrium with
solid CaCO,. The equilibria that have to be considered are listed in Table 3.
Table 3 CaCO and Carbonate Equilibria
Reaction No.
12
13
14
15a
15b
15c
16
Reaction
C02(g) +H20 = H2CO]3*
H2C03* = HC03- + H+
HC03- = C03 -z + H+
CaCO., (s) = Ca+2 + C03-2
CaC03(s) + H+ = Ca+2 + HCO,-
CaCO (s) + H0CO ' = Ca+2 + 2 HCO -
o J A A
CaC03(aq) = Ca+2 + CO.,-2
logK
- 1.5 (K)
6.3 (K^)
10.3 (K2)
8.3 (KB)
+ 2.0 (K9/K2)
- 4.3 (K^/K,)
3.0 (?)
Remarks: H2C03* refers to the sum of dissolved C02 and H2C03. In order to simplify the
writing of the equations in the text the following terms are introduced to
define the total concentration of dissolved carbonic species, CT; alkalinity,
[Alk].
CT = [rl.CO,*] + [HCOS-] + [CO-2] (17)
304 CHEMISTRY IN RELATION TO WATER QUALITY
-------
[Alk] = [HC03-] +2[C03-] + [OH-] - [H+] (18)
[Acidity] =2[H2C03*] + [HC03-]+ [H+] - [OH~] (19)
The following abbreviations are derived from Equations 13, 14, and 17:
a0= [H2CO?*]/CT = 1/(1 + K1/[H+] + KXK2/ [H+p) (20)
KI= [HC03]/CT =!/(!+ [H+J/K, + K/2 [H+]) (21)
a.2= [COs-2]/CT= !/(!+ [H+]/K2+ [H+p/K^K,) (22)
Furthermore, the ion product of water [H+] [OH""] = Kw is taken as 10"1*.
CASE 1: SYSTEM CLOSED TO THE ATMOSPHERE
As a first approximation, we assume that our system is not exposed to the atmosphere
and we treat H2C03'f as a non-volatile acid. Under these circustances all Ca+2 that
becomes dissolved must equal in concentration the sum of the dissolved carbonic species:
[Ca+2] = CT (23)
Furthermore, the solution must fulfill the condition of electroneutrality:
2[Ca+2] + [H+] = [HC03-] +2[C03~2] + [OH~] (24a)
or
2 [Ca+2] = [Alk] (24b)
i.e., in such a solution the calcium hardness is equal (equivalent) to the alkalinity.
Since, according to equations 15, 22, and 23, [Ca+2] or CT is equal to (Ks/a2)°-5,
we can, considering equation 21, rewrite the electroneutrality condition 24 in the following
way*:
(Ks/a2)°.5[ 2 - ffll - 2tt2] + [H+] - Kw/ [H+] = 0 (25)
This equation can most conveniently be solved for [H+] by trial and error. For the
given [H+], the equilibrium concentration of the additional dissolved species can readily
be accomplished. The result of such computation gives:
pH = 9.9; [Ca+2] = 1.2 x If}-*; [HCO,,-] = 9 x 10-";
[C03-2] = 4 x 10-5; [H2C03*] = 2.5 x IQ-S;
[Alk] =2.4x10-*; [Acidity] =0.
INFLUENCE OF ACID AND BASE
The equilibrium mixture (CaC03 + water) thus obtained is not well buffered (we
disregard for the moment the influence of Si02 and A1(OH)3), and small quantities of
acids or bases will change the pH and the solubility relations. We might visualize that
such pH changes could occur, for example, upon addition of acid or base waste con-
stituents, through dissolution of volcanic HC1 or through the influence of biological
reactions (e.g., H+ addition as a result of nitrification or OH~ addition as a result
* In this and many of the subsequent equations, some of the terms are (even in very
exact calculations) negligible. Generally, mathematically exact equations are given.
This "precision" might appear to be in contrast with the many otherwise uncertain
factors involved in these calculations, but the author believed it necessary to emphasize
that the quantitative evaluation of the systematic relations that determine equilibrium
concentrations of a solution constitutes a purely mathematical problem that is, without
the need for introducing a priori assumptions, subject to exact and systematic
treatment.10 A relatively simple way to survey the interrelationships of the equilibrium
concentrations of the individual solute species consists of a simultaneous graphical
representation of all the requisite equations (see Figures).11' 12
Stumm 305
-------
of denitrification). From a computational point of view the problem is analogous to the
deration of a CaC03 suspension with strong acid, CA, or strong base, CB. Such acid
or base addition will shift the electroneutrality condition expressed in equation 24
to the following:
CA-CB= (Ks/a2)0'6 (2-o1-2«ll) + [H+] -KW/[H+] (26)
With the help of this equation, it is always possible to compute the quantity of CA
or CB needed per liter of water in contact and in equilibrium with solid CaCO to reach
a given pH value (Figure 3). Of course, the addition of acid and the resultant lowering of
. 30
20
!°
9
PH
11
11
Figure 3 Titration of CaCO Suspension
With Acid and Base.
Figure 4 Dissolved Species of a CaCOs
Suspension.
pH will lead to an increase in dissolved Ca+2 and carbonic constituents (Ca hardness
> [Alk]), whereas base addition will result in deposition of CaC03 (Ca hardness
< [Alk]). Under our assumptions the condition of equation 23 still holds; thus, the
pH dependence of soluble Ca+2 and of the sum of carbonic species, CT, is determined by
[Ca+2] =CT= (Ks/«2)o.5 (27)
Equilibrium concentrations for Ca+2, HC03~, C03~2, H,C03, and alkalinity are depicted
in Figure 4. (Equation 26 and Figures 3 and 4 represent the essential quantitative
principles involved in the Ca+2 removal by lime-soda softening.)
CASE 2: SYSTEM OPEN TO ATMOSPHERE
In our calculations so far, we have neglected the influence of atmospheric CO, and
have treated H2C03* as a non-volatile acid. In order to approach more realistic condi-
tions, we open our system to atmospheric C02 and we assume that the partial pressure
of C02 is equal to approximately 3 x 10~4 atmosphere. On the basis of Henry's law
(equation 12), the equilibrium concentration of H2C03* is given by approximatelyKp002
= 10~5. The electroneutrality condition of equation 24a still applies; [Ca+2] is
no longer equal to CT, but equation 24b is still valid, i.e., the calcium hardness is
306
CHEMISTRY IN RELATION TO WATER QUALITY
-------
equivalent to alkalinity. The equilibrium condition of equation 24 can be rewritten in
the following form:
= - x (tt + 2az) + [OH-] - [H+] (28)
o «o
2
Solution of this equation gives :
[Ca+2] = 5 x 10-*; [C03-3] = 1 x 10~5;
[HCCg = 10-3; PH = 8.4; [H2COa*] = 10-=;
[Alk] = 10-3
Comparing this result with that of Case 1, we see that the influence of atmospheric C02
has depressed the pH markedly and that the concentration of [Ca+2] and [Alk] has
been raised to values very representative of those in natural waters.
CASE 3: WATER ISOLATED FROM SOLID CaCO3
In a water isolated from its sediments and mineral rocks, such as water in
epilimnetic layers of a lake, or in samples brought to the laboratory, the presence of
H2C03*, HC03-, and CO3"2 is primarily responsible for the maintenance of near neutral
pH conditions. Since the total concentration of carbonic species seldom exceeds a few
millimoles per liter, we deal with a system of very low buffer capacity. A few millimoles
of acid or base per liter are sufficient to change the hydrogen ion concentration by some
orders of magnitude. Thus, heterogeneous chemical equilibria (interaction of the solution
with carbonate rocks, cation-exchange reactions with silicate minerals) , as well as bio-
chemical processes (C02 removal by photosynthesis, C02 production by respiration), and
the C02 cycle between the atmosphere and the natural waters are more significant for
H+ ion regulation in natural waters than the buffer contribution of dissolved carbonic
species. The dissolved carbonate system is actually a mediator or indicator of the buffer
systems of fresh water rather than the sole, or even a principal, buffering agent. *-, ll.
For « water in which the concentration of C02 is governed only by an equilibrium
between the dissolved carbonate system and the C02 of the atmosphere, the following
equation describing the interrelation between [H+], [Alk], and partial pressure can
be derived:
[Alk] = (KKlPco /[H+]) U + 2K2/[H+]) +KW/[H+] (29)
Accordingly, for a water in C02 equilibrium with the atmosphere, the H+ ion concentra-
tion is defined solely by the alkalinity of the water and the partial pressure of C02
(i.e., the same pH should ultimately occur for equinormal solutions of NaOH, NaHC03,
or Na2COs ) . The pH of a solution containing 10~ 3 equivalents of alkalinity per liter
in contact with the atmosphere (Pco2 = 10~ 3-6) should have a pH of approximately 8.4.
Therefore, most fresh waters are oversaturated in C02 with respect to an equilibrium
with the atmosphere. Accordingly, aeration of fresh waters frequently leads to an
increase in pH, causing a closer approach to equilibrium conditions. The conclusion
that may be drawn from this is that reactions that tend to depress the pH of natural
waters, such as ion exchange, CaC03 deposition, and respiration, in general are kinetically
more rapid than the C02 exchange with the atmosphere.
The case for the addition or removal of dissolved carbon dioxide will be developed
in some detail since the effect of carbon metabolism upon pH causes this case to be of
particular interest. Following is a schematic generalized reaction for carbon metabolism.
Stumm 307
-------
CO, + 2 FLO
Photosynthesis
> >
<; *
Respiration
(CH,0)n + 0,
H20
Any addition of H,CO * to a carbonate solution increases both the acidity of the solution
and CT. The alkalinity, unlike the case for the addition of strong acid, remains unchanged
however. (Our assumption in this case does not consider any interaction with CaC03 or
precipitation of CaCO^.) The change in CT, as a result of addition of C02 (or removal
of C02), can be characterized by equation 30.
[Alk] - [OH"] + [H+]
a2(2+ [H+]/K2)
(30)
Equation 30 and its graphical representation (Figure 5) are convenient tools for the
evaluation of biochemical respiration and C02 assimiliation, and for the assessment of
metabolic activity from diurnal variations in pH.
_ < E
PH
Figure 5 Addition or Removal of CO2 to
or From a Homogeneous Carbonate Solution
(Alk = 10-3 = Constant).
ANALYTICAL IMPLICATIONS OF CaCO? SOLUBILITY EQUILIBRIUM
The CaC03 solubility equilibrium has been applied in water chemical interpretations
for over 50 years. Equations describing the CaCOg solubility equilibrium (equivalent
to equations 15a, b, and c, Table 3) were independently derived by van't Hoff (1890),
Tillmans (1912), Kolthoff (1921), Langelier (1936), and others. In many ground
waters and a large number of surface waters, the relation between analytically determined
concentrations of Ca+3, H+, and carbonic species (alkalinity or CT) is in very good
accord with the CaCO, solubility equilibrium. In water technology the same concept has
308
CHEMISTRY IN RELATION TO WATER QUALITY
-------
been used with analytical data to predict whether a water will tend to deposit or dissolve
CaCO,. Figure 6 gives a plot of maximum soluble [Ca+2] as a function of pH for
=. -2 -
Figure 6 Dissolved Species of a CT =
103M Carbonate Solution in Equilibrium With
CaCOs(s).
CT = 10~3M. Equations 15 to 15c can be rearranged in various ways to make them
more suitable for direct use with analytically determinable parameters; for example, if
maximum soluble Ca+2 is expressed as a function of [H+] and CT or [Alk] :
[Ca+2] = Ks/CTa, (3D
[Ca+2] = (Ks/Kj ([H+]/HCO-]) (32)
Stumm
309
-------
or if Ca+2 is expressed as a function of [H2CO3*] and [Alk] or CT:
[Ca+2] = (K^/K,) ([H,CO,*]/ [HC03-p) (33)
In both of these expressions, (HCOg-] can be substituted for the analytically readily
determinable [Alk] or CT by
[HC03-] = ([Alk] -Kw/ [H+] + [H+]) (1 + 2K2/ [H+]) (34)
or
[HC03 ] = CT ai (34a)
(At pH values below pH 9, [HC03"~] can be set equal to [Alk] ; similarly, within the
pH range 7 to 9, Cei is very close to 1.)
Since equations 31 to 33 are conceptually equivalent, one might wonder why 32 is
preferentially used in the United States while equation 33 is almost exclusively used in
continental Europe. From an operational point of view, equation 33 is analytically more
satisfactory for hard, high-alkalinity waters than for soft, low-alkalinity waters (smaller
relative error in analytic determination of [H,CO3*] than of [H+]). Correspondingly,
the elucidation of "stability" can be rendered more precisely for soft, low-alkalinity
waters if based on the analytical determination of [H+] and [Alk].
It has been suggested especially by Greenwald12 that Ca+2 interacts with
and C03~2 to form soluble complexes, e.g., CaHC03+ and CaC03(aq) (see equation 16,
Table 3). Although the formation of such complexes is entirely plausible, according to
carefully controlled experiments by the author, they do not appear to be of any signifi-
cance in controlling CaC03 solubility in the concentrations and pH range found in
natural waters. Since CaC03 solubility is rather dependent on temperature and ionic
strength, it is relevant that constants valid at appropriate temperature and proper activity
corrections (e.g., those suggested by Larson and Buswell14) be used.
OTHER ANIONS
We now add some sulfate, chloride, and nitrate in the form of their sodium or
potassium salts. These added ions will have very little influence upon the equilibria
already discussed. The solubility product of CaSOd is of the order of 10~6, so that no
PHOSPHATE
The phosphate concentration in natural waters seldom exceeds 0.3 milligram per
liter. Upon addition of 10~* mole of phosphate per liter (in the form of Na2HP04),
our aluminum and calcium equilibria are influenced. Some of the phosphate will form
soluble phosphate-aluminum complexes (AlHPOi+). The solubility product of AlP04(s)
is of the order of 10~21, and a calculation will show that AlP04(s) will be formed only
under slightly acid conditions (pH 5 to 6).
For the interaction of phosphate with calcium we have to consider the following
reactions:
Ca+z + HP04~2 = CaHP04(s) ; K = 107 (34b)
5 Ca+2 + 3 PO4-s + OH- = Ca5(P04)3OH(s) ; K = 10+S6 (34c)
In the pH range of natural waters, reaction 34c, i.e., the possible formation of
hydroxyl-apatite, Ca5 (PO4) 3OH(s), has to be considered; we might a?k ourselves
whether some of the added phosphate will convert some of the CaC03 into apatite. We
obtain the equilibrium constant for such a reaction in the following manner:
310 CHEMISTRY IN RELATION TO WATER QUALITY
-------
5 Ca+2 + 3 P04-" + OH- = Ca6(POJ3OH(s) ; K = W+ (34c)
5 CaC03(s) + 5 H+ = 5 Ca+2 + 5 HC03~; K = 10+ (15b)
3HP04~2 =3H+ +3P04-"; K = 10-^ (34d)
H20 = H+ + OH"; K =
5 CaC03(s) + H+ + 3 HPO^-2 + H20 = K = 10+i6 (34e)
Ca5(P04)3OH(s) + 5 HC03-;
If we now compute the free energy, AF, for the conversion of CaC03 into apatite
by means of the equation, AF = RTln Q, where Q is the quotient of the reactants.
K
In order to compute Q, we assume the following values for the reactants: [H+] = ID"8,
[HC03-] = 10-3, [HP04~] = 10-*. Then we obtain
corresponding to a AF of approximately 15 Kcal, i.e., reaction 34e will proceed from
left to right until a new state of equilibrium is reached. At pH 7, the total amount of
phosphorus in equilibrium with hydroxyl-apatite is of the order of 10~6M (0.03 mgP/1).
Of course, such a figure is only approximate since the constants applied are not known
with good precision, but the tentative result suggests that at the sediment and water
interface the phosphorus concentration will be buffered by the presence of hydroxyl-
apatite as a stable solid phase. This conclusion, if verified, is of utmost significance in
connection with the eutrophication of lakes, because it would suggest that the phosphorus
distribution in a lake can be interpreted as a heterogeneous distribution equilibrium
between sediments and the lake, i.e., any addition of phosphorus (sewage) would lead to a
progressive accumulation of phosphorus in the sediments.
BIVALENT METAL OXIDES OR CARBONATES
FeC03
We now add about 0.5 mole of FeC03 (per liter of solvent) to our system. Much
of the iron that occurs in the earth crust is available as Fe(III). But later, when we
open our system to atmospheric oxygen, most of the Fe(II) that we have added as
FeC03 will be oxidized to ferric iron. Thus, the solubility conditions we now describe
apply for ferrous iron only.
Table 4 Fe (II) Solubility
Reaction No.
35a
35b
36
37
38
39
40
41
Reaction
Fe+2 + 2H,0 = Fe(OH)2(s) + 2 H+
Fe(OH)2(s) = Fe+2 + 2 OR-
FeCO, (s) = Fe+2 + CO ,-*
d 0
Fe+2 + H20 = FeOH+ + H+
Fe+2 + 3 H,0 = Fe(OH), + 3 H+
£, O
FeS(s) = Fe+2 + S~2
H2S(aq) = H+ + HS~
HS- = H+ + S-2
logK
+ 12.9
15.1
10.6
8.3
32
17.4
7.0
12.9
Comparison of equation 15a (Table 3) and equation 36 (Table 4) shows that the
solubility product of CaCO is about 200 times larger than that of FeCO?. Thus, only
Stnmm
-------
about 1 x ID"6 to 2 x 2Q-6 mole of FeC03 per liter (0.056 to 0.112 mg/1 Fe+2) will
go into solution without causing any appreciable change in pH or in concentration of
carbonate species.
In the pH range of natural waters, soluble bivalent iron consists of Fe+2 and FeOH+.
The solubility of ferrous iron in all carbonate-bearing waters (CT >10~4M,) within the
common pH range (pH 6 to 9) is governed by the solubility product of FeC03 (equation
36) and not (as is frequently assumed) by the solubility of Fe(OH)2 (equation 35b).
The solubility product constants of FeC03 and Fe(OH)2 have different dimensions,
i.e., mole2/liter2 and mole/liter3, respectively; thus, in order to decide which of the
solubility products controls Fe(II) solubility, one must evaluate the pH dependence of
Fe(II) solubility by using both constants. Figure 7 gives a solubility diagram for
Figure 7 Solubility of Fe(OH)2(s) in a Non-
Carbonate Solution (CT = 0).
Figure 8 Maximum Soluble Fe(ll) in a Car-
bonate-Bearing Water (CT = 1CT3M) Only
at High pH is the Solubility Controlled by Solu-
bility products of Fe(OH) .
Fe(OH)2 in a non-carbonate water; Figure 8 shows maximum soluble Fe(II) for a
carbonate-bearing water (CT = 10~3M). A comparison of these two figures shows
that the solubility product of Fe(OH), governs the solubility of Fe(II) only in waters
that contain no carbonate, or are at very high pH. Thus, essentially the same type of
equations that have been used quantitatively to describe the solubility relations oi
CaC03 can be used to evaluate Fe+2 solubility in natural waters (substitution of
K FeC03 for K CaC03). The maximum soluble Fe(II) for a water that is in CaC03
saturation equilibrium is only about 0.5 percent its calcium content. These considera-
tions apply only up to a pH of about 8 or 9. Above this pH, hydrolysis of Fe+2 to
FeOH+ (reaction 37) will slightly influence the relations for total soluble Fe(II). Up
to about pH 10, soluble Fe(II), as a function of [H-f:] and [Alk] or CT, can be
estimated by means of the following equation:
[Fe(II)] = [Fe+2] + [FeOH+] =
K37/[H+])
(42)
In passing, we should be aware that the solubility of ferrous iron can also be con-
trolled by the solubility of ferrous sulfide. The presence of small quantities of S(II)
312
CHEMISTRY IN RELATION TO WATER QUALITY
-------
components (H2S, HS , S 2 and polysulfides), as they may occur in hypolimnetic
waters (e.g., through bacterially mediated reduction of sulfate), is inconsistent with
the presence of appreciable amounts of soluble ferrous iron. Frequently, under natural
conditions, ferrous iron controls the amount of soluble sulfur, S(II), constituents
rather than vice versa. A quantitative evaluation of metal sulfide solubilities is frequently
difficult, because solubility products are not known with sufficient accuracy, and the
existence of various polysulfide species makes a simple interpretation impossible. As
a first approximation, the Fe(II) solubility, as a function of the total sulfide,
[S(ID] = [H2S] + [HS-] + [S-2],
in sulfide-containing waters can be estimated by
[H+]/K4
K)
(43)
Since FeS is less soluble than FeCO , deposited FeCOu can be converted by low
concentrations of S(II) into black FeS(s) (or FeS2(s)) :
FeCO.(s) f HS- = FeS(s) + HCO,-; K =
MnCO.
= 105-5
(44)
In rocks manganese is less abundant than iron, but like iron it occurs in multiple
oxidation states. Upon addition of 0.1 mole of MnO per liter of our solution we would
observe that a similarly small quantity of Mn(II) would go into solution. The solubility
relations of Mn(II) are very similar to those of Fe(II), as is evident from comparison
of Figures 8 and 9. The latter figure has been constructed by using the following
equilibrium constants: log KMnOH2 = 13; log KMnOO3 =10.41; log KX (Mn+2 +
H20 = MnOH+ + H+) = 10.6; JogKMnHOO3 + (Mn+a + HCQ- = MnHCO,-f)
-3
-5
-7
-9
MnHCO,+
10
12
PH
Figure 9 Maximum Soluble Mn(ll) in a
CT=10-3M Water.
Stumm
313
-------
= 2. It is seen that the MnC03 solubility equilibrium controls the solubility of Mn+2
of most natural waters. Most of the MnO that has been added to our mixture will be
converted to MnCO . (This conversion will increase the pH of the solution somewhat.)
OTHER METALS
It is beyond the scope of this discussion to estimate solubility equilibrium relations
for all the significant cations in water. Metal carbonates do not seem to control the
solubility of Mg+2 and Cu+2. The solubility of magnesium can be calculated from the
solubility product of Mg(OH)2 and from the first hydrolysis constant of Mg+2. The
solubility of bivalent copper was estimated in Figure 10 as a function of pH; the follow-
ing constants were used: log KCll(OH)2 = 18.8, log KCuC03 = 9.6 log K± (Cu+2 +
H, 0 = CuOH+ + H+) = 6.8. It appears from this figure that the predominant
soluble Cu(II) species in most natural waters is CuOH+. The solubility of copper
increases again at high pH values, because of hydroxo and/or carbonato complex forma-
tion. In the author's laboratory, C. Schneider has determined a stability constant of
approximately 1010 for a soluble [Cu(C03) 2]~2 complex.
12
Figure 10 Maximum Soluble Zn(ll) in a
CT=10-3M Water.
Figure 11 Maximum Soluble Cu[ll} in a
CT=10-3M Water.
Maximum soluble Zn+2 is plotted as a function of pH in Figure 11 (log KZnOH2
= 16.0; log KZnC03 = 10.8, log Kt = 8.7, and log /33 (Zn+2 + 3 OH- =
ZnOH3-) = + 14], For a CT of 10~3M, the zinc solubility is controlled by ZnC03(s)
below pH 7.5, and by Zn(OH)2(s) above that pH. "With a 10-fold increase of CT, the
zinc solubility would appear to be solely controlled by the ZnC03(s) equilibrium.
It might be well at this time to remind the reader that the solubility predictions are
based on the selected constants, which might be in error, and all the constants are of
course subject to revision as more and better data become available. Frequently, not
yet identified species such as soluble Cu(OH)2 or Zn(OH)2 or carbonato complexes
might influence the solubility behavior drastically.
OXIDATION REDUCTION POTENTIAL
In our imaginary experiment, we now open the system to the atmosphere representing
314
CHEMISTRY IN RELATION TO WATER QUALITY
GPO 614-105-11
-------
a huge reservoir containing oxygen at a fixed partial pressure, PO2, of approximately
0.2 atmosphere. The dissolution of oxygen can be described by Henry's law
[02]=kP02
where k is the equilibrium constant (Henry's law constant) for the oxygen solubility.
The oxygen that becomes dissolved might react with some of the constituents in our
system. We can visualize for example the oxidation of ferrous iron and Mn(II) to ferric
oxide hydrate and to manganese dioxide, respectively:
2 Fe+s + 1/2 02 + 2 H30 = 2 Fe(OH)3 + 4 H+ (46)
and
Mn+2 + 1/2 02 + H20 = Mn02 + 2 H+ (47)
But any oxygen consumed incipiently in these redox reactions will be refurnished
from the atmospheric reservoir so that at equilibrium the dissolved oxygen concentration
will still be denned by equation 45. It is thus obvious that in any water system that is in
equilibrium with the atmosphere the redox potential is defined by the solubility of
oxygen at the given partial pressure of oxygen. All other redox couples, Fe(OH)3
Fe+2, Mn02 Mn+2, etc., will at equilibrium be adjusted in such a way that the ratio
of their activities corresponds to the redox potential of the 02 H.,,0 couple. By applying
the Peters-Nernst equation, this redox potential of the reaction
H20 = 1/2 02 + 2 H+ + 2 e- (48)
can be defined by
.-^-S"-2^
or for conditions in the model
EH = us + - log p. (49b)
where EH is the electrode potential of the half reaction (48) , as compared with the
standard hydrogen gas hydrogen ion couple:
H2 + 2 H+ + 2 e, E0 = 0 (50)
By this convention, the potential is negative if the reductant in the half reaction under
consideration is a better reductant than hydrogen gas and should reduce H+ to Hg; in
other words, a high EH can be interpreted as a high oxidation intensity, or more pre-
cisely, a low electron activity, and a low (negative) EH reflects a high electron activity.
On the basis of equation 48, the EH of our system is given by
EH = 1.23 + 0.0295 x V2 log 0.2 0.059 pH (51)
or
EH = 0.773 v; EH = 0.714 v; EH = 0.665 v
pH 6 pH 7 pH 8
These values show that EH is slightly dependent on pH. The presence of dissolved
oxygen is certainly a dominant factor in the oxidation intensity of a water, but it is
interesting to note that the potential is remarkably insensitive to changes in the dis-
solved oxygen concentration. Reducing the oxygen concentration 99 percent, i.e., from
10 to 0.1 milligram per liter, will lower the potential by only 30 millivolts.
THE ABUSE OF EH MEASUREMENTS IN NATURAL WATERS
Under certain circumstances, electrode potentials can be determined experimentally
by inserting an inert metal like platinum in combination with a reference electrode into
the solution. With the availability of such an experimental method for electrode potential
measurements, it becomes very tempting to use such a procedure for the investigation of
reduction and oxidation conditions in waters. For nearly 40 years sanitary engineers and
315
Stumm
-------
water chemists have based results on the misconception that they were able to evaluate
the total oxidative (reductive) capacity as well as oxidation (reduction) intensity by
such a comprehensive technique. Unfortunately, these measurements have, in the opinion
of the author, failed to yield results amenable to intelligible interpretation. Similarly,
anyone who attempts to verify equation 51 for the oxygen-water system by measuring the
EH in an oxygen equilibrated water soon becomes frustrated by the significant discrep-
ancy between observed and calculated data, and by his failure to obtain reproducible
EH readings.
CONCEPT VERSUS MEASUREMENT OF THE POTENTIAL
In textbooks, generally, the concept of electrode potential in oxidation reduction
processes is introduced by considering the thermodynamic properties of electrochemical
cells. It is necessary however to distinguish between the concept of the potential, as it
is employed by Latimer'' and others, and the measurement of an electromotive force in
an actual cell. Potentials quoted by Latimer and by others have been derived from
equilibrium data, thermal data, and the chemical behavior of a couple with respect to
known oxidizing and reducing agents, and from the direct measurements of cells. The
conceptual meaning of a particular potential, in the thermodynamic sense, is that it is the
equivalent free energy, i.e., the free energy change per mole of electrons associated with
a given oxidation or reduction:
TT _ AF (52)
EH~ -sr
where AF is the free energy, f is the Faraday, and n is the number of moles of electrons
involved per mole of reactant. There is no a priori reason to identify the thermodynamic
potentials with measurable electrode potentials in a given aqueous system.
The measurement of an electrode potential involves a question as to the electro-
chemical reversibility or irreversibility of the electrode reaction characterized by the
rate of electron exchange at the electrode (exchange current). It is realized that EH
measurements are of great value in a few systems for which the variables are known
and under control.
Some of the essential principles involved in the measurement of an electrode poten-
tial can be qualitatively described by a consideration of the behavior of a single
electrode (platinum) immersed into a Fe+2 Fe+3 solution. To cause the passage of
a finite current at this electrode, it is necessary to shift the potential from its equilibrium
value. One thus obtains a curve depicting the electrode potential as a function of the
applied current (polarization curve). At the equilibrium potential, i.e., at the point of
zero applied current, the half reaction
Fe+3 + e ^ Fe+2
is at equilibrium; but the two opposing processes, the reduction of Fe+3 and the oxida-
tion of Fe+2 proceed at an equal and finite rate that can be expressed by the exchange
current. As indicated in Figure 12, the net current can be visualized as the algebraic
summation of two opposing currents (cathodic and anodic). The rate of Fe+3 reduction
(conventionally expressed as cathodic current) generally increases exponentially with
more negative electrode potential values and is furthermore a function of the concentra-
tion of Fe+G and of the effective electrode area. Similar considerations apply to the
rate of Fe+J oxidation (anodic current), which is proportional to [Fe+2], electrode area,
and the exponential of the potential. It is obvious from the schematic representation of
Figure 12 that in the case of Fe+z Fe+3, provided that the concentration of these
316 CHEMISTRY IN RELATION TO WATER QUALITY
-------
ions is sufficiently large, e.g., 10^3 to 10^*M, an infinitesimal shift of the electrode
potential from its equilibrium value will make the half reaction proceed in either of
the two opposing reactions. Operationally, the measurement of the equilibrium electrode
potential in such a case is feasible. We might contrast such behavior with the condi-
tions we would encounter in attempting the measurement of the electrode potential in
distilled water containing dissolved oxygen. A schematic representation of the polariza-
tion curve for this case is given in Figure 13. The equilibrium electrode potential should
0, - H,0
Figure 12 EH Measurement in
FE+2 Fe+3 System.
H.0-0,
Figure 13 EH Measurement in H3O-O2
System.
3
Figure 14 EH Measurement in Fe+3
Fe+8 System Occurence of Mixed Poten-
tial Because of Low Concentration of Fe+2.
again be located at the point where the net applied current (i.e., the algebraic sum of
cathodic and anodic currents) is zero. The exact location is rendered very difficult. Over
a considerable span of electrode potentials, the net current is virtually zero; similarly,
the electron exchange rate, or the exchange current reflecting the opposing rates of
the half reaction,
H20 Pt i/2 02 + 2 H+ + 2 e
is virtually zero. Operationally, a remarkable potential shift must be made to produce
a finite net current and the current drawn in the potentiometric measurement is very
large compared with the exchange current. Even with modern instrumentation in which
the current drain can be made extremely low, the experimental location of the equili-
brium potential is ambiguous. Furthermore, because of the negligible exchange current,
the rate of attainment of the equilibrium potential is very low. The measured potentials
drift for hours or even days, and the steady state potential, which is essentially reached,
is neither reproducible nor indicative of the thermodynamic electrode potential. Such
a system is called an electrochemical irreversible system, and its redox reactants are
called non-electroactive. Many redox reactants encountered in natural waters behave
Stumm
317
-------
irreversibly at inert electrodes; these reactants include sunde-sulfur-sulfite-sulfate, NO"-
_ NO - N NH2OH NH3, C103~ OC1~ Cl", most organic redox couples, etc.
The measuring electrode is very easily contaminated by insidious trace quantities of
tensioactive materials. Although such a contamination does not necessarily affect the
equilibrium position of the potential, it generally leads through adsorption to a significant
reduction in effective electrode area and thus reduces markedly the exchange current,
which in turn results in a much more sluggish response of the electrode; thus systems
that are otherwise electrochemically reversible may become irreversible.
It is necessary to introduce an additional and possibly most important restriction
regarding the measurement of 1%: The point of zero-applied current in the polarization
curve is not necessarily the equilibrium potential. Figure 14 schematically depicts the
polarization curves for the electroactive Fe+2 Fe+3 system at various concentrations
of Fe+- and Fe+3. The measured equilibrium electrode potential is in accord with the
potential calculated according to the Peters-Nernst equation as long as the concentrations
of Fe+2 and Fe+3 are larger than about 10~5M. (This threshold concentration de-
pends on the effective electrode area.) Below these concentrations, the measured poten-
tial can no longer be interpreted in terms of the Peters-Nernst equation. If for example,
Fe+3 is larger and Fe+2 is smaller than 10~6M, respectively, the measured electrode
potential becomes independent of Fe+2.15 It is evident from Figure 14 that under
these conditions the measured equilibrium potential is defined by the point where the
equivalent rate of Fe+3 reduction is equal to the equivalent rate of H20 oxidation
(H,0 = % 02 + 2 H+ + 2 e). Such a potential is of course no longer characteristic
of the Fe+3 Fe+2 system. Such a potential is called a mixed potential and bears
no simple relationship to the activities of the reacting species. Correspondingly, in a
solution of Fe+2 containing less than 10~5M Fe+3, the measured potential would drift
to a value where the rate of reduction of H+ (or H2O) would just equal the rate of
oxidation of Fe+2 to Fe+3. Since Fe+3 ions are produced in the reaction, the measured
potential would slowly be shifted until eventually an equilibrium would be reached
(e.g., after days) in which both half reactions would be at the same potential. This
potential would have no bearing, however, on the incipient activities of Fe+2 and Fe+3.
It is obvious that minute trace quantities of oxidants other than H+, e.g., oxygen or an
oxide film at the electrode, that are reduced at less negative potentials than H+ might
significantly affect the potentiometric reading.
Of the redox reactants in natural water systems, Fe+2 and Fe+3 may be among the
most electroactive species. As is evident from the solubility considerations given in this
paper for Fe+2 and Fe+3, the concentration of free Fe+* and Fe+3 should very seldom
exceed 1Q-5M.
We must, therefore, conclude that most EH measurements carried out in natural
water systems represent mixed potentials. Under conditions of a mixed potential, a net
chemical reaction is proceeding at the electrode and the potential is not characteristic of
either half reaction. The measured value of the electrode potential cannot be interpreted
quantitatively by simple relationships such as that given by the Peters-Nernst equation.
It certainly might be expected that fresh waters containing primarily oxidizing agents
give high EH measurements and those containing predominantly reducing agents exhibit
low EH readings, but a quantitative interpretation does not appear to be justified. A trust-
worthy analysis of some of the pertinent constituents of the water, e.g., 02, HS~,
N02~, NHa~, Fe(II), and Fe(III), that can be carried out more precisely and usually
318 CHEMISTRY IN RELATION TO WATER QUALITY
-------
faster and simpler than an EH measurement is generally much more informative than
an EH reading.
FE(in) AND Mn(IV)
Fe(III)
After this digression into a discussion of the concept and measurement of the ORP,
we resume the discussion of aqueous iron. In oxygenated water, ferrous iron is oxidized
to the ferric iron. The solubility of Fe(III) in natural waters is controlled by the
solubility of ferric hydroxide or ferric oxide hydroxide, FeOOH. The equilibrium
constants (reactions 53-57) used in the construction of Figure 15 are listed in Table 5.
10
12
PH
Figure 15 Maximum Soluble Fe(lll).
Table 5 Fe (III) Solubility
Reaction No.
53
54
55
56
57
58
59
Reaction
Fe(OH)3(s) = Fe+s + 30H-
Fe(OH)3(s) = FeOH2+ + OH~
Fe (OH)3(s) = FeOH+2 + 2 OH
Fe(OH)3(s) + OH- = FeOH4-
FeP04(s) = Fe+3 + P04~s
Fe+s + HP04~2 = FeHP04
Fe+s + SiO(OH)3- = FeSiO(OH)3+2
logK
36.0
14.77
24.17
5.0
23.0 [16]
+ 8.4 [16]
+ 9.3 [16b]
According to Figure 14, at pH 7 the following constituents of soluble Fe(III) are in
saturation equilibrium with Fe(OH)3(s): Fe+s = 1(>-i6; FeOH+* = 6 x 10-";
Stumm
319
-------
_ 2 x 10_s. FeOH = 1012. Total soluble Fe(III) is thus in the order
of only 1 micro gram per liter.
For an air-saturated water with an EH of 0.717 volts (pH 7), the equilibrium
concentration of Fe+2 can be calculated by applying the Peters-Nernst Equation to
the reaction
Fe+2 + 3 OH- = Fe(OH)3(s) -|- e; E0 = 1.31 v (60)
0.717 = - 1.31 + 0.059 log [Fe+2] [QH-p (61)
For pH 7, the calculated equilibrium concentration of Fe+2 amounts to approximately
5 x 10-".
Virtually no iron should exist in solution in equilibrium with the atmosphere. This
does not appear to be in accord with the analytical findings for real systems. Real
systems may not, however, be in equilibrium with oxygen.18 Furthermore, the solubility
of ferric iron might be enhanced by complex formation with inorganic constituents, e.g.,
phosphate and silicate complexes, or organic constituents. Analytically, it is rather
difficult to distinguish between dissolved and suspended iron. Lengweiler, Buser, and
Feitknecht17 have shown that with very dilute Fe(III) solutions containing Fe59 as
tracer and brought to a pH between approximately 5 and 12 all the iron hydroxide can
be sedimented by ultracentrifugation (93,000 g, 180 min). The size of the Fe(OH)3
particles varies with the pH of the solution. The diameter can be as small as 100 A°-
It is obvious that nitration (even through membrane niters) does not always provide a
satisfactory operation for the distinction between the dissolved and suspended fractions
of a particular species.
As we have seen, both ferrous and ferric iron generally are not very soluble in
natural waters. Despite this low solubility, the capability of iron to undergo reversible
oxidation and reduction reactions plays a significant role in the chemistry and biology
of natural waters. In limnology, the redox reactions of iron are related to the metabolic
cycles of nearly all other important elements and to the distribution of oxygen in a body
of water.18 During the seasonal variations in an eutrophic lake, the continuous sequence
of circulation and stagnation is accompanied by oxidation and reduction as well as
precipitation and dissolution of iron. This leads to a progressive accumulation of iron
in the lake sediments. In many lakes, interesting correlations between the concentra-
tions of Fe(II) and Fe(III) and those of phosphates and silicates are observed. The
strong affinity of phosphates and silicates to Fe(III) (reactions 57-59) might provide
an important clue for a more quantitative interpretation of such correlations.
MANGANESE
Figure 16 gives a redox potential pH diagram for manganese. At the potential of
an air-saturated solution Mn+2 is thermodynamically unstable. In the absence of strong
complex formers, Mn(III) does not occur as a dissolved species. In Figure 16 it is
seen that Mn02(s) is the only manganese oxide phase that would be stable in oxygenated
waters. In deep-sea sediments Mn02 is indeed an abundant constituent. The manganese
concentration found in aerated fresh waters probably consists of a mixture of Mn(II)
and colloidally dispersed Mn02. The oxidation of Mn(II) to higher valent manganese
oxides has been found20 to be strongly pH dependent and autocatalytic. Below pH 8.5,
the rate of oxygenation is extremely low. The oxygenation does not lead to stoichiometric
oxidation products such as MnO2, MnOOH, or Mn304. The results of studies on the
oxidation of Mn(II) can best be interpreted by assuming that the oxidation products
consist of Mn02 onto which various quantities of Mn+2 have been adsorbed. Colloidal
320 CHEMISTRY IN RELATION TO WATER QUALITY
-------
aqueous manganese dioxide has been shown to have a remarkable ion-exchange capacity
for Mn(II) and other metal ions, since this property is strongly dependent upon pH.
Sorption capacities in excess of 0.5 mole of Mn+2 per mole of Mn02 are found in the
slightly alkaline pH range.21
Figure 16 Redox Potential pH Diagram
for Aqueous Manganese.
Aqueous manganese dioxide, to a pronounced degree, possesses some of the character-
istics that appear to be generally applicable to an interpretation of properties of poly-
valent metal oxide hydrates. In a similar way, ferric hydroxide exhibits cation-exchange
properties, especially at high pH. At high pH values, exchange capacities as high as
1 equivalent per mole of hydrous metal oxide (e.g., Mn+2 on Fe(OH)3) are not un-
common. Cation exchange on the hydrous oxides is comparable to the cation exchange
on clay materials. Such ion-exchange phenomena on hydrous metal oxides and other
precipatates (solid solutions) represent special cases of heterogeneous metal ion buffers.
The concept of solid solutions provides one possible explanation for the observed occur-
rence of certain impurities (e.g., metal ions) in sediments that have subsided from
solutions apparently (without considering the activity coefficients of the solid) un-
saturated with respect to the impurity.
FINAL REMARKS
The imaginary experiment could of course be continued at great length and many
of the cases that have been discussed should be treated in much more detail. But the
Stumni
321
-------
primary aim of this discussion was to show the simple methodological tools that the
chemist can use to arrive at conclusions on mineral relations in natural waters. All the
information gathered together in the examples discussed has been taken from standard
reference tables on the energies or on the relative stabilities of various compounds. It
is regrettable that this easily available information has not been sufficiently used in the
past to help answer many of the qualitative and quantitative questions involved in the
mineral relations of natural waters and to serve as a guide in the interpretation of
analytical results.
An attempt has been made to describe the stability relationships of the distribution
of the various soluble and insoluble forms through rather simple graphic representations.
The principle involved in elucidating the equilibrium relationships consists essentially
in writing down as many equations as one has unknowns and to solve them. A simul-
taneous graphical representation of all the requisite equations gives the means for
attacking even very complicated systems. Two types of graphical treatments have been
used in this discussion: first, equilibria between chemical species in a particular oxida-
tion state as a function of pH and solution composition; second, the stability of different
oxidation states (potential-pH diagram) as a function of pH and solution composition.
Diagram of the latter type require for their delineation an intensity factor as a variable
representing the stability of the various oxidation states. Since the redox potential, in
most cases, cannot be measured operationally (it can be computed, of course), the
potential-pH diagrams are somewhat less amenable to a simple and direct interpretation
than the log concentration pH diagrams.
From the few examples discussed, it has become apparent that there are considerable
gaps in our information. Many equilibrium constants are only approximately known
and some are missing. But we also lack information on the real systems. Many reported
analytical data of natural waters are unreliable. For example, the author would doubt
the reliability of most of the results that have been published on the Fe(III) and Fe(II)
content or on the phosphorus and sulfide concentration of natural waters. Many of the
analytical methods we use are not sufficiently specific; it is also difficult to distinguish
analytically between dissolved and suspended species. Extensive redox potential measure-
ments in natural media have failed to yield information that can be interpreted quantita-
tively. It is hoped that all these obvious shortcomings represent an incentive for careful
investigations in the future.
REFERENCES
1. Sillen, L. G., Proc. Inter. Ocean. Congress, Publ. No. 67, p 549, AAAS, Washington,
D. C., 1961
2. Bjerrum, J., G. Schwarzenbach and L. G. Sillen, "Stability Constants," The Chem.
Soc., London, 1958
3. Latimer, W. M., "Oxidation States," Prentice Hall, 1952
4. Goldschmidt, V. M., J. Chem. Soc., 655 (1937)
5. Cited from Hem, J. D., "Geological Survey Water-Supply Paper 1473" (1959)
6. Lagerstrom, G., Acta Chem. Scand., 13, 722 (1959)
7. Weber, W. J. and W. Stumm, J. Chem. Engr. Data, July 1963
8. Brosset, C., G. Biedermann and L. G. Sillen, Acta Chem. Scand., 8, 1917 (1954)
9. Matijevic, E., et. al., J. Phys. Chem., 65, 826 (1961)
322 CHEMISTRY IN RELATION TO WATER QUALITY
-------
10. Sillen, L. G., in Treatise on Analytical Chemistry, Part 1, Vol. 1, p. 277, Interscience,
New York (1959)
11. Weber, W. J., and W. Stumm, Jour. A.W.W.A., 55, Oct. 1963
12. Greenwald, L, J. Biol. Chem., 141, 789 (1941)
13. Larson, T. E., and A. M. Buswell, Jour. A.W.W.A., 34, 1667 (1942)
14. Morgan, J. J., Thesis, Harvard University, 1963
15. Coursier, J., Anal. Chim. Acta, 7, 77 (1952)
16b. Weber, W. J. and W. Stumm, to be published (1964)
17. Lengweiler, H., W. Buser and W. Feitknecht, Helv. Chim. Acta, 44, 805 (1961)
18. Stumm, W., and G. F. Lee, Ind. Eng. Chem., 53, 143 (1961)
19. Stumm, W., and G. F. Lee, Schweiz. Z. Hydrologie, 22, 295 (1960)
20. Morgan, J. J., and W. Stumm, Presented ACS Meeting, New York, Sept. 1963
21. Morgan, J. J., and W. Stumm, J. Coll. Sci., 19, 347 (1964)
Stumm 323
-------
Dr. Gerard A. Rohlich
Professor of Sanitary Engineering
University of Wisconsin, Madison
SUMMARY
The analysis of water measurement data for basic relationships among hydrological,
chemical, and biological parameters is discussed. The data to be assembled and in-
terpreted by the sanitary engineer concerned with environmental problems usually are
gathered by scientists in other disciplines; thus, the sanitary engineer mast rely heavily
on the validity of their interpretations. A properly planned program with well-defined
objectives is of paramount importance. Inadequate planning of sampling procedures is
more likely to lead to erroneous conclusions than are correlation and statistical handling
of data. Another source of error in interpreting results and drawing conclusions in the
study of water supply and water pollution control problems lies in the relating of
laboratory studies to field situations. The dynamic system in nature is frequently over-
simplified; adjustments of variables in laboratory experiments seldom parallel changes
in the natural system.
DATA INTERPRETATION DRAWING CONCLUSIONS
In giving consideration to the subject of our discussion, "Data Interpretation
Drawing Conclusions," I was reminded of the story that Professor E. B. Phelps told in
the preface to his book Stream Sanitation.1 After pointing out that it might appear that
the subject stream sanitation was rather specialized and could not be "contained within
definite boundaries such as scientists are so fond of laying down," he then went on to
relate that while serving as an expert witness in a stream pollution case he was
questioned at length during cross-examination concerning his title of Professor of
Sanitary Science and the scope of his expert qualifications. "Are you a biologist?" he
was asked, "a chemist? a botanist? Does your knowledge cover the physiology of fish,
and the geology of the area?" To all of these questions he felt he had to reply in
the affirmative, with qualifications, for his testimony had, in fact, as he states, "tres-
passed upon all these 'fields' of science.''
The sanitary engineer usually finds himself in this position in the interpretation of
data and in drawing conclusions, and in fact in his assessment of a situation he must
frequently consider many other facets such as flood control, power development, and
irrigation, as well as political and economic factors, when practical situations are
confronted.
Obviously the engineer relies heavily on the chemist and biologist to supply water
measurement data that may be integrated with physical data in order that the overall
evaluation can be made.
As has been mentioned on more than one occasion during this conference, there
is no substitute for a properly planned program, if this is at all possible, before em-
barking on an extensive, costly, and time-consuming project. In any research project
we are well aware of the need for experimental design in the planning stages, without
which the results may not be worth treating statistically.2 It has been said that to
get the right answers we must ask the right questions. Certainly the investigator must
ask "what is the objective?" and formulate simple, clear, specific aims that are as
refined as possible. If this is done, a good start has been made, providing the objective
is realistic in terms of time and resources available. The most important and most
Rohlich
325
-------
difficult task in any program is to know when to stop the easiest thing to do is to
continue to get more data. Frequently, more data are sought in the hope that perhaps
by some chance a key piece of information will appear that ''may unlock the puzzle
and conclusions will fall out and become self-evident." This is usually a forlorn hope.
In the laboratory research experimental method in which an event occurs under
known conditions where "as many extraneous influences as possible are eliminated and
close observation is possible, relationships between phenomena can be found."2 The
experimental method is not appropriate to all types of research, however. In the field
of environmental measurements, as has been emphasized at this meeting, the unknowns
and variables remain in many instances unknown; since we do not have the "controlled
experiment," we become purely observational investigators. This position is frequently
one to which we do not adjust readily. The principles of the experimental method are
not to be forgotten, however. The main difference, as Beveridge2 states, is that the
hypotheses are tested by collection of information from phenomena that occur naturally
instead of those that are made to take place under experimental conditions. Unfortunately,
although considerable useful data are available for formulating conclusions, there are
gaps and limitations, and it is unlikely that we shall ever fully understand the
ecological pattern involved in man's relationship to the water environment.
Perhaps the biologists are more aware of the complexity of the microcosm with
which we are concerned than are some of the rest of us, and, as in the past, will continue
to contribute to an understanding of the relationship of the parameters of water quality,
which we now know how to measure, to the environmental problem.
Despite the limitations that we may have in understanding the whole structure,
there is no escape from the fact that we are in many instances required to obtain
results that will have some practical application, initially to aid in understanding a
problem, and, through understanding, to arrive at a solution, however inadequate that
solution may be in the light of subsequent information. There is, of course, danger in
separating our activity from our contemplation. All too often we become too rigid in
the cataloging of existing knowledge; in particular those of us in engineering are
inclined to rely heavily on mathematical symbols and models (perhaps rather than an
understanding of mathematics). As was pointed out by Lord Kelvin, "Nothing can be
more fatal to progress than a too-confident reliance on mathematical symbols, for the
student is only too apt to take the easier course, and consider the formula, and not the
fact, as the physical reality." I don't mean to say that if we have all the data, stored,
and readily retrievable we won't be able to find out things that our common sense
would not lead us to. But the common sense approach is still useful and should not
be discarded.
We must ask ourselves whether the information we would like to have is really
going to be useful. If we are considering water quality and water pollution, we must
consider carefully the parameters involved in our specific problem. We are inclined to
speak glibly about pollution without defining it for the reason that it defies defnition.
We recognize that pollution is strictly a relative term and depends upon the particular
use a person wants to make of the water. What might be polluted to one user is far
from being polluted to another. Consequently, the parameters that might interest one
person are quite different from those that might interest another.
If we are to use intelligently the massive accumulation of data and avoid the
separation of activity from contemplation, we should refer frequently to the basic con-
siderations of and the reasons for the data gathering. We must ask ourselves critically:
326 DATA INTERPRETATION (WATER)
-------
For what purpose do we intend to use these data? What do we wish to find out or
define by these data?
As mentioned previously, the engineer concerned with environmental problems
usually is confronted with the assembly and interpretation of data gathered by scientists
in other disciplines. In pollution studies he must rely heavily on the chemist, biologist,
hydrologist, meteorologist, oceanographer, economist, and frequently on the political
scientist and lawyer before his final conclusions can be made. The engineer, like the
others, can not claim to be an expert in all these fields and so must rely heavily on
the validity of the individual expert's interpretations of his data. Frequently, there is an
imbalance in the kinds of data obtained, and the mistake of placing reliance on meager
data is a pitfall to be avoided. Although precision my be apparent with minimal informa-
tion, extension of the study may well indicate that the limited data at best were re-
flecting a low or high portion of a trend that was in fact related to some other en-
vironmental factor that may or may not have been properly considered.
Thus, in the interpretation of data, the qualitative factor should be evaluated before
the quantitative aspects are considered. The extent and replication of sampling in
relation to the complexity of the area under study are of obvious importance as guides
in determining the reliability of the conclusions drawn. In dynamic systems such as
lakes and streams, physical, chemical, and biological properties are related to time of
sampling, and misleading or erroneous conclusions frequently result unless careful
consideration is given to the representation of the particular samples. A knowledge of
the extent to which homogeneity exists in the body of water sampled is equally as
important as the analytical procedures used on the samples obtained. The correlation
and statistical handling of data, although not always simple procedures, are much less
serious problems and are less likely to lead to erroneous conclusions than the errors
resulting from inadequate planning of the sampling procedures. A useful reference in
this regard is the Geological Survey Water-Supply Paper 1473 on the study and interpreta-
tion of the chemical characteristics of natural Water.3
Another source of error in interpretating results and drawing conclusions in the
field of water supply and water pollution control lies in the relating of laboratory studies
to field situations. The relation of occurrences in the laboratory to the dynamic system
in nature is frequently oversimplified, and the conclusions drawn from adjustment of
variables in laboratory experiments seldom parallel similar changes that can be made
or might occur in the field.
Dr. Stumm has pointed to the fact that many gaps remain in our information re-
garding the chemistry of natural waters in relation to water quality and Dr. Hynes has
made reference to the complexity of interpretation of biological data with reference to
water quality. Although there seems to be little need to emphasize the statements of
these experts in chemistry and biology, their papers serve as reminders of the dangers
of oversimplification in interpreting data and in drawing conclusions.
REFERENCES
1. Phelps, E. B. Stream Sanitation. John Wiley and Sons, Inc., New York, 1944.
2. Beveridge, W. I. B. The Art of Scientific Investigation. 2nd edition. William Heine-
mann Ltd. London (1953).
3. Hem, J. D. Study and Interpretation of the Chemical Characteristics of Natural
Water. Geological Survey Water-Supply Paper 1473. U. S. Government Printing
Office, Washington, D.C. 1959.
Rohlich 327
GPO 81410512
-------
BIBLIOGRAPHIC: Robert A. Taft Sanitary Engineering
Center. ENVIRONMENTAL MEASUREMENTS: VALID
DATA AND LOGICAL INTERPRETATION. A Sym-
posium. PHS Publ. No. 999-AP-15 (or No. 999-WP-15).
1964. 327 pp.
ABSTRACT: This Symposium on Environmental Measure-
ments, held in Cincinnati in September 1963, was jointly
sponsored by the Division of Air Pollution and the Divi-
sion of Water Supply and Pollution Control of the Public
Health Service. The Proceedings contain 26 papers by
experts on the major operational steps that are part of a
measuring system: sampling, detecting, recording, vali-
dating, interpreting, and drawing conclusions. Discussions
are also included.
ACCESSION NO.
KEY WORDS:
BIBLIOGRAPHIC: Robert A. Taft Sanitary Engineering
Center. ENVIRONMENTAL MEASUREMENTS: VALID
DATA AND LOGICAL INTERPRETATION. A Sym-
posium. PHS Publ. No. 999-AP-15 (or No. 999-WP-15).
1964. 327 pp.
ABSTRACT: This Symposium on Environmental Measure-
ments, held in Cincinnati in September 1963, was jointly
sponsored by the Division of Air Pollution and the Divi-
sion of Water Supply and Pollution Control of the Public
Health Service. The Proceedings contain 26 papers by
experts on the major operational steps that are part of a
measuring system: sampling, detecting, recording, vali-
dating, interpreting, and drawing conclusions. Discussions
are also included.
ACCESSION NO.
KEY WORDS:
BIBLIOGRAPHIC: Robert A. Taft Sanitary Engineering
Center. ENVIRONMENTAL MEASUREMENTS: VALID
DATA AND LOGICAL INTERPRETATION. A Sym-
posium. PHS Publ. No. 999-AP-15 (or No. 999-WP-15).
1964. 327 pp.
ABSTRACT: This Symposium on Environmental Measure-
ments, held in Cincinnati in September 1963, was jointly
sponsored by the Division of Air Pollution and the Divi-
sion of Water Supply and Pollution Control of the Public
Health Service. The Proceedings contain 26 papers by
experts on the major operational steps that are part of a
measuring system: sampling, detecting, recording, vali-
dating, interpreting, and drawing conclusions. Discussions
are also included.
ACCESSION NO.
KEY WORDS:
------- |