EPA-R5-73-019
MAY 1973
Socioeconomic Environmental Studies Series
Stream Quality Preservation
through Planned Urban Development
                    £
  S3SZ
                              UJ
                              a
                               Office of Research and Monitoring
                               U.S. Environmental Protection Agency
                               Washington, D.C. 20460


-------
            RESEARCH REPORTING SERIES
Research reports of the  Office  of  Research  and
Monitoring,  Environmental Protection Agency, have
been grouped into five series.  These  five  broad
categories  were established to facilitate further
development  and  application   of   environmental
technology.   Elimination  of traditional grouping
was  consciously  planned  to  foster   technology
transfer   and  a  maximum  interface  in  related
fields.  The five series are:

   1.  Environmental Health Effects Research
   2.  Environmental Protection Technology
   3.  Ecological Research
   U.  Environmental Monitoring
   5.  Socioeconomic Environmental Studies

This report has been assigned to the SOCIOECONOMIC
ENVIRONMENTAL   STUDIES   series.    This   series
describes  research on the socioeconomic impact of
environmental problems.  This covers recycling and
other  recovery  operations   with   emphasis   on
monetary incentives.  The non-scientific realms of
legal   systems,  cultural  values,  and  business
systems  are  also  involved.   Because  of  their
interdisciplinary  scope,  system  evaluations and
environmental management reports are  included  in
this series.

-------
                                                   EPA-R5-73-019
                                                   May 1973
            STREAM QUALITY  PRESERVATION
         THROUGH PLANNED URBAN DEVELOPMENT
                          By

      Robert E.  Coughlin and  Thomas R.  Hammer



                  Project 16110 DYX
                   Project  Officer

                    Mr. Bart Hague
                       Region I
          Environmental Protection Agency
                   424 Trapelo Road
           Waltham, Massachusetts  02154
                     Prepared for
         OFFICE OF RESEARCH AND MONITORING
      U. S. Environmental  Protection Agency
            Washington, D.C.   20460
For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402
              Price $2.60 domestic postpaid or $2.26 QPO Bookstore

-------
                EPA Review Notice
This report has been reviewed by the Environmental
Protection Agency and approved for publication.
Approval does not signify that the contents neces-
sarily reflect the views and policies of the
Environmental Protection Agency, nor does mention
of trade names or commercial products constitute
endorsement or recommendation for use.
                       11

-------
                         ABSTRACT

The effects of a land use plan to restrict urban development
in areas critical to the water resource system are identified
through empirical studies.  Specifically,  relationships are
established between amount, density, type, and location of
urban development, on the one hand, and stream water quality
and stream channel enlargement on the other.  The amount of
open space with such a plan as compared to that with normal
development is determined.  Perception of water quality by
untrained observers in the field and its relationship to
overall site preference is studied and pilot studies are
reported concerning the preferences of a sample of observers
for various landscape characteristics.

The evaluation of these effects is approached through house-
hold surveys designed to  determine how use of and preference
for stream sites  is related to water quality of stream and
distance of residence from stream.  In addition, the effect
of preserved open space on adjacent land values is explored
empirically.

This report was submitted in  fulfillment of Grant No. 16110
DYX to the Regional Science Research Institute under the
sponsorship of the Office of  Research and Monitoring, Environ-
mental Protection Agency.
                             111

-------
                      TABLE OF CONTENTS
SUMMARY OF FINDINGS                                           1
  Effects of Urbanization on Stream Channel Enlargement       1
  Effects of Urbanization on Stream Water Quality at
     Low Flow                                                 2
  Effects of Urbanization on the Preservation of Open
     Space Use                                                2
  Evaluation of Effects of Urbanization:  Aspects of
     Environmental Preference                                 3
  Evaluation of Effects of Planning:  Fiscal Benefits         3
  Evaluation of Effects of Planning:  Non-Fiscal Benefits     4
  Capitalization of Effects of Environmental Preservation
     in Land Values                                           4
  Conclusions                                                 5

RECOMMENDATIONS                                               7

                          Chapters

I.    INTRODUCTION                                            9
        Planning and Management of Stream Systems             9
        A Planning Approach for the Preservation of
          Water Resources                                    10
        The Plan for the Upper East Branch of the
          Brandywine                                         11
        Research on the Physical Effects of Planned
          Urban Development and on their Evaluation          13

II.    HYDROLOGIC EFFECTS                                     19
        Introduction                                         19
        Research Procedure                                   24
        Results                                              29
        Summary Conclusions                                  42

III.   EFFECT OF URBAN DEVELOPMENT ON STREAM WATER QUALITY    45
        Introduction                                         45
  '     Data on Water Quality                                47
        Preparation of Watershed Variables                   50
        Form of the Analysis                                 58
        Analysis and Results                                 61
                              v

-------
                       TABLE OF CONTENTS - Cont.
          Preliminary Analysis                                61
          Results as a Whole                                  64
          Results for Individual Chemicals                    74
        Conclusions                                           84

IV.  OPEN SPACE EFFECTS                                       87
        Introduction                                          87
        Research Procedure                                    87
        Results                                               88
          Percent of Segments Having No Development           88
          Percent of Segments Developed at Over 0.25
            Dwelling Units Per Acre                           90
          Percent of Segments Developed at Over 1
            Dwelling Unit Per Acre                            92
          Factors Related to Density of Development           92
       Conclusions and Implications                           97

V.    EVALUATION OF EFFECTS:  ASPECTS OF ENVIRONMENTAL
     PREFERENCE                                              103
       Introduction                                          103
       Agreement Among Observers on Environmental
          Attractiveness                                     104
          Research Procedure                                 106
          Findings Concerning Agreement                      106
          Consistency of Ratings                             109
       Landscape Characteristics Relevant to Preference      109
          Research Procedure                                 110
          Findings Concerning Landscape in General           111
          Findings Concerning Stream Sites                   113
          Patterns of Preference                             117
       Quantitative Measurement of Landscape Quality         118
          Research Procedure                                 118
          Findings                                           121
       The Influence of Water Quality in the Evaluation
          of Stream Sites                                    124
          Research Procedure                                 125
          Findings                                           126
       Summary and Conclusions                               139
                            VI

-------
                      TABLE OF CONTENTS - Cont.
VI.   EVALUATION OF EFFECTS:  ESTIMATING THE VALUE OF
      ENVIRONMENTAL PROTECTION                               141
        Introduction                                         141
        Fiscal Benefits                                      143
          Avoidance of Reservoir Costs                       143
          Avoidance of Water Purification Costs              144
          Avoidance of Flood Losses                          144
          Non-Fiscal Benefits                                145
        Preferences and Choice of Residential Environ-
          ments                                              147
          Research Procedure                                 147
          Findings Based on Interviews                       148
          Findings Relating Interview Responses to
            Residential Variables                            152
        Perception and Use of Streams in Suburban Areas      158
          The Effect of Distance from Residence to Stream    160
          Findings Concerning Effect of Distance of
            Residence from Stream                            161
        The Effect of Water Quality
          Research Procedure                                 168
          The Data                                           170
          Analysis and Results                               172
            Individual Indicators of Chemical Water
              Quality                                        172
            Summary Indicators of Chemical Water Quality     176
        Conclusions                                          183

APPENDIX TO CHAPTER VI                                       188

VII.  EVALUATION OF EFFECTS:  CAPITALIZATION OF EFFECTS
      OF ENVIRONMENTAL PRESERVATION IN LAND VALUES           193
        The Expression of Environmental Benefits in
          Land Values                                        193
        The Determinants of Land Values                      194
          Research Procedure                                 195
          General Results                                    200
          Results Specifically Concerning Environmental
            Attractiveness                                   204
        The Effect of a Large Urban Park on Real Estate
            Values                                           208

-------
VIII.  CONCLUSIONS                                       212
         Evaluation of Effects                           212
         Implications for Project Planning               214
         Issues Arising from this Study                  216
                          vni

-------
                         TABLES


 1  Conceptual Scheme  for Evaluation of Effects of a Plan
      for Stream Valley Preservation                         16

 2  Regression Results:  First Phase                         30

 3  Regression Results:  Second Phase                        31

 4  Means and Standard Deviations of Chemical Variables      48

 5  Primary Sources of Chemicals Found in Surface Waters     49

 6  Watershed Data                                           51

 7  Summary of Regression Results:  Unforced Regressions     65

 8  Summary of Regression Results:  Forced Regressions       68

 9  Chemical Data for Watersheds Excluded from Analysis      72

10  Comparison of Regression Coefficients for Sodium and
      Chloride                                               75

11  Regression for Potassium with Population-Related
      Variables in Density Form                              76

12  Regression for Nitrates with Population-Related
      Variables in Density Form                              78

13  Comparisons of Nitrate Concentrations in High and
      Low Discharge Watersheds                               80

14  Unforced Regression for Bicarbonates                     83

15  Percent of Segments (0-300' Band) Having No Develop-
      ment of Given Type, by Density of Township             89

16  Percent of Segments Developed at Greater Density
      than 0.25 Dwelling Units Per Acre                      91

17  Percent of Segments Having One or More Dwelling Units
      Per Acre                                               93
                            ix

-------
                       TABLES - Cont.


18  Summary of Regression Results                          96

19  Description of the Tests                              105

20  Percent of Total Variance in Attractiveness
      Accounted for by Photographic Slides (Results
      of One-Way Analyses of Variance)                    108

21  Summary of Major Significant Correlations Between
      Objective Characteristics of Stream Sites and
      Judges' Ratings - Field Test                        114

22  Regression Equations Explaining Preference by
      Measurement of Landscape Quality                    122

23  Streams Grouped According to Level of Organic Load    127

24  Correlations between Chemical and Biological
      Measurements and Perceived Water Attributes         129

25  Correlations between Objective Ratings and Perceived
      Water Attributes                                    130

26  Sources of Total Variance in Responses to Basic
      Preference Questions (Variance Among Observers
      and Stream Sites)                                   132

27  Correlations between Preference Ratings and Per-
      ceived Water Attributes                             133

28  Basic Preference Questions:  Sources of Total
      Variance in Mean Responses for Organic-Load Groups  135

29  Correlations between Activity Ratings and Chemical
      and Biological Characteristics of Water             137

30  Variation of Activity Ratings Among Levels of
      Organic Load                                        133

31  Considerations in Choice of a Place to Live
      (Number and Percentage of Respondents Making
      One or More Responses of Each Type)                 149
                            x

-------
                         TABLES - Cont.
32  Summary of Residential Preferences, For Respondents
      Asked about Choice of Present or a Hypothetical Home
      (Percentage of Respondents Making One or More Re-
      sponses for Each Major Category, or Combination of
      Categories)                                            153

33  Inter-Areal Comparison of Residential Preferences
      (Percent of Respondents Making one or More Responses
      for each Major Category, or Combination of Categories) 154

34  Correlations between Response Variables and De-
      scriptive Variables                                    156

35  Summary of Regressions Explaining Survey Responses
      by Distance Along Public Ways Between Stream and
      Residence, Socio-Economic Index, and Stream-
      Specific Dummy Variables                               162

36  Estimated Straight Line Distance of Residence from
      Stream for Given Relative Use Level                    164

37  Estimated Distance of Residence (Along Streets to
      Point of Public Access) from Stream for Given Use
      Level                                                  165

38  Summary Measures of Water Quality                        173

39  Summary of Regression Results Using July Measure-
      ments for Individual Chemical Variables                174

40  Factor Loadings on First Component of Pollution          177

41  Significant Correlations between Summary Measures
      of Water Quality and Variables Expressing Perception,
      Usefulness, and Use of Stream Sites                    178

42  Variables Used in Analysis of Sales of Undeveloped
      Parcels                                                190

43  Means and Standard Deviations of Variables               193

44  Summary of Regression Results (Dependent Variable:
      Sale Price Per Acre)                                   195

-------
                       TABLES - Cont.
45  Correlation Coefficients Between Aesthetic Variables
      and Sale Price Per Acre                                200

46  Rough Evaluation of Effects of a Plan for Stream
      Valley Preservation                                    212
                          xii

-------
                         FIGURES
 1  Effect of Urbanization  on Mean  Annual Flood  for a
      1-Square-Mile Drainage Area                           23

 2  Effect of Residential Impervious Area on  Stream
      Channel Enlargement,  as a  Function of Average Land
      Slope and Proportion  of Street Length Sewered         37

 3  Impact of Impervious Development Relative to the
      Impact of the Same Type and  Intensity of Develop-
      ment in a One-Square-Mile  Watershed                   38

 4  Influence of Flow to Channel                           40

 5  Influence of Flow in Channel                           41

 6  Illustration of Population  "Sewered-In,"  "Sewered-
      Out," and "Unsewered."                                 54

 7  Density of Dwelling Units  in Houses Found in Bands
      Alongside Streams                                     98

 8  Density of Total Impervious  Area Found  in Bands  Along-
      side Streams                                          99

 9  The Concepts of Agreement  and Differentiation          107

10  Conceptual Relationships  Between Water  Quality of
      Stream, Distance of Residence from Stream, and
      Use of Stream                                        159

11  Percent of Base Level Use  of Stream Site  by Distance
      of Residence from Stream,  for Selected  Activities     166

12  Estimated Distance from Stream for 20%, 40% and  60%
      of Base Level Use Rate  of  Typical Activity           169

13  Effect of Water Pollution  on Use  of Stream Sites       180
                            xin

-------
                     FIGURES - Cont.
14  Effect of Water Pollution on Perception of Stream
      and House Value                                     182

15  Effect of Water Pollution and Distance to Resi-
      dence on Per Cent of Households which Use Stream
      for Taking A Walk                                   183

16  Effect of Water Pollution and Distance to Resi-
      dence on Percent of Households which Engage in
      A Typical Activity (Such as Sitting or Wading)
      at Stream                                           185

17  Estimated Value of Residential Properties by
      Distance from Pennypack Park                        205
                            xiv

-------
                   SUMMARY OF FINDINGS
Land use planning of urban development can have a signi-
ficant effect in preventing deterioration of stream water
quality and in preventing the enlargement of channels which
typically accompanies urban development.

Effects of Urbanization on Stream Channel Enlargement
Development near the stream and on steep slopes is de-
leterious to the water resource system.  The effects on
streamflow and channel enlargement of development in
these critical areas, however, are probably less important
than the effects on erosion and sewage effluent.

Statistical analysis of a large sample of watersheds in-
dicates that the most important consideration concerning
streamflow and channel enlargement is type of development.
For example, if a given amount of residential development
is storm sewered, it will cause much greater channel en-
largement than if storm sewers are not provided.  The
provision of storm sewerage appears to make much more
difference than the location of the development relative
to the stream.  As another example, large contiguous areas
with impervious surface may have a greater effect than
smaller scattered sites with the same total amount of im-
pervious surface.

The effect of a given amount of development on channel
enlargement is greater for steeply sloped watersheds than
for flatter ones.  Two findings suggest that development
should be kept to a minimum at headwaters areas.  First,
the location of development at upstream points will affect
a larger length of channel than will development at down-
stream points.  Second, the absolute amount of channel
enlargement resulting from a given absolute amount of
impervious surface decreases with increase in/xwatershed
area, which of course increases as one goes downstream.

-------
Effects of Urbanization on Stream Water Quality at Low Flow
Urbanization has a pervasive effect on water quality.  During
low flow conditions significant effects involving urbaniza-
tion have been  found for a number of chemicals which are
often considered as being geologically determined rather
than produced primarily by human activities.

The pattern of  effects associated with different types of
urbanization tends to be similar for a wide variety of
chemicals.  Specifically, relatively large increases in
chemical concentrations are associated with watershed pop-
ulation in unsewered dwelling units, and with population
in dwelling units whose sewage is disposed of within the
watershed.  Smaller, but nevertheless significant, increases
are associated  with population in dwelling units whose
sewage is sewered out of the basin.  The latter results
indicate that urbanization is likely to cause lower water
quality regardless of what is done with sewage.

The study has indicated that the contribution of population
in unsewered dwelling units to streamwater pollution is
related inversely to the distance of such units from the
stream channel.  Other findings of the study are that
chemical concentrations are systematically related to
manufacturing employment and to soil drainage characteristics.

Effects of Urbanization on the Preservation of Open Space Use
A study conducted to determine the amount of development
which typically occurs near streams with normal urbanization
indicates that  less development tends to occur near streams
than occurs some distance away.  The relative lack of devel-
opment near the stream is related statistically to the size
of the drainage area of the stream and the steepness of the
stream valley.  Thus, natural development forces appear to
be consistent with the objectives of stream valley planning.

Although a rather high percentage of streamside land segments
have no development at all, the amount of development in
such segments rises as township population density rises.
In fact,  it would appear that in the early stages of devel-
opment streamside sites are left undeveloped,  perhaps

-------
because of steep slopes or wet soil conditions.  Thus, in
the later stages of development streamside sites are among
the few sites still undeveloped, and unless they are pur-
chased for public open space they are then likely to be
developed more intensively than surrounding areas.

Evaluation of Effects of Urbanization:  Aspects of
Environmental Preference
A number of pilot studies indicate that untrained field
observers can identify water quality reasonably well.
However, no significant correlations were found between
chemical measures and preference ratings for the stream,
its surrounding area, or the area overall.

Although there is a general consistency between chemical
measures of water quality, perceived attributes, and pre-
ferences for the stream and surrounding area, the field-
observer study does not indicate that water pollution is
a strong determinant of preference.  Criteria other than
water quality were evidently more important to the obser-
vers, who were generally agreed on the relative attractive-
ness of a stream.

Other investigations carried out indicate that one can
expect to find a reasonably high level of agreement on
the attractiveness of alternative environments.  Generally
preferred landscapes are those which are park-like, with
mowed grass and scattered large trees rather than completely
natural or wild.  Specific natural features, however, are
less important than an illusion of naturalness, and this
can be most easily lost through the presence of negative
characteristics such as junk, ugly development, or other
man-made "misfits."

Evaluation of Effects of Planning:  Fiscal Benefits
Properly planned urban development would result in higher
ground water levels and in more constant streamflow than
would be expected under typical urbanization.  However, a
planned pattern of land use is likely to have little effect
on seasonal high and low flows.  Therefore, it is doubtful
that any substantial savings in capital or operating costs
of reservoirs could be realized.

-------
Somewhat higher water quality might be expected with
properly planned urban development.  Resulting savings
in water purification costs, however, would be minimal.
Higher water quality must be justified primarily on
aesthetic and recreational grounds.

A good land use plan would result in the avoidance of
flood losses in two ways:  through reduction in the in-
crease of peak flows associated with urbanization, and
through prevention of development in areas subject to
flooding.  The latter of these is probably the more im-
portant.

Fiscal benefits to the taxpayer do not appear to be the
major consideration in evaluation of land use plans.  Non-
fiscal benefits, which accrue directly to individuals
(such as environmental benefits) would appear to be more
important.

Evaluation of Effects of Planning:  Non-Fiscal Benefits
Actual use and perceived suitability for use of a stream,
and perceived land values near the stream, are greater if
the stream water is cleaner, and greater for nearby resi-
dents than for those living at'greater distances.  The
effects of changes in water quality and distance to
residence on the above variables can be determined from
regression equations and curves presented in the report.

Capitalization of Effects of Environmental Preservation
in Land Values
Nearly all of the effects of stream valley preservation,
and especially the non-fiscal or environmental benefits,
can be expected to be registered in higher land values.
This increase in value is enjoyed directly by land owners,
who themselves constitute a significant proportion of the
public, and secondly by the general public through higher
property tax revenues.

Analysis of property values around a large stream valley
park in Philadelphia indicates the existence of a land
value gradient, with higher values close to the park.  The

-------
land value effect of the park falls from $15,000 per acre
at 100 feet to $1,700 per acre at 2,500 feet (this being
an area for which the "base" land value is roughly $25,000
per acre).

Conclusions
The results of the study indicate that open space effects
of land use planning may exceed the stream quality effects
in generating public benefits.  The value which persons
derive from stream valley areas seems to be related more
strongly to the existence of open space per se than to
the condition of the stream itself.  In addition, the
major fiscal gains associated with preserving stream
quality appear to be minor relative to the amenity value
of open space.

-------
                    RECOMMENDATIONS
A land use plan which would restrict urban development
near the stream and on steep slopes would have beneficial
effects on stream hydrology and water quality, and would
provide a desirable environment with substantially more
open space than would be achieved through normal develop-
ment.

Such plans should be developed for a number of watersheds
and implemented as demonstration and research projects.
Only through studying actual examples can we gain a well-
grounded knowledge of expected environmental benefits and
the evaluation of them which would be made by residents
and visitors.  Actual demonstration projects would provide
experience on residents' acceptance of such plans and on
the actual effect on land values of the attractive and
guaranteed environment which such plans promise.

In such plans, development should be restricted near streams
and on steep slopes.  Large contiguous paved areas should
be avoided.  Also, where a choice exists among small water-
sheds, development should be restricted in areas containing
the headwaters of streams.

If significant impervious areas are unavoidable, allowance
for artificial detention and dispersal of storm water should
be provided to the extent possible.  This is in contrast to
most urban storm drainage engineering which speeds water to
the stream.

The natural association between the preservation of open
space and the preservation of stream quality should form a
basis for the development of environmental preservation
plans.  The choice of land to be preserved should represent
some mix of the two strategies of maximizing open space
value to residents and minimizing stream quality deterior-
ation.

-------
                       I.    INTRODUCTION
Planning and Management of Stream Systems
For many decades one of the major emphases in environ-
mental planning and management has been the management
of stream and river systems.  This effort has resulted
in a tradition of highly developed techniques for planning
and plan evaluation, which have been effective in identi-
fying and remedying a number of hydrologic and water
quality problems.  These analyses have concentrated on
the stream itself and have resulted in many well worked-
out systems of flood control, flow regulation, water
purification, and channel widening and straightening.

It has long been recognized that a stream is part of a
larger system which consists of its entire watershed,
including geological and vegetative characteristics and
rnan-made structures.  Programs of action to treat the
entire system have been most successful for rural areas.
The effects of rural land uses such as agriculture and
forestry have been studied extensively and are now quite
well understood.  In addition, these uses have not simply
been accepted as a given.  It has been considered approp-
riate public policy to vary rural land uses and their
location in order to achieve environmental objectives.

In urban areas, in contrast, the effects of different
types and amounts of development are not well understood,
and only the stream itself is considered an appropriate
object of management.  Streams are dammed, widened,
deepened, lined, culverted, and paralleled by levees, but
little consideration is given to changing amounts and
types of land use within the watershed in order to reduce
the deleterious effects of urbanization.  There are no
urban counterparts of the Soil Conservation Service's
policies of reforestation, strip cropping, and contour
plowing, which have become accepted practices in many
rural watersheds.  As a result of this inattention to
the entire system, water resource planning in urban water-

-------
sheds has led to corrective rather than preventive measures.

The lack of attention given to urban development  in the
past is understandable, since cities covered a very small
portion of any large watershed.  They tended to grow in a
very compact pattern, determined by technological necessities
such as the suitable locations for railroad lines and gravity
flow sewers.  As a result, the range of possible  choice over
the pattern of development was relatively small.

More recently, the technological constraints have been re-
leased, making development possible almost anywhere.  As a
result it is now more possible to guide urban development
into patterns which would be least harmful to water resources
Population growth rates of urban concentrations has been
covering much more land per person.  Since much more land
is now involved in urbanization, it is also now more neces-
sary to plan for the effects of urban development or water
resource systems.

A Planning Approach for the Preservation of Water Resources
Consideration of an urbanizing watershed as a total system
suggests that many of the deleterious effects may be
minimized so that fewer remedial measures are necessary.
Such an approach would consider certain aspects of urban
development in addition to the usual factors considered
in water resource system planning.  First, the total amount
of urbanization in a watershed might be limited in order
to keep stream quality within an acceptable range.  Second,
the location of development within a watershed, which may
be an important determinant of the effect of urbanization
on a stream, might also be controlled.  For example, severe
restriction of development close to streams might limit the
deleterious effects of stream water quality.  Third, the
type and design of development might be specified in order
to avoid unnecessary harmful effects.  For example  the
construction of very large areas of impervious surface
such as shopping centers, which may have particularly strong
effects on the peak flows of small streams, should perhaps
be restricted.
                              10

-------
The object of guiding the development of urbanized regions
in this way would be:  to maintain water quality; to
prevent the increase in frequency of flooding and the
increase of peak flows associated with moderate to heavy
rainstorms; to prevent enlargement and degradation of the
channel which accompanies increases in flooding; to pre-
vent reduction of the "base flow" of the stream during
periods of little rainfall; and to avoid losses due to
flooding of development on flood plains.

The above objectives are related to the economic value of
the stream, as, for example, the obvious importance of
water quality to its suitability for domestic use.  The
effects are also related to the amenity value of the
stream, that is, its value as an aesthetic and recreational
resource.  This relationship, and the fact that the planning
system described involves the allocation of land uses,
suggests that water resource planning can build upon people's
evident natural attraction to watercourses.  The preservation
of open space near the stream, therefore, should also be an
objective.

This natural link between planning for water resource
preservation and planning to preserve open space and
natural amenity suggests that the concepts of water
quality and the stream flow regimen should be enlarged
to include the setting of the stream as well as the water
itself.

The Plan for the Upper East Branch of the Brandywine
This broader approach to water resource planning has been
followed in but few plans.  One of the most complete is
the Plan and Program for the Brandywine [Institute for
Environmental Studies, RSR1 and USGS, 1968], which was
prepared for the 37-square-mile watershed of the upper
east branch of the Brandywine Creek in Chester County,
Pennsylvania.  This plan specified various restrictions
on development with respect to natural features, and
specified legal and administrative measures for enacting
the restrictions.  The plan was developed following study
of attitudes, land values, and population growth, as well
as the physical characteristics of the watershed and the
                              11

-------
 existing  legal  and  administrative power of government.
 Similar studies were  carried  out for the nearby Pickering
 Creek watershed with  a view to  comparing the future  ( i.e.,
 after urbanization  has taken  place) water quality and
 hydrologic  effects  in the  Brandywine watershed with  effects
 in  a  watershed  with typical planning and legal controls.

 The development of  the Brandywine Plan was based on  the
 delineation of  areas  where development would be critical
 to  the water resource system.   These areas consisted of
 flood plains, the area within 300 feet of all streams and
 swales, areas of over 15%  slope, and wooded areas of 10
 or  more contiguous  acres.   These critical areas constituted
 the Water Resources Protection  District.  Within this
 District, the following restrictions were placed on devel-
 opment :

 On  flood plains  - No  new development or additions to
                  existing  structures would be permitted.
 On  other areas within 300  feet  of all streams and swales
                  (known as the stream buffer area) -
                  No  new development except for additions
                  to  existing uses would be permitted.
                  However,  development would be restricted
                  to  2,000  square feet per parcel, or 5%
                  of  the parcel area, whichever is greater.
                  In  addition,  extensive impervious areas
                  such as parking lots would be prohibited
Woods  and slopes- No  new structure could be built on a lot
                  of  less  than  four acres.   Also, total
                  impervious  area would be limited to 2,000
                  square feet per parcel or to 5% of parcel
                  size, whichever is greater.

For areas outside the Water Resources Protection District,
where most urbanization would occur, "a system of works--
water  supply, erosion control,  surface runoff, and waste
disposal—is proposed  to reduce markedly the usual harmful
effects" [Institute for Environmental Studies, et al., 1968,
p.  III-B-9].
                              12

-------
The primary method proposed for establishing the restric-
tion was the purchase of conservation easements by the
Chester County Water Authority.  In turn for granting an
easement, each landowner would be paid an amount equal to
the loss in value of his land resulting from the restric-
tions imposed.  This amount would vary depending upon the
amount of his land in each of the categories of the Water
Resources Protection District, and the potential develop-
ment value of his land.  Following the purchase of easements
the landowner would retain all prior rights in the land
except those specified in the easement.  For example, no
one except the landowner would have the right of access onto
the land.

The Brandywine Plan was debated thoroughly in the community
over a period of two years, but because of local opposition,
was never carried out.  Among the many reasons for this
opposition were:- the development restrictions and legal
instruments proposed were unfamiliar, relatively complex,
and therefore difficult to understand; the effects of the
easements on the future value of land could not be deter-
mined with a sufficient degree of accuracy, especially
because  there is virtually no market experience with such
easements to be referred to; the imminence and inevitability
of urban development was not apparent to the residents of
the watershed; and distrust of government, based partly on
recent power line and dam projects, was endemic [Keene and
Strong,  1970].

Agreement with the objectives of the plan, however, was
generally asserted by local leaders who established a
regional planning commission to achieve the objectives of
the plan, but with methods acceptable to the residents of
the watershed and with operations fully under local control.
The regional planning commission has been meeting since
January  1968, but to date has not announced a plan of action.

Research on the Physical Effects of Planned Urban Development
and on their Evaluation
The Brandywine Plan received considerable attention through-
out the  country.  The physical solutions to environmental
                              13

-------
problems proposed in the Plan were based on inadequate
scientific knowledge, however.  It was not possible to
estimate with any precision the amount of improvement in
stream quality which would result from the Plan, relative
to what conditions would otherwise be.  In addition, it
was not possible to form an estimate of the value which
would actually be derived from these physical effects by
the present and future residents of the area.

When the Brandywine Plan was developed, very little was
known about the precise effects of amount, type, and
spatial patterns of urbanization on stream hydrology and
water quality.  For example, although it was known that
increases in the amount of land area rendered impervious
in a watershed tend to increase peak flows, there was little
information available concerning the relative effects of
different types of impervious area and the importance of
where the impervious area was located with respect to the
stream, steeply sloping land, and other natural features.
The same sort of information was lacking with regard to
water quality; little was known about typical effects
produced by leakage from sewer pipes, contamination from
septic fields, and other ways in which urbanization might
be expected to affect stream water quality, even if collec-
tion and treatment of sewage were practised.  In specific
cases, or under various assumptions, engineering-type
formulas might be applied to predict these effects.  But,
little was actually known about the typical magnitude of
impacts of urbanization on stream flow and water quality.

In the Brandywine Plan, the land use restrictions proposed
were based upon the best professional judgment available
[Leopold, 1970; Rogers and Associates, 1968].  With regard
to the single most important provision of the Plan, namely
the preservation of a 300-foot strip in natural uses along
all streams and swales, it was reasoned that keeping devel-
opment back from the stream would minimize both water
quality and streamflow effects due to the ability of the
intervening land to absorb excess runoff, and to purify
ground water moving towards the stream.  However, there
                              14

-------
was little information available to indicate what would
be the "proper" width of the buffer area around the stream.

In preparing the Plan, it was also difficult to estimate
the value to the public of the physical effects of the
Plan.  As with any plan, it is important to be able to
demonstrate that the value which the public places on the
effects of the plan outweigh the various costs incurred
in carrying out the plan.  Some of these value elements
could be expressed fairly readily in dollar terms such as
the value of reduced flood damages and reduced water
treatment costs, but others are much more difficult to
quantify.  The latter value components include the aesthetic
enjoyment of preserved stream valleys and the use people
would make of them.

The research summarized in this report was designed to gain
information regarding the questions raised by the Brandy-
wine Plan, which would be useful in the formulation and
evaluation of future plans.

Table 1 below provides an outline of the elements involved
in evaluation of a plan similar to the Brandywine Plan.
The rows of the table identify the three major physical
effects which would result from the plan: namely, more
open space, better water quality, and more stable stream-
flow than would result in the absence of the plan.  The
columns denote the value components which would be affected
by the physical changes.  These include:  more active use
and visual enjoyment of the stream and streamside areas;
various ecological benefits such as more diversified fauna;
and reduced costs resulting from flood damages and water
treatment.  The "x's" in the table denote probable relation-
ships between physical changes and value components.
                              15

-------
                        Table 1

CONCEPTUAL SCHEME FOR EVALUATION OF EFFECTS OF A PLAN FOR
                STREAM VALLEY PRESERVATION
                                                   Reduced
           More     More     More     Reduced      Need for
          Active   Visual  Ecological   Flood        Water
           Use    Enj oyment  Benefit  JDamages     Treatment
Physical
Changes

More stable                    ^         x
streamflow

Better
water       XXX                       X
quality

More open   x         x        x         x
space
The research summarized in this report involves eleven sep-
arate studies on the physical effects, the value-related
effects, and the relationships between the physical effects
and the value-related effects.  Three studies, reported in
Chapters II, III, and IV, are concerned with investigating
each of the three major physical effects.  Of necessity,
the studies have been of the effects which occur with typical
urbanization.  Urban patterns such as that called for in the
Brandywine Plan do not exist and therefore cannot be studied
directly.  The channel enlargement study (Chapter II) has
been concerned with the effects of urbanization on stream
channel enlargement, and by implication, on increased peak
flows; the water quality study (Chapter III) has involved
an investigation of relationships between various aspects
of urbanization and a number of water quality indicators;
and the open space study (Chapter IV) was designed to see
how much development tends to occur in stream valleys in
the absence of strong planning measures, such as conser-
vation easements.

                              16

-------
The other studies have been concerned with the first two
value-related effects listed in the Table, namely active
use and aesthetic enjoyment of the stream and its surroun-
ding area.  One study (reported in Chapter VI) deals with
the general topic of active use of stream sites by nearby
residents; in this study, the effect on use of distance
of residence from the stream is investigated for each of a
number of different uses, and the importance of various
socio-economic factors is also examined.  Another study
(also in Chapter VI) has been concerned specifically with
the relationship between use levels and stream water quality.

Several studies have been concerned with the general subject
of visual enjoyment and aesthetic preference.  These studies
have dealt with basic questions, such as whether there is
substantial agreement among individuals with regard to
preference (in Chapter V), what factors are important to  ,
preference (in Chapter V), and whether persons' actions
are consistent with their stated preferences (in Chapter VI).
An additional study has been concerned with the relationship
between aesthetic preferences for stream sites and water
quality (in Chapter V).

An important aspect of all these value-related components
is that they may be registered in terms of land values.
That is, land values will be bid up in the vicinity of pre-
served areas.  Two of the studies mentioned above provide
background information with regard to land value effects,
namely that dealing with residential preference (in Chapter
VI) and that dealing with the effect of distance on use.
Two additional studies have been carried out.  The first
relates land prices to characteristics and accessibility
of parcels in the Brandywine area itself.  The second provides
an estimate of the land value gradient surrounding a major
urban park in the City of Philadelphia (Chapter VII).

These studies have not succeeded in answering all the ques-
tions raised by the Brandywine Plan.  However, many answers
have been gained, and important conclusions can be reached
regarding the true value of such a plan.
                              17

-------
               II.  HYDROLOGIC EFFECTS


INTRODUCTION
The hydrologic effects of urbanization have been indicated
in the introductory chapter.  These are:

1)  increase in peak flow magnitudes and frequency of
    flooding due to rendering of land impervious and
    altering of the drainage systems,
2)  possible reduction of base flow of the stream in
    periods of low rainfall due to decreased infiltra-
    tion to ground water storage, and
3)  enlargement and degradation of stream channels due
    to changes in the stream flow regimen identified
    in (1) above.

In order to understand the interrelationships among these
effects, it is necessary to describe briefly the nature of
stream channels and "flood plains," and their relationship
to the flow regimen of the stream.

The typical stream valley in rolling terrain has the follow-
ing features:  valley sides, a "valley flat" near the stream,
and the stream channel itself.  The location of the stream
channel on the valley flat is usually not permanent; the
channel typically migrates back and forth over long periods
of time, and may in fact have occupied most or all of the
valley flat at one time or another in its history.  The
movement of the stream, and the maintenance of the form of
the channel, are described by the familiar "meander model"
of stream behavior.  Stream meanders are accepted as a
general phenomenon, although the degree of sinuosity
obviously varies tremendously among streams.  The signifi-
cant feature of meandering is the tendency for the stream
channel to move laterally, in a direction outward from
the direction of curvature.  The stream bank at the outside
of each bend tends to recede, due to erosion by the flow
of the stream and other factors, and the inner bank tends
to advance due to deposition of material by the stream.
Thus the form of the channel remains generally the same
while the channel moves laterally.
                              19

-------
The important point for our purposes is the nature and
origin of the equilibrium channel form which is maintained
during lateral motion.  Various observers have noted that
stream channels tend to be maintained at a size which is
in accordance with the higher flows of the stream (e.g.,
the flows following major storm events).  Specifically,
the stream channel tends to be maintained at a size such
that one or bbth banks are overflowed only about every 1.5
years on the average.  The existence of this common "bankful
frequency" of streams has been documented by Leopold [1964,
pp. 317-328] , and seems to apply to a wide variety of stream
sizes and conditions.  Thus, according to the common-bankful-
frequency hypothesis, the situation may be described as that
in which the stream, moving laterally and maintaining a
more or less constant width, builds up the bank on the side
away from the direction of motion to just that height at
which it is overflowed only every 1.5 years on the average,
given the flow regimen of the stream; or equivalently, the
bank is built to a height such that it "defines" a channel
whose size is sufficient to contain all flows occuring with
a recurrence interval less than 1.5 years.

The succession of these bank segments built over a period of
time would, if the channel-bottom elevation remained constant,
come to constitute a level plain, which is flooded every 1.5 '
years on the average.  This is the "flood plain" of the
stream.  The flood plain may constitute a large portion of
the valley flat; and portions of the flood plain may lie on
both sides of the stream at some time (e.g., if the lateral
motion of the stream has recently changed direction).  But
on the other hand, there may exist a valley flat without any
true "flood plain" area at all.

The relevance of these considerations to the phenomenon of
urbanization is that, given the change in streamflow regimen
brought about by urbanization, the state of equilibrium
between streamflow and stream channel is upset.  With the
existing channel size, a greater frequency of overbank
flows will occur.  Leopold [1968, pp. 7-11]  has calculated
that a 50% increase in the average annual flood of a small
basin (i.e., the flow with a recurrence interval "of 2.33
years), which might be associated with conversion of the
                              20

-------
basin from rural to 2070 impervious and 20% "sewered," would
result in a doubling of the number of overbank flows.

The response of the stream to increases in the magnitude of
peak flows is a gradual process of channel enlargement.
This process has been noted by many investigators, although
it does not appear to have been studied systematically.
Leopold has hypothesized that enlargement proceeds until the
frequency of bankful flow returns to 1.5 years.  That is, the
channel enlarges until a state of equilibrium with the new
streamflow regimen is reached, this equilibrium being the
same in nature as that which prevailed before urbanization.

The process of enlargement typically involves a major
deterioration in the aesthetic and recreational value of
the stream.  During the period of disequilibrium, the stream
tends to have unstable, scoured, and unvegetated banks,  and
nearby valley surfaces are likely to be marked by freshly
scoured flood channels and large areas of deposition.  These
characteristics greatly reduce the amenity value of the
stream.  The period of disequilibrium is likely to persist
for a long time, due to the fact that the intensity of urban
development tends to increase continually.  And even if a
stable situation is reached, the stream will be less attrac-
tive than originally, since it will still exhibit the negative
characteristics generally associated with streams having
widely variable flows.

The effect of urbanization on stream channel characteristics
does not appear to have been studied in detail, but its
effect on streamflow characteristics has been studied by a
number of investigators.  This research was summarized
recently by Leopold [1968, pp. 4-6] in a publication dealing
with the various hydrologic effects of urbanization.

Leopold assembled the results from six field studies, invol-
ving from one to eighty-one sample watersheds, and one study
in which flows were simulated by a mathematical model.  The
data yielded by these studies, which were basically similar,
were converted to a common form.  This form was the ratio of
stream discharge after urbanization to discharge before
urbanization for a one-square-mile basin, where "discharge"
referred to the mean annual flood  (the peak flow having a

                              21

-------
recurrence interval of 2.33 years).  Only two measures of
watershed urbanization were considered:  "percentage of
area made impervious," and "percentage of area served by
storm  sewerage."  Most of the studies dealt with little
more information  concerning urbanization than this, although
several provided  more complete descriptions of the nature of
the drainage  system [Anderson, 1968; Carter, 1961].

The H.-ita from these studies were summarized in the graph
shown  below as Figure 1, in which the percentage of area
sewered was plotted against percentage of area impervious,
with iso- lines used to show the ratio of peak discharge
under  urbanized conditions to the peak flow under unurban-
ized conditions.

According to  the  information summarized here, the ratio of
post-urbanization discharge to pre-urbanization discharge
would  approach 7.0 for a totally sewered, totally impervious
watershed.  The ratio for a 50% impervious, 5070 sewered
watershed would be 2.8 or 2.9.  The iso- lines tend to make
a  somewhat greater angle with the horizontal axis than with
the vertical  axis, indicating that area impervious is some-
what more important than area sewered, in percentage terms.
This graph represents roughly the state of knowledge
concerning the effect of urbanization on peak flow, as of
1968.

With only this amount of information, planning proposals
could  be justified only insofar as they involved limiting
the total amount  of development or the proportion of devel-
opment sewered.   It would obviously be desirable to have
information concerning different types of development, and
particularly  the  effects associated with development at
different locations within the watershed.  Location would
be relevant both  insofar as it involves location with respect
to natural features, such as land slope and soil type, and
also as it determined distance and path of flow from developed
areas  to downstream points.  For example, it might turn out
that merely by guiding the location of development much of  -
the effect on streamflow regimen and channel enlargement
could be avoided.

                               22

-------
       100
   
-------
RESEARCH PROCEDURE
The research carried out under this grant has included an
extensive study of the effect of urbanization on stream
channel enlargement [Hammer, 1971].  The study is also
relevant to streamflow due to the relationship between
stream channel and streamflow characteristics discussed
above.  The channel measurement program was designed expli-
citly to maximize the association between stream channel
size, measured as channel cross-section area, and charac-
teristics of its streamflow regimen.  Measurement criteria
are discussed in detail elsewhere [Hammer, 1970].  These
criteria made it possible to identify points at which one
bank was judged to be a true flood plain, so that the
channel was in "equilibrium" with flow.  The use of channel
measurements as surrogates for flow measurements enabled
us to utilize a broad sample of watersheds without the high
expense of a streamflow gauging program.  The large sample
has enabled us to examine a large number of factors simul-
taneous ly.

The study involved a sample of 78 small watersheds in the
5-county region comprising the piedmont portion of the
Philadelphia metropolitan area.  The basins range from 1
to 6 square miles in area, and vary in watershed condition
from natural to highly urbanized.  Twenty-eight of the water-
sheds contain only rural land uses; the remainder show some
degree of urban development, with 16 more than 25% impervious

A set of channel measurements was made pertaining to the
stream reach at the mouth of each watershed.  Each such
measurement consisted of a channel cross-sectional profile,
from which the width, depth, and cross-sectional area of
the channel were computed.

A wide variety of land-use and other information was obtained
for the sample watersheds.  Each watershed was subdivided
into 40-acre grid squares, and the amounts of land in the
various uses were recorded for each grid square;  similarly,
topographic and drainage system information was recorded
for each square.   More than a dozen land-use categories were
employed, including three types of impervious area:  streets,
                              24

-------
houses, and "other" (primarily large structures and parking
lots).  For each type of impervious area, three different
age categories were employed, so that account could be taken
of the fact that channel enlargement takes place gradually.
Other variables recorded for each grid square were:  the
length (and density) of sewered and unsewered streets;
average soil drainage score; average land slope; distance
and slope of flow-to-channel from the grid square; distance
of flow in pipe; and a complete profile of the channel flow
intervening between the grid square and the measurement
point.

The availability of data for small sub-areas of each water-
shed made it possible to formulate and test a wide variety
of "interaction" variables, in which the amount of land in
some use would be weighted by another factor (e.g., land
slope or street sewerage) and then summed over all grid
squares to form a single observation for each stream.  Thus,
full utilization of the information available required the
use of rather complex variables.

The analysis of these data has been based on the implicit
assumption that the comparison of different watersheds ex-
hibiting varying degrees of urbanization can indicate what
happens to a single watershed over time as urbanization
proceeds.  This implicit assumption has, of course, been
shared by other studies such as those summarized in Figure 1.
The making of such an assumption (which is common in many
fields of research) has received support from the very high
level of statistical explanation obtained in this study,
which can only be interpreted as meaning that, given topo-
graphic and other factors, urbanization affects different
watersheds in a similar fashion.  Hence, one can extrapolate
from the comparison of different watersheds in different
stages of urbanization to speak of the before-and-after
effects of urbanization on a given watershed.

Channel cross-section area was related, using multiple re-
gression analysis, to variables describing land uses and
other watershed characteristics.  The dependent variable in
these regressions was the "channel enlargement ratio" which
consisted of the ratio of observed channel area to the
channel area which would be expected under natural conditions
                              25

-------
for a watershed of the given size.  Natural channel area
as a function of watershed size was estimated using the
28 rural watersheds in the sample.  Over 100 different
independent variables were tested, including a large num-
ber of interaction variables involving both land uses and
natural features.

The analysis was conducted in two phases, each of which
involved somewhat different forms of certain variables.
For each phase of analysis, two different sets of final
results were obtained corresponding to the inclusion and
non-inclusion of variables in which "pre-war" residential
development is considered separately.  Thus, there are
four sets of "final results."  The variables appearing in
these results are listed below.  A full explanation of
these variables and other matters is given in Hammer [1971]
For all the variables except the watershed shape index, the
observation for a given stream consists of quantities
measured for individual grid squares which have been summed
over the whole watershed, in weighted or unweighted form.
For variables involving land uses, the above-mentioned
sum is divided by watershed area, so that the variable is
in the form of a proportion of watershed area devoced to
a given use.

The final variables, which emerged through a process of
repeated testing in regression analysis, are listed below:
Dependent Variable

Y    channel enlargement ratio.  This equals the ratio of
        observed cross-section area of channel in square
        feet to the "natural" channel area.  The latter
        was estimated to be 24.8 A -657^ where A is water-
        shed area in square miles.
Independent Variables
X,   land in forest

X«   land in cultivation
X,.   land in golf courses

X,   average soil drainage score (observed for individual
        grid squares and averaged over whole watershed)

                              26

-------
X   watershed shape index "moment ratio."  This variable
 -*     measures the elongation, or deviation from circu-
       larity, of the watershed as a whole.

X,  area of sewered streets, roads, and sidewalks which
       are more than four years old.
X7  area of houses fronting on sewered streets.  This
       variable consists of area of houses, driveways,
       outbuildings, and other impervious domestic area
       such as patios, for houses more than four years
       old which front on sewered streets.  (Does not
       include houses with direct gutter connection to
       storm sewer=)
XR  other impervious area (second phase).  This variable
       includes all impervious area other than streets,
       houses and driveways, etc., plus area of houses
       with direct gutter linkage to storm sewers.  (Applies
       to area greater than four years old; in regression
       number 4, area greater than 15 years old and area
       4-15 years old are weighted differently.)
XQ  area of post-war sewered streets and sidewalks.  Con-
       sists of area greater than four years old, with
       area 4-15 years old weighted less than area greater
       than 15 years old.
X,~ area of pre-war sewered streets and sidewalks.  Con-
       sists of area of streets constructed before World
       War II.
X, , area of post-war houses, driveways, etc., fronting on
       sewered streets.  Similar to Xy with houses built
       before World War II excluded, and with area 4-15
       years old weighted less than area greater than 15
       years old.

X.J2 non-impervious and unsewered residential area  (used in
       second phase only).  Consists of unsewered street
       and house area, plus tended land such as lawns in
       settled areas.  (In regression 4, includes pre-war
       houses fronting on sewered streets.)

X,o residential impervious area weighted by proportion of
       streets sewered (used in First Phase).  Includes

                              27

-------
       area of houses, driveways, etc., plus area of streets
       assigned to "residential" category times the propor-
       tion of street length  (i.e., in the grid square of
       observation) which consists of sewered streets.  Ap-
       plies to areas over four years old; does not include
       homes with direct gutter linkage to storm sewers.
X   residential impervious area weighted by average land
       slope.  Includes area  of houses, driveways, etc.,
       plus area of "residential" streets, times average
       land slope in the grid square of observation.  Applies
       to areas over four years old; does not include houses
       with direct gutter linkage to storm sewer.
X, ,. pre-war residential impervious area.  Similar to X^3 but
       includes only streets, houses, driveways, etc., con-
       structed before World War II.
X_ , other impervious area  (first phase).  Consists of all
       impervious area other than streets, sidewalks, houses
       driveways, etc., plus area of "non-residential"
       streets and area of houses with direct gutter linkage
       to storm sewers.  Applies to areas over four years
       old.

X, _ slope-of-f low interaction variable (second phase).  Con-
       sists of an impervious area index "l" weighted by
       the average slope of the flow path from the grid
       square of observation to the channel measurement
       point.  (In regression 3,1= 3.25X, + 0.83X_ + 3 . 79X0 ;
                                          O        /        O
       in regression 4, I = 2 . 74X  + 1.95X10 + 2 .Ql^
       + 4.48Xg.)
X, o watershed-size interaction variable (first phase).  Con-
       sists of an impervious area index "l" weighted by
       watershed area to the  exponent -1/2. (In regression 1,
       I = 2.02X13 + 0.18X14 + 4.71X16; in regression 2,
       I = 3.15X13 + 0.14X14  - 1.59X15 + S.OOX^.)
X, q flow-to-channel interaction variable (first phase).  Con-
       sists of an impervious area  index  (as in X]^) weighted
       by the distance of flow --excluding flow in pipe- -from
       the grid square of observation to the stream channel,
       divided by the square  of the average slope of the flow
       path to the channel.

                              28

-------
X n flow-in-channel interaction variable (first phase).
      Consists of an impervious area index (as in X..Q)
      weighted by a function of the lengths and slopes of
      channel intervals over which flow from the grid square
      must pass, this function being a positive function of
      distance and a negative function of slope.  The func-
      tion does not consider intervals of piped channel.

The primary difference between the first and second phases
of analysis was that impervious area involving streets and
houses was tested differently.  In the first phase of anal-
ysis, the area of streets and sidewalks in each grid square
was apportioned into "residential" and "non-residential"
categories, according to the area of houses (and other im-
pervious area) in the grid square.  The "residential" street
area was then added to the area of houses to form a single
"residential impervious area" variable.  Area of "non-resi-
dential streets" was added in with "other impervious area."
This was done to minimize the problem of intercorrelation
between variables.  In the second phase of analysis, street
and house area were considered separately.

Another major difference between the first and second phases
of analysis was that much more elaborate interaction variables
were used in the first phase.  These were replaced by a single
variable relating to the overall slope of flow, as it was
found that equally good statistical explanation could be
achieved using much simpler variables.

In both phases of analysis, separate versions of the regression
were obtained for the cases in which special consideration was
and was not given to "pre-war" residential development.  It
emerged during the analysis that older development—constructed
before World War II--appeared to have a relatively low effect
on the stream channel (see below).

RESULTS
The regression results are given in Tables 2 and 3.  Regres-
sion coefficients which are statistically significant at the
5% level are underlined once; coefficients significant at
1% are underlined twice.

Each of the regression coefficients pertaining to a land-use
variable represents the influence of the given land use on
                             29

-------
                           Table  2

           REGRESSION  RESULTS:   FIRST  PHASE
"14
V19
                      REGRESSION 1

                      Pre-war area not
                   considered separately

                    Coeffi-   Standard
                    cient
     Land in forest  -0.2691
     Land in culti-
     vation

     Soil drainage
     score
                 0.1679
                -0.1128
     Watershed  shape
     index          -0.2817
     Residential
     impervious area
     times  street
     sewerage
                 2.0196
Residential
imperv. area
times land  slope  0.1799

Pre-war resid.
impervious area    	

Other imperv.
area             4.7087

Watershed-size
interaction
variable         0.2972

Flow-to-channel
Inter, variable  -1.9796
    Flow-in-channel
    inter, variable -0.1206
                            Error
                           .1043
.1108
.0291
                           .0877
.4735
                               .0846
.3645



.0879


.7280

.0261
       CONSTANT TERM     176413

       MULTIPLE - R2      .9813
                REGRESSION 2

                Pre-war  area
            considered separately
              Coeffi-    Standard
              cient         Error
              -0.2869      .1007
 0.1704     .1070


-0.1076     .0281
              -0.2587      .0847
 3.1522
                                              0.2459
                                             -1.3358
.6723
              0.1414      .8186


              -1.5937      .7131
                                              5.0017      .3829
            .1042
                                                    .6802
                                        -0.1291     .0245
                                             1.6003
                                               .9829
                                30

-------
                            Table  3
            REGRESSION  RESULTS:   SECOND PHASE
xr
 10
 17
                          REGRESSION 3
                         Pre-war area not
                      considered separately
                         Coeffi-  Standard
                         cient     Error
     Land in forest     -0.1518
     Land in cultivation  0.3896
     Land in golf courses 1.6416
Soil drainage score -0.1072
Watershed shape
index              -0.1990
Area of sewered
streets             3.2499
Area of post-war
sewered streets       	
Area of pre-war
sewered streets       	
Area of houses
fronting on sewered
streets             0.8291
Area of post-war
houses fronting on
sewered streets       	
Other impervious
area                3.7855
Non-imperv. and un-
sewered resid. area 0.1870
.1151
.1291
.5538
.0247

.0776

.8998
                                   .7462
                                   .4486
                                   .0890
Slope-of-flow inter-
action variable     0.2966    .0350
           CONSTANT TERM      1.4487
           MULTIPLE -  R2       .9813
                                            REGRESSION 4
                                            Pre-war area
                                        considered separately
Coeffi-
 cient
-0.2004
 0.3213
 1.2799
-0.0994

-0.1966
                                                1.9518
            4.4797
            0.1459
                                               0.2386
Standard
  Error
  .1115
  .1264
  .5412
  .0237

  .0750
                                               2.7429    1.0636
                       .9027
                                               2.0392    1.2012
            .5388
            .0867
                       .0326
                                                1.4575
                                                 .9829
                               31

-------
stream channel size, expressed in terms of the channel en-
largement ratio.  This influence is expressed as the differ-
ence between the effect on channel size of the given land
use  (land use being expressed as a proportion of watershed
area) and the effect on channel size of "open land," which
consists of land not forested, cultivated or otherwise
developed and which is not included in any of the indepen-
dent variables.  The effect of "open land"--i.e., the value
of the channel enlargement ratio for a watershed consisting
entirely of open land—can be estimated from the regression
results.  It equals the constant term plus the means of the
variables not involving land uses (namely the soil-score
and watershed-shape variables) times their respective re-
gression coefficients.  The effect of open land is estimated
at approximately 1.0 in the first phase of analysis, and
approximately 0.9 in the second phase of analysis.  Thus, a
regression coefficient of 1.5 for some land use (say, in
the  second phase of analysis) would indicate that a watershed
devoted entirely to this land use would have a channel enlarge-
ment ratio of equal to 1.5 + .9, or 2.4.  The channel would
be approximately 2.4 times the size which would be expected
under normal conditions.

The  situation is a bit more complicated for impervious land
uses, however, due to the fact that these land uses also
appear in weighted form in the interaction variables.  Each
of the interaction variables was based on an impervious area
index which involved multiplying each of the land uses by a
factor which was equal to the regression coefficient of the
given land use when appearing as a separate variable in the
regression.  This index was formed for individual grid
squares, then weighted by topographic or other watershed
characteristics and summed over the whole watershed.

To continue the example give above, we may imagine a water-
shed in which the average slope of flow to the channel meas-
urement point for all points in the watershed was equal to
2%.  For an impervious land use with a regression coefficient
of 1.5, the effect on channel enlargement of interaction of
this land use with slope of flow would be equal to (1.5) (2)
(0.29) = 0.87, where 0.29 is the regression coefficient
obtained for the slope of flow interaction variable (in
                              32

-------
phase 2).  Thus, the channel enlargement ratio associated
with a watershed devoted entirely to this land use would
actually be equal to .9 plus 1.5 plus 0.87 = 3.27.  Ordi-
narily, of course, a watershed is not devoted entirely to
a single land use, particularly an impervious use.  Thus,
effects are usually calculated on the basis of the propor-
tion of watershed devoted to each use.

The largest effects obtained were associated with the "other
impervious" category.  For a watershed such as that described
in the example above, complete coverage by "other impervious
area" would result in a channel enlargement ratio equal to
0.9 + 3.8 + 3.8(2)(Q.29) = 6.9.  (This is close to the
maximum increase in peak flows predicted by other investi-
gators, as summarized by Leopold [1968].)

The results of the first phase differ greatly from the re-
sults of the second phase in that the first-phase results
involve interactions between impervious area and specific
flow components.  These interaction effects are in the form
of negative impacts on the influence of impervious area
attributed to flow to channel and flow in channel, with
these negative impacts increasing with distance of flow
and decreasing with slope of flow.   An important point is
that the effects of drainage system alterations, namely the
piping of flow-to-channel and flow-in-channel, are expressed
as the absence of negative effects associated with these
flow components.  (Stream reaches in which the channel had
been "improved," such as by lining the channel with some
material, were treated as if half the length of channel had
been piped.)  In the second phase of analysis, none of the
variables which proved significant incorporated any infor-
mation regarding drainage alterations (other than the
presence of street storm sewers).

The following paragraphs will consider each of the factors
involved in the relationship between urbanization and stream
channel enlargement.

Streets.  Area of sewered streets appears to have a greater
effect on channel enlargement than the area of houses, but
a lesser effect than other impervious area.  The effect of
street area appears to be heavily dependent on the existence
                              33

-------
of storm sewerage.  The first phase regressions attributed
only a small effect to unsewered residential streets (through
the interaction of residential impervious area with land
slope); and the effect attributed to unsewered streets in the
second-phase results was miniscule (this being the effect
attributed to "non-impervious and unsewered residential
land").

Houses.  Houses appear to have the smallest effect upon
channel enlargement of the three impervious land uses (al-
though houses were not distinguished from residential streets
in the first phase of analysis).  This would presumably be
due to the fact that each house constitutes only a small
concentration of impervious area, so that the runoff leaving
its surface represents only a relatively small concentration
of flow and thus is likely to infiltrate into the surrounding
pervious soil.  As in the case of streets, the effect of
house area seems to depend heavily on the existence of
sewerage, with only a small effect attributed to houses not
fronting on sewered streets.  The importance of sewerage to
the effects of street and house area may be somewhat exagger-
ated in the regression results, for various reasons.  It is
possible that there is an association between the extent of
sewerage and the quality of sewerage, i.e., areas with low
proportions of streets sewered tend to have relatively poor
sewer systems for the remaining streets.  This would give
a downward bias to estimates of the effect of unsewered
streets (and houses), and an upward bias to estimates of
the effect of sewered streets.

Pre-war streets and houses.  Houses and (residential) streets
constructed before World War II were found, in the first
phase of analysis, to have a significantly lower effect on
channel enlargement than similar development constructed
more recently.  In the second phase, it was found that the
difference between effects of pre-war and post-war homes
was greater than the difference between effects of pre-war
and post-war streets.  The effect of post-war homes was in
the vicinity of two, whereas the effect of pre-war houses
was not found to be significantly different from zero.  (Re-
call that "effect," as the term is used here, refers to the
amount by which the influence of a given land use on channel


                              34

-------
size differs from the influence of "open land."  An effect,
or regression coefficient, of zero would imply that the land
use in question has an influence on channel size equal to
that of "open land.")  It was initially hypothesized that
the lesser effects attributed to pre-war development had to
do with sewerage:  the older residential areas in the sample
either were provided with relatively poor sewer systems to
begin with, or else have experienced deterioration of sewer-
age facilities due to clogging, etc., over time.

An alternative hypothesis was suggested by the results of the
second phase, namely the possibility that the nature of the
land surrounding older houses has changed over time.  Due to
the addition and growth of trees, shrubs, gardens, etc., the
ability of the surrounding land to retain storm water may
have increased, perhaps enough to nullify the effect of the
house itself.

Other impervious area.  Impervious area other than streets and
houses was found to have the greatest effect upon channel
size.  The impervious area in this category consisted pri-
marily of commercial and industrial buildings and their
accompanying parking lots, although land uses as diverse as
greenhouses and airport runways were also included.  The
most significant characteristic of this impervious area was
the large average size of individual concentrations; nearly
half of the area of this type in the sample watersheds was
found in concentrations exceeding 5 acres in size (with each
"concentration" consisting of a single building or shopping
center and its accompanying parking lot).  It was hypothesized
that the great effect attributed to "other impervious area"
was due to the large average size of impervious parcels,
which would cause runoff to occur in large concentrations of
flow so that subsequent infiltration or storage would be
minimal.  Rather surprisingly, no significant effect was
observed for "other impervious area" weighted by the density
of sewered streets.  This may have been due to various
spurious associations, however.

Land in golf courses.  A quite large effect was attributed,
in the second phase of analysis, to land in golf courses
(excluding wooded areas between fairways).  This effect was


                              35

-------
larger than  that  of any other non-impervious land use.  The
large effect could be  due  to the type of vegetation, the
design of  golf  courses for maximum drainage, and the watering
of  fairways  in  summer  months.

Average  land slope.  Average land slope, as observed for 40-
acre grid  squares, was found to have a significant positive
influence  on the  effect of "residential" impervious area
(i.e., houses and residential streets).  This influence of
land slope is illustrated  in the diagram presented as Figure
2.  The  diagram uses iso-  lines to indicate the effect of
residential  impervious area on channel enlargement for given
values of  land  slope and proportion of streets sewered, based
on  the results  of the  first phase of analysis.  The effects
shown in the figure are comparable to regression coefficients,
i.e., express effects  in excess of the effect of open land.
Thus, for  example, the channel enlargement ratio for a water-
shed containing only residential impervious area might be
calculated as follows  (ignoring other factors such as flow-
to-channel).  If  the watershed is 3070 impervious, with 80%
of  its streets  sewered and an average land slope of 10%, the
per-unit effect of its impervious area would be 3.4 (from
Figure 2), and  its channel enlargement ratio would be:
.9  + (.3)(3.4)  =  1.9.

Watershed  size.   Watershed area was found, in the first phase
of  analysis, to have a negative influence on the effect of
impervious area.  That is, the effect of a given intensity
and type of  impervious development on the stream-channel en-
largement  ratio would  be less, the greater the size of the
watershed.   This  influence of watershed size is illustrated
in  Figure  3, which shows,  for example, that a given percent
coverage would  have only 8570 as much effect in a 5-square-
mile watershed  as in a 1-square-mile basin.  The estimated
influence  of watershed size, while important, was somewhat
smaller  than had  been  hypothesized prior to the analysis.
Watershed  size  was not attributed a separate influence in
the second-phase  regressions; this was also true for the
flow-to-channel and flow-in-channel factors discussed in the
next two paragraphs.   The  reason is that the overall-slope-
of-flow  variable  used  in the second phase subsumed all of
these influences, including the influence of land slope (see
below).

                              36

-------
           Net Effect of Residential Area
  16%


  14%


  12%


  10%
CO
0)
W)
CO
Li
   4%
   2%
      0
     .0         .2         .4         .6         .8        1.0
              Proportion of Street Length Sewered

                           Figure 2

             EFFECT OF RESIDENTIAL IMPERVIOUS AREA

          Isolines express the effect of residential imper-
     vious area, over and above the effect of "open land,"
     as a function of average land slope and proportion of
     street length sewered.
                               37

-------
M-I a
O 0)

•P D-
0 O
CO rH
CX 
6 >
M  CO
•H P
•U O
CD -H
 
   LI
00
95
90
85
     .80
            1.0      1.5    2.0       3.0   4.0  5.0  6.0

                Watershed Area in Square Miles
                           Figure 3


          IMPACT OF  IMPERVIOUS DEVELOPMENT RELATIVE
        TO THE IMPACT  OF THE SAME TYPE AND INTENSITY
        OF DEVELOPMENT IN A 1-SQUARE-MILE WATERSHED
                               38

-------
Flow to channel.  For impervious development at a given loca-
tion, the effect upon stream channel enlargement was found to
be negatively related to the distance of flow to the stream
channel and positively related to the average slope of this
flow.  The influence of flow to channel is illustrated in
Figure 4, which gives the effect of impervious area for
various distances and slope of flow-to-channel, relative to
the effect of impervious area located immediately adjacent
to the stream channel.  For example, impervious area located
one-half mile away from the stream channel with its flow path
having an average slope of 2%, would have slightly less than
807o as much effect as impervious area located at the stream
channel.  The negative effect associated with flow-to-channel
is considered to be absent when this flow component is piped.
For the sample of watersheds studied the average distance to
channel was only just over 1/4 mile and the average slope of
flow to channel was approximately 2-3%.  Thus, the average
reduction in effect of impervious area (where this flow is
not piped) was only about 1070 (that is, the influence on im-
pact as given in Figure 4 was about 90%)-  The smallness of
this reduction of effect indicates that local restrictions on
location of impervious development--such as prohibiting de-
velopment within 300 feet of the stream--would have relatively
little effect on peak flow magnitudes and channel enlargement.

Flow-in-channel.  The effect on channel enlargement (at the
channel measurement point) of impervious development at a
given point in the watershed was found to be negatively re-
lated to distance of flow-in-channel and positively related
to the slope of flow-in-channel.  (Channel slope was expressed
as an index in which it was adjusted for drainage area.)  The
influence of flow-in-channel is illustrated in Figure 5.  As
in the case of flow-to-channel, large negative influences on
the effect of impervious development can occur when the dis-
tance of flow is great and when the slope is low; but the
influence for average points in a typical watershed is fairly
small.  The average distance of channel flow for a one-square-
mile watershed, for example, is about 3/4 of a mile, and the
average value of the slope index is, by definition, 1.0, so
that the typical reduction in effect of impervious area is
only about 10?0.  For a watershed with channel slope typical
of our sample, the overall location of development--i.e.,
whether it is concentrated near the measurement point or near
                               39

-------
    .0         .2         .4         .6        .8
     Distance of Flow to Channel,  in Miles

                    Figure 4

          INFLUENCE OF FLOW TO CHANNEL

     Isolines express the effect on stream chan-
nel cross-section area of impervious development
located the given distance from the channel,
having the given slope of flow to the channel,
relative to what the effect would be were there
no distance of flow to channel (i.e., were the
development located adjacent to the channel,  or
were the flow path to the channel piped).  The
data pertain to a 1-square-mile watershed.
                     40

-------
  4.0-,
A
  2.0
 
-------
the headwaters  of  the  stream--does not  seem to make a drama-
tic difference.  Also,  the  piping of channels  (which in our
formulation  eliminates  the  negative effect of channel flow)
does not  appear to be  of  particularly great importance.

Overall-slope-of-flow.  In  the  second phase of analysis none
of the previous  four factors was expressed explicitly.  Instead,
the results  included a  single variable  consisting simply of
the average  slope  of the  overall flow path from a measurement
point.  This variable,  to which a strong positive effect was
attributed,  consisted  simply of the difference in elevation
by the length of the flow path  between  these two points.  The
important feature  of this variable is that it is associated
with most of the previous factors:  negatively associated
with watershed  size and distance of flow-in-channel; and
positively associated with  land slope,  slope of flow-to-
channel,  and slope of  flow-in-channel.  Thus this overall
slope variable  was able to  capture the  effects which were
attributed in phase 1  to  separate variables.

One major difference between the two sets of results, however,
is the fact  that no information pertaining to the piping of
flow-to-channel .and flow-in-channel is  contained in the
second phase results.  The  fact that equally good statistical
explanation was  obtained  in spite of this omission could be
due to the fact  that such drainage alterations are closely
associated with the extent  of impervious development, so that
the effects  of  the former were  spuriously attributed to the
latter in this  phase.  The  second phase results make possible
the simple calculation of the effect of development in a
given basin, whereas the  first  phase results are more useful
for planning purposes.

SUMMARY CONCLUSIONS

Location  within  a  watershed.  The results obtained concerning
flow are  generally consistent with the  proposals of the Brandy-
wine Plan, which would restrict development near the stream
and on the steep slopes.  The effects on stream flow and
channel enlargement of development in these critical areas
are probably less  important, however, than the effects on
erosion and sewage effluent.  The most  important consideration
concerning stream  flow and  channel enlargement seems to be
                              42

-------
type of development.  Whether or not a residential area is
storm-sewered, for example, appears to make much more dif-
ference than its location relative to the stream.  This
suggests that serious consideration should be given to the
possibility of purposely underdesigning storm sewer systems,
providing catchment basins, or of finding some way to dis-
perse the water which is collected by storm sewer systems.

Due to the large effect attributed to "other impervious area,"
it appears advisable to prohibit or severely restrict the
construction of large parcels of impervious area such as
shopping centers in watersheds which are to be protected.

Allocation of development among watersheds.  It is clear
that when there is a choice between allowing development in
a steeply sloped watershed and a relatively flat watershed
(referring to all aspects of slope including channel slope),
development should be apportioned primarily to the less
steeply sloped watershed.  Also, it is apparent that location
of development upstream of a relatively flat reach of channel
can result in a considerable reduction of this effect, rela-
tive to its effect if located further downstream.  Significant
reduction of effect requires that the channel slope be quite
gentle, however; and in addition, it must be remembered that
location of development at upstream points will affect a
larger length of stream channel.  Therefore, the best general
policy actually appears to be that of preventing development
in the headwaters of streams.  This would be consistent with
the fact that the absolute amount of channel enlargement in
square feet resulting from a given absolute amount of im-
pervious development decreases markedly with increase in
watershed area.

The equations developed in this study have been used to esti-
mate the hypothetical effects of development at various sub-
basins within the Brandywine Creek watershed upon the stream
channel at all downstream points [Hammer, T. R., 1971, p. 279-
314].   The amount of channel enlargement, in cubic feet,
associated with one unit of development in a given sub-basin
was then computed for all downstream points and summed to give
a total impact figure.   The lowest total impacts were obtained
for sub-basins near the watershed mouth.  These impacts were
substantially lower than those calculated for the few sub-
basins which lay upstream of relatively flat portions of
channel.  These results would tend to support a policy of

                              43

-------
preserving upstream portions of watersheds, as a means of
minimizing overall environmental degradation due to urban-
ization.
                              44

-------
  III.  EFFECT OF URBAN DEVELOPMENT ON STREAM WATER QUALITY
INTRODUCTION
It is generally accepted that urban development causes a de-
terioration in stream water quality.  The effect, however,
is not a simple one, and methods of predicting it are not well
developed.  Urban development is inherently complex since it
includes a wide variety of activities, involving many types
and patterns of physical development.  In addition, the man-
ner in which a particular activity or type of development
causes change in water quality may itself be complex.  For
example, residential areas served by sanitary sewers may
affect the stream in several ways:  by contamination of ground
water from scattered small leaks in the system; by more or
less direct contamination of the stream by leakage of main
sewers running along it; and finally by direct disposal of
treated sewage into the stream.  Because of the complexity of
the phenomena, very few attempts have been made to make quan-
titative estimates of the effects typically associated with
various amounts, locations, and types of urbanization.

This study represents an attempt to estimate, utilizing a
large sample of watersheds with varying types and degrees of
urbanization, the stream water quality effects typically pro-
duced by various types of urbanization.  It is an attempt to
see whether, despite the complexity of urbanization and the
variability of water quality parameters, regularities can be
observed between urbanization and stream water quality.  Al-
though the relationships can give only a general indication
of how the effects are produced, they should be useful in
providing some general direction for land use planners.

The study involved analysis of one water quality sample for
each of a large number of streams.  These samples were taken
at the time of season low flow, and the results of the ana-
lysis refer to the effects of urbanization on base-flow
conditions.  The study has focused upon base-flow  conditions
because the concentrations of pollutants from urban sources
are frequently highest at low flow, due to the minimal
                              45

-------
dilution of effluents which are introduced into the stream.

The contribution of street dirt and other surface debris to
runoff and stream pollution has been documented [American
Public Works Association, 1969] .  Relationships between
various characteristics of urban development and stream
water quality at typical flow conditions has also been the
subject of recent research by Economic Systems Corporation
[July 1970].  Their findings indicate that stream water
quality decreases with increasing population density and
with the amount of development, particularly as measured by
land devoted to streets, the type of streets, the amount of
main covered storm sewer, and the ratio of covered storm
sewer to total length.

The low-flow study conducted as part of this research project
complements these two studies which refer to the effects of
storm runoff in urban areas.  Since only one set of chemical
observations was made in the low-flow study, only those
chemicals were investigated which were expected to be reason-
ably stable over time for a given stream.  Therefore, many
of the water quality parameters which are commonly considered
most important, including dissolved oxygen, biochemical
oxygen demand, and coliform count, were not investigated.  A
list of the chemicals studied is presented in the next
section.

The sample of streams, all of which are located in the pied-
mont portion of the Philadelphia metropolitan area, consisted
of watersheds ranging between 1 and 6 square miles in area.
Due to the small size of the streams studied, the watersheds
contained relatively few water-using industries.  Therefore,
the study focuses more on water pollution caused by human
residence than on industrial pollution.  Since only a few
of the watersheds contained major sewage disposal facilities,
the study maybe said to deal primarily with "local" effects
of urbanization, as opposed to overall effects which typically
involve disposal of some or all sewage into bodies of water
larger than the streams considered here.

The study was based on detailed urbanization data for each
watershed, including direct counts of numbers of dwelling
units.  Most of the sample watersheds coincided with water-
sheds used in a related study of the effect of urbanization
                              46

-------
on stream channels which is summarized in Chapter II [Ham-
mer, 1971].
DATA ON WATER QUALITY
Water quality samples were taken as part of this study for
58 streams on two consecutive days in September, 1969, at
a  time when there had been no rainfall in the area in
several weeks.  The samples thus represent water quality
at "base flow."  For all but two of the streams, discharge
was less than one cubic foot per second per square mile of
watershed area.  In addition to the water quality data
gathered explicitly for this study, water quality data for
10 watersheds were taken from an earlier study of Brandywine
Creek [Institute for Environmental Studies, RSRI and USGS,
1968] , which is also located in the piedmont portion of the
Philadelphia metropolitan area.

Both sets of water samples were analyzed by the U.S. Geo-
logical Survey, which performed its standard analysis.  The
chemicals analyzed, with their means and standard deviations,
are given in Table 4.  An additional variable measured in
connection with the water-quality samples was stream dis-
charge, which was measured and recorded at the time each
sample was taken.

For 8 of the 10 streams in the Brandywine area, more than
one chemical sample was available.  The total number of
Brandywine samples was 46 (all representing base-flow con-
ditions).  Not all of these contained data for all of the
chemicals, however; the number of Brandywine observations
available ranged from zero (for dissolved solids--sum) to
46 (for chlorides).  For most of the chemicals, the number
of observations obtained from the Brandywine samples was
between 15 and 30.

In addition, 7 of the 58 streams for which samples were taken
as part of this study were deleted from the statistical
analysis.  These were omitted because of their tendency to
"dominate" the statistical results, as is discussed below.
Thus,  the sets of observations used, to which the means and
standard deviations given in the previous section apply,
ranged between 51 and 97 in number.  The number of obser-

                              47

-------
                            Table  4
 MEANS AND STANDARD  DEVIATIONS OF CHEMICAL VARIABLES
                                        Units        Mean      S.D.
Chloride (Cl)                            pptn         16.8      10.4
Sodium (Na)                             ppm         10.8       4.9
Potassium (K)                            ppm          3.2       1.3
Calcium (Ca)                            ppm         21.6       8.9
Magnesium (Mg)                           ppm          7.6       3.4
Hardness (Ca,Mg)                         ppm         85.2      34.3
Hardness (non-carbonate)                 ppm         40.0      21.4
Sulfates (SO,)                           ppm         30.3      17.0
Nitrates (N(>3)                           ppm         11.2       7.8
Phosphates (PO,)                         ppm           .27       .64
Bicarbonate (HCO-)                       ppm         60.7      23.3
pH                                  -log H+          7.6        .40
Manganese (Mn)                           ppm           .02       .04
Fluoride (F)                            ppm           .23       .15
Iron (Fe)                               ppm           .04       .03
Silica (Si02)                            ppm         19.0      11.8
Total dissolved solids: sum              ppm        179.00     40.00
         :  residual on evaporation       ppm        160.00     59.00
         :  specific conduc.      micro mhos        298        72
                            48

-------
                          Table 5

    PRIMARY SOURCES OF CHEMICALS FOUND IN SURFACE WATERS

                     Geologic   Domestic
                     Influence  Waste       Industry   Agriculture
Chloride                 XX                       X
Sodium                   X                                 X
Potassium                X                                 X
Calcium                  X                     X
Magnesium                X
Hardness                 XX           X
Sulfate                  X         X
Nitrate                  XX           X           X
Phosphate                          X                       X
Bicarbonate              X                     X
pH                       X
Manganese                X
Fluoride                 X                     X
Iron                     X                     X
Silica                   X
Total diss. solids       XX           X           X
                                49

-------
vations used for each chemical is given below with the
statistical results.

The following  table  summarizes very briefly the primary
sources of each chemical, as  found in surface waters.  The
purpose of the table is  to  indicate in a rough way which
chemicals one  might  expect  to be correlated with indices of
human activity.

Geological influence is  mentioned for all the chemicals ex-
cept phosphates.  A  number  of the chemcials, such as mangan-
ese and silica, are  commonly  thought to be primarily geolo-
gical in origin.  With regard to the chemicals for which more
than one source is indicated  in Table 5 (above), many of
these are known to be contained in the effluents from various
activities, but the  relative  importance of human versus geo-
logic influence is difficult  to specify.
PREPARATION OF WATERSHED VARIABLES
All of  the variables used  to  "explain" the observed chemical
concentrations were, with  the exception of stream discharge,
variables relating  to watershed conditions, primarily con-
ditions relating  to human  occupancy and activity.  These vari-
ables were all prepared using a watershed grid system, with
grid squares 40 acres in size.  The values of the quantities
in question were  measured  and recorded for each individual
grid square, and  then summed  over the entire watershed.  In
some cases, variables were weighted by other quantities ob-
served  for individual grid squares, before being summed.

The watershed data  for this study were prepared in conjunction
with data for the study of effects of urbanization on channel
enlargement.  Although the latter study was concerned with
somewhat different  variables, in particular the amounts of
various types of  impervious area, there were several variables
which were employed in both studies.  The following table
contains a list of the quantities measured for each grid square,
and a statement of  the "basic variables" computed from these.
The basic variables pertain to whole watersheds; each involves
a summation (denoted by "2")  of some quantity over all grid
squares.  The paragraphs which follow then discuss the various
quantities in detail.

                              50

-------
                            Table 6

                         WATERSHED DATA

Quantities measured for each grid square:

X«      Area of grid square (in square miles)

X,      Total population of grid square

X9      Proportion of grid square population served by sani-
        tary sewer system involving disposal of effluent
        within the watershed

X~      Proportion of grid square population served by sani-
        tary sewer system involving disposal of effluent
        outside the watershed

X,      Proportion of dwelling units in grid square built
 ^      before 1940

Xc-X,,   Manufacturing employment in grid square, by 4-digit
        SIC code (up to 4 industries)

Xn      Land area in cultivation (in square miles)
 y

X,,.     Land area impervious (in square miles)

X.. 1     Average soil score for grid square (1-4 scale)

X,~     Average distance of land in the grid square from the
        stream channel (in feet)

Basic variables computed for watersheds:

Postwar population sewered out: S X,X~X,

Prewar population sewered out:  S X,X~(1-X,)

Population sewered in:  £ X-X^

Population unsewered, no gradient:  £ X, (1-X^-X.,)
                             51

-------
Population unsewered, 2,000' gradient: SK
     where K = Xl-XX   (1-X   /2000) if X   < 2000,

     and K = 0 otherwise.  (Similarly  for variables incor-
     porating 1500-foot and  1000-foot gradients.)

Average soil score: Z X-.X-.-./SX..

Land in cultivation:  SXq

Percent of watershed impervious:   S X1f)/ 2 X,

Total manufacturing employment:   S (X- + X, + X  + X_)

Manufacturing employment figures  for  subcategories , including
     individual SIC groups and a  single "high water-using and
     water-polluting industries"  category, were also prepared,
     by simple summation of  the relevant quantities.
                              52

-------
Area and total population of grid square.  Total population
of each grid square was estimated by counting the number of
single-family houses from aerial photographs, and determin-
ing in the field the number of dwelling units in multi-
family structures by counting mailboxes.  A total population
figure was obtained by multiplying the number of single-
family dwellings by 3.6 and the number of dwelling units in
multi-unit structures by 2.4 (figures based on average con-
ditions in the study region), and summing these two components.

Proportions of population "sewered in" and "sewered out."
A factor which was considered to be of primary importance was
existence of sanitary sewerage.  In this regard, a significant
factor was whether or not the treated effluent yielded by a
sewer system was disposed of inside or outside of the water-
shed in question.  For small urbanized watersheds, a typical
situation is that the sewage from most dwelling units is ex-
ported from the watershed, frequently by sewer mains running
alongside the stream, and disposed of at some downstream
point.  In such a watershed, the sewage produced by the res-
ident population can only affect water quality insofar as
there is leakage from the sewer system within the watershed.
On the other hand, some watersheds contain the disposal
facilities themselves.  Sewage from dwelling units served by
these facilities affects water quality both by the content of
the effluent from the disposal plant and by leakage from the
sewer system.

In this study, population in dwelling units served by sewer
systems which export sewage from the watershed for disposal
elsewhere is referred to as "population sewered out."  Popu-
lation in dwelling units served by sewer systems which in-
volve disposal within the watershed is termed "population
sewered in."  It should be noted that this distinction is
merely a result of our choice of stream measuring points;
the population "sewered out" for a given watershed will almost
invariably be "population sewered in" for another measuring
point.  Population "sewered in," "sewered out," and "unsewered"
are illustrated below in Figure 6.

It will be noted that population in dwelling units outside a
watershed may be relevant to water quality in the watershed
if it is served by a sewer  system which involves disposal
                              53

-------
                            Figure 6

            ILLUSTRATION OF POPULATION "SEWERED IN,"
                 "SEWERED OUT," AND "UNSEWERED"
                                Stream
LEGEND
I	i
  O
Sewer district
(e.g., a borough
or township)

Sewer main

Sewage treatment
and disposal plant

AREA CONTAINING
POPULATION
"SEWERED IN"

AREA CONTAINING
POPULATION
"SEWERED OUT"

AREA CONTAINING
POPULATION
"UNSEWERED"
                                  Watershed
                                    mouth
                                                   Watershed
                                                    boundary
                            54

-------
within the watershed.  This complication arose in only two
of our watersheds, both of which were eliminated from the
sample for statistical reasons.

Sewerage information for the sample watersheds was obtained
from the excellent maps provided by the Delaware Valley Re-
gional Planning Commission.  The data recorded for each grid
square consisted of estimates of the proportions of the
population in the square which were sewered out or sewered
in.  These proportions were either 0 or 1 for the great
majority of grid squares.

Proportion of dwelling units in grid square which were built
before 1940.  It was reasoned that leakage from sewer systems
might typically increase over time.  Although it was impos-
sible to obtain data relating to age of sewer lines themselves,
it was thought that this age should be highly correlated with
age of dwelling units.  Therefore, utilizing U.S. Census of
Housing information for census tracts plus aerial photographs,
an estimate was prepared of the proportion of dwelling units
in each grid square which were built before 1940.  These are
hereafter referred to as "pre-war dwelling units."

Manufacturing employment.  All manufacturing establishments
in the sample watersheds were noted in the course of the
field survey.  Then an industrial directory was used to de-
termine the employment of each establishment and its classi-
fication according to the Standard Industrial Classification
(SIC) detailed (4-digit) code.  The employment of each estab-
lishment and its SIC code were recorded by grid square.
(There were never more than 4 establishments in a given grid
square; therefore, this information is denoted by four employ-
ment variables, Xc through X0).
                 _> ...         o
Land area in cultivation.  Land in cultivation is defined as
land that was under tillage at the time the air photographs
were taken.  This variable was measured for each grid square
using the aerial photographs; it was not considered feasible
to conduct field surveys for this purpose.  This variable was
rather unsatisfactory in expressing the overall impacts of
agriculture on water quality, due to the fact that it pro-
vides no indication as to the intensity of cultivation or the
number of livestock being grazed.
                              55

-------
Land area impervious.  The area of land in each grid square
covered by impervious surface, such as streets, buildings,
parking lots, etc., was measured using aerial photographs
and field survey, as part of the channel enlargement study.

Average soil score.  This variable, which was originally pre-
pared for the channel enlargement study, expresses the drain-
age characteristics of the soil on a one-dimensional scale.
Each of the soil types mapped by the Soil Conservation Service
[1963] is described as being either "well drained," "moder-
ately well drained," "somewhat poorly drained," or "poorly
drained."  In forming a one-dimensional scale, "soil scores"
of 4, 3, 2, and 1 were assigned to soils having the above
descriptions, respectively.  The average soil score for each
grid square was computed by averaging the scores for soil types
found at four randomly located points within the grid square.

Average distance of grid square from stream channel.  The
average distance of flow from each grid square to stream chan-
nel was estimated by measuring the distance from the central
point of the square to the channel along a line running per-
pendicular to the topographic contour lines (i.e., the most
probable course of surface or sub-surface flow).  Occasion-
ally, a situation was encountered in which portions of a
given grid square would drain in quite different directions.
In such cases, the distance to channel was obtained as a
weighted average of two or more distances measured separately.

The computation of each of the basic variables pertaining to
population was accomplished by multiplying total population
in each grid square by one or mor.e factors of proportion,
relating to sewerage and/or age characteristics of housing
in the grid square, and summing over grid squares.  It was
known that the straightforward use of proportionalities would
involve errors for individual grid squares; for example, a
grid square with 50% of its housing sewered out and 5070 of
its housing pre-war need not have approximately 25% of its
housing "pre-war sewered out."  However, the errors should
cancel whole watersheds.

As indicated in Table 6, the basic variables relating to
population were:  population in post-war dwelling units
sewered out, population in pre-war units sewered out, popu-
                              56

-------
lation in units sewered in, and population in unsewered
dwelling units (in various forms; see below).  Pre-war and
post-war dwelling units "sewered in" were not considered
separately since leakage from sewer lines was presumed to
account for only a minor portion of the effect produced by
these units (i.e., minor relative to the effect produced by
the effluent disposed directly into the stream).

It was thought that the distance of dwelling units from the
stream channel might be particularly important for unsewered
units whose sewage is disposed of in the soil.  The distance
from the dwelling unit to the stream would represent the
distance within which the purifying action of the soil might
remove contaminants originating at the dwelling unit.  Thus
several "gradient" forms of the variable involving unsewered
dwelling units were prepared, which attributed lower effects
to dwelling units farther from the stream.  In each case, a
linear form was used, with the effect reaching zero at a
specified distance.  For example, in the "2000-foot gradient"
variable the effect of a house adjacent to the stream would
be 100%, the effect of a house 1000 feet from the stream
would be 50%, and the effect of a house 2000 feet or more
from the stream would be 070.

Variables incorporating 2000 feet, 1500 feet and 1000 feet as
the zero point were tested.  The 2000-foot gradient variable
was almost always more significant in explaining chemical
concentrations than either of the variables incorporating
shorter gradients.  Thus, in the discussion which follows
only the 2000-foot gradient variable is mentioned.

Numerous variables were prepared and tested which related to
manufacturing employment.  These included sums of manufactur-
ing imployment for individual standard industrial classifi-
cations, and also a variable expressing total manufacturing
employment.  In addition, a vairable was prepared expressing
the sum of employment in certain industries which are consid-
ered to be high water-using and high water-polluting industries
[Isard and Romanoff, 1967, pp. 7-10, 21-22].
                              57

-------
FORM OF THE ANALYSIS
As described above, the collection of watershed data was
quite straightforward.  Each basic variable for a watershed
consisted of a magnitude--e.g., manufacturing employment,
or population in dwelling units having a certain type of
sewer age--summed for the watershed as a whole.  It was rec-
ognized, however, that the basic form of the variables was
not appropriate for the job of explaining the concentrations
of chemical constituents in stream waters.  In order to ob-
tain meaningful results, it was necessary to "scale" these
absolute magnitudes by other factors, such as watershed size,
which would be expected to interact with these magnitudes
in determining stream conditions.

The importance of watershed size arises from the fact that
one generally would expect the density^ of development in a
watershed to be more relevant  than the absolute amount of
development in determining water quality.  For example, a
stream draining a one-square-mile basin containing 10,000
people would be expected to have higher concentrations of
urbanization-related cherncials than a stream with a 100-
square-mile watershed containing 10,000 people; the former
is "urban" whereas the latter  is not.  Thus, each of the
basic land-use variables was transformed by dividing it
through by watershed area (i.e., the value for each stream
was divided by watershed area  for that stream).  This yielded
variables of the form:  population per square mile, manufac-
turing employment per square mile, etc.

The transformation of land-use variables to a per-area basis
is quite common in all sorts of research, and usually goes
without comment.  In this study, however, a similar transform-
ation was also made using another factor, which resulted in
variables of a less familiar nature.

This second factor was stream  discharge, measured at the time
the chemical samples were taken.  Since all the samples were
taken at base flow, discharge was fairly highly correlated
with watershed area; but the correlation was far from perfect,
The use of discharge rather than watershed area to "scale
down" the land-use variables would be appropriate in certain
circumstances such as the following:  Suppose that there is
a source such as a factory which introduces some chemical to

                              58

-------
a stream at a constant rate; and suppose that this is the
only major source of that particular chemical in the water-
shed.  Then the concentration of that chemical in the stream
would be determined by the discharge of the stream, i.e.,
by the extent to which the chemical received is diluted by
the water in the, stream.

When dealing with a chemical for which this general sort of
situation is felt to be typical, a possible mode of analysis
would be to multiply the observed concentration of the chem-
ical times the stream discharge, to yield a figure for the
absolute amount of the chemcial being discharged by the
stream.  This figure would then simply be related to a vari-
able expressing the absolute magnitude of the suspected
source of the chemical.  That is, the relationship being
investigated would have the following form:

     CD or X
where C = observed concentration of the chemical in question
            in the stream
      D = stream discharge
      X = magnitude of the source (or some surrogate variable;
            e.g., employment in a certain manufacturing in-
            dustry)
Since it is generally more convenient to retain the dependent
variable in the form of a chemical concentration, the follow-
ing sort of relationship was employed in this study (which
relationship is obtained from the above by dividing by dis-
charge) :
      r   X
      C * D
Thus, a second set of independent variables was derived in
which each of the basic magnitudes, such as population in
a given type of dwelling unit, was divided by discharge.

One will note that the form of variables involving division
by discharge should be the more relevant for "point" sources
of pollution, and the form involving division by watershed
area should be more relevant to "area" sources.  For point
sources, at which contaminants are introduced more or less
directly into the stream, the only important factor would


                              59

-------
be the amount of dilution  (as discussed above).  For area
sources, which tend to involve contamination of the ground-
water at points located more or less randomly throughout
the watershed, the important factor would be watershed
areas, which determine "how much" groundwater is involved
and hence determine the intensity of contamination.  Stream
discharge, which is related to the rate of outflow from
groundwater storage, would be important for area sources
only to the extent that chemical concentrations in the ground-
water are correlated over  time with the rate of outflow.

These two forms of the independent variables were thus inten-
ded to embody two archetypes of possible ways in which chem-
ical concentrations might be determined.  As such, they were
taken to represent "extreme cases."  Therefore, it was felt
that an additional set of  independent variables should be
formulated which expressed an intermediate case, in which
the effect produced by a source of a given magnitude would
depend on both watershed area and discharge.  This interme-
diate form was obtained by calculating the geometric mean
of watershed area and discharge, and using this as a divisor
for the magnitudes of the various pollution sources.

Thus, for each of the basic variables discussed in the pre-
vious section (except the  "soil score"), three independent
variables were prepared:
     "Density form"     X../A.

     "Discharge form"   X../D.

     "Mixed form"       X../  (A.D.)^
                         ij     11

where X.. = magnitude of pollution source j in watershed i
        -*     (e.g., some component of population or employ-
               ment)
      A.  = area of watershed i in square miles

      D.  = discharge of stream i at the time the sample was
       1       taken, in cubic feet per second

In the analysis, the procedure was not to make an a^ priori
judgment as to which form of a variable would be most appro-
priate for a given chemical.  Rather, the plan was to see
                              60

-------
which form would prove most successful as an independent
variable in the regression analysis.

The results obtained, with regard to which of the three forms
is most successful for a given source and chemical, have
been rather interesting.  However, in interpreting these re-
sults, one must keep in mind the following facts:  First,
the statistical tests which were carried out related only
to the "significance" of each variable in itself, not its
significance vis £ vis other forms; thus, no confidence level
can ben attached to the choice of one variable over another.
And second, there are actually a number of reasons why one
form might be more successful than another, not involving
the behavior of "point" sources versus "area" sources as
discussed above.
ANALYSIS AND RESULTS
Preliminary Analysis
The procedure followed in this study was to attempt to explain
each of the water quality variables statistically using multi-
ple regression analysis.  Except in the preliminary phase of
analysis, each of the basic watershed variables was entered
in the regression in three forms as discussed in the previous
section.  (This was not true for the soil score variable.)

A computer program for stepwise regression was utilized in
the analysis.  However, considerable experimentation was con-
ducted for each of the chemicals, in the form of forcing in-
dependent variables into, or deleting variables from, the
regression, so that the results do not necessarily represent
simply the "stepwise solution."

An extensive preliminary analysis was conducted, utilizing
both the full sample of streams (for a given chemical) and
also samples with some of the streams deleted.  A major out-
come of the preliminary phase of analysis was the finding
that it was not possible to utilize the information concern-
ing manufacturing employment at the level of detail which
was available.  The reason for this was that a given 4-digit
industrial category [Bureau of the Budget, 1967] would
typically be represented in only one or two watersheds; or


                              61

-------
at least,  the employment  in one or  two watersheds would be
highly dominant.  This problem was  encountered even when
the manufacturing data were aggregated to the 2-digit stan-
dard industrial classification level.  Therefore, it was
decided  simply to use two summary variables, namely, total
manufacturing employment  and  total  employment in high water-
using and  water-polluting industries.

It is recognized that the results obtained for these two vari-
ables can  be attributed only  very general significance.  That
is, a positive coefficient obtained  for, say, total manufac-
turing employment indicates only that perhaps one industry
in ten or  one industry in a hundred  is actually contributing
the effluent in question.  This would have been a problem
regardless of the level of detail employed, but it is clearly
worsened by aggregation.

A second major consideration  which was dealt with in the pre-
liminary analysis was the problem of heteroscadasticity, i.e.,
dominance  of the regression by a few observations.  This
problem  occurs when  one is dealing with variables having
skewed distributions:  typically, most of the values of the
observations are small but a  few are relatively large.
(Technically, heteroscadasticity refers to certain charac-
teristics  of the "error term" of a regression, rather than
the distribution of  the dependent variable per se; but a
highly skewed dependent variable will almost always imply
heteroscadasticity,)  The difficulty is that under these
circumstances the regression  will be inordinately influenced
by the values of the "large"  observations.  This means that
the chance of erroneous results is increased because the
results are being determined  by so few observations.

An extreme example of heteroscadasticity is provided by the
case of phosphate concentrations as  measured in this study.
For the full sample  of streams, the  variation (i.e., sum of
squared deviations around the mean)  in phosphate concentra-
tions was greater than 800.   Removal of two of the 80 obser-
vations reduced the variation to 43.  Clearly, the regression
would be totally dominated by these  two observations; any
factor which explained the deviation of one or both of these
from the mean would be found  highly  "significant."

                              62

-------
It was found that the streams whose observations were domi-
nant tended to be the same for most of the chemicals.  Thus,
it was decided to deal with heteroscadasticity by eliminating
certain streams in all of the analyses.   At first, two streams
were eliminated.  When heteroscadasticity problems were found
to persist, five more streams were eliminated.  Of the seven
streams eliminated, five contained major sewage disposal
facilities; in addition, five contained  major concentrations
of industry.  The water quality data for these omitted streams
are presented and discussed below in this section.

In the initial phases, the soil score variable was not included
in the analysis.  Two related variables  which were included
were:  stream discharge, and discharge per watershed area,
both in simple form (i.e., neither was being used to weight
some other factor). The latter of these, discharge per water-
shed area, was found to be significantly related to many of
the chemicals.  This occurrence led to speculation as to
whether the importance attributed to discharge per se  was
due either to some action of discharge itself or was instead
an indirect reflection of permanent watershed characteristics,
such as those related to geology.  Thus, it was decided to
test the soil score variable, which was  available from the
earlier channel enlargement study and which measured soil
drainage characteristics, as described in the previous sec-
tion.  The outcome was that the soil score variable was
attributed major significance for many of the chemical vari-
ables, greater than the significance formerly attributed to
discharge per watershed area.  Thus, the soil score variable
was allowed to replace the discharge-per-watershed-area
variable in the regressions.

One rather disappointing, but not unexpected, outcome of the
analysis was the failure of land in cultivation to appear as
a significant explanatory variable in all but one of the re-
gressions.  (This was a regression for bicarbonates, which
is listed separately below.  A significant relationship in-
volving agriculture was also found for nitrates, using a
special sub-sample of observations.)  The general non-
significance of land in cultivation was  due both to inad-
equacies of the variable as a measure of agricultural
activity, and to statistical factors involving the strong


                              63

-------
negative correlation between  land  in cultivation and the
urbanization-related chemical sources.

The percent-of-watershed-impervious variable, when tested,
was marginally  significant  for several of  the chemicals.
However, the  appearance  of  this variable as an explanatory
factor  seemed to be due  just  to its close  association with
other aspects of urbanization,  rather than any causal rela-
tionships  involving impervious  area itself; thus this vari-
able was deleted from  the analysis.  Neither land in culti-
vation  nor percent-of-watershed-impervious area are mentioned
in the  tabular  summaries of regression given in the next sub-
section.

Results as a  Whole
Following  the preliminary analysis described above, in which
the statistical sample was  adjusted and various decisions
were made  with  regard  to invidual  variables, a "final" set
of regression results  was obtained.  In these regressions,
the urbanization variables  were allowed to enter in whatever
form worked best--i.e.,  whether the basic  quantity was divided
by watershed  area, discharge,  or the geometric mean of these
quantities.

In a few cases, more than one form of an independent variable
would enter the regression  at a given time.  For example, a
variable might  enter in  "discharge" form with a positive co-
efficient, and  also in "density" form with a negative coef-
ficient.   In  such cases, the  weaker form of the variable was
suppressed (this was always the one with the negative coef-
ficient).  However, the  "unsewered houses" variable was
allowed to enter both  in 2000-foot gradient form and in the
form with  no  gradient.

Since considerable experiementation was carried out after
obtaining  the stepwise regression  solution, the results should
be near the "best" results  obtainable.  However, no formal
test was applied to assure  that the results were, in fact,
best according  to some statistical criterion.

This set of regression results  is  summarized in Table 7.  The
table lists only the forms  of  independent variables which
were significant in the  regression, and not the actual coeffi-


                              64

-------
                                                      Table 7

                        SUMMARY OF REGRESSION RESULTS:   UNFORCED  REGRESSIONS
                                     Form of Independent Variable and Significance Level
Ln
Soil
Chemical Score
Cl
Na
K
Ca
Mg
Hardness (Ca.Mg)
(non-carb.
S04
N03
P04
KC03
pK
Mn
F
Fe
Si02
IDS : SUM
R.O.E.
Specific Cond.
X
X
>:
X
X
X
>*
>C
3C
X_
X
X
Sewered Out
Postwar Prewar
Density
Density
Density
Density
Density
Density
Mixed
Mixed
Density
Mixed
Mixed
Mixed
Density
Mixed
Disch.
Density
Disch.
Mixed
Mixed
Sewered
In
Mixed
Density
Density
Mixed
Mixed
Density
Mixed
Mixed
Density
Mixed
Mixed
Mixed
Unsewered
2000 ft. No
Gradient Gradient
Density
Mixed
Mixed
Mixed
Mixed
Density
Density
Mixed Disch.
Mixed
Disch.
Density
Density
Mixed
Manufacturing Employment
Effluent-Related
Total Industries R2
Density
Density
Mixed
Mixed
Disch.
Density
Density Mixed
Density
Density
Density
Mixed
.68
.71
.53
.80
.60
.80
.55
.64
.35
.91
.65
.24
.19
.29
.48
.16
.57
.77
.66
n
97
71
61
84
84
84
7J
81
85
73
74
7b
60
61
60
61
51
76
51
                                                                     All listed variables are significant at
                                                                     the 0.05 level;  underlined variables are
                                                                     significant at the 0.01 level.

-------
cients obtained.  Variables significant at the 1% level are
underlined.  Significance of the soil score variable is in-
dicated by an "x."

The  level  of statistical explanation for most chemicals was
fairly high, perhaps  suprisingly high, in view of the fact
that  the determination  of many  of  the chemical concentrations
is commonly considered  to be geological.  For 13 of the 19
chemicals, more  than  half of the total variation was ex-
plained by the significant  independent variables in the re-
gression.  The average  value of R^ was 0.56.

An interesting feature  of the regression results, when viewed
as a  whole, is the  considerable similarity in the forms of
the  independent  variables which best explained the various
chemcials.  For  population  in post-war dwelling units sewered
out,  the density form was chosen in all six cases where the
variable was statistically  significant.  For the other four
variables  involving population, the mixed form was more pre-
valent.  The totals were as follows:  for "pre-war sewered
out," there were 8  in mixed form,  3 in density, and 2 in
discharge; for "sewered in," there were 9 in mixed form, and
3 in  density; and for "unsewered," including both 2000-foot
gradient and no  gradient versions, there were 7 in mixed
form, 5 in density  form, and 1  in  discharge form.  For total
manufacturing employment, the density form was prevalent.

It is difficult  to  compare  the  results for various chemicals
when  the form of the  independent variables differs among
chemicals.  Since it  was recognized that a variable might
work  nearly as well in  one  form as in another, it was decided
to rerun the regressions forcing each independent variable
into  a common form.   The form chosen in each case was the
form most  prevalent in  the  set  of regression results listed
in Table 7.  As  theoretical justification for forcing into
a common mold the explanation of a wide variety of chemicals,
one may argue that  the  relative importance of the various
forms of the variables  had  to do primarily with the manner
in which effluent from  a given  source reaches the stream.
The hypothetical description on point and area sources given
earlier in this  chapter, for example, might apply equally
well  to a wide variety  of chemicals.

                              66

-------
An additional change made in the regressions was that the
imsewered no-gradient variable, which was found significant
for only three chemicals, was eliminated from the regres-
sions.  This was done in the interest of greater uniformity
of results; it was suspected that the 2000-foot gradient
variable would serve nearly as well in these cases.

This second set of "final" regression results is presented
in Table 8, in which the actual regression coefficients are
given.  The average value of R.2 for these regressions is
0.54, down only 0.02 from the previous set.  Still inter-
preting the regression results as a whole, there is a very
interesting overall pattern of regression coefficients.
(The reader should note that a coefficient which is not
included in the table does not necessarily have a value of
zero; such coefficients simply have failed to prove statis-
tically significant.  Thus, in comparing regression coeffi-
cients for different regressions, it is best to consider
only pairwise comparisons between variables which are both
significant.)

The coefficients obtained for the sewered-in variable and
the unsewered 2000-foot gradient variable tended to be sim-
ilar, with some variation.  For each of the 9 chemicals for
which both variables were significant, the coefficient for
one of these variables never exceeded twice the coefficient
of the other.  The coefficients for pre-war sewered-out were
much lower than these, with values lying between 1/14 and
1/4 the size of the coefficients for the sewered-in and the
unsewered gradient variables (except in the case of phos-
phates).  Finally, the coefficients for post-war sewered-
out tended to be lower than for pre-war sewered-out.  Although
these coefficients cannot strictly be compared due to the
fact that these two independent variables are in different
forms, it is possible to make a general comparison by noting
that, given the average values of discharge and watershed area
observed in this sample, the effect expressed in a mixed-form
variable is roughly twice the effect expressed in a density-
form variable (since the denominator in the latter variable
is on the average twice as great).  Thus, the post-war-out
coefficients should be divided by a factor of approximately
2 for comparison with the others.  This would indicate that
population in post-war dwelling units sewered out tends to


                              67

-------
                                                        Table 8

                         SUMMARY OF  REGRESSION  RESULTS:   FORCED  REGRESSIONS
                                                    Independent Variable
00
Form of Indep. Variable

Cl
Na
K
Ca
Mg
Hardness (Ca.Mg)
      (non-carb.)
S04
N03
P04
HC03
PH
Mn
F
Fe
       IDS : SUM
           R.O.E.
           Specific Cond.
Constant Soil
Term Score
28.126 -5.76048
19.840 -3.99000
5.228 -.90732
42.405 -8.73838
15.125 -3.07161
162.119 -32.76747
102.374 -22.94466
97.74 -23.08936
8.390
.0637
77.380 -11.55862
8.879 -0.43112
.01591
.190
.0199
16.492
227.010 -28.42377
269.115 -50.06776
419.137 -61.45015
Sewered
Postwar
Density
.00230
.00042
.00124
.00051
.00480
.00453
.00223
.00729
Out
Prewar
Mixed
.00196
.00143
.00223
.00101
.00992
.00357
.00533
.00008
.00730
.01335
.01852
.02312
Sewered
In
Mixed
.02253
.01726
.00329
.01710
.05833
.02355
.00592
.05635
.00008
.10598
.12656
. 14420
Unsewered
2000 ft.
Gradient
Mixed
.02652
.00867
.02487
.00871
.10216
.8532
.00008
.02290
.07807
.14554
.10722
Manuf. Employment
Effluent-Related
Total Industries
Density
.00701
.00285
.00481
.01448
.00789
.00003
.00008
.00003
.02421
.02750
.04635
Mixed
.16622
.00164

                                                                                                           R2
.69
.70
.42
.80
.60
.80
.52
.64
.29
.90
.63
.24
.19
.29
.46
.07
.59
.79
.62
                                                                      All listed coefficients are significant at
                                                                      the 0.05 level;  underlined variables are
                                                                      significant at the 0.01 level.

-------
have between 2070 and 60% as much effect on water quality as
population in pre-war units sewered out, and between 2?0 and
7% as much effect as population in sewered-in or unsewered
dwelling units.

These relationships between coefficients are quite in accor-
dance with expectations.  Population in post-war dwelling
units which are sewered out of the watershed would be expec-
ted to have the lowest effect, since the only effluent
reaching the stream would be leakage from sewer pipes which
have been fairly recently constructed.  Population in pre-
war sewered-out dwelling units would be expected to have a
higher effect since leakage would probably have increased
over time.  Population in unsewered dwelling units and
dwelling units sewered to disposal plants in the watershed
would be expected to have much higher effects since large
amounts of effluents are added directly to the groundwater/
surface-water system of the watershed under study.  In both
cases there may be considerable purification of effluents,
by the soil in the case of houses with septic systems, and
by direct treatment  in the case of effluent passing through
disposal plants.  However, the potential effect upon stream
water quality would obviously be much greater than for
dwelling units sewered out of the watershed.

With regard to the forms chosen for the population-related
variables, the use of density form for the post-war-out
variable would be reasonable if one assumed that leakage
from relatively new sewer systems tend to involve only small
local leaks which are randomly located throughout the water-
shed, as opposed to leakage from sewer mains running along
the stream.  This situation would suggest a "density" effect
as discussed above.  A similar argument could be given,
however, with regard to unsewered dwelling units.  The fact
that the unsewered-gradient variable tended to be slightly
more successful in mixed form than in density form may be
due to the incorporation of the distance-to-stream gradient,
which gave greater weighting to some extent with a dilution
effect.  The use of mixed form for population in pre-war
dwelling units sewered out is logical if it is true that the
sewer leaks which develop over time tend to involve sewer
mains which run along the stream.  Finally, the use of dis-
charge form is most logical for population sewered to dis-

                              69

-------
posal points in the watershed, if it is the case that the
volume of treatment plant effluent is uncorrelated with stream
discharge at base flow.  In any case, the use of mixed form,
which does incorporate discharge, is more logical than the
use of density form.

A somewhat surprising result of the analysis was the fact
that the variable expressing total manufacturing employment
proved significant for many more chemicals than the variable
expressing employment in high water-using and water-polluting
industries.  This result may be due more to the inadequacy
of the latter variable than the strength of the former vari-
able.

To some extent, the significance of the total manufacturing
employment variable may be attributed simply to human pre-
sence, i.e., to the production of domestic-type sewage by manu-
facturing employees.  This would be consistent with the use
of the manufacturing variable in density or mixed form rather
than discharge form, since industries are unlikely to dis-
charge domestic-type effluent directly into a stream.

The manufacturing employment variable was the only informa-
tion in the study concerning non-residential human presence.
The suggestion that such presence per se_ may be an important
influence on stream water quality indicates that it might
have been worthwhile to have a total employment variable for
each watershed.  Data on non-manufacturing employment (or
some other index of human presence) would have been much more
difficult to obtain, however.  A second weakness of the study
in this regard was the fact that manufacturing employment was
not categorized by type of sewage facilities available in the
area, as was done for residential population.

On the whole, the strongest of all the independent variables
was the "soil score" variable; this was significant at the 1%
level for 13 of the 19 chemicals.  In general, soil character-
istics are expected to be important since they relate both to
the capacity of soils to filter and purify effluents in the
water passing through the soil, and to chemicals in stream
water which originate from the soil; i.e., the geologic in-
fluence on stream water quality.  As described earlier, the
soil score variable measures in a rough way the drainage

                              70

-------
characteristics of a soil, with scores ranging between 4.0
for a "well-drained" soil and 1.0 for a "poorly drained"
soil as described by the U.S. Department of Agriculture's
soil survey.  The results obtained from this variable in-
dicate in every case that higher chemical concentrations in
stream water are associated with lower values of soil score,
that is, with more poorly drained soils.  This result is
difficult to interpret for two reasons: first, because it
is impossible to determine to what extent the effect is
actually due to drainage and to what extent it is due to
geologic factors; and second, because it is not entirely
clear how soil drainage characteristics relate to the average
amount of purification and filtration which has been under-
gone by water entering the stream.  In general, it would
seem likely that stream water from watersheds with poorly
drained soils will have penetrated less deeply into the soil
than water from well-drained watersheds, and therefore, will
have undergone less cleansing effect; but this may not always
be true.

The significance of the soil score variable is probably due,
to a considerable extent, to basic differences in geology
between the watersheds in our sample characterized by well-
drained soils and those characterized by poorly drained
soils.  Most of the variation in soils scores in our sample
is due to differences between broad geographic areas; the
watersheds with low average soil scores are all located in
a well-defined region lying to the north of Philadelphia.
Thus, the importance attributed to the soils score variable
may be due partially to causal factors involving geology
rather than to soil drainage itself.

Before discussing results for individual chemicals, it is
worthwhile to consider briefly the data for the seven streams
which were excluded from the sample.  These data are presented
in Table 9 below.  As is shown by comparison with the means
for the various chemicals given in Table 4, the concentrations
for these seven streams tend to be very much higher than the
concentrations in the sample actually analyzed; if left in
the sample they clearly would have dominated the regression
results.  (Most values of R^ would have been much higher if
these streams had been included, since all the watersheds
except stream 49 were very high in at least one of the in-
dexes of urbanization.)


                              71

-------
                                                         Table  9

                            CHEMICAL DATA  FOR WATERSHEDS  EXCLUDED  FROM ANALYSIS
N5
No population
Sewered in:


N7o: 49
   90
      Average:

Population Sewered in
 (Major disposal plants);
   46
   57
   58
   59
   67
      Average:
        No Population
         Bewared in:

           49
           90
              Average:

        Population Sewered in
         (Major disposal plants);

           46
           57
           58
           59
           67
              Average:
                                                    Chemicals for which "Sewered in" is Significant
Fe

02
01
015
04
08
,07
07
09
07
Ca

46
122
84
60
69
52
54
41
55
K

4.5
2.0
3.25
18.0
20.0
15.0
12.0
13.0
15.6
Chemicals
SiOn
10
27
19
18
22
20
24
31
23









Mn
.00
.04
.02
.04
.11
.25
.00
.09
.10
Na

44
84
64
69
126
122
89
69
95
for
Mg
16
33
25
19
23
23
19
11
19
HC00

120
267
194
283
128
198
207
102
184
Cl

80
134
107
105
181
167
115
104
134
which "Sewered in1'
§P-4
76
226
151
67
138
103
83
45
87
F
1.0
.4
.7
.5
.7
.7
.5
.7
.6
NO,

6.5
3.0
4.8
5.5 23
75.0 2
38.0 2
34.0 2
70.0 19
44.5 9
SS*
.10
.79
.45
.00
.20
.10
.70
.00
.80
SUM

363
763
563
500
718
638
532
435
565
ROE

336
812
574
490
755
674
557
461
587
KxlO"

601
1,230
916
847
1,190
1,130
910
724
960
Ca. MI
181
440
311
228
267
224
213
148
216'
is not Significant
Non-carb.
83
221
152
0
162
62
43
64
66


















PH
8.4
7.9

7.0
6.7
7.0
7.9
7.2






















-------
It is of interest to consider these data simply to see if
they are more or less consistent with the results obtained
for the reduced sample.  The most important feature of this
set of streams is the presence of large amounts of population
(and industry) which are "sewered in."  Five of the seven
watersheds contain major sewage disposal plants, each serving
a far greater population than any disposal facilities in the
watersheds left in the sample.  For this reason, it is of
particular interest to see whether the importance attributed
to population sewered in seems to hold also for these streams.

In Table 9, the two streams without disposal plants and the
five streams with disposal plants are segregated, and, for
each group, separate means are computed for each of the 19
chemicals.  The data pertaining to the 12 chemicals for
which population sewered in has been found significant in
the main analysis, with these streams omitted, are presented
in the upper portion of the table; the data for the remaining
chemicals are presented in the lower portion.  It is noted
that for the former set of chemicals, the mean for the five
streams containing disposal plants is higher than the mean
for the other two streams for 9 out of 12 chemicals.  For
the remaining chemicals (excluding pH), the mean for the
streams with disposal plants exceeds the mean for the others
in only 2 out of 6 cases.  This provides a very rough indi-
cation that population sewered in is indeed more important
for the first set of chemicals than for the second.  Due to
the considerable within-group variations the difference be-
tween means is statistically significant at the 5% level only
for two chemicals, namely iron and potassium.

The fact that not all the relationships between means are as
expected, and the fact that high levels of within-group vari-
ation are observed, presumably reflect the great differences
in amounts of industry present in the various watersheds.
The watershed of stream 90 (one of the two without disposal
plants) contains a large amount of high water-polluting in-
dustry; on the other hand, the watershed of stream 46, which
contains the largest sewered-in population, contains virtually
no industry.   Finally, it should be noted that for both ni-
trates and phosphates, the chemicals most prominently related
to human activities, the means for streams with disposal
plants are strikingly higher than the means for streams without


                              73

-------
disposal plants.  However, these differences between means
fail to be statistically significant because of high within-
group variation.

Results for Individual Chemicals
With the results of the analysis as a whole in mind, we shall
now consider the results for specific chemical indicators of
water quality.  In this section we will not attempt an exhaus-
tive review of the results for each of the chemicals, but
instead will concentrate our comments on the chemicals or
groups of chemicals which are of greatest interest.  The
statistical results for each chemical are,  of course,
summarized in Tables 7 and 8.

Sodium, Chloride, and Potassium
Most of the urbanization-related variables were found to be
statistically significant in explaining both the sodium and
chloride concentrations.  Only employment in high water-using
and water-polluting industries failed to be significant for
chloride; for sodium, population in post-war dwelling units
sewered out was also not significant.  The strong association
with urban-related variables was somewhat expected for both
of these chemicals.

Sodium and chloride tend to occur in similar concentrations,
since they are most frequently derived from the single com-
pound, sodium chloride.  In solution sodium chloride tends
to dissociate into sodium cations and chloride anions; in
the soil, sodium cations can be absorbed onto colloidal
surfaces and thus be removed temporarily from the soil-water
solution, whereas the chloride anions tend to remain in the
solution.  Thus, one might expect sodium concentrations to
be affected more by travel through the soils than chloride
concentrations.  In this regard, it is interesting to compare
the four population-related variables considered here with
respect to the probability that effluent must travel an
appreciable distance through the soil to reach the stream.
In the following table, the variables are arranged in in-
creasing order of this probability (under the assumption
that leakage from post-war sewer systems tends to be local
whereas leakage from pre-war systems is more likely to in-
volve the sewer mains running near the stream).  It is
notable that the ratio of the sodium regression coefficient
to the chloride coefficient decreases with the probability


                              74

-------
                         Table 10

        COMPARISON OF REGRESSION COEFFICIENTS FOR
                   SODIUM AND CHLORIDE

                                             Ratio of Coeff.
Variable                 Regression Coeffs.   	Na/Cl

                           Na        Cl

Population sewered in    .01726    .02253          .766

Pre-war population
   sewered out           .00143    .00196          .730
                                       /

Post-war population
   sewered out            	     .00230           	

Population unsewered     .00867    .02652          .330
that the effluent must travel an appreciable distance through
the soil.

Potassium is an alkali metal, similar to sodium; however, it
has a tendency to be selectively returned to the solid phase
in the soil by recombination with clay minerals or by base
exchange.  As in the case of sodium and chloride, the soil
score variable was found to be statistically significant in
explaining potassium concentrations.  This is consistent with
the function of the soil of selectively removing potassium
from solution.  Of the urbanization variables, population
sewered in and population in post-war dwelling units sewered
out were found to be significant for potassium; population
in pre-war dwelling units sewered out was also significant
in the unforced regression.  The failure of population in
unsewered dwelling units to be significant may reflect the
ability of the soil to remove potassium from sewage affluent
within a short distance of travel.

As in the case of nitrate concentrations, potassium is much
better explained by the urbanization variables in density
form than in the forms used in the forced regression.  The
results obtained with all population-related variables in
density form are shown below.
                              75

-------
                            Table  11

     REGRESSION FOR POTASSIUM WITH POPULATION-RELATED
                VARIABLES  IN DENSITY FORM
Variable                    Coefficient            t-statistic

Soil  score                      -.9370                3.74

Post-war  sewered  out             .0031                3.45
                      \
Pre-war sewered out              .0027                2.25

Sewered in                       .0104                4.44
               2
The value of R for this regression was 0.53, as opposed to
0.43  in the forced regression.

Calcium,  Magnesium and Hardness
All of the urbanization-related  variables with the exception
of employment  in  high water-using and water-polluting indus-
tries were statistically significant at the 170 level in
explaining calcium concentrations.  For magnesium, population
sewered in and total manufacturing employment were not sig-
nificant, and  the overall  level  of explanation was somewhat
lower  (R^ = 0.60, vs. R2 = 0.80  for calcium).  These results
are generally  consistent with the expectation that these
chemicals might be related to human activities, both directly
through their  presence in  industrial effluent and domestic
sewage, and indirectly, through  displacement by other cations
such as sodium which are contained in sewage.  The signifi-
cance of  the manufacturing variable for calcium was predic-
table since its presence in manufacturing effluent is common.
Magnesium is not  as commonly found in industrial effluents,
so that the non-significance of  the manufacturing variable
might have been expected in this case.

The regression results for carbonate (or temporary) hardness
are virtually  identical to those obtained for calcium, with
each regression coefficient roughtly four times its value
for calcium.   This was to  be expected since the two variables
were almost perfectly correlated (R = 0.987).

                              76

-------
For non-carbonate (permanent) hardness only two variables
beside the soil score variable were significant, namely popu-
lation in pre-war and post-war units sewered out.  These re-
sults indicate that non-carbonate hardness does bear some
relationship to overall intensity of urbanization.

Sulfates, Nitrates, and Phosphates
Sulfates, nitrates, and phosphates are among those constituents
expected to be most influenced by man's activities.  Somewhat
paradoxically, in this study each of these chemicals has been
found to be significantly related to relatively few of the
independent variables which deal with man's activities.  Only
one of the independent variables, besides the soil score
variable, was significant in explaining sulfate1 concentrations;
two variables were significant for nitrates and two for phos-
phates (in forced form).  The level of statistical explana-
tion was high for phosphates but was quite low for nitrates
and would have been low for sulfates without the soil score
variable.

Sulfates, nitrates, and phosphates are all present in fully
degraded sewage.  However, sulfates and nitrates are derived
from intermediate products of oxidation of sulfur and nitro-
gen, namely sulfides, sulfites, ammonia, and nitrites, which
were not measured in this study.  There may be a trade-off
between amounts of nitrates and sulfates observed and the
amounts of these intermediate products.  This might explain
partially why the concentrations of nitrates and sulfates
alone are not correlated as highly with urbanization vari-
ables as we would expect for sulfur and nitrogen compounds
on the whole.

Sulfates.
The strong contribution of the soil score variable in explain-
ing sulfate concentrations would suggest that different soils
vary greatly in their ability to filter sulfates.  An alter-
native explanation would be that geologic factors are rela-
tively more important as a source of sulfate than was anti-
cipated.  The appearance of population in pre-war sewered-
out dwelling units as the sole urbanization-related explan-
atory variable for sulfates is difficult to explain.  The
failure of population in unsewered dwelling units to be
significant might be explained by rapid filtering action of

                              77

-------
the soil, which would be most relevant to the effects of
this population.  The failure of population sewered in to
be significant is surprising, although it is possible that
the sulfur contained in the effluent from these disposal
plants is concentrated in intermediate forms rather than in
the stabilized sulfur form.  With respect to population sew-
ered out, it is likely that population in post-war as well
as pre-war units is contributing to the the effect.

Nitrates.
Three urbanization-related independent variables were signi-
ficant in explaining nitrate concentrations in the unforced
regression.  These three variables, all in density form,
were:  population in post-war dwelling units sewered out,
population in dwelling units sewered in, and population in
unsewered dwelling units (2,000-foot gradient).  The last
of these three variables was not significant when changed
to mixed form in the forced regression.  Nitrates was one
of two chemicals, the other being potassium, for which the
use of independent variables in density form was markedly
more successful than the use of the forms chosen in the
forced regression.  The results of the regression with all
the variables in density form are shown below; R^ for this
regression is 0.35 as opposed to 0.29 for the forced regres-
sion.

The failure of population in pre-war housing units sewered
out to appear significantly is not considered particularly
noteworthy.
                             Table 12

    REGRESSION FOR NITRATES WITH POPULATION-RELATED VARIABLES
                          IN DENSITY FORM
Variable

Post-war sewered out

Sewered in

Unsewered (2,000-foot
   gradient)
Coefficient

  .00248

  .05854

  .02036
t-statistic

   5.06

   3.87

   2.08
                              78

-------
The significant features of the results obtained for nitrates
were the greater success of variables in density form than
in mixed or discharge form, and the failure of land in cul-
tivation to appear as a significant variable.  The latter
finding was surprising because nitrate fertilizers are
commonly considered to be a very important source of nitrates
in surface waters.  In examining these findings it was noted
that nitrate concentrations seemed to have a positive rela-
tionship with stream discharge.  Specifically, quite low
values of N03 were observed for streams with very low dis-
charge, even though these included some watersheds with
very high percentages of land in cultivation.  (The greater
success of independent variables in density form, rather
than a form involving discharge in the denominator, would
also be consistent with there being some sort of positive
effect associated with discharge.)

A further analysis of nitrate concentrations was conducted,
using only the rural streams, with population densities
below 250 people per square mile.  This sub-sample was par-
titioned into two parts:  the observations for thirteen
streams to the north and east of Philadelphia, and 36
observations from the Brandywine area to the west of Phila-
delphia.  For each of these groups taken separately, a "t-
test" was performed to test for differences between nitrate
concentrations where  discharge was low.  The dividing line
between "high" discharge and "low" discharge was set at 0.2
cubic feet per second per square mile.  (There were in fact
no observations with discharge between 0.16 and 0.26 cubic
feet per second per square mile.)   In both cases, the means
for observations with high and low discharge were found to
be significantly different at the 170 level.  These tests
are summarized in Table 13.

These results would indicate that there is some sort of
positive association between nitrate concentrations and dis-
charge at base flow, at least for rural streams.
                              79

-------
                         Table 13

COMPARISONS OF NITRATE CONCENTRATIONS IN HIGH AND LOW
                   DISCHARGE WATERSHEDS


               Brandywine Observations   Observation N & E of Phila,
               Number    Mean N0~(ppm)      Number  Mean NO-Cppm)


High Discharge
 watersheds      31          6.42             4       20.00

Low Discharge
 watersheds       5          2.58             9        4.54

   Results of
    t-test            t = 2.73                   t = 4.99
Next, all of the observations with low discharge, including
both Brandywine and non-Brandywine streams, were pooled to-
gether; and similarly for streams with high discharge.  Within
each of these groups of observations, the correlation between
nitrate concentration and percent of watershed area in culti-
vation was computed.  For both groups the correlation was
positive, and for the high-discharge group it was highly sig-
nificant  (R = 0.86).  The correlation for the low-discharge
group (R = 0.27) was not statistically significant, but there
was relatively little variance in nitrate concentrations in
this sample.

Since the results for the high-discharge group were dominated
by three non-Brandywine streams, with very high percentages
of land in cultivation, the correlation was also computed
for Brandywine streams alone.  This correlation was 0.65,
still statistically significant.  Thus there did appear to
be a strong positive relationship between nitrate concentra-
tion and the proportion of land in cultivation (which here
may be a surrogate for intensity of cultivation as well as
extensiveness of cultivation).  This relationship was appar-
ently obscurbed in the main analysis by the postive relation-

                              80

-------
ship existing between nitrate concentrations and discharge.

Although the study was eventually successful in demonstrating
a relationship between nitrates and land in cultivation for
a sub-sample of observations, the nitrate results on the
whole were disappointing because of the low level of explan-
ation achieved for the entire sample.   This low level of
explanation occurred because    many of the highest nitrate
concentrations pertained to streams in "urban fringe" areas,
where the proportion of land in cultivation is fairly low
but housing density is not yet high.  The high concentrations
observed for these streams remain a mystery; perhaps they
are related to the presence of specialized types of agricul-
ture (e.g., nurseries) which are often found in urban fringe
areas.

Phosphates.
Phosphate concentrations in stream waters are known to be
heavily affected by the presence of domestic waste, and in
particular by household detergents.  The results obtained
are consistent with this; our results indicate that the
determination of phosphate concentrations is attributed
overwhelmingly to population in dwelling units sewered to
disposal plants within the watershed.   Since secondary
sewage treatement does not remove phosphates from sewage
effluent, municipal disposal plants simply serve as con-
centration points for the introduction of phosphates into
streams.  The failure of population in unsewered dwelling
units to be an important influence on observed phosphate
concentrations should perhaps have been expected, in view
of the fact that phosphates introduced into the soil in
small quantities are likely to be adsorbed onto soil par-
ticles  or used as nutrients by plants.  The failure of the
manufacturing variables to be significant could perhaps also
have been expected.  On the other hand, the appearance of the
pre-war sewered out variable but not the post-war variable
probably has little significance in itself.  As in the case
of nitrates (where the opposite occurred), the situation is
probably that populations in both pre-war and post-war
dwelling units are contributing moderately to stream concen-
trations, with the effect of population in pre-war units
being somewhat the greater.


                              81

-------
As was mentioned earlier, the sample of observations for phos-
phates presented extreme problems of heteroscadasticity, i.e.,
dominance by large observations.  For the original sample of
80 observations, phosphate concentrations ranged up to 19 and
23 parts per million  (both for streams with major disposal
plants), whereas for  all but 9 of the observations concentra-
tions were below 1 part per million.  Even with 7 observations
eliminated, as was done for all the chemicals, the remaining
sample was overly dominated by a few observations.  Therefore,
an additional regression was run with all watersheds contain-
ing any population "sewered in" eliminated from the sample.
The result was that the pre-war sewered out variable remained
significant, and the  soil-score variable also entered the
regression (with a negative coefficient that was barely signi-
ficant at the 5% level).  R2 was drastically reduced, from
.90 to .40.  It is notable that the mean phosphate concentra-
tion for this sample  was only 0.14 ppra.

Consideration of the  full set of data for phosphates suggests
strongly that the coefficient value for the population-
sewered-in variable given in Table 8, namely .00592, is some-
what too large.  A regression of phosphate concentrations on
population sewered in, using just the watersheds with non-
zero values of the latter variable, yielded a coefficient
estimate of .00300.   An additional regression was also run
using these observations, with both variables converted to
logarithmic form in order to eliminate heteroscadasticity;
the outcome of this was a coefficient estimate of .00269.
Thus, it appears likely that a more accurate estimate of
the coefficient relating phosphate concentrations to popu-
lation sewered out is around .003 rather than .006.

Bicarbonate and pH.
The amount of bicarbonate present in a water body depends
upon pH in the sense  that bicarbonate is a dominant ion only
in the range of pH from 4.5 to 8.2.  However, since nearly
all the streams in our sample had pH values well within this
range, and since, in  fact, there was little variation in pH,
bicarbonate and pH levels were not highly correlated in this
sample.

All of the urbanization variables were statistically signifi-
cant in explaining bicarbonate concentrations, with the
                              82

-------
exception of population in post-war dwelling units sewered
out.  The omission of this variable probably has no causal
significance.  On the whole, a rather strong relationship of
bicarbonate to urbanization is evident.

One of the unforced regressions for bicarbonates contained
land in cultivation as a significant variable.  These regres-
sion results are as follows (R.2 for this regression was
.65):
                            Table 14

              UNFORCED REGRESSION FOR BICARBONATES
Variable
Pre-war sewered out (mixed form)

Sewered in (mixed form)

Unsewered,gradient (mixed form)

Total manufacturing employment
  (density form)

Employment in high water-using
  and water-polluting industry
  (discharge form)

Land in cultivation (discharge
  form)
Coefficient   t-statistic
   .00738

   .06495

   .08669


   .00830
   .13148

   .86547
6.36

3.84

5.83


2.39
4.37

2.53
Only the soil score variable was significant in explaining
pH.  In this case, the soil score was probably serving as a
surrogate for the chemical effects of geologic factors.

Manganese, Fluoride, Iron and Silica.
The levels of statistical explanation of manganese and fluoride
concentrations were quite low.  Only variables relating to
manufacturing employment were significant.  For fluoride, both
total manufacturing employment and employment in high water-
using and water-polluting industries were significant, whereas
                              83

-------
for manganese  only the total employment was significant.

Two variables relating to population, namely population
sewered in and population in unsewered dwelling units, were
signficant in explaining iron concentrations.  The results
indicate a general relationship with urbanization; but the
choice of these two particular variables is not accorded a
great deal o£ importance.  In addition, total manufacturing
employment was significant, as might have been expected.

The statistical explanation of silica concentrations was
extremely low, with the only significant variable being popu-
lation in unsewered dwelling units.  No meaning is attached
to these results, since silica is rarely found in domestic
and municipal waste.

Total Dissolved Solids.
As discussed above, total dissolved solids was measured in
three different ways: as the sum of individual components,
as the residual on evaporation, and as conductivity.  The
statistical results obtained were very similar for the
three cases, despite the fact that somewhat different samples
were employed in each case.  For all three measures of total
dissolved solids, the following variables were significant:
soil score, population in pre-war units sewered out, popula-
tion sewered in, population in unsewered units, and total
manufacturing employment.  In addition, population in post-
war dwelling units sewered out was significant for dissolved
solids measured as residual on evaporation.  Since total
dissolved solids is a summary measure, the results are
generally similar to those obtained for most individual
chemicals.  The strong effects of urbanization on this
summary measure of water quality are evident.
CONCLUSIONS
Rather than summarizing all of the results presented above,
we will present here  several overall conclusions which have
emerged from this study.  First, significant effects involving
urbanization have been found for a number of chemicals which
are often considered as being geologically determined rather
than produced primarily by human activities.  Some of these
chemicals are:  calcium, carbonate hardness, magnesium,


                              84

-------
sodium, and bicarbonate.  Second, the relative effects of
urbanization variables upon chemical concentrations are
quite similar for a large proportion of the chemicals.  As
was noted in the discussion of the results as a whole, there
is a common pattern of regression coefficients (when signi-
ficant) , with large effects being attributed to population
in unsewered dwelling units and to population in dwelling
units whose sewage is disposed of within the watershed.
Smaller effects are attributed to population in pre-war
"sewered out" dwelling units, and still smaller effects to
population in post-war dwelling units sewered out.  Third,
the small but nevertheless significant effects for popula-
tion sewered out and for total manufacturing employment
(most of which was in areas "sewered out") indicate that
urbanization is likely to cause lower water quality regard-
less of what is done with sewage.  The partial futility of
sewage treatment is also demonstrated by the fact that
regression coefficients for population sewered in are often
as high as those for population in unsewered dwelling units.
Fourth, although the importance of location of development
within a watershed was not examined for sewered dwelling
units, this location factor was found to be an important
determinant of the effects of population in unsewered
dwellings.  Specifically, the effects produced were inver-
sely related to distance from the stream channel.  The
results generally indicate that unsewered houses located
1/4 mile or more from the stream channel may have very
little effect.

In general, the results indicate that urbanization has a
pervasive influence on water quality.
                              85

-------
                IV.  OPEN SPACE EFFECTS
INTRODUCTION
Almost by definition, urban development results in a re-
duction in the amount of land devoted to agriculture,
forestry, and other natural uses.  Generally, the only
land left in open space uses is land which is set aside
permanently as public parks, or is held in relatively open
uses by institutions or by private owners.  Some land is
withheld from development by speculators who are waiting
for the time when demand will justify high density develop-
ment.  Although their primary intention is not speculation,
institutions and private house-owners also respond to
development pressure, sell their land for more intensive
redevelopment, and move on to newly developing areas else-
where .

The loss of open space to development has serious consequences
for stream hydrology and water quality, as detailed in the
preceding sections.  It also has consequences for the eco-
logical health of the area and reduces the recreation oppor-
tunities and amenities available to residents.

A land use plan for water resources protection would maintain
sizeable areas in open uses.  However, it cannot be assumed
that those areas would necessarily become fully developed in
the absence of a plan.  In fact, urban development in all
but the most fully developed areas tends to be somewhat
haphazard, resulting in "scatteration."  Land near streams,
in particular, tends to be either too wet or too steep to
be prime building land.  Therefore, in evaluating a land use
plan for water resources protection, one must be careful to
compare the amount of open space with the plan to the amount
which would be expected without the plan.

RESEARCH PROCEDURE
In order to determine the probability that development would
occur near a stream, we have examined a large sample of
streams in the Pennsylvania suburbs of Philadelphia [Coughlin,
Sheldon, & Hammer, 1971].  Heavily urbanized areas, such as
                              87

-------
the City of Philadelphia, were excluded from the sample
since such heavy urban development is not likely to be
typical in the future.

Most of the observations were made on air photographs (1" =
400') taken in the spring of 1970 for the Delaware Valley
Regional Planning Commission.  On these maps marks were
made every 1000 feet along each stream and lines were drawn
parallel to the stream at distances of 300 feet, 600 feet,
and 900 feet.  Each observation, then, referred to a segment
300 feet wide and roughly 1000 feet long.  The final sample
consisted of 1,926 observations: 856 segments lying within
300 feet of a stream, 606 segments lying between 301 and
600 feet, and 464 lying between 601 and 900 feet from a
stream.  The sample contained different numbers of segments
in the three categories because only the closer segments
were studied for many of the small streams.

For each segment the number of houses (that is, single-
family and two-family houses) was counted, as was the number
of apartment units.  The amount of impervious area resulting
from houses was estimated, and the amount of land covered by
streets and by other impervious surface was measured.  The
size of the watershed at the location of each segment was
noted and the average land slope (transverse to the stream)
of each segment was measured from U.S. Geological Survey
topographic maps.  Finally, the 1970 population density of
the township containing each segment was computed from
Census data.

RESULTS

Percent of Segments Having No Development
The most extreme plans of land use control for water resources
protection would prohibit all development near streams.  For
example, the Brandywine Plan proposed the public purchase of
conservation easements which would prevent development in
all flood plains and within 300 feet of all streams [Institute
for Environmental Studies, et al., Section III-B p. 10-21].
The data collected in this study make it possible to determine
to what extent such a development pattern has occurred
"naturally" in suburban Philadelphia, in the absence of a
conservation easement program.
                              88

-------
                  Table 15

PERCENT OF SEGMENTS (0-300'  BAND) HAVING NO
         DEVELOPMENT OF GIVEN TYPE,
           BY DENSITY OF TOWNSHIP
Type
of
Development
Houses
Apartments
Other imper-
vious area
Streets
Total dwellings
Total imper-
vious area
All
Locations
54.91
97.31
85.40
39.14
53.04
27.69
Density
0-999
68.54
100.00
98.88
53.93
68.54
43.26
of Township
1,000-1999
61.86
98.31
89.83
36.44
61.02
26.27
(persons per
2,000-2,999
52.42
95.42
82.95
37.66
49.11
26.21
sq. mi.)
3,000+
41.32
98.20
73.65
28.74
40.12
15.57
                        89

-------
Of all segments within 300 feet of the stream, 53.04% were
found to have no dwelling units at all (neither houses nor
apartments), as can be seen in Table 15.  In addition, 27.69%
did not have any measurable impervious area:  that is, no
streets, dwellings, or other structures were observed.
These would appear to be remarkably high percentages in
view of the fact that no general programs have been carried
out to preserve green space along streams in the Philadelphia
region, and that streams with major valley parks (most of
which are in the City of Philadelphia) have been omitted from
the analysis.

As would be expected the percent of segements with no devel-
opment falls as township density rises.  The percent with
no dwelling units falls from 68.54% in townships whose den-
sity is less than 1000 residents per square mile to 40.12%
in townships whose density is 3000 or over.  Similarly, the
percent of segments with no impervious area falls from 43.26
to 15.57.

Percent of Segments Developed at Over 0.25 Dwelling Units
Per Acre
A second development standard which has been suggested for
water resources protection is the limitation of development
to one house per four acres (or 0.25 dwelling units per
acre).  For example, in the Brandywine Plan [Institute for
Environmental Studies, et al., Section III-B, pp. 21-24]
this less stringent limit of 0.25 dwellings per acre was
set for wooded areas and areas with slopes of 15% or more
lying futher than 300 feet from streams and swales.

In our sample as a whole, only 53% of segments in the 301-600
foot band and 56% in the 601-900 foot band were more heavily
developed than 0.25 dwellings per acre (Table 16).  Intensity
of development, however, increases with development density
of township.  Thus, in townships with overall densities
between 3.0 and 5.0 dwelling units per acre, 87% of all
segments-in the 301-600 foot band and 10070 of all segments
in the 601-900 foot band are developed at more than 0.25
dwelling units per acre.

Not all the land within 301 and 900 feet of a stream would
have the characteristics of "woods and slopes" and, there-
fore, not all would be placed under the restrictions approp-


                              90

-------
                         Table 16

PERCENT  OF SEGMENTS  DEVELOPED AT  GREATER DENSITY THAN
             0.25 DWELLING UNITS PER ACRE
Township Density
   (du/ac)

  .001 -  .250

  .251 -  .500

  .501 -  .999

 1.000 - 1.999

 2.000 - 5.000

 All townships
                            Distance from Stream
0 - 300'
5.66
24.00
27.96
43.68
50.32
37.50
301 - 600'
15.79
47.66
45.87
55.69
77.96
52.81
601 - 900'
18.19
28.77
50.47
58.98
88.38
55.80
                           91

-------
riate to that land type.  Therefore, under a Brandywine-type
plan which limited development of woods and slopes areas to
0.25 dwelling units per acre, density of development in the
301-900 foot band would be greater than 0.25 dwelling units
per acre.  This fact makes the interpretation of Table 16
difficult, but it remains evident that development in sub-
urban Philadelphia typically exceeds the densities envisioned
in the 301-900 foot band by the Brandywine Plan.

Percent of Segments Developed at Over 1 Dwelling Unit Per Acre
It is instructive to examine one more density limit--one
dwelling unit per acre.  This is about the lowest density
which can be maintained through zoning, since ordinances
specifying a lower density are likely to be judged uncon-
stitutional.  It should be noted, however, that as a prac-
tical matter, one-acre zoning is not likely to withstand
the pressures of development.  As these pressures mount, re-
zoning can be expected which will enable development at
higher densities.

Table 17 indicates that as long as the average density of
dwelling units in a township remains below one dwelling unit
per acre, the probability is rather small that land near
streams will be developed at over one dwelling unit per acre.
However, the farther a segment is from the stream, the greater
its development is likely to be.  For example, for townships
in the density class of 2.0-5.0 dwelling units per acre, the
probability is very great that development near streams will
be greater than one dwelling unit per acre; it increases
from 29% in the 0-300 foot band, to 64% in the 301-600 foot
band, and to 77% in the 601-900 foot band.

Tables 15-17 give a rough indication of how development
density of streamside segments varies with township density.
However, it is not clear from the tables whether the vari-
ation in density is due solely to variation in township
density, or whether it is related to other factors, such as
valley slope, width of flood plain, or watershed areas.  A
simple cross-tabulation is not adequate for showing how the
various factors interact in affecting density of development.

Factors Related to Density of Development
In order to take into account more than one variable at a
                              92

-------
                        Table 17

       PERCENT OF SEGMENTS  HAVING ONE OR MORE
               DWELLING UNITS PER ACRE
TownshipDensity
   (du/ac)

  .001 -  .250

  .251 -  .500

  .501 -  .999

 1.000 -1.999

 2.000 -5.000


 All townships
                               Distance from Stream
0 - 300'
0
6.40
5.08
24.57
28.66
18.46
301 - 600'
5.26
30.22
13.76
35.77
63.78
32.84
601 - 900'
4.55
15.07
24.75
41.75
76.75
38.35
                               93

-------
time in explaining density of development, a set of multiple
regression analyses was performed.  The dependent variable
for each analysis was a specific measure of density of devel-
opment of each segment.  The independent variables consisted
of density of township and a number of measures of each seg-
ment.

More specifically the dependent variables are:

Y,    Density of Houses (du/acre)

Y2    Density of Apartment Units  (du/acre)
Y~    Density of Streets (mi/sq mi)

Y,    Density of "Other Impervious Area" (ac/sq mi)

Y5    Density of Total Dwelling Units (du/acre)

Y,    Density of Total Impervious Area (ac/sq mi)

These variables were computed separately for segments in the
0-300, 301-600, and 601-900 foot bands.

The independent variables used in all analysis are:

Z,    Overall Valley Slope (within 900 feet from stream)
Z2    Valley Slope in 0-300 foot band

Z3    Valley slope in 301-600 foot band
Z,    Valley slope in 601-900 foot band
2.     Width of Flood Plain (feet)
I.,    Population Density of Township (persons/sq mi)
 6
Z?    Drainage Area of Watershed  (sq mi)

Note that all variables except Z$, Zg and Zj refer to the
segments observed.  Z$ refers to the flood plain at the
given segment, Zg to the entire township in which the segment
is located, and Zy to the watershed of which it is the lowest
point.

On the basis of logical considerations, one would expect
that the independent variables act in combination, rather
than singly.  For example,  a gentle valley slope should tend,


                              94

-------
to encourage development, but this effect would only be im-
portant for areas (townships) with substantial overall
densities; a gentle slope by itself is not enough to induce
development.  Therefore, a multiplicative regression model
was employed.  The form of the regressions was as follows:


        Yi = kZ!ZjZ5Z6Z7
or      Y. = log k + a log Z. + c log Z  + d log Z  + e log Z


         i is one of the dependent variables, Y , ..., Y,
                                               1        o >
           measured for one of the 3 distance intervals from
           the stream, and
where   Z. is the valley slope variable, Z , Z,. , or Z,, per-

           taining to the given distance interval, and
                                                       '1
           Zc  Z, and Z_ are as defined above.
            5,6      7

The results of the regression analysis are summarized in
Table 18.  Only significant variables are included, i.e.,
variables for which the t-statistics exceeded 2.0.

It will be noted that only three of the independent variables
appear in the final regressions: overall valley slope, town-
ship population density, and drainage area.  The more specific
slope measurements for each band, and the variable expressing
an approximate measurement of width of flood plain, did not
prove to be statistically significant.

Population density of township, which appears in each equation,
is by far the single best explanatory variable.  This is con-
sistent with the notion that urban development tends to engulf
all in its path, without paying respect to natural features.

Overall valley slope appears significantly in six of the 18
equations, in four of the equations for the 0-300' band,
and twice in equations for the 301-600' band.  These results
indicate that the greater the valley slope, the less develop-
ment is likely.  It is of interest that this relationship
holds for houses, but does not hold for "other impervious
area."  This supports an impression experienced during data
gathering:  that single houses tend not to be built near

                              95

-------
                                                    Table 18

                                          SUMMARY 01- UKCRESSIOK RESULTS
                                 Independent Variables

i

Log of Overall
Valley Slope
Log of
Township Popu-
lation Dcnsi ty
Dependent VariableJ

Name
Log of
Density
of Houses
Log of
Density of
Apartment
Units
Log of
Density of
Streets
Log of
Density
of Other
Impervi-
ous Area

Band
0-300'
301-600'
601-900'

0-300'
301-600'
601-900'
0-300'
301-600'
601-900'


0-300'
301-600'
601-900'

R
.3150
.3990
.4062

.0783
.1356
.1759
.2169
.3625
.3515


.2510
.2969
.2175
t-Sta-
Coeff. tistic
-.0546 3.049

	

	
	
	
-.0606 2.529
-.0811 3.420
	



	
	

Cooff .
.1408
.2339
.3076

.0386
.0797
.0740
.1029
.1745
.1988


.2336
.3166
.2624

t-Sta-
ti.stic
9.481
10.695
9.327

2.280
3.361
2.828
5.186
9.507
8.00C


7.569
7.635
4.785
Log of
brainap.c Area


Cocff .
-.0438

.0583

	
	
.0409
-.0311
	
.0281


	
	
	

t-Sta-
tistic
6.300

3.316



2.949
3.346

2.121






F
Ratio
46.4
114.4
45.5

5.2
11.3
7.3
14.0
45.6
32.4


57.3
58.3
22.9

Level of
Significance
17.
17.
1%

57.
17.
17.
17.
17.
17.


17.
17.
17.
Log of
Density of
Total         0-300'  .3224
Dwelling    301-600'  .4007
Units       601-900'  .4480
-.0702   2.720
.1838     8.532
.3058    10.747
.3790    10.358
-.0444   4.438

 .0847   4.347
 32.9
115.5
 57.8
17.
17.
17.
Log of
Density of
Total Im-     0-300'  .3412
pervious    301-600'  .4452
Area        601-900'  .4052
 .3824   4.628       .8056    9.586      -.1536   2.899       37.4
 .3404   4.088      1.0957   12.163        .1290   2.112       49.6
                    1.0556    9.526        	       90.75
                                                      17.
                                                      17.
                                                      17.
 Multiple R's, Coefficients, F's and F ratios on last step of regression where all t's  were  greater  than  4.0


'"In order to avoid the problem of taking the logarithm of zero, arbitrary constants were added  to  the  dependent
 variables before taking logarithms.  Thus, dependent variables were constructed using  constants from  the set:
 0.01, 0.1, 1.0, and 10.0.  Ihe resulting dependent variables which yielded the nost  significant statistical
 results are those to which the regression results of this table refer.   Specifically,  before taking logs,  1.0
 was added to each of the above variables except density of streets, to  which 10.0 was  added.
                                                       96

-------
streams, but that apartment houses and some non-residential
uses often appear there.  The implication is that in the
course of development single houses are first built on the
readily developable land; then as development continues,
there is a market for higher intensity uses, and at the
same time, relatively few readily developable sites remain.
Streamside sites which in the earlier stage of development
were uneconomical to develop, and which in some cases may
have been protected by zoning, now can be developed econo-
mically for high density uses, and the zoning restrictions,
which were adequate to withstand light and moderate develop-
ment pressure, fall before the higher land prices created
by the increased pressure for development.

The effect of drainage area is consistent, for the most part,
with our impressions during data gathering: the larger the
stream, the less likely is development within 300 feet of
it.  On the other hand, development in the 601-900 foot
band is positively associated with size of watershed.  It
would appear that larger streams have a more obvious amenity
value, which results from increased valley slope as well as
increased size of stream.  This amenity value is often pro-
tected by public or institutional land purchase, and is then
capitalized on by private developers who build houses, apart-
ments, and other structures as close to the protected (and
physically difficult-to-build-on) area as they can get.

The regression equations for houses are summarized graphically
in Figure 7.  Since the effect of drainage area is significant
for the 0-300 foot band, a separate curve is presented for
drainage areas of .25, 2.5 and 25.0 square miles.  The re-
gression results for total impervious area are depicted in
Figure 8.

In these curves, one can observe the rising development density
of Streamside segments with increased township population
density, and the tendency of segments in outer bands to be
more densely developed than segments nearer the stream,
particularly for higher township densities.

CONCLUSIONS AND IMPLICATIONS
The general impression gained from the data is that stream
valleys are relatively lightly developed.  Remarkably high
percentages of the segments studied have no development at
                              97

-------
                           Figure 7
           DENSITY OF DWELLING UNITS  IN HOUSES
            FOUND IN  BANDS ALONGSIDE  STREAMS
                                                     :•!-.!..I: ••;-. i:.:
                                                        900 Ft.Band !
                                                       -600 Ft.Band
                                                        . I . . .

                                                     -0-300 Ft.'BandJ
                                                       .25 Mi.2  .,: j
                                                     '. Drainage Area
                                                       2,5: Ml.
                                                    i ; Drainage Area!
                                                       25: Mi.2;
                                                     . Drainage; Area:
±i±250:x; n ri  500 :! -;-1000 •  i  _:.2.ppO;.j _;; j:; L^OOO: i [. ;:^80pO;::.:';::.;
2-TpWnship!- Popuiafcibnl-DensityiJ(Ih-Eersons^^-PertSqr-Mi^^RatiolScale)—- -J
                               98

-------
                       Figure  8


          DENSITY OF  TOTAL IMPERVIOUS AREA

          FOUND IN BANDS ALONGSIDE STREAMS
RJ
•u
o
H
awnsh]Lp: Population Dfensltjr (In| Persons'
                           99

-------
all, except for streets.  However, as township density rises,
the amount of development in streamside segments increases
also.  Therefore, land use control programs are necessary if
stream valleys are to be kept in an undeveloped state.  Such
programs are especially important if it is desired to retain
a continuous green strip along streams, since even if natural
market forces result in a small total amount of development,
that development may change the visual impression of the
landscape dramatically, and may interrupt streamside walkways.

The net effect of a land use control program depends upon two
factors:  the extent of development in streamside land when
the program was put into effect, and the subsequent urban
growth of the townships.  An impression of the effect of
these factors can be gained from examination of Figure 8.
Assume that a control program which will prevent all future
development in the 0-300 foot band is instituted at a time
when total impervious area in the band is 4 acres per square
mile.  The amount of total impervious area in the band will
not be expected to increase subsequently even though the
population of the township grows substantially.  This amount
of expected impervious area is given by the dashed line in
Figure 8.  The net effect of the controls, then, is given
by the difference between the dashed line and the curve
showing expected amount of total impervious area in the
0-300 foot band under normal conditions.  Thus, when township
population density reaches 1,000 persons per square mile the
net effect of the control program would be to have prevented
the construction of 8 acres of impervious area per square
mile--677o of what would normally have been built.  By the time
the township density had increased to 8,000 persons per square
mile, the net effect would be to have prevented construction
of 50 acres of impervious area--93% of what would normally
have been built.

Alternatively, the same control program might not have been
instituted until total impervious area in the band amounted
to 12 acres per square mile.  The net effect of such a pro-
gram when the township reaches a density of 1,000 persons
per acre would be estimated to be negligible, since on
average a 0-300 foot streamside band in a township with a
density of 1,000 persons per acre would have 12 acres of
impervious area per square mile.  By the time the township
                             100

-------
density had increased to 8,000 persons per square mile,  how-
ever, the net effect of the program would be to have prevented
construction of 42 acres of impervious surface per square
mile.  The prevention of so much impervious area would have
strong effects on streamflow and channel enlargement.

It must be borne in mind that conclusions and implications
drawn from the equations of Table 18 and the curves of Figures
7 and 8 should be treated as only rough indications of the
amount of development which might be expected in any parti-
cular streamside segment.  The data, however, are broadly
representative of the Pennsylvania portion of metropolitan
Philadelphia.  It is believed that they would also be repre-
sentative of rolling terrain in other metropolitan areas of
the eastern United States.
                             101

-------
          V.  EVALUATION OF EFFECTS:  ASPECTS OF
                 ENVIRONMENTAL PREFERENCE
INTRODUCTION
The previous three chapters have identified the major adverse
effects of urbanization on hydrology, water quality, and open
space.  These consist of increased flooding, stream channel
enlargement, deterioration of water quality, and reduction
in the amount and continuity of open space.  The objective of
land-use planning for environmental protection is to minimize
these adverse effects.

To determine whether a given plan is justified it is neces-
sary to know not only the physical effects of the plan, but
also the value of these effects to those who would benefit.
Many of the benefits of the environmental effects do not have
a market value.  Instead, they consist of the personal enjoy-
ment of the preserved environment received by individuals.

In order to determine the valuation which people place on
environmental preferences, we must first consider some basic
questions involving environmental preference.  First, it is
necessary to know whether persons tend to distinguish between
environments and to form consistent preference orderings
among environments.  The second question has to do with
whether different individuals tend to agree in their environ-
mental preferences.  In certain cases, it would appear
obvious that people would have the same preferences for
given environments:  for example, one would expect persons
to agree that an unpolluted stream is preferable to a severe-
ly polluted stream, all else being equal.  However, it is
not clear that substantial agreement can be expected in most
environmental comparisons, since these typically do not
involve extreme situations.

Third, given that there is substantial agreement among obser-
vers on environmental preferences, it is necessary to know
the characteristics of an environment which elicit prefer-
ences.  If we can identify the characteristics which determine
preference, we can gain some understanding of how much a plan
is going to effect things which are important to people's
preferences.

                             103

-------
Since there is very little literature dealing with environ-
mental preferences (see, for example, Craik [1968; 1969],
Jacobs and Way [1968], Lansing and Marcus [1969], Michelson
[1966], Peterson and Neumann [1969], Shafer, Hamilton, and
Schmidt [1969]), it was necessary for us to undertake some
relatively basic research in order to deal with the questions
listed above.  In the initial stages of research the focus
was on the natural environment in general; in the later
stages, it was possible to focus on stream sites and then
on water pollution.

These investigations described in this chapter were in the
form of pilot studies focused on differences between envir-
onments rather than on differences between judges.  Therefore,
no attempt was made to obtain observers who were fully
representative of the entire population.  In fact, in some
of the investigations the observers were chosen to be
representative of rather specific groups:  of upper-middle-
class housewives in one instance and of lower-middle-class
housewives in another.  Therefore, the results of these
studies should not be ascribed to the population at large.

Preferences of middle-class housewives, at any rate, may be
more relevant than preferences of the entire population to
the evaluation of stream valley preservation plans.  This
point is discussed at greater length below, where it is
noted that the persons who will receive the primary benefits
of stream valley preservation are those who will move into
the portions of the valley where development is allowed.
These persons, given the current characteristics of the
housing market, are likely to come from the middle class
rather than, say, from the lower class.
AGREEMENT AMONG OBSERVERS ON ENVIRONMENTAL ATTRACTIVENESS
Before one can speak analytically about scenic benefits of
conservation projects, one must have confidence that the
public will show a substantial level of agreement on the
attractiveness of a given scene.  A corresponding question
is whether the public will identify some scenes as substan-
tially more attractive than others.  In order to explore
these fundamental questions, five pilot tests were conducted,
                             104

-------
o
(Jl
                                             Table 19


                                  DESCRIPTIONS  OF  THE  TESTS
                                                                              ENVIRONMENTAL DISPLAY


. TEST


1. Environmental
Abs trac t Ion
Test


2. I ocaclonal
Character:
Field Test





3. 1. ocaclonal
Character:
Photo Test







4. Use and
Attractive-
ness Test




5. Consistency
Test




MAJOR OBJECTIVES
1. To determine whether judges can
abstract the natural elements
from a scene composed of both
natural and man-made elements
2. To establish whether judges
need specific instructions as
to the definition of the na-
tural environment
1. To determine extent to which
field observers agree in
rating locations for attrac-
tiveness

1. To gather evidence on consis-
tency of field ratings and
photo ratings

2 . To determine whether a group
of photographs can represent
one location, and If one of
those photos represents its
group





1. To determine whether Judges
rate locations differently
for residential use and for
scenic enjoyment


1. To ascertain if responses
change significantly after
in-depth analysis of the
slides by the judges
2. To determine whether Judges
are consistent In their re-
sponses after a time-lapse

NUMBER NUMBER METHOD OF
OF PRE- OF PRESENTA-
PROCEDURE JUDGES TEST UNITS TYPE RANGE TION
A. Rate attractiveness of ^
each scene 10 yes 20 slides rural- hand-
slides urban viewer
B. Rate attractiveness of
the natural environment
In each scene
C. As B, but with more de-
tailed and specific
instructions
Rate the attractiveness
of the natural environ- 2 no 92 actual rural- field
raent in the field loca- envi- urban trip
tions ron-
ments

11 no 337 slides rural- pro-
slides urban jec-
tion
Rate the attractiveness
of the natural environ-
ment as seen In each
slide

A. (Assuming that each en-
vironment has the same 10 yes t,g slides rural pro-
services available to It) slides Jec-
Rate each slide as to tlon
1) desirability for li-
ving 2) desirability for
driving through (or by)
B. Rate desirability for li-
ving. On the first 25,
give short explanation of
response
A. Rate attractiveness of
each slide 11 yes 14 slides rural hand-
„ „ . , slides viewer
B. Rate attractiveness of
each slide after verbal
analysis 11
C. Rate attractiveness of
each slide after wait
of one month 7


RESPONSES


ratings



ratings
and short
written
comments


ratings







ratings









ratings
(and oral
comments
between
Phases A
and B)

          35 mm color transparancles

-------
Research Procedure
The objectives, procedures and other characteristics of each
of the five tests are outlined in Table 19.  Basically, the
tests consisted of asking a number of judges to rate the
attractiveness of a number of environments on a scale ranging
from 1 to 7.  In all but one test, which was conducted in the
field, the environments were depicted by color photographs.
With the exception of the first test, which included some
midwestern scenes, all of the environmental displays consisted
solely of scenes depicting the watersheds of the Upper East
Branch of Brandywine Creek in Chester County, and the North
Branch of Neshaminy Creek in Bucks County, Pennsylvania.

The major findings of the tests are summarized below.  A
full report on the tests has been published elsewhere [Coughlin
and Goldstein, 1970].

Findings Concerning Agreement
Differentiation among environments on the basis of preference
or attractiveness rating requires (1) that judges tend to
agree in the ratings they assign to a given slide, and (2)
that mean ratings differ significantly among slides.  A
rough measure of the first property is the standard deviation
of ratings by individual judges on a given slide.  The
smaller the standard deviation, the higher the level of
agreement on a given environment.  The second property is
indicated by the distribution of mean ratings for each
slide.  The standard deviation of mean ratings provides one
measure of this distribution:  the larger the standard
deviation, the more "spread out" or differentiated are the
environments with regard to their perceived attractiveness.

As can be seen in Figure 9, various combinations of agreement
and differentiation are possible.  This diagram, of course,
is equivalent to the concept of an analysis of variance,
which deals with "within-slide" variance  (difference in
ratings by different judges on a given slide) and "between-
slide" variance (difference in mean ratings).

Analyses of variance of the ratings given in the five tests
indicate that judges are able to discriminate among environ-
ments on the basis of their attractiveness and tend to have
a relatively high level of agreement on a given environment.
In seven out of ten test situations, differences in the

                             106

-------
                          Figure  9

        THE  CONCEPTS OF AGREEMENT AND DIFFERENTIATION
                           Differentiation Among Environments
                              on Basis  of Attractiveness
                            Low  Standard  De-
                            viation  of Mean
                             Attractiveness
                                 Ratings
                                High Standard
                                Deviation of
                                Mean Attrac-
                               tiveness Ratings
           CO
           CO
           0)
           C -«J
            0)
           •H e
           *j c
           o o
           (0 J-t
           fc •«-<
           -P >
           *J P!
           < W

           C C
           O QJ
           C O
           0)
           Q) «
           n
           60
•D 0)
^4 >
cd o
T3
C G w
«0 O a)
W -H M
CO 4J -XJ
. « 3
  tu
  o
C C M
CO O (!)
4J «rl W)
W 4J 'D
  cd 3
A 1-1 *«n
60 >
i-l 0)
S Q
high agreement
low differenti-
ation

low agreement
low differenti-
ation

high agreement
high differenti-
ation

low agreement
high differentia-
tion

Note:  Each  distribution refers to the  distribution of attrac-
       tiveness ratings by a number of  judges on a given
       environment.
                                107

-------
                                                       Table 20
                  PERCENT  OF  TOTAL VARIANCE IN ATTRACTIVENESS  ACCOUNTED  FOR BY  PHOTOGRAPHIC  SLIDES
                                      (Results of One-Way  Analyses  of  Variance)
o
oo
                                           Number of:
                                       Slides
                                     (Treatment
                                       Groups)
1.  Environmental Abstraction
   Test
    A.  Entire Environment         20
    B.  Natural Environment
       only                       20
    C.  Natural Environment
       given specific de-
       finition                   20
2.  Locational Character:
    Field Test1
3.  Locational Character:
    Photo Test                   337
4.  Use & Attractiveness Test
    A-l Desirability for
        Residence                 48
    A-2 Desirability for
        sightseeing               48
    B   Desirability for
        Residence                 50

5.  Consistency Test
    A. Before commentary          14
    B. After commentary           14
    C. One month later            14
 An analysis of variance  was  not
Judges



 10

 10


 10
                                                     11


                                                     10

                                                     10

                                                     10

                                                     11
                                                     11
                                                      7
                                                        Percent Variance
                                                         accounted for
                                                           by slides
                                                                   51.2%

                                                                   69.5%


                                                                   69.2%
               53.6%


               35.9%

               35.3%

               35.2%

               51.3%
               51.1%
               53.3%
Significance level for
rejection of hypothesis
of no difference in
    mean ratings	
     p < 0.001

     p < 0.001


     p < 0.001
     p < 0.001


     p < 0.001

     p < 0.001

     p < 0.001

     p < 0.001
     p < 0.001
     p < 0.001
                                            performed  on the  field  test  because  there were only 2 judges.
       However other measures of agreement were found  between the judges,  and  are reported in the section
       on this test.

-------
slides themselves accounted for over 50% of the variance
(Table 20) as determined by analysis of variance tests.   The
remainder of the variance is, by definition, accounted for
by inter-judge differences in ratings on individual slides
and by unexplained random variation.

The conclusion that there is a substantial level of agreement
among judges on environmental attractiveness is impressive,
considering the fact that the slides were relatively similar,
consisting of typical scenes from suburban and rural areas
near Philadelphia.  No extremes of beauty or ugliness were
included, at least as these would be judged by the researchers

Consistency of Ratings
A preference rating is a subjective measurement and therefore
cannot be replicated by asking any other observer to view
the landscape under the same conditions and with the same
instruments used by the initial observer.  However, if
preference ratings have any meaning it is necessary that
there be no significant difference between the ratings given
to a particular landscape by an individual on repeated obser-
vations.  Similarly, mean ratings given by a group  of judges
should not vary significantly on repeated observations.   It
was found in Test 5 (Consistency Test) that mean ratings
were not affected significantly by either an intervening in-
depth analysis or a time lapse of one month.  However, after
the in-depth analysis, the judges tended to disagree with
each other somewhat more than at the time of their first
observation.

In summary, analysis of the five preliminary tests indicates
that enough agreement among judges can be expected on the
environmental attractiveness of landscapes for the use of
average ratings to make sense.
LANDSCAPE CHARACTERISTICS RELEVANT TO PREFERENCES
Given evidence of significant levels of agreement among
observers as to the attractiveness of different landscapes,
it is possible to proceed to the second stage of the analysis
the identification of specific objective characteristics of
those landscapes which could be shown to have significant
relationship to preference ratings.  This type of identifi-


                             109

-------
cation is highly important in any attempt to measure the
important non-fiscal benefits of a proposed conservation
project [Stevens and Coughlin, 1966].

Research Procedure
The data used in this stage of the analysis consisted of the
337 slides used in Test 3  (Locational Character--Photo Test),
brief written comments by observers on the 14 of those slides
which were used in Test 5  (Consistency Test), and other data
obtained in the field as part of a much more extensive
project (Field Test—Stream Sites).  The Field Test involved
15 stream sites in the Philadelphia area, ranging in character
from rural to completely urbanized.  Data on water quality,
channel enlargement, and other characteristics of streams
which are related to urbanization of the watershed area had
already been collected for most of these sites [Kawashima,
Hammer, and Coughlin, 1970]; further information was gathered
on 183 other variables, primarily measures of visual charac-
teristics such as size and shape of the stream bed, roughness
of surrounding topography, area covered by trees, grass, or
other ground cover, and elements of visual pollution such as
junk or discoloration of the water.  Ten observers were then
taken to these sites, and asked for brief written descriptions,
as well as ratings of the sites on a 1-5 basis for 29 differ-
ent preference scales.  These observers were selected for
their similarity in background and education, rather than
for their diversity or "representativeness"; the rationale
being that in this exploratory study the researchers wished
to focus upon differences between sites rather than differ-
ences between observers.

Several different methods of analysis were employed.   For the
337 slides of Test 3, extreme high and extreme low-scoring
groups in terms of preference were selected, dominant charac-
teristics of slides in each group were identified by inspec-
tion, and the extreme groups were then compared with each
other and with the "average" group.

For the 15 photographs used in Test 5, the method was basi-
cally the same, but the analysis was much more thorough.   For
each photograph, a 17 x 23 descriptive matrix was completed,
dealing with such aesthetic characteristics as Variety,  Color,
Brightness,Spaciousness, and Naturalness, applied to individual
elements of the scene as well as to the whole.  All of these
                             110

-------
characteristics were measured on a 1-5 scale.   Three RSRI
employees participated, each completing a matrix for each
of the photographs; the results were then averaged for
each photograph, and analyzed to determine similarities
and differences between the five highest-scoring,  the five
lowest-scoring, and the "average" photographs.

For the Field Test data, a matrix of simple correlations
(241 x 241) was produced, containing the 183 "objective"
variables as well as means and standard deviations of the
29 preference scales for each stream.

Finally, the brief comments of the observers on the 14
photographs of Test 5 (Consistency Test), and  the  longer
"objective descriptions" of 15 sites,  which were provided
by the 10 observers in the Field Test, were analyzed in
terms of the frequency of occurrence of certain descriptive
elements, and the evaluative connotations of these elements.

Findings Concerning Landscape in General
In general, the results of all stages  of this  analysis suggest
that there does exist a reasonabley strong, consistent pattern
of preference, with high agreement among judges as to the
essentials of a "good" landscape.  The findings for the photo-
graphic tests agree substantially with those for the Field
Test.  A brief summary of major conclusions follows.

1.  There are some landscapes which everyone agrees are good.
       a.  Preference and Agreement -  For all  tests, the
       correlations between mean preference scores and
       standard deviations, the latter of which measure the
       amount of disagreement on preference scores for any
       given landscape, are negative.   This suggests that
       people tend to agree more about what they like than
       about what they dislike.  These correlations are in
       no case significant, however; the highest is -.488
       for Field Test Question 3, "How would you like to use
       this area for recreation?"  (With this  number of obser-
       vations, a correlation would have to reach  .52 to be
       significant at the .05 level of probability.
       b.  Preference and Objective Content -  When the 337
       photographs from Test 3 (Locational Character - Photo
       Test) were analyzed as to objective content and
                             111

-------
       dominant characteristics, the most highly preferred
       107o of photographs were found to be strikingly
       similar.  This similarity was even more pronounced
       for the sample of 14 photographs which were analyzed
       extensively as part of Test 5 (Consistency Test).
       Furthermore, 3 of the 6 highest-scoring Field Test
       sites contained the same general type of landscape
       which predominated among the highest-scoring photo-
       graphs.  For all three tests, the lowest-scoring
       landscapes were less similar, and those scoring in
       the middle range least similar of all.

2.  These generally preferred landscapes have a strong tendency
to be "parklike" or obviously man-influenced.  Mowed grass and
scattered large shade trees seem to be the determining factors.
Judges may say, "This is nice because it looks natural, away
from civilization."  However, the scenes to which they are
referring are not in a wild or natural state but are clearly
"landscaped."

3.  This preference for parklike landscapes is partially ex-
plained by the fact that, when the instructions on a test do
not differentiate between preferences for "use" and for "ab-
stract attractiveness" of sites, judges, in rating those sites,
seem to think primarily in terms of their usefulness for
recreation.  "Recreation," moreover, seems to be defined
primarily as "picnics," and only secondarily as "hiking" or
"active sports."

The participating judges were able to discriminate between
sites on the basis of active use to a greater extent than on
the basis of scenic value.  Thus, in the Field Test, there is
considerably less variation among stream sites in the answers
to the question, "How would you rate this site as a place to
pass through and enjoy the scenery?" than in answers to ques-
tions concerning use of sites for living and recreation.

4.  There are significant "minority preferences," relating
primarily to extreme spaciousness, extreme seclusion, and
extreme naturalness.  These characteristics inspire strong
emotional reactions among a few of the judges, and landscapes
containing them tend to have disproportionately high standard
deviations.  This finding holds true over all aspects of the
tests.
                             112

-------
5.  The characteristics common to those landscapes with high
mean preference ratings and low standard deviations are largely
synthetic, having to do with patterns and arrangements rather
than with individual elements.  Dislike, on the other hand,
seems to focus on individual elements, and primarily on man-
made ones.  Certain types of man-made objects or conditions
provoke an almost universal, strongly negative reaction:  the
tendency for low-scoring stream sites to show high disagree-
ment among preference ratings is contradicted in the case of
sites which show excessively polluted water, excessive amounts
of trash, or excessively loud automobile noise.  For such sites,
ratings are uniformly low.

6.  Elements mentioned as important by judges on high- and low-
scoring photographs and stream sites tend to be the same as
those which objective analysis shows to be strikingly different
between the two groups.  The preference for "parklike" land-
scapes is seldom explicitly verbalized,but in other respects,
the things the judges say they like or dislike are, as far as
we can tell, much the same as those to which they are reacting.
The exception is the high proportion of the Field Test comments
which were devoted to non-visual stimuli such as stream noise
and grass smells:  almost no significant correlations were
found between preference ratings and pleasant non-visual
characteristics.

Findings Concerning Stream Sites
The details of the Field Test, which was concerned with the
evaluation in the field of a sample of stream sites, is of
particular interest.  Therefore, we will present its results
in some detail.

The major significant correlations between the variables ex-
pressing objective characteristics of the streams and their
settings and the 29 rating scales which express the prefer-
ences of the field observers are summarized in Table 21.
Scales 1 through 12 and scale 17 represent answers to ques-
tions of the form "How much would you like to use this area
for	?"  Responses were on a 5-point scale ranging
from dislike very much (1) to like very much (5).  The
semantic differential scales (13 through 16 and 18 through 29)
involve ratings, on a 1-5 scale, of the site on a continuum
defined in terms of two opposite concepts.  For example, in
                             113

-------
                                          Table  21
SUMMARY OF MAJOR  SIGNIFICANT  CORRELATIONS  BETWEEN  OBJECTIVE
       CHARACTERISTICS  OF  STREAM  SITES  AND JUDGES'   RATINGS  -
                                        FIELD  TEST
                                        Variables Expressing  Objective  Characteristics:
        Mean  rating

         by judges for:
         'l  Ltvln*                     •+ +  *++      & '

         t. SUS. ,hrou«h               . 2 .  a, .    ? * ^ t " 1 ®
         1. IxraatUa                  • • •  « •    + • « O  Q
           uiiiktitf.                    •«+**••    a
           Hcolrkln»                  €>*«*» +   +  e
           CaiplQf                    « + +  +   + *    0
           naylot kail                 '«««*«»*
           XaaUtatUl                        •» »      - -
        U t»a«-I»s                    • «   * * +     - 9

        lajantU OlffacantUl iealai
O  - -     *  +
e - - ©     *  +
e  e    * *  «
   .0   - + +
        11. Tlaltlat atona                »            O Q O  8 -  «
        12. TUltlnz wLch ona ethar                 *      +GOQ  ....
        13. Onattraetlva - Attractlva            •<•«•*      « 0 £ C - £ - -
        U. lo«lou> - Haalthy              0   A       * O « O « O -  6
        li. (arran . lu«h                 '   f
        U. la(lacta« - Uall-kapt            *   00*4   « - C S  S •
        U. Vllltlat vltti «aciy othati          « «? ff 9 « 9 »
        U. Confute - SpaciMil             0 0 • * * * *
        It. Claaaa- • Opan                 0 + C+ + e$+  9

        20. Dull . Mctviraaqvia                +  •         - -
        21. taw mlatr - Htlh «rlaty           0  +                  0       +    »
        It. Nonocooa of color - Klth contrite
           •f color                    4-0        CO-     00    0*0    -  »  • » *

        U. Satatfarad with - Natural                      •OOGG-                          0.
        24. Xaceogruoua coatolnatton .
           tKpactad coad>lnatlon>                        » B   O O   •  •                       C •

        2S. Saparacad pattern - Hlxad pattarn                                        .                  + •
        2t. ardarad pattam « lUndoat pactarn     -OQ.-a.                                    009
        27. Sivpla pattarn - Cooplax pattarn     -Q..+                   .                a^4.+
        21. fattara cypa                          8      »           +                        +0
        2*. bpotad - Sacluila4                       ..+*•+      0       -                00

          . * ac . IxlUatai cocralatlon altnlClcant at .0} laval; 8 or Q tndleataa corralatlon al|nlllcanc at .01 laval.
                                               114

-------
scale 14, a stream site which was judged noxious would receive
a rating of (1), and a stream site which was judged healthy
would receive a rating of (5).  A "+" or a "-" in any column
indicates a positive or negative correlation of .50 or higher
(significant at the .05 level); a 'V or a "9" indicates a
correlation of .62 or higher (significant at the .01 level).

Rows 11-29 in this table, containing three preference and 14
bipolar "semantic differential" scales, are given in five
clusters, each of which shows extremely high correlation among
its component scales.   These clusters were derived from a
linkage analysis of the inter-correlations among rating scales;
they may be regarded as factors or dimensions relevant to
preference.  The three rating scales for "Visiting alone,"
"Visiting with one other," and "Visiting with many others"
were included in these clusters because of their strikingly
close relationship to specific semantic differential scales.

These clusters, although distinct, are not in the least inde-
pendent of each other.  Nearly every rating scale in the first
four clusters is positively and significantly correlated at
the .05 level with most of the other rating scales.  The fifth
cluster, composed of scales which seem to describe a wooded,
secluded stream valley, is the only one which is not signifi-
cantly related either to other semantic differential scales
or to the basic preference scales  (the first three scales
in the table).

Inspection of the table shows a few objective characteristics
with consistently significant relationships to preference.
They are:
     Positive: percent of area covered by mowed grass, percent
of foreground visible, "open" banks, percent of foreground
area that was firm ground, and, surprisingly, several variables
denoting number and size of houses.  These last correlations
may have been due to a tendency for preferred sites to be in
"high-status" rural areas, where open lawns and meadows pro-
vided views of the large nouses of the owners.
     Negative: all variables denoting presence of junk, percent
of area covered by twigs, area of spongy ground, number of
medium-sized rocks in and around streams, and height of berms
(secondary banks).  The last correlation is probably due to
the extremely high berms of the two sites having large amounts
                             115

-------
of junk and trash.  Unpleasant noise, though mentioned fre-
quently  in the comments, is important only for "passing
through to enjoy the scenery," not for "living" or "recre-
ation. "

A striking feature of these important variables, especially
those with positive correlations to preference, is that they
seem to involve primarily characteristics which have been
created or influenced by man, rather than "natural" features
of the site.  The preference for open, parklike areas, noted
in the previous section, is quite evident here, and there is
some indication that neighborhood status is also a factor.
Some of the less significant negative correlations are related
to relative degree of urbanization of the watershed, but it is
difficult to tell if this is a reflection of the noxious in-
fluence of urbanization on the stream itself, or simply of
the low status of the surrounding urban neighborhood.

Among the 183 variables tested there were several variables
relating to characteristics of the stream whose relationships
to urbanization have been studied in other parts of this
project.  These were channel enlargement and various aspects
of water pollution.  None of these variables had an appreciable
number of significant correlations with either preference
scales or the semantic differential scales.  Thus, these
variables are not included in Table 21.  Indicative of the
relative unimportance of these variables is the fact that
the site most preferred of all was one in which the stream,
which drained a highly urbanizaed area, had generally poor
water quality and a greatly enlarged channel.  (Its channel
enlargement ratio exceeded 3.0.)  The stream, however, ran
through a well-kept park with large trees and grassy expanses.

One finding, which is not evident from inspection of Table 21,
involves the dimensions of the area which is relevent to
preference for a site.  Since many characteristics were scored
separately for banks, foreground, mid-ground, and far-ground,
the highest correlation for a particular characteristic would
indicate the most important location for that characteristic.,
For most variables, these important locations were foreground
and mid-ground:  in other words, their influence on prefer-
ence does not extend beyond a distance of about 200 feet.
For the major positive and negative factors, namely grass,
large junk and twigs, only the banks and the foreground area  ;
                             116

-------
showed significant correlations.  Comments on the sites,
however, indicate that the "atmosphere"  of an area, in-
cluding all of the visible and audible surroundings, is
also an important factor.

Patterns of Preference
Some other very interesting patterns are evident in Table
21.  Some of the concepts involved in preference, for in-
stance, seem to be basically negative:  The absence of
certain detracting elements is what is important.  Prefer-
ence ratings for "Passing through to enjoy the scenery" and
the first four scales of the first semantic cluster are the
prime examples.  The large majority of significant correla-
tions between these scales and objective variables are
negative correlations with variables which express unpleasant
characteristics.  Relatively few positive correlations exist
with variables expressing positive characteristics.  This
indicates that the primary difference between a good site and
a bad site for a typical observer is simply the presence of
negative characteristics in the latter, rather than the re-
quirement that a "good" site possess particular positive
features.  For "use" (recreational preferences) questions, on
the other hand, and for all scales in the second cluster, both
positive and negative characteristics are extremely important.
This could be interpreted to mean that judges tend to be more
demanding when rating the "use" aspects of a site; that is,
judges require that certain positive features be present for
a site to be good, in addition to requiring the absence of
negative characteristics.

One possible explanation for this pattern is that persons are
in a sense starved for naturalness, or the illusion of natur-
alness, so that nearly any stream site will be considered
attractive   as long as it lacks certain obviously detracting
man-made characteristics.  Another possible explanation could
be that this group of judges has more uniform and well-arti-
culated preference standards for use of sites than for "attrac-
tiveness" in the abstract.

The preceding paragraphs have related judges' ratings to
objective characteristics of site.  A different approach to
the analysis of preference patterns is to relate the judges'
                             117

-------
preference ratings to their own judgments concerning the
basic aspects of the site, as expressed in the semantic dif-
ferential scales.  Preference ratings of the sites as places
for "living," places to "pass through" and to "enjoy the
scenery," and as places to visit for "recreation" were most
highly correlated with semantic differential scale ratings
for "Attractive" and "Healthy"; preference ratings for
"Living" and "Recreation" were also highly correlated with
ratings for "Spacious," while preference ratings for "Passing
through" were highly correlated with ratings for "Picturesque."

Although several psychological studies have demonstrated that
both animals and humans of all ages show a preference for a
complex over a simple stimulus pattern,and though our analysis
of photographs has indicated that high variety is associated
with high preference, the judges in the field experiment
consistently preferred sites which they saw as "ordered" and
"simple" in pattern.  The correlations are seldom significant--
the highest is .52 for the correlation between "ordered pattern"
and preference rating of sites for "Living" and "Recreation"--
but the signs are positive for correlations between all pre-
ference variables and "ordered," and negative for correlations
between nearly all preference variables and "complex pattern"
or "pattern type."  This does not mean that the preferences
of these field judges were different from those of the judges
of photographs; judgments of "high variety" are still signifi-
cantly associated with high preference ratings.  Rather, it
would seem to indicate that these judges—as well as the judges
in the earlier test using photographs--have difficulty appre-
ciating variety unless it is in some way logically or aesthet-
ically ordered.
QUANTITATIVE MEASUREMENT OF LANDSCAPE QUALITY

Research Procedure
Based on the above findings, objective measures of "landscape
quality" were defined.  These measurements were applied to
photographs of the Field Test sites, the results averaged
over all photographs depicting a given site, and the result-
ing variables were entered in regression analysis with normal-
ized mean preference ratings as the dependent variables.   A
full report from which the following summary was made can be
found in Rabinowitz and Coughlin [1971].
                             118

-------
The landscape quality variables are listed below.

1.  Variety of Elements.  Hypothesis:  a high variety of
elements in a photograph will be associated with high pre-
ference.  Measure:  number of different element categories
in each photograph, element categories being defined in terms
as simple and intuitively obvious as possible:  trees, bushes,
streams, mowed grass, high grass.

2.  Variety of Brightness.  Hypothesis:  high variety of
brightness will be associated with high preference.  Measure:
difference between the highest and lowest readings obtained
by moving a photographic light meter back and forth over the
projection on a viewing screen of each photograph.

3.  Size of Largest Trees.  Hypothesis:  a large tree domin-
ating the photograph will be associated with high preference.
Measure:  total area covered by largest tree in photograph.
The area in the photograph with the highest measurement is
used as the measurement of a group of photographs depicting
a site.

4.  Length of Longest View.  Hypothesis:  an unusually long
view will be associated with high preference.  Measure:  log-
arithm of estimated distance of longest view, in feet.  The
measurement in the photograph with the longest measurement
is taken as the measurement for a group of photographs depict-
ing a site.  The log scale emphasizes differences between
relatively short distances, and correspondingly deemphasizes
equal differences between longer distances.

5.  Spaciousness-Seclusion.  Hypothesis:  observers prefer
neither too much open space nor too little; a low score on
spaciousness-seclusion will reflect the moderate degree of
openness desired by subjects, and will, therefore, be assoc-
iated with high preference.  Measure:  the absolute value
of the difference between the figure for variable 7 (Actual
Open Space) for a given site, and the mean of all measurements
for variable 7 over all slides.
6.  Misfits.  Hypothesis:  presence of misfits will be assoc-
iated with low preference.  Measure:  total area of photograph
covered by misfits.  A "misfit" was defined as any man-made
object neither insignificant nor obviously attractive.
                             119

-------
7.  Actual Open Space.  Hypothesis:  a large amount of open
space will be associated with high preference.  Measure:
area of "open space" in photograph times measurement for X4
(Length of Longest View).  This measurement provides an
index of the actual area of open space at the pictured site.

8.  Area of Water in Photograph.  Hypothesis:  a large area
of water in a photograph will be associated with high prefer-
ence.  Measure:  area of water in each photograph, measured
for all photographs for any site which showed any water at
all, and then averaged.  In order to approximate the actual
amount of water present at the site, those slides which were
taken looking away from the streams were eliminated from the
calculations.

9.  Area of Frequent Flooding.  Hypothesis:  a large area of
frequent flooding will be associated with low preference.
Measure:  total area of sand or pebble bars or denuded ground
within the stream banks in photograph.

10.  Number of Visible Rocks.  Hypothesis:  a large number
of rocks will be associated with low preference.  Measure:
number of rocks greater than one foot in diameter.

Several other "quality" characteristics were also measured
for use in these equations, for instance: "Variety of Color"
and "Presence of Water."  The fact that all the highly pre-
ferred photographs in the Consistency Test showed open
expanses of lawn with scattered shade trees suggested that
total area of open space (defined  as anything that can be
walked on or through without difficulty), area of mowed grass,
or arrangement of open space (unbroken space vs. space
interrupted by bushes or trees) might also be important.
But none of these measurements turned out to be significantly
associated with preference ratings.

The 10 variables described above were measured for color slides
(as viewed on the screen of a projector-viewer) of the stream
sites which were the subject of On-Site Observations.  Most
of these slides were also used for a set of Home Interviews,
in which 18 middle-class suburban housewives and mothers
participated.  Their social and economic characteristics
were remarkably similar to those of the subjects in the On-
Site Observations.
                             120

-------
Findings
Table 22 gives the equations which were derived, using site
ratings as the dependent variables and site characteristics
measured from slides as the independent variables, for the
On-Site Observations and the Home Interviews.

In the equations for the On-Site Observations, as in the
equations for the Home Interviews, the independent variables
were derived from the slides which were used in the Home
Interviews--in other words, the two slides which together
provided the best representation of each of the actual slides.

The set of equations for the On-Site Observations (Table 22)
is quite interesting.  We obviously cannot claim that the
characteristics of the slides, which were chosen many months
after the On-Site Observations had taken place, had any causal
relationship to the ratings which subjects assigned to the
sites themselves.  Nevertheless, these equations, which were
initially derived merely to provide a basis for comparison
with the equations for the Home Interviews, show highly
significant correlations between the ratings in the field
and many of the slide characteristics.

It should be kept in mind that these multiple regression
analyses all involve extremely low numbers of degrees of
freedom.  Although this fact is taken into account in the
tests of significance, nevertheless, having this few degrees
of freedom is highly undesirable.  The reason for this is
that the estimates are highly sensitive to values of variables
for individual observations, so that a large error component
in a single observation can have a major effect on the re-
gression results.  Thus, the results should be viewed with
extreme caution.   They are published in spite of these
limitations, since the study is admittedly an exploratory
one and because the levels of significance are unusually
high.

Three variables appear as consistently significant in the
equations for the On-Site Observations: X2, Variety of
Brightness, X3, Size of Largest Tree, and Xg, Area of Water.
(The negative sign on the correlations with X2, Variety of
Brightness, unfortunately runs counter to our earlier results,
and to our hypothesis that high variety will be preferred.)
                             121

-------
                                 REGRESSION
ro
                                                       Table 22

                               EQUATIONS EXPLAINING PREFERENCE BY MEASUREMENTS  OF LANDSCAPE QUALITY


                                           Dependent Variables (Ratings  by Observers)
Independent variables
  (measurements on
   photographs)	

X.  Variety of Elements

Xj  Variety of Brightness

Xj  Size of Largest Tree

X^  Length of View

X.  Spaciousness-Seclusion

Xg  Misfits

X,  Area of Open Space times
      Length of View

X..  Area of Water

Xg  Area of Frequent Flooding

Xlfl Number of Rocks > 1ft.  in
      Diameter
             R

             F Ratio

             Significance Level
On-Site Observations


Question
A



u
c
Q}
Tl
U
1-1
M-l
H
U
o
u

-.820
.091

-7.583 2
-.336
.028

-.646
.257

b
0
li
p
Id

•o
M
(9
•a
C
a
u
C/l

.155
.018

.145
.030
.006

.050
.042
.9884
73.288
r < .001


Question
B
^
o
M
c u
u
•r< •D
O H

>H -D

41 a
O u
U M
-10.054 1.387
-3.247 .377
.259 .034

-28.456 2.726
-.475 .036
.024 .007
.222 .026
-.295 .049
.215 .052
.9906
47.328
p < .001
Question
C
b
o
H
U VJ
C H
H -0

-------
The equations which were derived for the Home Interviews
experiment (later columns of Table 22) are very similar to
those for the On-Site Observations, with an even higher
percentage of variance explained by the slide characteristics
In all three of these equations the multiple R^s are .97 or
higher, at an inordinately high level of significance.
Furthermore, all but one of the "photo quality" variables
are significantly related to preference ratings in at least
two of these equations; and only three of these relation-
ships are significant at a lower significance level than .02.
According to our original hypotheses, X£ ,  Variety of Bright-
ness and X4, Length of Longest View, would be expected to
show positive correlations with preference ratings.  However,
they appear in Table 22 as negatively related to preference.
This suggests that, if the Home Interviews method did indeed
emphasize the "real world" aspects of the sites rather than
the aesthetic qualities of the slides, then subjects in a
"real world" situation evidently prefer a fairly secluded
site (short length of view), but not one which is heavily
wooded (heavily wooded sites scored highest on X«).

For both the On-Site Observations and the Home Interviews,
the negative correlation coefficient for "Variety of Bright-
ness" is lowest in the equation for Question A--MHow would
you like to live here?"--an indication that the seclusion
provided by thick woods has some positive value in a resi-
dential situation.

The characteristics of the desired landscape are further
illuminated by the large, and highly significant, negative
correlation for X5, Spaciousness-Seclusion, which appears
in the majority of the equations in Table 22.  The high
significance of this correlation provides added support for
the hypothesis that the "generally preferred" type of site
conforms to a fairly precise degree of spaciousness, neither
too open nor too confining.

It will also be noted that in both sets of equations the
level of explanation for Question B, "How would you rate this
site as a place to pass through and enjoy the scenery?" is
higher than for either of the others, and that the Home
Interview ratings were most highly correlated with the On-
Site Observation ratings for this question.  It would seem
that the cultural stereotype of the "attractive landscape"
                             123

-------
for scenic enjoyment is more clearly defined, and more closely
agreed upon, than the conception of a landscape that is useful
or liveable.

Most of the other correlations which appear in Table 22 are as
expected.  There is a positive correlation for X3, Size of
Largest Tree, and there are small positive correlations for
Xy, Area of Open Space times Length of View, and Xg, Area of
Water.  Negative correlations for XQ, Area of Frequent Flood-
ing, in the equations for the Home Interviews, suggest that
some of the effects of urbanization on a stream may be per-
ceived as deleterious.  The large negative correlation for
X;L, Variety of Elements, in two of these equations (contrary
to the direction of correlation expected), suggests the
hypothesis that visual complexity high enough to result in
visual confusion would be disliked. However, it is doubtful
that any of the sites were complex enough to account for such
a response.

These equations cannot be interpreted as a "model" of prefer-
ence patterns, which could be used to predict preferences of
different subjects for different landscapes.  Our samples of
both subjects and landscapes have as yet been far too small.
But the general consistency of these measurements in explain-
ing these sets of preference ratings convinces us that we
have succeeded in uncovering certain very important aspects
of landscape quality.
THE INFLUENCE OF WATER QUALITY IN THE EVALUATION OF STREAM SITES
The previous sections have indicated that individuals do appear
to prefer some natural environments over others, and to be
reasonably consistent in their preferences.  A substantial
level of agreement among individuals has also been observed.
Preferences were found to be related to a number of environ-
mental characteristics.  Subjects tended to respond negatively
to obvious noxious human influences such as junk, but tended
to prefer parklike scenes to more natural landscapes.  Most
of the environments studied, either in the field or by photo-
graph, consisted of stream sites; however, the preferences
expressed for these sites do not appear to have been dominated
by reactions to the stream itself.  Therefore, a series of
studies was conducted which focused on how stream preferences
are influenced by water quality.
                             124

-------
Research Procedure
In the first of these studies, the Pollution Perception Study
[Scherer and Coughlin, 1971] , which is summarized in this
section, a number of subjects were taken to observe a dozen
stream sites.  The other studies, which are reported in the
next chapter, involved household surveys in which subjects
were asked questions about the stream nearest their residence,

Unlike the Field Test reported in the previous section, the
subjects in the Pollution Perception Study were asked a num-
ber of questions relating specifically to the condition of
the stream, the surrounding area, and the area as a whole.
Questions concerning pollution were mixed in with other envir-
onmental questions so that the respondents would not realize
that the main focus of the study was water pollution, and,
therefore, accord pollution excessive importance in their
answers in order, perhaps subconsciously, to please the
researchers.  The respondents' answers were analyzed in
relation to a wide variety of other objective measures of
stream condition.

The general procedure followed in the Pollution Perception
Study was as follows:  Twelve stream sites were selected
which were as similar as possible physically, but which
varied in water quality.  Twelve observers, lower-middle-
class housewives for the most part, were taken to these
sites over a period of two consecutive days,  Each filled
out a questionnaire on each site, with respect to the area
within about 100 feet from the observation spot.  While the
respondents filled out their questionnaires, two investi-
gators rated the stream on a number of attributes referring
to physical characteristics and water quality.  In addition,
water quality samples were taken for all the streams within
an hour of each stream site interview.

Laboratory Analysis was performed for the following chemicals:
Chemical Oxygen Demand (COD)
Total Coliforms
Fecal Strep
Total Phosphates
Ortho Phosphates
Organic Nitrogen
Total Dissolved Solids at 103*
Water Temperature
Acidity-Alkalinity (pH)
Organic Nitrogen
Ammonium Nitrates
Nitrites
Nitrates
Sulfates
Chlorides
Total Dissolved Solids
   at 179°C
Dissolved Oxygen
                             125

-------
Using the resulting data, the streams were grouped roughtly
by level of organic load, as shown in Table 23.  An overall
Water Quality Index was also calculated for each of the 12
streams using the methodology of Brown, McClelland, Deininger,
and Tozer (see Chapter VI, p. 171, for a description of this
method).  Since data were available on only seven out of the
nine variables used by Brown et al., the indices were computed
on the basis of the seven variables, which allow a possible
high score of 80, for best water quality and were then
expanded to a scale with a possible high score of 100.

The range of water quality as shown by the Water Quality Index
was 20.5 to 63.5 on a possible scale of 0-100.  Although the
range was rather wide, none of the streams could be described
as having unusually good water quality.  Also, dramatic
evidence of pollution, such as severe discoloration or
accumulations of foam or scum, was not present.

Subjects were asked to answer the following three basic ques-
tions on a scale of 1 (dislike) to 5 (like):  "How much do
you like the area within about 100 feet which you can see
from this spot?", "How much do you like the stream itself?",
and "How much do you like the surrounding area?"

Two other important sets of questions were asked:  The first
set concerned how much the respondent would like to engage
in each of a number of activities at the site.  The second
set concerned the subjects' perceptions of the site; these
asked the respondent to state the extent to which the site
could be described by terms such as "pleasant," "healthful,"
"inviting."  A number of the questions concerning perception
related specifically to the conditions of the water; that is,
is the stream rapidly flowing, transparent, clean, smelly,
pleasant, full of trash, inviting color, polluted, healthy?

Findings
The responses to questions involving perception of stream
conditions indicated that people are able to identify water
pollution reasonably well.  Table 24 gives the statistically
significant correlations between perceived water attributes
and various chemical and biological measurements of water
quality.  The greatest number of significant correlations
occur for the following four perceived attributes:  transparent,
                             126

-------
                  TABLE 23




STREAMS GROUPED ACCORDING TO LEVEL OF ORGANIC LOAD

Heavy
Organic
Load
Moderate
Organic
Load
Streams
6, 8, 12
Measures of
Organic Load
Total P (ppm)
Ortho P (ppra)
N02 (ppm)
N03 (ppm)
Measures of
Sewage-Related
Pollution
Total
Coliforms
fecal strep
TDS 103° (ppm)
IDS 179° (ppm)
W.Q.I.
15.20 -
19.40 -
.74 -
25.72 -
1270 -
100
434 -
342 -
20.5 -
23.60
28.00
1.76
33.33
4000
275
520
508
29.0
Streams
7, 9, 10
.47 -
.43 -
.16 -
11.00 -
4000 -
70 -
418 -
342 -
37.9 -
12.20
13.20
2.67
21.20
24000
95
567
522
46.6
Low
Organic
Load
Streams
3, 4, 11
.18 - 1.88
.09 - 2.08
.04 - .07
5.15 - 6.08
3000 - 9500
240 - 1645
118 - 274
75 - 208
48.3 - 59.5
Very Low
Organic
Load
Streams
1, 2, 5
.07 - .24
.00 - .06
.02 - .05
1.72 - 13.96
1000 - 1825
340 - 1070
75 - 123
43 - 88
48.0 - 63.5
                    127

-------
clean, polluted, and healthy.  Most of the significant corre-
lations involve the following chemical characteristics:  COD,
Fecal Strep, Total Phosphates, Ortho Phosphates, Nitrates,
Chloride, Total Dissolved Solids, and the Water Quality Index.
Most of the correlations show the expected signs.

However , four desirable attributes are positively correlated
with fecal streptococci; this apparently spurious result
occurred because the concentration of fecal strep was high
for rural streams where water quality was generally good.  The
negative correlations between various chemical measurements
and the "rapidly flowing" attribute are probably spurious also
since there is no logical relationship between water quality
and speed of flow.  This result could be due to particular
characteristics of the sample of streams, that is, the
polluted streams happened to be slow-flowing.  It is notable
that the highest correlation observed is the correlation of
-.72 between the perceived attribute "polluted" and the overall
Water Quality Index.

The perceived water attributes were also significantly corre-
lated with a number of objective, but non-chemical, character-
istics of the stream (see Table 25).  The highest correlation
involved the percent of the stream bed covered with mud.
These correlations reflect a dislike of mud on the part of
the subjects plus the fact that other stream characteristics
such as speed of the flow of the stream tend to be associated
positively or negatively with the existence of a muddy stream
bed.  Of the seven objective characteristics of the stream
which might be related to water quality--algae, rooted aquatics,
floating solids, scum, bubbles, clarity, and discoloration--
only discoloration was significantly correlated with a perceived
water attribute.  The lack of significant correlations with
these aspects of pollution is somewhat puzzling, particularly
in view of the substantially higher levels of correlation
between perceived attributes and the chemical measurements
which were performed with rigorous laboratory procedures.  It
may lie in the imprecision of the "objective" ratings.  At any
rate, Tables 24 and 25 provide some evidence that the respon-
dents were able to perceive levels of water quality.

The questions which are most relevant to the value of a stream
to an individual are the basic questions relating to how much
                             128

-------
                                      Table  24

                  CORRELATIONS BETWEEN CHEMICAL AND BIOLOGICAL
                   MEASUREMENTS AND PERCEIVED WATER ATTRIBUTES
Perceived
Water
Attributes
rapidly
flowing
transparent
clean
smelly
pleasant
full cf
trash
inviting
color
polluted
healthy
Biological and Chemical
fecal Total
COD Colifortns strep P
-.60 -.53 -.54
-.67 .56 -.6S
-.65 .54 -.61

.55


-.57 .52 -.55
.69 -.53 .59
-.63 -.57
Measurements
TJ o f* Qf
Ortho IDS TDS Water •>• f
P NH3 N02 N03 S04 Cl 103° 179° Temp. D.O. pH TnH,=v
-.51 -.56 -.74 .61
-.62 -.69 -.58 -.58 -.65 .60
-.56 -.62 -.55 -.59 -.65 .63



61 -.56
-.56
.56 .57 .67 .61 .64 .70 58 -64 -.72
-.51 -.54 -.51 -.53 -.60 .63
Underline indicates p is less than .01
N=12
other coefficients: p is less than .05

-------
     CORRI
                        Table  25

BETWEEN OBJECTIVE RATINGS  AND  PERCEIVED WATER  ATTRIBUTES
t-1
LO
O
Perceived
Water
Attributes
rapidly flowing
transparent
clear
smelly
pleasant
full of trash
inviting color
polluted
healthy
Objective Ratings
Rapidity
of flow Riffles
.73 .82
.62 .63
.57 .56
.59 .61
.60 .70
.71
Discolor- Trash on
Pools Sediment ation stream
-.57 -.53
-.51
-.56
.58
.79
-.54
.54
-.54
Trash on % %
banks Rocks Mud
.77 -.87
.52 -.82
-.78
.64
-.66
-.81
-.51 .73
.56 -.83
Rural-Urban
-.62
^69
.52
.71
-.61
single underline indicates  that p is less than  .01
double underline indicates  that p is less than  .001
N-12
                         all other listed coefficients: p is  less than .05
                         Correlations between all other combinations were not significant
                         (were less than  .05).  No significant correlations were  found
                         between perceived water attributes and algae, rooted aquatics,
                         floating solids, scum,  bubbles, clarity, percent bottom  covered with
                         sand, or percent bottom covered with leaves.

-------
the respondent likes the stream, the surrounding area, and
the site as a whole, and the questions pertaining to various
specific activities he would enjoy at the stream site.  The
nature of the responses given to the three basic preference
questions can be examined by considering the proportions of
the total variation in responses which are attributable to
various sources.  Table 26 lists for each of the three sets
of responses the percentages of variation attributable to:
differences in the mean responses of observers, differences
in the mean responses for streams, and residual variation.
The variation due to difference in observer means indicates
the extent to which the variation in responses is due simply
to a tendency for some observers generally to give higher
ratings than others (i.e., on a 1 to 5 scale).  It is inter-
esting that observers showed much greater differences in
average ratings of streams than in average ratings of sur-
rounding areas; this would indicate that some observers like
streams in general more than others.  On the other hand,
observers show greater agreement for streams than for the
surrounding area when rating sites relative to one another.
This is shown by the fact that a much greater proportion of
total variation is explained by difference between site means
for streams than for the surrounding areas.  (That is, persons
tend to agree as to which streams should receive relatively
high or low ratings; thus the mean ratings for sites differ
considerably.  For ratings of the surrounding areas, dis-
agreement among observers causes the high and low ratings
of different observers to cancel each other out, so that
the overall means for the various sites to not differ greatly.)
Equivalently, the amount of residual variation, which is a
measure of disagreement among observers, is greater for
ratings of surrounding areas than for streams.

Table 27 presents correlations between responses to the three
basic preference questions with the perceived attributes dis-
cussed earlier.  In addition, the responses to the question
"Would you like to come back here?" are also correlated with
the perceived attributes.  These correlations should indicate
the extent to which general liking is determined by the
perception of water characteristics.

Most of the possible correlations are statistically signifi-
cant.  The number of observations involved in these correla-
tions is 144 (12 observers x 12 streams) so that even small
                            131

-------
                                 TABLE 26

                   SOURCES OF TOTAL VARIANCE IN RESPONSES TO
                          BASIC PREFERENCE QUESTIONS

                   (Variance Among Observers and Stream Sites)
                                    OVERALL   RATING OF
                                    RATING    STREAM
                                    (Basic    (Basic
                                     question  question
                                      1)	2)
                          RATING OF
                          SURROUNDING
                          AREA
                          (Basic
                           question
                            3)	
PROPORTION OF TOTAL
VARIANCE DUE TO:

   Difference between
   observer means
   31.67.
49.9%
 22.9%
   Difference between
   site means
   30.6%
32.4%
 17.2%
   Residual
   37.8%

  100.0%
17.7%
                                               100.0%
 59.9%

100.0%
   Significance of variance
   due to difference between
   site means
   [ F .01 (11,121) = 2.40 ]
F = 8.90  F = 20.13
             F = 3.16
                                All significant at the 1% level
                                     132

-------
                    TABLE 27

       CORRELATIONS BETWEEN PREFERENCE RATINGS
          AND PERCEIVED WATER ATTRIBUTES
PERCEIVED
WATER
ATTRIBUTES
rapidly
flowing
transparent
clean
smelly
pleasant
full of
trash
inviting
color
polluted
healthy
PREFERENCE RATINGS
Overall
Rating
.25
.42
.45
--
.64

-.22
.44
-.22
.46
Rating for Rating for
stream surrounding area
.32
.48 .22
.52 .29
--
.67 .49

-.19
.59 .26
-.28
.54 ^38
Desire to
return
.25
.38
.40
.24
.59

--
.47
--
^38

N = 144
Single Underline indicates p is less than 0.01
Double Underline indicates p is less than 0.001
All other coefficients:   p is less than .05
                     133

-------
correlation coefficients are statistically significant.   A
number of correlations between basic preferences and water
attributes are quite high.  However, these tend to involve
those attributes which have been seen to be least closely
related to pollution.  The highest correlation in each of
the four columns is with the attribute "pleasant," which
was correlated significantly with only one chemical measure-
ment in Table 24.  (The latter was the apparently spurious
positive correlation with fecal streptococci.)  On the other
hand, correlations involving the perceived attribute "polluted"
were quite small.  The correlation between liking for stream
and "polluted" was only -.28.  The correlation between desire
to come back and "polluted" was not statistically significant.
On the other hand, higher correlations were observed for
several other attributes which might be associated with
pollution, namely:  transparent, clean, inviting color,  and
healthy.

The low association between pollution and basic preference
was borne out in direct correlations between the responses
to the basic preference questions and the chemical measures.
No significant correlations were found.

The data presented in Table 28 refer to the average prefer-
ence rating of each respondent for the three streams in each
of the four organic load groups shown in Table 23.  Table 28
shows the proportions of total variation for each of the
three basic preference questions which are due to difference
between observer means, difference between organic load groups,
and residual variation.  The variation due to difference be-
tween organic load groups, which indicates the extent of
agreement among judges as to relative preference among the
groups, is quite small for all three basic questions, although
it does comprise a statistically significant component for
question 2.  This provides additional evidence that, for this
group of observers and streams, water pollution was not an
important determinant of preference.

The results obtained for the basic preference question relating
explicitly to streams indicate that although observers may be
able to discern pollution reasonably well, they may not con-
sider it of great importance to the attractiveness of the
site or even of the stream itself.  Liking of the stream does
                             134

-------
                             Table 28

 BASIC PREFERENCE  QUESTIONS:   SOURCES  OF TOTAL VARIANCE
          IN MEAN RESPONSES FOR ORGANIC-LOAD  GROUPS
PROPORTION OF TOTAL
VARIANCE DUE TO:

   Difference between
   observer means
                                 OVERALL
                                  RATING
                                 (Basic
                                  question
                                    1)
   80. Q4
             RATING OF
               STREAM
             (Basic
              question
                2)
   73.04
               RATING OF
               SURROUNDING
                 AREA
               (Basic
                question
                  3)
   57.43
   Difference between
   organic-load groups
     .66
    6.22
    1.09
   Residual
   Significance of variance
   due  to difference  in organic-
   load group means
   [  F  .05 (3,33)  = 2.89 ]
   19.30
                                  100.00
F - 0.38
   20.74
               100.00
F = 3.30
   41.48
                 100.00
F = 0.29
                                         Question 2 significant at 5%
                                135

-------
appear to be related to perception of the stream, but seems
to involve characteristics such as "pleasantness," which may
not be directly related to pollution.  It is significant
that although pollution is not important to these observers,
they exhibit a high level of agreement on the relative attrac-
tiveness of streams, implying that some criteria are important
to them.

Another way of determining people's preference for given stream
sites is to ask what activities they would enjoy doing at the
site.  Tables 29 and 30, which relate answers to this question
to chemical and biological measurements of water quality,
separate the activities for which the site is deemed suitable
into water-related and not-water-related activities.  Water-
related activities consist of wading- playing in the water,
and fishing.  Not-water-related activities consist of relaxing-
meditating,  enjoying the scenery, and picnicking.  In Table
29, no significant correlation was observed for not-water-
related activities; the only significant correlations were
for water-related activities.  All the correlations have the
expected sign—more pollution being associated with less desire
to undertake an activity at the site—except for the correla-
tion with fecal streptococci, which appears to be spurious as
before.  Correlations significant at the 0.05 level occur with
total phosphates and nitrates, with sulfates, with dissolved
solids (both at 103°C and 179°C), and with the Water Quality
Index.  This pattern is quite similar to that found in Table
24, which shows correlation between chemical measures and
perceived water attributes.

Table 30 presents the mean values of each of the activity
ratings for each of the four organic load groupings of streams.
The analysis of variance was conducted to identify significant
differences between organic load groups.  The differences
between organic load groups were significant for the two water-
related activities; at the 0.05 level for fishing and at the
0.01 level for wading - playing in the water.  The group means
in both of these cases show the expected pattern of differences,
namely lower ratings for high levels of organic load.  Signi-
ficant differences at the 0.05 level were also observed for
picnicking.  However, the group means in this case do not
display the expected pattern.  Instead, the lowest mean rating
occurs for the stream group with the lowest organic load.
                             136

-------
                                        Table 29

            CORRELATIONS BETWEEN ACTIVITY RATINGS  AND CHEMICAL
                    AND BIOLOGICAL CHARACTERISTICS OF WATER
ACTIVITIES AREA            Not-water-related                             Water-related
SUITED FOR:       	activities	   	activities	

                 Relaxing, meditating   Enjoying the  Picnicking   Wading, playing in     Fishing
                                      scenery                   the water
Chemical & Biologi-
cal Variables:

  COD
  Coliforras
  fecal strep                                                           .68             .69
  Total P                                                             -.52
  Ortho P
  NH3
  NO;,
  N03                                                                 -.59
  S04                                                                                -.51
  Cl
  TDS 103°                                                            -.54            -.52
  TDS 179°                                                            -.58            -.51
  Water temp.
  DO

  Water Quality
    Index                                                              -52

  Underline indicates p is less than  .01
  For all other coefficients p is less than .05

  N=12

-------
oo
                                   Table 30
       VARIATION OF ACTIVITY RATINGS AMONG LEVELS  OF ORGANIC LOAD1
                       (Mean Answers Per  Stream Group)

                           Level  of OrganicLoad

Hot-Water Related
Activities:
relaxing, meditat-
ing
enjoying the
scenery
picnicking
Water-Related
Activities:
wading, playing
in the water
fishing
Heavy Moderate Low Very Low F
2.38 2.28 2.25 2.28 .95
2.33 2.16 2.31 2.30 .89
1.86 1.92 2.00 1.58 3.35
1.19 1.36 1.58 1.64 6.89
1.33 1.33 1.61 1.64 4.14
P
N.S.
N.S.
.05
.01
.05
    ^Answers to question: "Assuming you enjoy all  the following activities, and
     you had come here for recreation, how much would you like to do each of
     them here?"   scale: l(not at all), 2(somewhat), 3(very much).

-------
The responses pertaining to what activities a subject would
engage in if he returned to the site (which are not presented
in a table) show no significant differences between streams
grouped by organic load.  This is consistent with findings
to be reported in the next chapter, which indicate that
persons are more apt to consider a site suitable for a given
activity than to say that they would actually engage in the
activity there.

In response to another question in this study, the observers
did not identify water quality as a criterion for choosing
an area to be made into a park.  Definite criteria emerged,
but water quality was not one of them.
SUMMARY AND CONCLUSIONS
The studies reported in this chapter provide a basis for
identifying the preferences citizens may have for the envir-
onmental preservation of alternative stream valleys or parts
of a stream valley.  The studies indicate that one can expect
to find a reasonably high level of agreement on the attrac-
tiveness of alternative environments.  Generally preferred
landscapes are those which are parklike, with mowed grass
and scattered large trees, rather than completely natural
or wild.  In addition, a minority can be expected to have
strong positive emotional responses to areas characterized
by extreme spaciousness, extreme seclusion and extreme
naturalness.  Variety, particularly if it is in some way
logically or aesthetically ordered, adds significantly to
preference.  However, specific natural features, in general,
are less important than an illusion of naturalness, and this
can be most easily lost through the presence of negative
characteristics such as junk, ugly development, or other
man-made "misfits."

A field study focused on the perception of water quality
indicated that persons are able to identify water pollution
reasonably well.   Their perception of the degree to which a
stream is transparent, clean,polluted or healthy was corre-
lated significantly with a number of chemically measured
characteristics.   However, no significant correlations were
found between chemical measures and preference ratings for
the stream, its surrounding area, or the area overall.  Thus,
the connection between objective level of water quality and
                             139

-------
liking for a site appears to be weak.  However, preference for
a stream and its surrounding area is related to the observer's
perception of whether the stream is transparent, clean, in-
viting, or healthy.  These perceived attributes, in turn, are
correlated with the chemical measurements.  Observers' judg-
ments of whether a stream is suitable for certain water-
related activities are also correlated significantly with a
number of chemical variables.  Therefore, there is a general
consistency between chemical measures of water quality,
perceived attributes, and preferences for the stream and
surrounding area.  However, the study does not indicate that
water pollution is a strong determinant of preference.  Cri-
teria other than water quality were evidently more important
to the observers, who were generally agreed on the relative
attractiveness of a stream.
                             140

-------
        VI.  EVALUATION OF EFFECTS:  ESTIMATING THE
               VALUE OF ENVIRONMENTAL PROTECTION
INTRODUCTION
A land use planning and control program such as has been
considered in this study would consist of a plan indicating
critical areas in the watershed where no development would
be allowed.  Typically, these would be flood plains, marshy
areas, steep slopes, and certain wooded areas.   Development
would be severely restricted in such areas through a combi-
nation of outright public purchase of some lands, which would
then become public parks and recreation areas,  and of public
purchase of conservation easements on other land which would
severely restrict its development by its private owners.
The public would not have the right of access to these latter
areas, unless hiking, fishing, or other right-of-way easements
were specifically purchased.  The program would also include
a well-thought-out set of sewage treatment policies to serve
permitted development.

The studies presented in Chapters 2, 3, and 4 indicate the
major expected effects of typical urbanization on stream
channel enlargement, peak flows, water quality, and stream
valley open space.   For a given amount of expected typical
development in a watershed, these results could be used to
calculate the effect of urbanization in the absence of a plan
for environmental protection.  The difference between these
predicted indicators of environmental quality and the level
of these indicators before urbanization provides a measure
of the maximum deleterious effects which might be avoided by
better planning and management.  It is not expected, however,
that all these effects could be avoided, since some deterior-
ation of the natural environment is inevitable with urbaniz-
ation.

It should be noted that the results of the studies described
in Chapters 2 and 3 would have to be modified considerably
in order to describe environmental effects on quality of
proper planning.  The urban development which would exist
                             141

-------
under planned conditions would differ from unplanned develop-
ment not only in location and overall intensity, which have
been dealt  with in our results, but also in the design of
physical features.  Planned development should include pro-
visions for improved treatment of sanitary sewerage, strict
regulations regarding on-site sewage disposal, and design
of storm sewerage (and perhaps the impervious surfaces them-
selves) to minimize hydrologic impact.  The results of our
studies, which are based on typical urbanization as it now
exists, would not be directly usable to predict urbanization
impacts under these conditions.

The environmental impacts associated with fully planned devel-
opment, incorporating improved design of facilities as well
as restrictions regarding location of development, would
certainly be smaller than the impacts of development restricted
as to location but not as to design.  (Impacts resulting from
variations in location could be calculated using our results.)
Although we cannot say how much less the impacts would be,
there is some possibility that the improvements brought about
by design changes may exceed the improvements obtainable by
restricting location, for a given overall intensity of devel-
opment.  This question is discussed further in the concluding
chapter.

Certain effects of urbanization have not been studied as part
of this research, such as the effect of construction activity
on sediment production.  Although the location of development
might have some bearing on these effects (e.g., restriction
of development on steeply sloping land should reduce sediment
production), it is suspected that these effects depend pri-
marily upon factors other than location.  Thus, as far as
sediment is concerned, the improvement brought about by the
plan relative to unplanned conditions would depend largely
upon the efficacy and enforceability of regulations of con-
struction practices which might be devised.

The previous chapter has described a series of studies con-
cerned with gaining information which would allow us to
estimate the potential value of land use planning to residents
of an area, i.e., to translate the physical elements into
value terms.  The research reported there has dealt with
basic issues regarding environmental perception and preferences
This chapter will consider more specifically the question of
plan evalution and the calculation of benefits accruing as
a result of the plan.
                             142

-------
Planning benefits can be considered in two categories:  the
"fiscal benefits" of environmental protection, including
direct economic benefits to taxpayers, and "non-fiscal
benefits" which accrue to individuals.

FISCAL BENEFITS
The fiscal benefits include avoidance of increased water
supply costs and avoidance of flood losses.  The two major
relevant aspects of water supply costs are reservoir costs
and the costs of water purification.

Avoidance of Reservoir Costs
A Brandywine-type plan would result in higher ground water
levels and in more constant streamflow than would be expected
under typical urbanization.  With such a land use plan, it
would be possible to use the stream as water source for a
larger percentage of the year than with ordinary urbanization.

It is doubtful, however, that this implies that reservoir
capacity could be appreciably smaller.  Design capacity for
a reservoir depends upon the relationship between seasonal
high flow and seasonal low flow, given a constant demand
for water [Babbit and Doland, 1955, pp. 170-172].  A planned
pattern of land use is likely to have little effect on the
amounts of water transported during the periods associated
with seasonal high and low flows, since these amounts are
determined primarily by variation in rainfall.  The effect
of a planned land use pattern would be to lengthen the lag
time of the effects of small and medium size storms and spread
out their hydrographs.  (Relevant here are the findings
reported by Richard W. Hawkins [1969].  He found that timing
delays of the order of half a month would have moderate
effects for reservoirs with intermediate storage capacity.
However, as the basic streamflow volume is not increased,
the upper limit of yield is not raised.)

Even assuming that a substantially smaller reservoir would
be adequate with the more constant flow of a Brandywine-
planned watershed, it is doubtful that major cost savings
would be enjoyed.  There would be savings in capital costs of
land acquisition and construction for a smaller reservoir, but
operating costs would be affected relatively little by a
moderate reduction in reservoir size.
                                  143s"

-------
Avoidance of Water Purification Costs
Somewhat higher water quality might be expected in a watershed
developed according to Brandywine principles as compared with
one developed in the ordinary pattern.  Therefore, if the
water downstream of such a watershed were to be used for water
supply, it would be necessary to remove somewhat smaller
amounts of impurities.

However, studies indicate that the savings in water treatment
costs would not be large.  Kneese and Bower state, "Some results
are available for the petroleum refining, fruit and vegetable
canning, thermal power, and beet sugar industries.  In all
these instances, industrial costs turn out to be surprisingly
insensitive to intake water quality within comparatively wide
ranges — especially in regard to aspects of quality that are
usually influenced by prior uses and discharge of effluents.
Sensitivity is greater to wastes which in most cases are of
natural origin, such as chlorides and magnesium " [Kneese
and Bower, 1968, p.125].

Continuing to draw from Kneese and Bower: for example, it
turned out that downstream water withdrawals to be treated
for municipal water supply had to be anywhere from 10 to
250 times as large as the watste discharge (after treatment)
upstream before additional waste treatment costs upstream
could be justified.  Again it appears that the need to prepare
potable water cannot justify particularly high standards of
quality in watercourses.

This leads to the conclusion that higher water quality must
be justified primarily on aesthetic and recreational grounds .

Avoidance of Flood Losses
The plan would result in the avoidance of flood losses in two
ways:  through reduction in the increase of peak flows assoc-
iated with urbanization, and through prevention of development
in areas subject to flooding.  The latter of these is probably
the more important.  The significant increase in flows due to
urbanization is associated primarily with small watersheds and
with floods of moderate frequency, i.e., floods occurring every
five years or less.  Most severe flood damage, however, is
associated with large streams and with floods of relatively
low frequency (i.e., floods so great that they occur only
every 20 or more years).  Thus, the most important step in
                             144

-------
avoiding flood losses is simply to prevent development at
locations where flood losses are likely to occur.
NON-FISCAL BENEFITS
These considerations suggest that fiscal benefits of envir-
onmental preservation are not the major consideration in
evaluation of land use plans.  In order to justify the public
expenditure on land and easement acquisition and the private
costs incurred by restrictions ,on development, it is clearly
necessary to demonstrate that major non-fiscal benefits will
occur.  These non-fiscal benefits will have to comprise the
major portion of the value contributed by the plan.   There-
fore, the remainder of this chapter is concerned with the
evaluation of non-fiscal benefits.

In order to evaluate benefits, they must be expressed in some
common measure so that they can be compared with costs.  There-
fore, estimates of the dollar value of non-fiscal benefits
are desired.  A major issue is the extent to which the benefits
accruing from the plan will be evaluated through economic
transactions in the market.  The primary market in question
here is the market for residential land.  Many of the benefits,
such as those related to various outdoor activities and to
general aesthetic enjoyment, may not be registered in the
market directly, and there is some question as to what extent
they are registered indirectly.

Techniques have been developed for placing values on various
activities, such as hunting and fishing, which are not priced
in the market place; these have been applied in many cost-
benefit studies.  Theoretically, assuming that appropriate
values can be determined, it does not matter whether benefits
are registered in the market place or not.  In the case of
evaluating plans for environmental protection, however, it
may be necessary to depend largely upon benefits registered
in the market place and to demonstrate that these will be
large enough so that a major share of the costs will be
recovered through higher real estate prices or other mechanisms
Otherwise it may be politically difficult to enact such plans.
It is assumed in this discussion that the costs of environmen-
tal preservation, i.e., the costs of easements and outright.
land acquisition, are to be borne within the watershed itself,
through some scheme of redistribution of real estate value
                             145

-------
increments.  Very large-scale environmental preservation
programs would not be subject to this constraint.

In order for the values created by planning for environmental
protection to be reflected in the real estate market, the
following two conditions must hold:  first, persons must
consider it important to have a natural and uncontaminated
environment, and second, persons must be aware that their
enjoyment of the natural environment depends on its not
deteriorating over time, and that the plan, in fact, guaran-
tees this.

There is evidence, contained in the studies reported in the
previous chapter and to be reported in this chapter, that
many persons are rather indifferent toward environmental
conditions other than obvious noxious influences such as
junk or extreme pollution.  Furthermore, it appears that many
persons who do care about the environment do not place
corresponding economic values on its enjoyment, perhaps
because they perceive the maintenance of environmental
quality to be beyond their control.

A second major issue concerning the evaluation of effects is
the question of to whom the values will accrue.  Since the
type of plan discussed here is concerned primarily with future
urbanization, the present residents of the area constitute
only a portion, perhaps a very small portion, of the persons
who will ultimately benefit from the plan.  In estimating the
value of the plan to future residents, one can either assume
that the future residents will constitute a rather typical
cross-section of the population of the region, or else that
the future residents will be a somewhat untypical group who
have chosen the protected area specifically because of the
environmental benefits which it offers.

The results of the studies reported in this chapter and the
previous chapter indicate that the typical cross-section of
suburban residents may not place a sufficiently high value
on environmental quality to justify the sort of plan envisioned
here.  A case will be made in this chapter that it may be
necessary for a considerable recruitment process to take place
in order to induce residents to the planned area who are willing
to pay  the costs of the plan, primarily through higher property
taxes which would result from higher real estate values.  With
                             146

-------
respect to environmental planning in general, this may be a
temporary situation, in that the existence of planned areas
is likely to generate additional demands for such planning
in the future, particularly as continued urbanization will
give a scarcity value to those areas in metropolitan regions
which do retain a measure of natural beauty.  In addition,
it is possible that the preferences of society are changing,
so that in the future more value will be attributed to envir-
onmental quality.  For the present, however, the possibility
remains that resident-consumers of environmental quality
will have to be recruited from a relatively small segment
of the population.

Two major research studies will be summarized in this chapter,
The first is concerned with the evaluation which residents
make of the natural environment, as it affects their decision
to purchase a house.  The second study relates the use of
nearby streams by residents to the distance of their homes
from the stream and to the water quality of the stream.
PREFERENCES AND CHOICE OF RESIDENTIAL ENVIRONMENTS
(A full report on the materials presented in this section may
be found in Menchik, Mark D., "Residential Environmental
Preferences and Choice:  Some Results Relevant to Urban Form,"
RSRI Discussion Paper Series: No. 46, March 1971.)

Research Procedure
In order to study preferences and choice of residential envir-
onments, a home interview questionnaire survey with 457 re-
spondents was conducted by RSRI [Menchik, 1971] which is
summarized below.  Using a cluster sampling design, samples
were drawn at different sampling rates in each of five areas
in southeastern Pennsylvania: (1) the Upper East Branch
Brandywine Creek area, a predominantly rural area in western
Chester County; (2) the Neshaminy Creek area, a predominantly
rural area in upper Bucks County; (3) the Northern Chester
County area, rural and suburban;  (4) the Chester County Towns
area, several relatively densely settled towns; and (5) the
Main Line area of northeastern Chester County, a relatively
wealthy and recently subdivided suburban area.  Both Bucks
and Chester are suburban counties within the Philadelphia
Standard Metropolitan Statistical Area.
                             147

-------
Respondents were asked one of two series of questions about
the choice of a place to live.  When the respondent had lived
at his present home for ten years or less and that home had
been chosen by the respondent or spouse, then the respondent
was asked about the choice of the present home.  The question
was phrased:  "When deciding on a new home in a new area,
some people judge a house and area by one set of characteristics,
while others judge them by other characteristics.  When you
were trying to decide on a place to live, what were all of
the chatacteristics that were important to you?"  The respon-
dent's open-ended reply was written down verbatim by the
interviewer.

The remaining respondents, those whose present residence had
not been chosen by them, or had lived there more than ten
years, were asked about a hypothetical move:  "Suppose you
were trying to decide on a place to live.  What would be all
the characteristics that would be important to you?"  Again,
the response was copied word for word.

The residential preference data used below were obtained from
a content analysis of the replies to these questions which led
to the established response types.  The response types are
listed in Table 31.  There are four types:  responses concern-
ing the natural environment, the non-natural environment,
accessibility, and chatacteristics of home and lot.

The respondent's preference for any item (or combination of
items) is operationally defined by the number of times the
relevant response type(s) was mentioned.  The questions
were designed to elicit considerations important in the choice
of a home (present or hypothetical) and, given the questions'
unstructured form, it was assumed that the content of the
response would indicate the degree of importance or preference
for any locational consideration.

Findings Based on Interviews
Table 32 shows, for a given response-type category, the percent
of respondents mentioning considerations from that category
in choosing a place to live.  For example, as shown in the
table, 49% of those asked about choice of present home
mentioned at least one consideration involving accessibility
and 52% of those asked about hypothetical choice of a home
mentioned accessibility.  The table also gives the percent
                             148

-------
                          TABLE 31

             CONSIDERATIONS IN CHOICE OF A PLACE TO LIVE
             (Number and Percentage of Respondents making
                One or More Responses of Each Type)
                                     No. of          % of
                                     Respondents     Respondents
Natural Environment  (N)
1.  Ruralness, woodsiness, country-
    like character of the area         103              22.9

2.  Specific mention of an area's
    prettiness, beauty of surround-
    ings, view                          23               5.1

3.  Other specific considerations
    referring to natural environment:
    climate, weather, topography, etc.  46              10.2

Non-Na tur a1 Envir onme n t  (M)

Density:  presence and number of
persons in the vicinity, in general,
without references to any of their
characteristics

4.  Good: Wanted to be near people       6               1,3

5.  Bad:  Didn't want to be near (too
    many) other people                  44               9.8

6.  Neutral:  Just interested in whether
    there are many persons nearby        2               0.4

Characteristics of people nearby

7.  Good:  Wanted to be near a particular
    kind of person, near my kind of
    people, etc.                        78              17.4

8.  Bad:  Don't want to live near some
    stated kind of person, disrespect-
    able, unfriendly, etc.              16               3.6

9.  Neutral:  Wanted to know what kinds
    of persons around                    3               0.7
                            149

-------
                          TABLE 31 (cont'd.)


                                      No. of          % of
                                      Respondents     Respondents

Presence of traffic and non-residential
land uses like stores, factories, etc.

10. Good:  Wanted to live near a
    store, etc.               „           4               0.9

11. Bad:  Didn't want stores, factories,
    etc., wanted purely residential
    uses, little traffic                21               4.7

12. Neutral:  Interest in presence
    of stores in vicinity                2               0.4

13. Other considerations referring
    to the non-natural environment:
    nice neighborhood, good section of
    town, quiet, noise, privacy, well-
    paved streets, etc.                 87              19.4

Ace ess ib i 1 i t y (A)

14. To work                             63              14.0

15. To schools                          73              16.3

16. To churches                         46              10.2

17. To transportation facilities        45              10.0

18. To shopping                         69              15.4

19. To other specified activity or
    convenience, in general             70              15.6

Characterististics of House and Lot (H)

20. House size, number of rooms, size
    of rooms                           112              24.9

21. Price of house and land, purchas-
    ing or renting arrangements, mort-
    gage, etc.                          54               12.0
                                150

-------
                          TABLE 31 (cont'd.)
                                     No.  of            % of
                                     Respondents       Respondents
22. House design characteristics,
    specific and general, room
    layout, heating, plumbing, etc.    189              42.1

23. Lot size                            94              20.9

24. Other natural environment lot
    characterisitcs: stream in back-
    yard, soil, landscaping, etc.(for
    lot, itself, not vicinity)           39               8,7

25. Other non-natural environment lot
    characteristics: distance from
    road, sewage facilities, orienta-
    tion, etc.                          30               6.7

Other

26. Respondent or spouse used to area,
    lived here before                   12               2.7

No. of respondents with usable answers 449

Respondents with no usable answers       8
                             151

-------
of respondents mentioning various combinations; for example,
the percent of persons mentioning both accessibility and
house and lot.

Of the four response types, the one mentioned least frequently
was that involving the natural environment:  34% of respondents
asked about the choice of their present home and 37% of those
asked about a hypothetical home mentioned the natural envir-
onment.  The greatest frequencies of responses were observed
for those involving house and lot.  The middle ranking response
types, in terms of frequency of responses, were accessibility
and the non-natural environment, which received roughly an
equal number of responses.  For none of the response types
were the results significantly different for those respondents
asked about the actual choice of their own house as opposed
to those asked about the choice of a hypothetical house.
Therefore, in subsequent analysis the respondent groups are
considered together.

Table 33 presents totals for each of the four areas in which
the survey was conducted.  Variation in the response rate over
the five areas is most striking for responses in the natural
environment category.  The response rate was more than twice
as high for the Brandywine area, which is the most rural and
natural of the four areas, as for the Main Line, which is
heavily suburbanized.

Additional tabulations of the responses were prepared which
considered variation with age of household head, total family
income, sex of respondent, and chilhood residence of respondent
No significant variation in the percent of respondents mention-
ing the natural environment was observed for any of these
classifications.  A most unexpected result was the fact that
mention of the natural environment did not increase with
income.  The lowest percentage of respondents mentioning the
natural environment was observed for the $15,000 and over
income group.

Findings Relating Interview Responses to Residential Variables
In order to compare respondents  residential preferences with
their actual choice of a place to live a detailed study was
conducted in which data were collected pertaining to the area
near each respondents' residence, using variables paralleling
the preference variables.  (The 457 respondents were drawn
from 92 small areas.  The descriptive variables applied to
                             152

-------
                            Table J32
SUMMARY OF RESIDENTIAL PREFERENCES,  FOR RESPONDENTS ASKED
      ABOUT CHOICE OF PRESENT OR A  HYPOTHETICAL HOME
 (Percentage  of Respondents  Making One or  More Responses
  for Each Major Category, or Combination  of Categories)
                             Present     Hypothetical
        Category               Home          Home	    Total
   A  (Accessibility)           497.            52%          50%
   H  (House and Lot)           677.            697.          687.
   N  (Natural Environment)      347.            377.          357.
   M  (Non-natural En-
        vironment)              467.            507.          487.
   NA                         177.            187.          177.
   NH                         197.            207.          197.
   NM                         207.            207.          207.
   AH                         267.            327.          287.
   AM                         237.            297.          257.
   HM                         247.            287.          257,

   Total No. of Respondents      257            151          408
   No Usable Answers             2             35
                              153

-------
                          Table 33


         INTER-AREAL COMPARISON OF RESIDENTIAL PREFERENCES
(Percentage of Respondents Making  One  or More Responses  for  each
 Major Category,  or Combination of Categories)
                               Northern
Preference
Category X2
A 6.368+
H 5.273
N 12.950**
M 6 . 440+
NA
NH
NM
AH
AM
HM
Total No. of
Respondents
No Usable Answers
Brandywine
56%
63%
45%
55%
24%
25%
23%
30%
34%
27%
163
2
Neshaminy
42%
67%
32%
39%
12%
14%
18%
23%
16%
18%
107
2
Chester Co.
52%
78%
31%
48%
16%
18%
21%
35%
23%
34%
88
1
Main Line
42%
66%
20%
44%
10%
12%
14%
20%
20%
22%
50
0
Total
50%
68%
35%
48%
17%
19%
20%
Z8%
25%
25%
408
5
          + Statistically significant  p  < .10
         ** Statistically significant  p  < .01

-------
each of these areas as a whole, rather than to the respondents'
own houses.)  Thus, descriptive data were obtained pertaining
to the actual level of the natural environment in the neighbor-
hood of the respondent's residence, as rated according to
various criteria, and to the levels of the other three response
types.  The measurements concerning the natural environment
included the total perimeter of ponds and length of streams
within both a half-mile and one mile of the center of each
residential area, the maximum difference in elevation within
each area, the percentage of the area in woodland, and a sub-
jective rating of the attractiveness of the natural environ-
ment in each area.  These measures were combined to form a
single natural environment rating.

The descriptive variable corresponding to the accessibility
response type is the log of the individual's commuting time.
Corresponding to the house and lot response types is the log
of the size of the individual's lot.  Corresponding to the
non-natural environment response type are measurements of the
density of structures and traffic within half-mile and one-mile
circles of the center of each residential area sample (with
these variables expressed in negative form).

Each of the response variables, which are related to these
descriptive variables, consisted of the number of responses
of a given response type made by an individual divided by
that individual's total number of responses.

The correlations between the four response variables and the
four descriptive variables were generally very low (Table 34).
Only four of the 16 correlations are significant at the 5%
level.  Two of these correlations were between the natural
environment response variable and the descriptive measures
of house and lot and the non-natural environment.  Considering
that the latter two variables both relate to the density of
development, these correlations are not surprising.  However,
the correlation between the natural environment  response
variable and the natural environment descriptive variables was
only 0.057.  Taken literally, this result indicates that there
is very little relationship between expressions of the natural
environment in choice of a residence and actual quality of
the environment near the residence one chooses.  This result
is somewhat contradicted by the result, described above and
shown in Table 33, that more than twice as many people living
                             155

-------
                                Table  34

               CORRELATIONS  BETWEEN  RESPONSE  VARIABLES
                       AND DESCRIPTIVE VARIABLES
   Response
  Variables
            Descriptive Variables^
                Accessibility
           House
           & Lot
        Non-Natural
        Environment
           Natural
           Environment
Accessibility
 .133*
-.121
.019
-.057
House and Lot
-.058
 .078
.010
-.015
Non-Natural
 Environment
-.026
 .069
.137*
 ,031
Natural
 Environment
-.043
 .175*
.226*
 ,057
               * - Significant at  the  .05 level
                                   156

-------
in the Brandywine area mentioned the natural environment than
was the case for respondents living in the Main Line area.
The environment in the Brandywine area is certainly more
natural than in the Main Line area.

The concept of "naturalness" is only partly measured by the
variable N (Natural Environment).  An extremely important
part of this concept, perhaps the most important, is a low
density of development and activity.  Low density is measured
explicitly by variable M (Non-natural Environment).  This
variable is significantly correlated with Preference for
Natural Environment.

The implications of these data for plan evaluation are some-
what difficult to determine, partly due to the consideration
just discussed, namely that natural environment as it relates
to the plan involves both variables M and N.

Also, it is difficult to compare response rates involving
different aspects of residential choice.  The critical factor
in the residential choice decision is determined not only by
the size of the response rate for each factor, but also by
the range over which choice on each factor can be made.  For
example, house and lot receives the highest response rate,
but since housing types are fairly standardized, other factors
may be more important in determining the location actually
chosen for a residence.  Housing, and also accessibility,
might be viewed as constraints which must be met, but within
which persons optimize other considerations.

The variables under analysis can also be looked at from the
point of view of a community which desires to direct its
future growth.  The community can do little about accessib-
ility since that is determined largely by its location within
the metropolitan area.  It can determine housing types to
only a limited degree through zoning and subdivision controls,
but the home building industry will continue to make the
major decisions concerning style and design of house.  The
community, however, can have a major effect on overall
density of development and pattern of preserved natural
amenity.
                             157

-------
PERCEPTION AND USE OF STREAMS IN SUBURBAN AREAS
The second major study relevant to the estimation of the value
of environmental programs is concerned with how persons per-
ceive and use the streams they live near.  Specifically, it
is concerned with their judgment of the stream's attractive-
ness, the extent to which nearby residents visit and use the
stream, the effect they believe it has on their property
value, the extent to which they are aware of the severity
of pollution of the stream, and the actions they would be
willing to take to abate the pollution.   The research carried
out on these subjects is discussed fully in Coughlin, Hammer,
Dickert  and Sheldon [1972] and is summarized below.

The focus of the analysis is on how these attitudes and uses
vary with the level of water quality of stream, and with
distance of residence from the stream.  The concern with water
quality arises from the fact that maintenance of relatively
high water quality is one of the major effects expected from
a plan of land use and waste management.  The perception of
water quality levels by neighbors and the use that they
actually make of the stream given their perceptions and the
objectively measured levels of water quality must be known
if one is to evaluate the benefits resulting from the plan's
effect on water quality.

The concern with distance of residence of respondent from
stream makes it possible to determine how the benefits due
to recreational and aesthetic functions provided by a stream
fall off with distance from it.  With this information it is
possible to estimate how total recreational and aesthetic
benefits vary depending on the pattern and amount of nearby
development, and thus to define a service area for the
recreational and aesthetic functions of a stream.

The effects of water quality and distance to residence are
conceptualized in Figure 10, which indicates that the more
polluted a stream is, the less a neighbor will use it for
recreation or value it as an aesthetic asset.   Also, the
further away he lives, the less he will use or value the
stream.
                             158

-------
                   Figure 10

CONCEPTUAL RELATIONSHIPS BETWEEN WATER QUALITY
     OF STREAM, DISTANCE OF RESIDENCE FROM
           STREAM, AND USE OF STREAM
      Use of
      Stream
                                              Distance from
                                                Stream to
                                                Residence
     Water Quality
       of Stream
                        159

-------
The Effect of Distance from Residence to Stream
In order to concentrate on the effect of distance from house
to stream, interviews were conducted with a total of 222
households, located in the vicinity of 5 stream reaches.
Respondents were asked which of a number of activities they
considered the stream to be suitable for and which of those
activities they actually engage in at the stream.  They were
also asked a number of questions concerning the effect of
the stream on property values.
                                                              
-------
A small constant (c) was added to the socio-economic indices
in order to avoid indices close to zero which in log form
would dominate the regression results.

It would have been desirable to have a single variable expres-
sing stream characteristics similar to the single socio-
economic variable.  However, with only five streams in the
sample, it was impossible to determine the appropriate weight-
ings for the various water quality and other stream character-
istics.  Thus, the procedure was to utilize a set of "dummy"
variables to assign an "effect" to each stream without providing
any indication as to why differences in effect appear between
the various streams.

Therefore, it was hypothesized that responses would be related
to distance to streams, times socio-economic index, times a
stream-specific factor; with each of these three factors
raised to various exponents.

Findings Concerning Effect of Distance of Residence from Stream
The equations expressing the effect of distance, socio-economic
index, and stream characteristics are given in Table 35.  It
will be seen that the socio-economic index is significant at
the 0.05 level for every dependent variable.  The fact that
socio-economic characteristics provide a large proportion of
the explanation of residents' attitudes toward, use of, and
evaluation of nearby streams is not unexpected.  In fact, most
analyses concentrate on such characteristics to the exclusion
of other independent variables.  Knowledge of such relationships,
however, is of limited value to the planner who is concerned
with delineating land development patterns which are intended
to be relatively permanent and cannot be easily changed.  For
the planner, relationships involving distance from the stream
are the most useful.

Distance berween residence and stream is significant at the
0.05 level for nearly all of the activities.  However, it is
significant only in scattered instances for the remaining
questions.  For example, distance to residence appears to have
little effect jUpon a person's knowledge that the stream exists,
upon the reasons why the stream is not visited, upon the degree
to which people would rather live closer to the stream, and
upon whether they think their house would be worth more were
it located nearer the stream.  However, the dollar amount of
increase in value estimated by the respondents does fall off
significantly with distance of present residence from the


                             161

-------
                                        Table  35
S'J>Bt\RY OF kEi:!t'-:V.!i*S FXfUlHm; f.UUVKV  UK KIN-JEN I'.V  mslAXiX AU>.,V H'lU.IC UV,VS »r.TU!:KN STREAM AXU
              K::.-.n:i:;ici:, SOCIO-UON.MIC  IND.;X, asn siKi:-\M-si-n.in

F.;;-;~r.!,e to Survey
Question:

1 Know strc.-.a is no.ir
1 Kr. j.: Its r.o-o
3 Cvjr.llx a.-£i-£ of it
4 Co lo s t v ear,
b jicr*:.;:. gc-cc for
ht-.-.tlnt
7 Strcu.i cs-xi tcr
/lihirj
9 S(.rt..T gcod for
i.:=.nc
IX ;;troiii f,oo=i good for
walking
19 Str>j=. Biod for bird
watching
*JL Strpatj good for ice
skatine
29 Streao goad for ct
least one activity
6 Er.gajs in hunting
8 trjage in fishing
10 £r*;ga£a in waging
12 Eaguje in sitting
14 Engage in picnicking
16 Engage in play. faxes
IS Engage in miking
23 Engage in bird
watching
22 Engage in lee skating
30" Engage In at least
one activity
31 Don't gn there: go
elsewhere
32 Don't go there: sticax
polluted
33 Don ' t go there : ether
reasons
34 Veuld use stream if
publicly ovr.ed
3$ Vj-jld like tc live
cljs-ir
36 Vo-jld Ilka to live
fronting
37 Would like to live
with view
33 Vould like to live
-..•irhln 5-alirjte
valk
39 House be vorth core
40 How much ntorr
41 Notice polluticn
42 Others consented on
pollution
43 Sua of actions:
letters
44 Su*n of actions:
direct action
45 V 9U Id take no action
46 Accept mre taxes to
abate pollution
47 Environmental concern
SOCIO-KCO:W.IC IKW>
STU
1
1.000
.97$
.597
.673

.38/1

.450

.322

.450

.431

.362

.601

.336

.538

.978
.018
.201
.056
.232
.201
.161
.468

.100
.380

.974

.020

.208

.158

.436

.245

.150

.095


-.001
.221
4.872
.637

.287

.882

.442 '
.116
.460
.309
EA>:-s!ri:cmc iKKt VA-IIABIFS
lfir,et*tl vi rci>ff:rlci:ra)
2
.595
.994
.56.'
.9>4

.345

.360

.226

.256

.270

.264

.411

.255

.382

1.015
.'.07
.064
.053
.189
.104
.071
.284

.092
.285

.993

.072

.010

.106

.457

.025

.171

.034


-.005
.028
3.607
.665

.240

1.318

.904
1.239
.442
.246
3
1.007
.Vl-j
.S09
.7b*

.348

.C16

.485

.542

.C09

.583

.589

.307

.596

.992
.108
.436
.259
.485
.533
.469
.551

.137
.444

.892

.069

.149

.183

.394

.013

-.014

.027


.002
.040
-.220
.714

.293

.691

.535
.325
.334
.292
4
.550
.973
.936
.622

.113

.28*

.168

.164

.031

.114

.373

.245

.404

.947
.063
.284
.124
.131
.076
.073
.401

.151
.447

.929

.167

.126

.481

.531

.055

-.001

.084


.003
.100
2.051
.865

.469

.745

.573
.503
.435
.295
5
.657
.820
.ETC
.475

.408

.564

.371

.540

.332

.366

.560

.340

.310

.889
.094
.253
.188
.190
.162
.142
.370

.148
.072

.908

.064

.018

.434

.573

.252

.210

.088


.025
-.012
2.351
.313

.127

1.027

.849
.299
.421
.281
lac (.-He)
Cncf flclrnt
	
J.2515
2.1S51
1.7S.01

3.2059

	

0.5461

	

	

1.0370

	

0.8550

3.6932

4.7330
....
J.0411
	
1.7902
2.9492
2.9683
2.3143

2.9962
0.7635

2.1265

	

	

3.7682

3 . 9043

	

	




. 0.2829
2.7635
	
1.7010

3.6936



4.1427
	
2.4664
3.7731
Strnd.4L'd
Kirror
	
(1.6041)
(0.1*610)
(O.SSiS)

(0.8735)

	

(0.3821)

	

	

(0.4467)

	

(0.4240)

(1.4324)

(2.4757)
	
(1.0222)
	
(0.5950)
(0.6960)
(0.7379)
(0.5783)

(0.7193)
(0.1887)

(0.9674)

....

	

(1.2555)

(1.0692)

	

	




(0.1136)
(1.2724)
	
(0.72/.5)

(1.3869)

....

(1.4049)
. 	
(O.M28)
(0.8736)
best Valuo
of fi
....
.5
.u
.0

1.0

	

.1

	

	

.0

—

.1

1.0

1.0
—
1.0
—
.5
1.0
1.0
.5

1.0
.0

.0

	

	

1.0

1.0

	

	




.1
1.0
—
.0

1.0

....

1.0
	
.5
1.0
DISTANCE ALSNC
I'L'llI.lC WAYS MHI.TTI'li.-
IOK D
Coefficient
.0410
-.OC73
.01,73
-.1880

-.2133

-.3048

-.2023

-.2784

-.2293

-.1343

.2361

-.0652

-.2337

-.0809
-.0570
-.2429
-.0554
-.2605
-.2495
-.1525
-.2204

-.1464
-.3074

-.1282

.0635

.0750

-.0061

-.2323

-.5334

-.0846

.0453


.0088
-.0524
-3.1238
-.1005

-.1554

-.4458

-.4682 .
.3101
-.1080
.o*n
Er- or
(.041.4)
(.0406)
(.0625)
(.0866)

(.0961)

(.0939)

• (.0542)

(.0963)

(.0936)

(.0953)

(.1014)

(.0954)

(.0991)

(.0350)
(.0568)
(.OS21)
(.0690)
(.0844)
(.0764)
(.0737)
(.0983)

(.0656)
(.0886)

(.0471)

(.0519)

(.0619)

(.0853)

(.1003)

(.3S>}>

(.0539)

(.0506)


(.0138)
(.0524)
(1.6535)
(.0952)

(.1132)

(.2..04

(.1801)
(.1768)
(.1012)
(.0709)
R
Arnvi.il X.-.-n,
.5217
.3734
.2370
.4243

.3230

.3370

.3155

.3681

.3981

.3680

.2608

.1703

.3373

.3131
.1498
.4241
.2831
.4165
.5207
.4797
.3416

.3109
.4353

.2718

.1S27

.2615

.4207

.3019

,i-',v

.2751

.1425


.Wi
..3595
.2314
.3Wb

.2190

.iO"9

.3157
.4706
.2253
.30i6
                                    162

-------
stream.  Similarly, whether the respondents notice pollution,
and whether they are willing to take action or pay additional
taxes, is unrelated to distance—except that whether they are
willing to take direct action does fall off significantly with
distance.  This is reasonable in that it takes a greater
commitment to take direct action, such as legal action,
participation in clean-ups, or picketing than it does simply
to write letters; and people who live close to the stream
and experience it more frequently are more likely to have
such a commitment.

In order to make it easier to interpret the equations listed
in Table 35, these equations have been computed on a relative
scale in which the activity level (the estimated value of the
dependent variable) at 100 feet from the stream is set at 1007»
for every activity (that is, for each dependent variable).
This level may be called the base-level use rate, that use
rate by residents which is unaffected by distance of residence
to stream [Coughlin, Taieb, and Stevens, 1972].  The distances
at which the activity level is 90%, 80%, ... 10% of the base-
level use rate are computed and summarized in Tables 36 and
37.  Table 36  relates to distances measured along the straight
line between residence and point of access on the stream;
Table 37 refers to distances measured along public rights of
way.  Curves corresponding to Table 36 are given in Figure 11.
For convenience, distances greater than a mile are not plotted.

Examination of Figure 11 shows that the number of residents
who take a walk along the stream drops off relatively slowly
with distance.  People living a mile from the stream are only
half as likely to walk along the stream as are people who live
only 100 feet from the stream.  The curve   for more specific
activities drop off more quickly than does the curve  for
walking.  At the extreme, persons living a mile from the stream
are only 5% as likely to go to the stream to picnic as are
persons living 100 feet from the stream.

The rate at which each use-rate curve falls off with distance
can be summarized in at least two ways.  The first is by the
slopes of the curves in Figure 11.  These are determined by
the coefficient of distance from the stream (D), and the
activity level of residents living 100 feet from the stream.
These coefficients may be read directly in Table 35; the
activity level at 100 feet may be estimated roughly as the
                             163

-------
                       Table 36

SUMMARY  OF REGRESSIONS EXPLAINING SURVEY RESPONSES
 MEASURING DISTANCE BETWEEN STREAM AND RESIDENCE,
            ALONG SHORTEST STRAIGHT LINE

          (Other Variables as in  Table 35)
 Response to Purvey Oucs11 on
                                 LOR of nistance Alonj*. Shortest Slr.ilf.ht Line

1
2
3
4
5
7
9
11
13
15
17
19
21
29
6
t
10
12
14
16
18
20
22
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47

Know stream Is near
Know its name
Usually aware of. it
Go to stream
Stream good for hunting
Stream good for fishing
Stream good for wading
Stream good for sitting
Stream good for picnicking
Stream good for playing games
Stream good for walking
Stream good for bird watching
Stream e,ood for ice skatine
Streao good for at least one activity
Engage in hunting
Engage in fishing
Engage In wading
Engage in sitting
Engage in picnicking
Engage in playing games
'Engage In walking
Engage in bird watching
Engage in ice ska tine
Engage in at least one activity
Don't go there: go elsewhere
Don't go there: stream polluted
Don't go there: other reasons
Would use stream if publicly owned
Would like to live closer
Would like to live fronting
Would like to live with view
Would like to live within 5-minute walk
House be worth more
How much more
Notice pollution
Others commented on pollution
Sun of actions: letters
Sum of actions: direct action
Would tako no action
Accept more taxes to abate pollution
Environmental concern
Coefficient
-.0273
-.0059
-.0706
-.1463
-.1684
-.3334
-.2231
-.2297
-.2814
-.1843
-.2264
-.0913
-.32S7
-.0134
- . 0644
-.2078
-.1448
-.2614
-.3138
-.1970
-.1719
-.1341
-.7615
-.1249
.1264
.1201
-.0585
-.1812
-,0265
-.0891
.0431
.0031
.0112
-.5625
-.0203
-.0896
-.2979
-.4580
.4488
-.1340
.0072
Standard Error
(.0482)
(.0423)
(.0651)
(.0902)
(.0589)
( . 1023)
(.0975)
(.1006)
(.0965)
(.0987)
(.1054)
(.0988)
(.1022)
(.0367)
(.0589)
(.0856)
(.0712)
(, 0877)
(.0783)
(.0761)
(.1021)
(.0681)
(.0927)
(.0498)
(.0533)
(.0639)
(.0885)
(.1047)
(.0684)
(.0653)
(.0525)
(.0143)
(.0545)
(1.7288)
(.0989)
( . 1176)
(.2505)
(.1877)
(.1821)
(.1055)
(.0940)
Multiple R
.5202
.3585
.2370
.4150
.3144
.3434
.3234
.3510
.4096
.3667
.2542
.1753
.3636
.2760
.1526
.4114
.2991
.4136
.5369
.4884
.3284
.3035
-4178
.2646
.2266
.2777
.4226
.2850
.3275
.2800
.1405
.2148
.3544
.2019
.3629
.2479
.1796
.3105
.4816
.2060
.3021
                         164

-------
                                           Table   37
7
•1


5

v
' I
11

15

71
23
*> ^
"*?


ft
10
12
14

JO
22
26




32




37
38
«0





47
K-..»- i i s "i- •- >



St;«-Jj: f^o.: *- - t----UT.I)


-
fi-h->«
wad in;
sitting
picnicking

bird watching
ice spacing
at least. 1 leisurely



, . ,
Dor.*r go chfrc - stream polluted




within view
vitSin 5 ninutc walk
How amch more





Envirorr»er.:al concern
100
100
1JO

100
)00
100

100
100
10J
100
100
100
lOu
100


100
100
100
100

100
100





Positive




Positive
Positive
100





Positive


31?

i->:

1 9.',


•"0
221
3f 6
211
170
189
21:


167
1S4
167
165

169











161




215



1008

3-.o
360
"^77

377
4^9


/i44
1'oS
323
444


280
339
280
272

259
2&5











259


33S

462



3''0i

710

731

732
920
lOfrO

935
5&0
936


463
623
468
448

416
430











417


g-»2







13'-i





2^v"7

1970
8^7
1042
1973


7S2
1146
783
735

663
811











672




2138





-,-->! * -,-,-








1403 23£0 ;036



1?08 2107 3655 	
:ilO 36F2 	
1310 2191 2665
1217 2005 33^ 	 --

1075 i:?s ::79 44^*. 	
1368 2300 3S94











1031 1741 2803 4 $12






• Denote? 
-------
                Figure 11

PERCENT OF BASE LEVEL USE OF STREAM SITE
 BY DISTANCE OF RESIDENCE FROM STREAM,
         FOR SELECTED ACTIVITIES
(distance is Straight-Line distance from
  residence to nearest point on stream)
            : from residence
             \    i  .  i
                  166

-------
average of regression coefficients for the dummy variables
which are found in the same table.

The second is the "effective service radius," that distance
at which use of the stream drops to a given fraction of the
use rate by residents who live near the stream.  If effective
service is defined as, say, 20% of base-level service (that
is, each resident at the effective service radius is getting
one-fifth as much use out of the stream as those who live next
to it), then we can observe that the effective service radius
for picnicking is about 2,500 feet, for wading about 3,300
feet, for sitting about 3,900 feet, for fishing about one mile,
and for the remaining activities, including walking, more than
one mile.

The effective service radius can be mapped directly to indicate
the residential locations for which the existence of the stream
is important.  Such a map, presented in Figure 12 for a portion
of Bucks County, shows boundary lines corresponding to 20, 40,
and 60% of the use rate by residents who live 100 feet from
the stream.  The data for drawing these lines is taken from a
curve typical of those in Figure 11, specifically the average
of the curves for sitting, wading, and bird watching.  Service
boundary lines are also drawn for the downstream portion of
Poquessing Creek.  This considerably smaller creek does not
fall within the size range of our sample, and therefore, our
estimated curves (Figure 11) may not be appropriate for it.
In general, one might expect a somewhat lower total level of
use, and a more rapid decline of use with distance than would
be expected for a larger stream typical of our sample.  There-
fore, less confidence should be placed in the service area
boundary lines drawn for Poquessing Creek than in those drawn
for Neshaminy Creek.

It is evident from Figure 12 that the majority of residential
locations lie beyond the effective service radius, and therefore,
receive relatively little use benefit from the existence of the
stream.  However, the total use which is made by the many
residents who live beyond the effective service radius is very
substantial.  Since the extent of our household survey was
limited and since we did not do a direct survey of users at
the stream, we cannot present data on the percent of users
who live beyond the effective service radius.  However, it is
possible to estimate the number of users by projecting the use
                             167

-------
 curves,  multiplying the resulting projected use  rates by  the
 number of people living at each location and summing.

 Since the use curves tend to flatten out beyond  the  effective
 service radius,  this area may well generate a significant
 proportion of all stream users.  For example, the "sitting"
 curve lies in the midst of all the specific activity curves.
 Therefore, it may be thought of as an average case.  If we
 assume uniform population densities as we go away from the
 stream,  we can calculate directly from that curve that 158
•visits (in arbitrary units) are generated by persons living
 within the effective service (20%) radius.   (This is computed
 by taking the average use rate in each 1,000-foot distance band
 and summing,  i.e., 63 + 42 + 30 + 22 = 158.)  In order to
 generate another 158 visits, one would have to go out nearly
 another 3 miles--that is, a total service area of somewhat
 less than 4 miles.  Thus we can hypothesize that if  the streams
 are 8 miles or more apart on the average, more than  half  of
 the users will come from beyond the 20?0 effective service
 radius.   Study of the map of Bucks County will indicate that
 although there are a number of localities where  streams are
 eight or more miles apart, typically streams are somewhat
 closer.   Therefore, substantially more than 5070  of the users
 of a Bucks County stream can be expected to live within the
 20% effective service radius.

 The Effect of Water Quality
 Research Procedure
 Observations  on five streams,  which were designated  in the
 previous section, are not sufficient to provide  a statistically
 convincing explanation of relationships between  water quality
 and the perception and use of streams.   Therefore, a mail
 questionnaire was administered to residents living near thirty
 of the water  quality monitoring stations operated by the  Bucks
 County Planning Commission.  In contrast to the  survey described
 in the previous  section, in which residents living at various
 distance up to a mile were chosen, in this survey residents
 were chosen who lived as close as possible to the stream.  The
 primary purpose of the analysis of the previous  section was to
 determine how use and perception vary with distance  of resi-
 dence, while  the primary purpose of this section is  to deter-
 mine how they vary with water quality.
                                   168

-------
           Figure 12

ESTIMATED DISTANCE FROM STREAM
 FOR 20%, 40% and 60% OF BASE
  LEVEL USE RATE OF TYPICAL
           ACTIVITY
            169

-------
A total of 312 usable questionnaires were returned out of
731 which were believed to have reached their addresses.  The
resulting response rate of 42.7% is unusually high for a
mail questionnaire, and may even understate the actual
response rate, since a number of the questionnaires had to
be sent out with incomplete addresses, and, therefore, may
never have reached the intended respondents.  Because of the
high response rate, it is felt that sampling bias, which is
inherent in mail surveys, is probably not a major problem.

The Data
Two basic sets of data are involved, that derived from the
questionnaire survey described above, and a set of chemical
measurements of stream water quality.

Chemical data on water quality were furnished by the Division
of Natural Resources, Bucks County Planning Commission.  These
data consisted of measurements of the following chemicals and
chemical properties which had been made, generally speaking,
on a monthly schedule.

Conductivity at 25°C
Dissolved Oxygen (7» saturated)
C02 (mg/1)
Total alkalinity (mg/1)
PH
Nitrates as N (mg/1)
Nitrites as N (mg/1)
Phosphates as PC>4 (mg/1)
Chlorides as Cl (mg/1)
Turbidity (mg/1)
Hardness as CaCCL (mg/1)
Color
Total Dissolved Solids (mg/1)
Loss on Ignition (mg/1)
Fixed Solids (mg/1)
In order to summarize the data, and obtain indicators of the
typical condition of each stream, the following measures were
computed for each chemical:

Average of all measurements during the year
Median of all measurements during the year
                             170

-------
To obtain indicators of extreme conditions the following were
chosen for each stream:

Second worst condition observed
July measurement (or average of July measurements)

The second worst reading was specified instead of the worst,
since the worst was more likely to fall outside of the normal
range of conditions and, therefore, not be typical of its
stream for most years.  Worst condition is usually associated
with the highest concentration of a given chemical; however,
for some measures such as dissolved oxygen over most of the
range, the worst conditions are indicated by the lowest numbers.
For a number of indicators (such as pH, total solids, and
dissolved oxygen over extreme ranges) a moderate reading is
associated with high water quality, whereas, both low readings
and high readings are associated with low water quality.  In
order to identify the second worst reading for such indicators
the two largest and the two smallest readings were plotted on
curves relating chemcial readings to water quality [Brown,
McClelland, Deininger  and Tozer, 1970], and the reading
corresponding to the second lowest water quality was identified.

July readings were analyzed since July tends to be a month of
relatively high pollution, and since July data were available
for all stations.

For purposes of analysis, and for ease of communication to
persons who are not well-versed in the chemistry of water
pollution, a list of 16 chemical measurements is awkward.  It
would be much more desirable to have a single indicator of
water quality.  Because of the complexity of water chemistry
and the many uses for which the quality of water might be con-
sidered, there is no generally accepted indicator of water
quality.  However, work to develop and refine a Water Quality
Index is being done by Brown, McClelland, Deininger  and Tozer
[1970].  Their Index has been developed by a carefully
iterated series of questions between the four researchers
and a panel of 142 regulatory officials, public utility
managers, consulting engineers, academicians, and others.  The
panel members identified 11 characteristics which they consid-
ered to be most important with respect to overall water quality.
Each panel'member was then "asked to draw a curve, which in
(his) judgment, represented the variation in level of water
                             171

-------
quality produced by the various possible measurements of each
respective  (characteristic)." The results were then averaged
to produce  a curve indicating the significance for water
quality of  given concentrations of each characteristic.  The
respondents were also asked to weight the importance to over-
all water quality of each characteristic.  The rating curves
and importance weights were applied to the available data on
the 26 Bucks County streams to produce the Water Quality
Indexes listed in Table 38.  Because the available data lacked
some of the characteristics specified in the Water Quality
Index, the  Index was computed on only 7 specified characteris-
tics, which would have constituted 80% of the possible high
score had all 9 characteristics been included.  (One of the
characteristics for which data were lacking was "Departure
from equilibrium temperature," a measure which was designed
to capture major effects such as are caused by a large infu-
sion of warm water from an electric generating plant.  Streams
in the sample would not register on this characteristic.  The
other characteristic lacking data was BOD-5.  This is a somewhat
more serious deficiency, since BOD levels are likely to vary
among the 26 streams.)

An additional "index" of water quality was derived by taking
the first principal component of the set of 16 chemical
measurements, as will be discussed below.  The score of each
stream site on this component was calculated to serve as an
index; these values are listed in Table 38.

Analysis and Results

Individual Indicators of Chemcial Water Quality

In order to determine the relationships between level of water
quality and the use and perception of the streams, regressions
were run in which each of the 16 chemical measurements appeared
separately.  They were of the following form:
    X. = a  + a (log Distance to Stream)  + a (No.  of Children)
     JL    J-    fc«                            «_J
            + a,(log Drainage Area) + a.  (Chemical j)
                   ,+ a,7 (Water Quality Index)
    where  X.  (i =  1 ...  50)  is  a dependent  variable  derived
    from the  questionnaire responses,  and

                             172

-------
                       Table 38
        SUMMARY MEASURES  OF WATER QUALITY
Stream Site
I
2
3
5
7
8
9
10
11
12
13
16
15
16
17
18
19
20
22
23
24
25
26
28
29
30
Year
M2I
74.4
67.6
75.8
68.1
60.0
72.0
75.6
74.5
67.6
69.2
45.8
53.1
60.1
57.1
60.0
62.4
61.8
45.6
61.9
66.1
57.9
69.8
63.6
72.2
71.1
71.9
Average
Score
on 1st
Component
-2.094
.116
-2.073
-1.8C4
.126
-1.283
-1.409
-2.927
-2.516
-2.239
9.122
4.454
.602
2.014
.968
.791
.802
4.238
2,083
.058
5.601
-2.597
-1.293
-2.269
-1.685
-3.290
Median
WOT.
75.9
70.1
72.3
69.6
66.5
71.6
75,9
76.9
72.9
73.8
45.3
58.9
62.8
61.1
62.8
64.1
64.6
47.1
65.0
64.5
57.6
74.6
65.4
75.1
74.0
73.5
Score
on 1st
Component
-2.161
.216
-2.007
-2.077
.544
.895
-1.294
-2.872
-2.890
-1.197
10.208
3.908
.390
.368
.517
.394
.505
4.496
1.404
.518
6.516
-2.535
.512
-2.196
-1.721
-3.271
2nJ Worsr
«21
72.8
64.8
63.9
63.8
55.4
64.1
66.5
65.8
53.8
57.6
52.2
51.7
54,6
56.7
48.0
55.6
54.6
47,0
56.7
60.0
45.5
58.5
50.5
56.4
61.8
65.1
Score
on 1st
Component
-1.748
.561
-2.080
-1.924
.277
-1.216
.987
-2.106
-2.882
-1.765
6.448
3.957
.953
1.683
1.209
-1.289
-1.062
3.204
4.542
-1.177
6.359
-2.351
- .979
-2.395
-1.228
-2.883
J
Ugl,
73.0
61.1
76.4
67.1
64.0
71.5
77.6
75.1
66.2
67.1
38.0
48.6
59.0
50.6
56.3
55.0
59.6
45.5
60.1
63.0
50.4
72.4
57.1
72.1
65.3
69.2
uly
Score
on 1st
Comnoncn C
-1.397
.070
-1.086
-2.070
-1.492
.950
-1.234
-1.938
-2.632
-1.392
9.825
2.344
1.027
.704
.246
-1.577
- .448
2.648
1.721
.068
5.699
-1.695
-1.534
-1.766
- .972
-1.790
See Table 8 for stream names and locations of water quality monltorlnj stations.
                            173

-------
                                                                  Table   39
                             SOt-LAXY C: REC:;«SIO>i .NKS'.'LIS USTN'C ji'iY y; .>'...-..L-.-.-TS I'OS IXIiKIId'Al. CIID'.ICAL VARIAM.£S


                                                        Co^ff ici.-nr:. f-r  1--:, •>v"di'Tir V.'riaMes"1'-
L.r.=.-.----r: Vrsrlibl*
1
2
3
4
5
6
7
6
5
10
11
n
13
:4
15
16
17
IS
19
23
21
22
23
24

25
25
27
23

29
30
31
*-. rc-;-.~s rollucion
1 -~;"h ?:-l!ucio-.
:r--- C'H -"^llcilcn
t-r.'i viriL: too polluted
/ :zr -.: ^ivcntss Ir.dex
C^C-' :' ^r 7.shln~
W:'d lr 7
Picr.icScir.5
Sitilng
C.^s
•A) king
BizJ w^cchin^
Sk:-ci:i;
Other
Er.;ie-e i;: TtsMng
Wading
Picnlcklns
Sitting
Ceces
Walking
Eird uatchlr.i
Skating
Othir
Tcl»I Acriviclas Enr.£3e8
.118
.257
.465
.292
.555
.020
4.250

.209

.934

.762
-.043

2.221
.453 1.963
.004
.630
.001
,075 --2i!
.355 -.002 .0:;
.036 .O!'.
.038 -.019
-.24', .036 .110 -.019

.045 .098 -.005
-.C01
-.243 .Oil -.035 .004

.078 -.001 .185 -.034
-.151 .047 .00£
T.001
-.2V5 .051 .104 .OH
-.074
.050 -.004
.001
-.154 .062_ -.003
.002
-.9f.4 .369 -3.217

.1^7 -.002

.251 -.355

-.009
.002

-.037 "°03
5 - 303    All listed coefficients si«nlfic«nc «t the 0.05 level.   Undi-.rllr.ed coefficients si-nificant at the 0.01  level.

           * In addition to the independent  variables  listed, the* folJo*-5i^ were tested for each dependent  variable end  fcund not significant:
             Dissolved oxygen (percent saturated),  total  alkalinity, phosphates as PO., turbidity,  loss  on  ignition.
                                                          174

-------
     a.  (j = 1  ... 16) are the regression coefficients corres-
      *J
ponding  to the  16 separate chemical indicators of water quality.

Distance to stream, measured as the shortest straight-line dis-
tance, was put  in logarithmic form so that a given increment
in distance would have less effect for residents far away from
the stream than for residents close to the stream.  Size of
drainage area of the stream, which has direct implications
for the  size and character of the stream itself, was also
expressed in logarithmic form, since an  increase in drainage
area from, say, one to two square miles would have a much
greater  effect  on perceptions than an increase from 11 to 12
square miles.

The results of  this regression analysis are given in Table 39.
They indicate that the higher water quality actually is:  the
less the stream appears to the residents to be polluted, the
better opinion  they have of its general attractiveness, the
more suitable they consider it for nearly all activities listed,
and the  more they engage in those activities.  The respondents'
desire to live  closer to the stream and their belief that their
own houses and  houses right next to the stream are worth more
because  of the  stream, also tend to be stronger if the stream's
water is cleaner.

In most  of the  equations of Table 39, number of children is
significant, and in many of them the log of distance between
stream and residence is significant.  Generally, at least
one chemical variable is significant in each equation, with a
wide variety of chemicals appearing in the various equations.
The summary Water Quality Index is significant in 5 equations--
in more  equations than any single chemical.  However, the
Water Quality Index falls short of acting as a general indicator
of water quality which provides explanation of most of the
variables concerning perception and use of streams.  The
potential value of such a general indicator in summarizing
and communicating results has already been mentioned.

Summary  Indicators of Chemical Water Quality
In an attempt to derive a suitable summary measure, the prin-
cipal components were computed for the set of 16 chemical
indicators.  The object of principal component analysis is
to identify the major underlying characteristics of a given

                           175

-------
phenomenon by determining the common relationships among
the variables available to describe it [Rummel, 1967].  Thus,
the underlying patterns among the 16 chemical indicators
were deduced and identified as a number of basic character-
istics or components.

For this analysis, the annual average, median, second worst,
and July chemical measurements were analyzed separately.  The
fact that many of the chemical indicators "move together" is
indicated by the fact that the first principal component
accounted for roughly half of the variance of all the chemical
indicators (Table 40).  The first four components taken
together account for about 80% of the variance, and the first
five taken together account for about 90%.

The loadings, listed on Table 40, indicate that nearly all
chemical variables are correlated moderately strongly with
the first principal component.  Only pH, nitrates, and tur-
bidity—and to a lesser extent color and alkalinity--are not
represented strongly on this component.

Because the first principal component accounts for such a
major proportion of the variance in the individual chemical
variables, it was considered suitable for use as a summary
variable measuring water quality.  Scores for each of the
stream sites on the first component are given in Table 38.
The streams with the highest scores have the worst pollution
(or the lowest water quality).  It must be kept in mind, in
considering these scores, that the first principal component
involves a weighting of chemicals only on the basis of the
extent to which each chemical is correlated with the others,
not on the basis of which chemicals are judged to be most
critical to "water quality."  The collection of component
scores must be viewed in somewhat the same way as one would
view a simple average of (standardized)chemical concentrations,
in that both are free of subjective factors.

Correlations between these scores and the variables expressing
perception, usefulness, and use are given in Table 41.  It
will be seen that although the correlations with the Water
Quality Index are higher for some variables than the correla-
tions with the first component, a substantially greater numb,er
of significant correlations occur with the first component.

                          176

-------
                    Table 40
FACTOR LOADINGS ON FIRST COMPONENT OF POLLUTION
Coliform Bacteria
Conductivity at 25°C
Dissolved Oxygen
(% sat.)
co2
Alkalinity
PH
Nitrates
Nitrites
Phosphates
Chlorides
Turbidity
'Hardness
Color
Total Dissolved Solids
Loss on Ignition
S04
% total Variance
explained
by first component
by components 1-4
by components 1-5
Year
Average
.246
.317
-.263
.270
.288
.110
.127
.299
.133
.303
.032
.271
.254
.322
.261
.276

.57
84
90
Median
.267
.309
-.268
.244
.291
-.059
.034
.265
.305
.301
.151
.265
.265
.290
.210
.261

.61
86
90
Second
Worst
.150
.347
-.132
.266
.182
-.186
.094
.299
.231
.342
-.057
.326
.122
.352
.307
.299

.47
80
87
July
.238
.328
-.208
.322
.051
-.107
.038
.298
.341
.346
-.008
.255
.053
.344
.248
.319

.47
77
83
                       177

-------
                                                            Table  41

                     SIGNIFICANT CORRELATIONS BETWEEN SUMMARY MEASURES OF WATER QUALITY AND VARIABLES EXPRESSING
                                           PERCEPriON, USEFULNESS AND USE OF STREAM SITES
                                                Scores on First Component: of Pollution
00
1   Perceived Pollution
2   Trash Pollution
3   Sum of Pollution
4   Don't visit: too polluted
5   Attractiveness Index
6   Good for Fishing
7            Wading
8            Picnicking
9            Sitting
10           Games
II           Walking
1.2           Bird Watching
13           Skating
14           Other
15  Engage in Fishing
16            Wading
17            Picnicking
18            Sitting
19            Games
20            Walking
21            Bird Watching
22            Skating
23            Other
24  Total Activities Engaged in
25  Want to live Closer
26    "  "   "   Farther
27  House value morn
28  Houses near stveam worth more
Year
Average
.29_
•24.

-.39
-.28
-.24
-.25
-.27.
-.22
-.18
-.16.
-.21
-.19.

-.26_
-.18
-.13

Median
.28.
.27
.15
-.39
-.29
-.24
-.26
-.28
-.25
-.23"
.15
-.18
-.17
.14
-.28
-.20
-.18"
Second
Worst
.27.
.25
.14
-.3J.
-.27
-.26
-.28
-.22.
-.21
-.20
.13
-.18
-,2J)

•30.
.15
-.16

July
•2£
.27
.14
-.33.
-.28
-.24
-.27
-•2JL
-.22
-.21
.13
-.17
-.18
-.12
-•29.
-.16
-.17
                                                     -.25
                                                      .19.
                                                      .20
-.26
-•I!
-.24
-•II
-.18
        -.25
-•22.
-.20
                                   Water Quality Index

                             Year               Second
                            Average   Modian    Worst
                             -.23      -.25      -.24
                                                                                           -.26.

                                                                                            .24.
                                                                                            .29
                                                                                            .29.
                                                                                            .23
                                                                                            .13
                                                                                            .14
                                                                                            •24:

                                                                                            .27
.11


•20.


.16
                                       -•24.

                                        .27.
                                        .26_
                                        .2_5
                                        .23.
                                        .18.

                                        .13
                                        .12
                                        .22_

                                        .25
                       .18
.14
                                        -,26_

                                         .17.
                                         .24.
                                         •28.
                                         .16.
                                         .15
                                         .17
          .13
                                        July
                                        -•23.

                                        -•21

                                          .24_

                                          •29
                                          .30
                                          •21
                                          .20
                                          .13
                                          •I5.

                                          .26
.11


.2JL



.14
              All correlations listed are significant at the 0.05 level; underlined correlations are significant at the
                0.01 level.

-------
This is especially true for the variables describing what
activities the stream is considered good for (variables
6-14), and what activities are actually engaged in (vari-
ables 15-24).  The correlation results using the first
component are also more satisfying than those reported in
Table 39 in that they indicate that the more polluted a
stream is, the less residents think that houses bordering
on it would be worth more because of their closeness to
the stream.

(Note:  As can be seen from Table 38,  a few streams have
extremely high scores on the first component.  In order to
eliminate the possibility that these scores would dominate
the correlations, all scores larger than 3.0 were assumed
to be 3.0 and a second set of correlations were run.  The
correlations of this set were generally slightly lower
than those presented in Table 41, which is based on the
factor scores as given in Table 40.)

Of the four sets of data used for computing the first
component, the set comprised of year-average measurements
provides the highest levels of correlation with the depen-
dent variables.  Therefore, the correlations of the first
component based on the year averages are used to derive
curves expressing how pollution level is related to each
of the dependent variables of the average individual in
the sample.  (The average individual is one who has 1.53
children and lives a distance of 813 feet from the stream,
which has a drainage area of 18.2 square miles.)

Curves relating to use of stream site are presented in
Figure 13.  Curves relating to perception of pollution,
attractiveness, and effect on house value are given in
Figure 14.

The probability of using a stream site falls with increase in
water pollution for nearly all activities.  Thus the probabil-
ity of wading and fishing, for which the relevance of water
pollution is direct, and of walking, sitting, bird watching,
and picnicking, for which water pollution is relevant through
its effects on aesthetics, all fall at a roughly similar rate
with increasing water pollution.  The two activities which
do not fall with increasing water pollution are ice skating
and playing ball or other games.  One would not expect water
pollution to have much effect on these activities; instead

                           179

-------
          Figure 13

EFFECT OF WATER POLLUTION ON
     USE  OF  STREAM SITES
                                        To
     Water Pollution Component
          180

-------
their major determinant is the availability of the necessary
physical facilities — smooth and solid ice for skating, and
the appropriate playing field for ball and other games.

Water pollution as perceived by the residents is consistent
with water pollution as measured by chemical analysis (first
two graphs of Figure 14).  On average, the residents consid-
ered the streams in our sample with the worst water quality to
be "somewhat polluted."  This perception can be compared
roughly with measurements using the Water Quality Index, which
ranges from low water quality (or very polluted); 0,to high
water quality (not at all polluted), 100.  The worst streams
in our sample had Water Quality Indexes of 45-55, which
would appear to fall in the category of somewhat polluted,
although the developers of the Water Quality Index do not
interpret their numerical indices in qualitative terms.  The
best streams in our sample had Water Quality Indices of around
75, and were judged to be "a little polluted" by the respon-
dents .

The attractiveness rating given to the stretch of stream is
related to the level of water pollution, the most badly
polluted streams being judged on average to be somewhat unat-
tractive and the least polluted to be somewhat attractive.
There is a corresponding result with regard to the perceived
value of real estate.  Houses near streams with good water
quality are perceived as being worth between "a little more"
and "much more" because of their closeness to the stream.  On
the other hand, even houses bordering the most polluted
streams in the sample are considered to be worth "a little
more" because of their closeness to the stream.  People are
probably somewhat more reluctant to make statements about the
value of their own houses than about the value of other
people's houses.  Thus, the probability of believing~that one's
house is worth more because of the closeness of the stream is
only 0.2 for the cleanest streams in the sample, and 0.025
for the dirtiest.  However, the relationship is clear:  higher
water quality is perceived as having a positive effect on
house' value.
CONCLUSIONS
This study has made it possible to fit specific curves to the
relationships hypothesized generally in Figure 10.  In general,

                         181

-------
                                       Figure  14
                          EFFECT OF WATER POLLUTION ON
                    PERCEPTION OF STREAM  AND HOUSE VALUE
No.  of Items
  Mentioned
"Very I ,-luted"
"Somewhat Polluted"
"A Little Polluted"   1 (±^r7-L_._'....
"Not  at All" or
"Don't Know"
"Very Attractive"
"Somewhat Attractive"
Neither/Nor,  or
"Don't Know"
"Somewhat Unattrac-
 tive"
"Very Unattractive"  -
"A Little More"
Neither More nor Less  0	
"A Little Less"
Probability of Be-
lieving House Worth
More
           utLoit_'I.tems_Men.tione'd '
•3. .-Number, of....P.O.]
.-Polluted-,is_£he..StrearaZ!
                     ... How .Attractive .do iyou "Consider .'this .Stre'Ect
                      ::Do :Houses  Near Streams  Tend :to Be Worth::.U
                      "More..or 'Less JDue to  the -.Stream? ;
                                                         to_PreaeiiC£
                                         2.         "4          6
                                       WATER  POlXinTON COMPONENT
                                               182

-------
                Figure  15

     EFFECT OF WATER POLLUTION AND
 DISTANCE TO RESIDENCE ON  PER CENT OF
       HOUSEHOLDS WHICH USE  STREAM
            FOR TAKING A WALK
1000
2000     3000
     Distance (fee
40QO
t)
5000
                 183

-------
the empirical work has borne out the hypothesized relation-
ships :  actual use and perceived suitability for use of a
stream, and perceived land values near the stream, are
greater if the stream water is cleaner, and greater for
nearby residents than for those living at greater distances.

In Figure 15, two of these relationships have been brought
together in a single graph which shows how the probability
that a resident will use a stream site for taking a walk
varies with the distance of his home from the stream and with
the water quality of the stream.  (This graph--and the graph
of Figure 16--was prepared by combining the information on
the relationship between use and distance from Table 36 and
the information on the relationship between use and water
pollution from Figure 13.)

Take, for example, a resident who lives 1,000 feet from a
stream whose score on the first component of pollution is 6.
The probability is 36% that he actually uses the stream site
to take a walk.  Consider what would happen if a pollution
abatement program were undertaken with the result that the
score of the stream on the first component of pollution was
lowered to 2.  Under those conditions the probability that
the typical resident at 1,000 feet would walk at the stream
site would be raised to 47%.  This is an increase of 11%,
which results in about one-third more use than would have
been enjoyed before the abatement program.

The increase in probability that the total resident population
would use the stream site for walking is also indicated by
Figure 15, on the assumption that the density of settlement
does not vary with distance from the stream.  In that case,
the total increase in probability is indicated by the shaded
area on the figure.  This increase may be related to the
total probability before the change in water quality, which
is given by the area beneath curve WPG 6.

A second approach to increasing the benefits derived from'the
stream would be to increase the density of settlement near
the stream.   The effect of such a change in density of settle-
ment (and any variation in density) on the total probability
of using the stream for walking could be determined by weight-
ing the curve by the density found at each distance.

                           184

-------
                Figure 16

EFFECT  OF WATER POLLUTION AND DISTANCE
TO RESIDENCE ON PERCENT OF HOUSEHOLDS
 WHICH  ENGAGE IN A TYPICAL ACTIVITY
(SUCH AS  SITTING OR WADING) AT STREAM
                                       WPC  -T_
                                      -, WPC! - 0-4
                                      -.WPC:.. 2-^.
                                      -WPC! 4"r
                                      .:.WPCl.:-6-^r
   1000
2000
 3000
Distance
  4000
(feet)
5000
                   185

-------
The effects of combinations of changes in water quality and
density of settlement can be explored using Figure 15 for
walking or Figure 16, which is more typical of all of the
activities given earlier in Figure 11.
                            186

-------
VII.  EVALUATION OF EFFECTS:  CAPITALIZATION OF EFFECTS OF
           ENVIRONMENTAL PRESERVATION IN LAND VALUES
The previous chapter was concerned with the evaluation of the
range of fiscal and non-fiscal benefits which result from
stream valley preservation.   Nearly all of these effects, and
especially the non-fiscal or environmental benefits, can be
expected to be registered in higher land values.  This in-
crease in value is enjoyed directly by landowners, who them-
selves constitute a significant proportion of the public, and
secondly by the general public through higher property tax
revenues.

THE EXPRESSION OF ENVIRONMENTAL BENEFITS IN LAND VALUES
It is worthwhile to consider at a general level the manner in
which environmental benefits may be translated into land
values.  Basically, differential land values are created as
a result of the importance of distance to an environmental
amenity or other facility.  Persons residing near the area
in question are more easily able to derive benefits from its
existence than persons living farther away.  Presumably, there
is some distance beyond which persons derive virtually no
benefit at all.  Assuming that persons recognize the value of
living close to the amenity, the price or rent of land close
to the amenity should be bid up above what it would otherwise
be.  The amount that it is bid up comprises a location rent
similar to the rent on a larger scale associated with proximity
to a large city.

At a theoretical level, a case can be made that most or all
people-oriented value-related effects created by an environ-
mental amenity will be expressed in terms of "location rent."
Given that the critical factor is distance, and that distance
is determined by residential location utilizing some minimum
amount of land, then market forces should cause land value at
each point to be bid up by an amount just equal to the envir-
onmental benefits obtained at that point.  This location rent
would be measurable as the difference, all else being equal,
between the land price or rent at a given point where some
                           187

-------
benefits can be enjoyed and its price at distances greater
than that at which any benefits at all can be enjoyed.

This analysis assumes that environmental features which yield
benefits such as discussed here are not a free-good in an
economic sense, that is, they are not ubiquitous and are not
perceived as ubiquitous.  The scarcity of environmental
amenities is becoming increasingly obvious, at least in urban-
ized areas.
                                          *
In order for a location rent gradient to be created in assoc-
iation with some amenity, the following must be true
of at least a portion of persons in the market for housing:

1)  Persons considering a residential site near some amenity
    must be able to imagine what use and enjoyment they would
    derive from the amenity by living there.

2)  Persons must be willing to attach, at least implicitly, a
    money value to this use and enjoyment.  This implies that
    persons recognize that environmental amenities are in
    scarce supply.

Two research efforts concerning land value effects will be
reported in this chapter.  The first is concerned with devising
a statistical  model which would make it possible to estimate
the market value of a parcel of land given information on its
characteristics, the characteristics of its neighborhood, and
its distance from certain amenities and population centers.
The second effort is concerned with estimating the effects on
nearby land values of a large urban park.
THE DETERMINANTS OF LAND VALUES
The effects of environmental characteristics, and their preser-
vation, upon land values are not immediately evident, since
many factors affect land value.  Therefore, if these effects
are to be identified, it is necessary to design a statistical
model which includes as variables all the major factors which
affect land prices.  Research was carried out for the purpose
of constructing such a model [Coughlin and Fritz, 1971].   This
research, which is summarized here, has the ultimate purpose
of providing a tool for estimating the effect of a conservation
easement program on land values.  With this ultimate purpose
in mind, data were gathered on sales of unimproved parcels

                           188

-------
located in the watershed of the Upper East Branch of the Brandy-
wine Creek in Chester County, Pennsylvania.  Extensive plans
involving the purchase of conservation easements were developed
for this watershed in 1967-1968 [Institute for Environmental
Studies, et a_l. ] , but were not carried out [Keene and Strong,
1970; Leopold, 1970; Thompson, 1969].  Therefore, the ultimate
purpose of measuring the effect of conservation easements on
land values could not be achieved through the study.

Research Procedure
The 37-square-mile watershed of the Upper East Branch of the
Brandywine, from which data on sales of unimproved parcels
were gathered, lies about 35 miles from the center of Phila-
delphia, and about 25 miles from Wilmington.   Thus, it is
beyond the area of present suburban expansion and only now is
beginning to experience the growth of an exurban population.
In 1950, the watershed had about 67 persons per square mile;
by 1966 the population density had grown to only 100 persons
per square mile.

Unimproved (that is, undeveloped) parcels were studied in order
to avoid the difficult problem of estimating and adjusting for
the value of buildings and other structures which are found on
improved or developed land.

Of the total number of sales recorded, many could not be in-
cluded in the analysis because they obviously did not reflect
market prices.  A limited number of other transactions had to
be eliminated from the analysis because a complete set of
data was lacking (e.g., price, size of parcel, or location of
parcel was not available).  The resulting sample  which was
analyzed consisted of 106 transactions.

The variables measured for each of the parcels are identified  .
in Table 42 and their means and standard deviations are listed
in Table 43.   The dependent variable is price per acre.  The
independent variables include those which describe the charac-
teristics of the parcel itself (its size, its limitations for
sewage disposal, its use, its slope, and whether or not it
includes a stream), characteristics of the parcel and its
surrounding area (residential density of neighborhood and
roughness of land), the accessibility of the parcel to urban
activities and services and to natural amenities (distance
                            189

-------
                            Table  42

  VARIABLES USED  IN ANALYSIS OF SALES OF UNDEVELOPED PARCELS



DEPENDENT VARIABLE
X_   price per acre  (data  from Chester County Assessor's Office)

INDEPENDENT  VARIABLES
Variables Measuring  Characteristics of Parcel

X~   size of parcel  (acres)

X-   percent of  land having  only slight or no limitations for
        on-site sewage disposal

X,   percent of  land having  moderate limitations for on-site
        sewage disposal

Xc   percent of  land having  severe limitations for on-site sewage
        disposal

X,   soils limitation index  2(XJ + X^


          (Data  for X~-X6  from U.S. Soil Conservation Service,

           Soil  Survey-Chester and Delaware Counties, Pennsyl-

           vania, 1963).


X    percent of  land in Flood Plain District (measured from
        map 27, Plan and Program for the Brandywine).

     percent of  land in Stream Buffer District (measured from
        map 27, Plan and Program for the Brandywine).

X    percent of  land in Woods and Slopes District (measured from
 ^      map 27, Plan and Program for the Brandywine).

Xin  percent of  land in forestry and pasture (measured from map
 10     prepared  by RSRI staff)
                              190

-------
                        Table 42- Cont.

   VARIABLES USED IN ANALYSIS OF SALES OF UNDEVELOPED PARCELS
X    percent of land in farming (measured from map prepared by
 A     RSRI staff)

X,2  percent of parcel wooded (measured from U.S. Geological
       Survey Map).

X-o  average slope of land in parcel (horizontal distance for
       20' change in elevation;  measured from U.S. Geological
       Survey Map).

X,,  feet of road frontage per acre (total number of front feet
       divided by acreage)

X,,.  existence of stream on parcel (0, 1)

X ,  existence of road frontage on parcel (0, 1)

X_ _  residential density of neighborhood (number of dwellings
       in 1 square mile square centered on parcel)

XIR  maximum slope of land in immediate neighborhood (larger
       of N-S or E-W slope over 2600 feet centered on parcel)

X, Q  roughness of land (standard deviation of 25 elevation
       measurements equally spaced over 1 mile square centered
       on parcel)

     roughness of land (approximation of surface area of 1 square
       mile of land centered on parcel).
Variables Measuring Accessibility of Parcel

     distance to King of Prussia industrial and shopping centers
       (miles)

     accessibility to manufacturing employment in surrounding
       locations weighted by distance squared:
                           191

-------
                        Table 42- Cont.

   VARIABLES USED IN ANALYSIS OF SALES OF UNDEVELOPED PARCELS
          E.         E. = manufacturing employment at location
         — 2              J in year of land sale, t.  (Locations
      j   cL.             are Coatesville, Downingtown, Honey
                          Brook, Potts town, Phoenixville)


                    d.. = distance along highways from parcel
                      J   i to town location j

     accessibility  to manufacturing employment in surrounding
       locations weighted by distance:
X«,  distance to nearest elementary or secondary school (miles)

X2c  distance to nearest perennial stream

X2fi  distance to nearest major road (route 322 in all cases).


Variable Expressing Characteristic of Township

X27  Property tax rate on market value of property.  (Source:
       Table IX, Local Government Financial Statistics, De-
       partment of Internal Affairs, Commonwealth of Pennsyl-
       vania) .


Variable Expressing Date of Sale

X    year of sale (1962 = 1, 1963 =2, ... 1968 = 7).
                           192

-------
                                      Table 43
                  MEANS AND STANDARD  DEVIATIONS  OF VARIABLES
                          Parcels under 3 Acres
Parcels of 3 Acres or More
Ml Parcels
•>o
Variable
Dependent
Xl
& Units
Variable
$/acre
Mean

1,287.40
S.D.

740.33
Mean

628.04
S^D.

416.43
Mean

1,019192
S.D.

706.26
Independent Variables
X2
x3
•X4
X5
X6
X7
X8
x9
X10
Xll
X12
X13
X14
X15
X16
X17
X18
X19
X20
X,.
21
X22
X23
X24
X25
X26
X27
X28
acres
7.
7.
T-
index
7.
7.
7.
*
7.
7.
7,
ft/ac.
0,1
0,1
DU/sq.mi
7.
ft.
ft.
mi.
,2
empl x d
empl x d*
mi.
ft.
ft.
DillS
1....7
1.28
14.29
47,94
37.78
123.49
1.59
17.30
22.22
37.30
31.83
29.06
0.748
160.97
0.048
0.778
46.65
3.50
53.50
52,050.27
27.10

3,202.64
370.23
2.75
1,119.10
8,011.14
1.041
5.86
0.72
34.11
45.47
42.89
62.76
12.60
35.11
40.25
48.33
45.83
44.4*8
5.79
193.57
0.215
0.419
30.09
2.02
22.56
46.75
2.82

516.90
236.76
1.45
769.27
5,761,16
0.655
1.50
17.77
6.16
43.58
48.05
139.67
14.09
18.77
20.47
55.12
44.65
37.44
9.40
45.09
0.302
0.744
23.05
4.14
69.47
52,082.98
29.54

2,818.13
249.28
2.99
975.84
13,639.60
1.037
5.77
19.28
18.02
35.53
34.28
44.25
30.43
26.26
35.99
46.39
46.13
41.25
5.37
42.62
0.465
0.441
13.36
2.68
31.47
65.39
4.12

421.03
109.13
1,53
971,37
7,867.59
0.270
1.76
7.97
10.99
46.17
41.94
130.06
6.66
17.90
21.51
44.53
37.03
32.46
8.26
113.96
0.151
0.764
37.075
3.76
59.98
52,063.40
28.09

3,046.65
321.16
2.85
1,060.93
10,294.39
1.040
5.82
14.67
28.86
41.60
39.77
56.32
22.41
31.69
38.41
48.14
46.17
43.20
5.68
161.62
0.360
0.427
27.235
2.32
27.54
57.16
3.59

' 514.44
203.53
1.48
858.21
7,215.86
0.531
1.60

-------
along radial expressway to King of Prussia industrial area and
shopping center, distance to nearest elementary school, gener-
alized accessibility to manufacturing employment, and distance
to nearest stream), and a variable characterizing the township
in which each parcel is located (tax rate on market value).
The final variable is year of sale.

Availability of public water and sewer does not appear because
virtually none of the land in the watershed is yet supplied
wi.th these services.  Zoning classification of the land is not
included either, since zoning ordinances were not passed in
any of the townships until the latter part of the period for
which sales data were collected.  In 1968, three out of six
townships still did not have zoning ordinances.  Therefore, it
was judged that zoning probably was not an important consider-
ation for most sales in the sample.

As can be seen in Table 42, the characteristics are rather
general ones and refer to topography, vegetative cover, and
land use.  In addition to scenic amenity, these characteristics
also relate to ease of development.  Unfortunately, many of
the characteristics implying enhanced scenic value (which
shbuld enhance sales price) also imply difficulty of develop-
ment (which should reduce sales price).  For example, a parcel
which consists largely of flood plain or woods, or has steep
slopes, is likely to be quite scenic, but it is also likely
to be difficult and expensive to develop.  Therefore, it is
difficult to predict how the market will price such a parcel.
As long as major subdividers are not operating in the area,
it is possible that the parcel will have a relatively high
price because of its scenic quality.  But as soon as cost-
conscious developers appear, other less scenic but more readily
buildable land may bring a relatively higher price.

General Results
Earlier analysis [IES, RSRI, and USGS,  1968],  as well as
discussion with appraisers and general knowledge of the real
estate market, has indicated that the market for building
lots is distinct from that for large tracts.  In general, it
Is believed that the price per acre for land in large tracts
is substantially lower than for land in small building lots.
Therefore, before beginning the analysis, the sample of trans-
actions was divided into those for parcels of less than 3
acres and those for parcels of 3 acres or more.

                           194

-------
                                         Table 44

                             SUMMARY  OF  REGRESSION  RESULTS
                    (Dependent Variable:   Sale Price  Per Acre)
                          Parcels under 3 Acres
                          Parcels of 3 Acres or More
                                                                                     All Parcels

t-1
vO
Ln


Multiple R .668
Scd. error of estimate 564.76
No. of observations 63
Constant Term 834.54
Coef- t-
Independent Variables ficient Statistic
Characteristics of Parcel
2. Parcel size
3. % slight limitations
for on-site sewage
12. % parcel wooded -5.04 -3.1
14. road frontage per acre
15. stream on property -793.18 -2.3
Characteristics of
Neighborhood
.709 .751
308.93 477.66
43 106
1,884.92 49;
Coef- t- Coef-
ficient Statistic ficient
-6.46
3.95
-3.05 -2.0 -3.19
0.82


7.06
t-
Statistic
-1.9
2.3
-2.7
2.6


17. residential density
19. roughness of land

Accessibility of Parcel
21. dist. to King of
    Prussia
26. dist. to nearest stream
13.66
5.6
                           -5.89
                          -28.00
                           0.10
                            -2.4
                            -2.1
                             1.4
                                                       14.59
                                                         7.57

-------
Stepwise regression analys'.s was carried out both for the
subsets of small and large parcels considered individually,
and also for the entire set of transactions.  The results
are summarized in Table 44.  The equations and their indiv-
idual component variables were generally significant at the
0.05 level.  All variables appear with the expected sign,
with the possible exception of those variables relating to
streams (variable 15 for small parcels and variable 25 for
large parcels).  Given the view that a stream is an amenity,
one might hypothesize that the presence of a stream would
add to property value.  However, for both large and small
properties, accessibility to a stream was found to be assoc-
iated with lower land values.  The reason for this is under-
standable for small properties:  the existence of a stream
on a property of less than 3 acres implies that at least a
significant proportion, if not all, of the property is in
flood plain, marsh, or is steeply sloped.  It is less clear
why the price of large parcels should be affected adversely
by stream accessibility, since even if a stream crossed the
parcel, most of the land in the parcel would remain buildable.
At any rate, the effect for large parcels is weak and the
coefficient fails to attain statistical significance.  Neither
"stream'Variable appears in the equation for the combined
sample of large and small parcels combined.

The results for large and small properties indicate that value
probably is determined differently for each group.  Only one
variable--percent wooded--appears consistently and significantly
for each sub-group and for the entire sample.  The stream
variables appear consistently, if not always significantly,
for the two sub-groups.  In addition to these variables, the
price per acre of small parcels is influenced by density of
residential development in the neighborhood.  Increased den-
sity of residential development in the neighborhood implies
the existence of certain urban services in addition to proximity
of neighbors--which is especially important to families new to
the area desiring playmates for their children.  Therefore,
although it is hypothesized that many people move into the
Upper East Branch Brandywine area because they are attracted
to the natural countryside, the availability of neighbors is
also a consideration, especially for those who buy relatively
small lots, which are more characteristic of suburban develop-
ment than of rural development.

                            196

-------
 For large parcels,  a different  set  of  variables  came  into  the
 regression equation:  distance  to King of  Prussia  commercial
 and industrial area (which is a proxy  for  distance to the
 center of the metropolitan area) and topographic roughness
 of the neighborhood.  Both of these variables can  be  inter-
 preted as measures  of the development  potential  of the land,
 and their presence  in the regression equation suggests that
 large tracts  are being bought with  an  eye  to their eventual
 development.   An alternative interpretation of topographic
 roughness is  that it is a measure of amenity, rolling or
 hilly land being more desirable than flat  land.  However,  to
 be consistent with  such an explanation,  topographic roughness
 would have to appear with a positive sign  in the equation.

 Although the  prices of large and small parcels appear to be
 determined by different sets of variables, the best overall
 statistical explanation is achieved when parcels of all sizes
 are combined.   Multiple R increases to .75 at the  same time
 that the sample size is approximately  doubled, resulting in
 a substantial increase in level of  significance.   In  the equa-
 tion for the  combined sample, parcel size  appears  as  an inde-
 pendent variable, reflecting the different market  forces
 acting on large and small parcels.  Residential  density, which
 appeared in the equation for small  parcels, also appears in
 the equation  for all parcels.   In addition, frontage  per acre
 and percent of land with slight limitations for  on-site sewage
 appear in the combined equation.  These variables  would have
 been the next to enter in the stepwise regression  analysis
 of small parcels.   However, missing from the combined equations
 are:  topographic roughness, distance  to King of Prussia,  and
 distance to nearest stream, all of  which appeared  in  the
 equation for  large  properties,  and  existence of  a  stream on
 the parcel, which appeared for  small parcels.

 In that these variables do not  appear  in the combined equation,
 it is less appealing than the equations for sub-groups.  The
 lack of distance to King of Prussia, or any other  variable
 expressing accessibility,  is particularly  disappointing.
 However, there may  be a reason  to believe  that differences
 in accessibility among locations in the Upper East Branch
 basin would have relatively little  effect  on price.   The
 basin lies 17 miles from the King of Prussia Interchange and
 between 30 and 35 miles from the center of Philadelphia, and
 is well beyond the  ring in which subdivision activity is now
-most intense.   At such distances from  major activity  centers,
 an additional few miles is likely to have  little effect on
 v
                            197

-------
perceived distance and correspondingly little effect upon land
price.  Accessibility to nearby medium-sized urban centers
(Downington, Coatesville, Honey Brook, Phoenixville and
Pottstown), which is measured in variables 22 and 23, is quite
strongly correlated with distance from King of Prussia.  It is
of interest that year of sale did not appear in the regression
equation at a significant level.  This indicates that the land
market in the basin was not affected during the period by
rapidly rising land prices, which typically precede urbaniza-
tion.  This is consistent with our interpretation of the
distance-to-King-of-Prussia variable.  If the zone of suburban
development had reached the inner edge of^ the basin, one would
expect to find a land price gradient dropping sharply with
distance from King of Prussia and distance to the center of
the metropolitan region.

The most significant independent variable explaining price per
acre is residential density in the neighborhood.  It appears
at a far higher significance level than does any other variable
in the small-parcel sample and in the complete sample.  It does
not appear, however, in the equation for  large parcels, even
though its correlation coefficient with price per acre is
relatively high (0.496).  This is evidently because residential
density has a high negative correlation with roughness of land,
which did enter the equation.

In summary, the equations for price per acre are statistically
adequate, and consistent with the market, locational, and
topographic conditions existing in the Upper East Branch
Brandywine basin.  The variables which appear in the equations
should probably be considered in any attempt to construct a
statistical model of the land market for  any comparable area
in the urban-rural fringe of a metropolitan area.  However,
in any such investigation it is likely that some of these
variables would be found to be not-significant, while other
variables, not in our equations, would be found to be signi-
ficant.  The equations derived here should be used as a guide
for studies of other areas, but it would be improper to apply
the equations as they stand to another area.

Results Specifically Concerning Environmental Attractiveness
The benefits of a land use and water resource plan, such as
the Brandywine Plan, consist in part of "scenic" benefits.
Following such a plan, certain areas would be selected in
which urban development would be prevented.  Thus, natural
or semi-natural scenery would be retained for public enjoyment,
ev'-Ti though :.he public could not actually enter on and use

                              198

-------
the land.  An approximation of the value of the "scenery"
provided by such a parcel of land is desired.  Fundamental
to an adequate approximation of the value of the "scenery"
is the relationship between price per acre and variables
which express certain aspects of environmental attractiveness.

In order to identify such relationships a number of "environ-
mental attractiveness" variables were conceptualized and
measured for the 28 properties which were sold in 1967 or
1968 and were still sufficiently undeveloped so that observa-
tions could be made in the field which would correspond, at
least roughly, to observations which might have been made at
the time of sale.

The conceptualization of the variables, which are listed in
Table 45, is based on research on environmental perception
and preference reported in Chapter V above.  That research
showed that a sample of middle-class suburban housewives
preferred areas with a parklike landscape:  generally open
areas with low ground cover and a relatively small number of
relatively large trees.  Various aspects of this preferred
landscape are measured in variables 1, 2, and 3.  The view
from a lot is generally considered to contribute to its
desirability (variables 4 and 5).  The hypothesized effect
of the number of houses or mobile homes visible from the
parcel (variable 6) is not clear, but our findings in the
previous section concern-ing residential density would indi-
cate a positive effect upon price per acre.  However, resi-
dential density within sight of a parcel (i.e., variable 6)
has only a very weak association with density within a one-
mile square (correlation coefficient of -0.097 for all 28
properties).  Therefore, it is reasonable to argue that num-
ber of houses visible and number of houses within an acces-
sible distance have different functional and perceptual
meanings for a prospective land buyer, and are not strongly
associated as measures of physical density.

The last two variables listed in Table 45 are both at a
somewhat higher level of abstraction and are measured less
objectively than the earlier variables.  The number of dif-
ferent elements or items observed by three field observers
provides a measure of the "complexity" of a given landscape.
(The resulting ratings have been adjusted for the systematic
tendency of some observers to list more things than others.)
                            199

-------
                             Table 45

                CORRELATION COEFFICIENTS BETWEEN
           AESTHETIC VARIABLES AND SALE PRICE PER ACRE
    Aesthetic Variable
1.  Average girth of trees

2.  Average height of
      ground cover (in.)

3.  Percent of area in
      open space

4.  Log of length of
      longest view (ft.)

5.  Log of average length
      of view (ft.)

6.  Number of houses or
      mobile homes visible

7.  Average number of
      items observed

8.  Average attractive-
      ness rating (3
      observers)
             -.560


             -.538


              .346


              .028


              .086


             -.046


              .293



             -.577
-.615
                                                               All
Expected   Less than   3 Acres   Trans-
  Sign      3 acres     & over    actions
          -.322


 .343     -.306
 .092
-.137     -.144


-.199     -.134


 .037     -.158


-.201      .228
 .171
-.485
Number of Transactions
               18
  10
  28
Note:  Correlations significant at the 0.05 level are underlined
                               200

-------
The attractiveness rating (variable 8) is a purely subjective
rating  (on a 1 to 5 scale) assigned by three observers and
averaged.

Comparison of the expected sign of the correlation between
each of  these variables and sale price per acre with the
signs of the actual correlations for our sample indicates
few consistencies (Table 45).  Although we hypothesized that
environmental attractiveness should bear a positive relation-
ship to  land value, the majority of the correlations indicate
a negative association.  Only average height of ground cover,
percent  of land in open space, and average number of items
observed have the expected sign both for the entire sample
and for one of the sub-samples.  Only one of these correla-
tions --that involving average height of ground cover for
small parcels—is significant at the 0.05 level.

In general, the level of significance of the coefficients is
very low.  The only coefficients significant at the 0.05
level for all transactions as a whole are those involving
average girth of trees and the average attractiveness rating,
and as was indicated above, these signs are in the opposite
direction from that hypothesized.  Because of the generally
low significance levels of the environmental attractiveness
variables, regression equations including them are not
reported here.  Suffice it to say that the only two such
variables to enter the regression equations were:  average
number of items observed in the equation for large parcels,
and average attractiveness rating in the equation for all
parcels.  Each of these appeared with a negative sign, indi-
cating that the more attractive a parcel is, the cheaper it
is.

We hesitate to draw any firm conclusions from our attempt to
define and analyze variables describing environmental attrac-
tiveness, because of the small size of the sample of transac-
tions.   Our conclusions, also, are limited by the narrow range
of variation in environmental attractiveness of the parcels
in the sample, all of which are relatively attractive.  Had
our sample included parcels of breathtaking beauty or extreme
ugliness, it is likely that the environmental variables would
have been of greater statistical significance.  However, our
results are consistent with the apparently contradictory
notion that land which is more scenic or attractive tends to

                           201

-------
be cheaper than land which is less attractive.  Specifically,
they suggest that a parcel which has large trees, or a large
percent of its area in open space, or receives a high attrac-
tiveness rating is likely to sell at a lower price than other
land.

Of these characteristics, the presence of large trees may add
to the difficulty of development.  The analysis in the earlier
part of this paper also indicated that certain characteristics
which suggest environmental attractiveness also suggest diffi-
culty or expense of development and therefore are associated
with lower prices per acre.  As examples are the variables:
percent of parcel wooded, topographic roughness of land, and
existence of a stream on the parcel.  Therefore, both parts
of this analysis of land values indicate that environmental
attractiveness, as we have defined it based upon our related
research, has a negative association with land price.  This
association is not to be explained by saying that people do
not like attractive land but by saying that they .are more
concerned with the fact that attractive land is generally
more difficult or expensive to develop than flatter, less
wooded land which is free of watercourses.

Since most of the land in the Upper East Branch watershed is
undeveloped, and the parcels analyzed were all undeveloped,
it is not possible to draw strong conclusions from this
analysis and apply them to the question of the effect of a
preserved stream valley and its open space upon nearby
property values in urbanized areas.  Environmental benefits
conferred by such preservation are not perceived until urban-
ization begins.  As urban development proceeds, and private
open space disappears, the value of the preserved open space
becomes more and more evident.  The following study is more
appropriate to such a situation.  However, since it is con-
cerned with only one open space, it does not include variables
which describe the characteristics of open space, and there-
fore, is difficult to generalize from.
THE EFFECT OF A LARGE URBAN PARK ON REAL ESTATE VALUES
The household surveys reported in the last chapter indicate
that a use gradient can be expected around an environmental
amenity (see especially Figure 11).  Associated with the
benefit gradient one would expect to find a location rent
gradient.  (Although this study was not reported by FWQA, it

                            202

-------
is reported here because of its obvious relevance to estimating
environmental benefits.  The study was supported by the Public
Health Service under Grant No. 1 R01 EM 00519-01.)  The subject
of this study was Pennypack Park, a 13,000 acre stream valley
park in northeast Philadelphia.  The area surrounding this
park has been developed over the past twenty years—mostly with
twin houses on 1/12 acre lots.

One of the major difficulties in identifying location rent lies
in isolating of land value from the market value of property,
which includes the value of buildings, utilities, and other
structures on the land.  Because much of the development around
Pennypack Park was done at a large scale, relatively few basic
house-types are represented.  Therefore, it appeared possible
to identify each house-type by a dummy variable, which would
capture value of house and a base value of land, leaving the
value of location rent to be explained by distance from the
park.

A total of 16 house-types were identified from air photographs,
field trips, and information in the Philadelphia Real Estate
Directory.  As an example, one house-type set consisted of
houses which looked identical on the air photograph, were
semi-detached with basement garage, had two stories, were of
brick and frame construction, and were assessed at $9,200.

Such a house-type classification cannot reflect improvements
which have been made to the inside of a standard builder's
house and which are neither visible from the outside nor
accounted for in the assessed value of the house.  Obvious
outside additions, such as swimming pools, were observed
and such properties were dropped from the sample.

Dummy variables were also entered for year of sale, for
whether or not the house was on an irregularly shaped lot,
and for whether it abutted the park directly.

In order to eliminate influences other than the park, proper-
ties studied were limited to those which were closer to the
park than to any other open space or did not adjoin any
retail area or major highway.  In addition, corner lots were
excluded from the sample unless they bordered on the park,
and these were identified specifically by dummy variables
in the analysis.

                           203

-------
The major focus of interest was on variables expressing dis-
tance to the park.  Distance to the park was measured in two
alternative ways:  as the straight line from the house to the
nearest point in the park, and as distance along public ways
to the nearest point of public access to the park.  The vari-
ables expressing distance were in the form of negative expon-
ential functions of distance (to which a small constant had
been added to prevent domination by small distances).  Two
different exponents, -1 and -2, were tested;. -1 was found to
be more satisfactory.  The negative exponential gives a
gradient asymptotic to the distance axis, which was considered
a desirable property.  The dependent variable (property value)
was expressed in both log form and linear form, with similar
regression results in each form.  Only the linear form will
be discussed here.

In the regression analysis the sale price for 303 properties
was related to a set of independent variables consisting of
one or more distance variables,and numerous dummy variables.
The public access form of the distance variable proved to be
the more significant form.  The overall level of statistical
explanation was quite high (multiple R = .902) but this was
due primarily to the dummy variables.  The distance variable
was significant at the .05 level.

The magnitude of the estimated property value gradient can
be appreciated from Figure 17.  The location rent can be
calculated as the difference between the estimated property
value at a given distance and the "base" property value (i.e.,
the extrapolated value when distance from the park is infinite)
The location rent gradient as estimated drops off rapidly with
distance.  The rent ranges from about $1,240 per property at
100 feet from the park to $138 per property at 2,500 feet from
the park.  At first glance, this amount does not appear to
be very large.  It is only 7.2% of total property price at
100 feet from the park, 1.9% at 1,000 feet, and 0.9% at 2,500
feet.

Location rent, however, must be considered in relation to land
value, not to total property value.  This is clear on a theo-
retical level, since the attribute of location is intrinsic
only to land, not to the structures which may be built on it.


                              204

-------
                           Figure  17

      ESTIMATED  VALUE OF RESIDENTIAL  PROPERTIES  BY
               DISTANCE FROM PENNYPACK PARK
$18,000

 17,000

 16,051
     Location Rent per House Lot
 13,638
                   Value of House  Lot
 10,000
  5,000
       -r
       4-,
           -fri-
-I-S ,-
                   Value of House
                                           ....._!	1.
                -4-4
TIT


                      EHT
                                   -I-]_!_„ f	r_f
                                      ::_...i_i.T
                                                i _
                                              r •
                                     Tii .[_
                                       - ^—i-1   '--
                                                               !:::[•/
                                                               .. .-. j
          200  400  600  800  1000                    2000
                  Distance from Pennypack Park (feet) 	3
                                    205

-------
By subtracting a rough estimate of average house price from
total property price, land price may be identified.  This was
done by observing that for the entire sample the tax assess-
ment of the house was 83.2% of total assessment.  The value
of the average property, $16,051, multiplied by 83.2%, yields
the average house value of $13,638, which is plotted on Figure
17.  The remaining distance under the curve represents land
value, part of which is location rent.  Relative to total
land value, location rent due to the closeness of the park
is substantial.  At 100 feet it accounts for 34% of land
value; at 1,000 feet for 11%, and at 2,500 feet for 5%.  In
dollars per acre—recall that on average there are 12 houses
per acre—location rent due to the park falls from $15,000
at 100 feet to $1,700 at 2,500 feet.  These values would
appear to lie in an expected and reasonable range.

An estimate of aggregate value of location rent generated by
Pennypack Park can be made by multiplying the values per
dwelling unit found in Figure 17 by the numbers of dwellings
found at corresponding distances.  This computation yields
a net increase in aggregate real estate value of $4,400,000
which is attributable to the park.

The results of this case study are indicative of what one
might expect for a stream valley preservation project which
involved retention of major areas of open space.  However,
they cannot be applied directly to such a project, since the
estimating equations do not contain variables which explicitly
measure characteristics of the retained open space and the
developed area.  A more general estimating equation, similar
in scope to the equations developed in the previous section,
would be necessary.  Such an equation can be derived only
by studying a large sample of open spaces and streams surroun-
ded by built-up areas.
                            206

-------
QUESTIONNAIRE FOR  SURVEY RELATING TO THE EFFECT OF WATER QUALITY
                        Please answer  all  questions       Appendix  to Chapter VI
This questionnaire is concerned with  the  stretch of
Creek near your home.
l.a)  How familiar are you with this  creek?
      I know it very well
      I know it fairly well
      'I know it is there, but that's  about all I know about it
      I didn't even know it was there
  b)
3.a)
Do you consider this stretch  of creek to be generally
attractive or unattractive?
  Very attractive
  Somewhat attractive
  Neither attractive nor unattractive
  Somewhat unattractive
  Very unattractive
  No opinion
Are you usually aware of it when you drive by?
  Yes 	
  No
Is the stream and its sur-
rounding area a good place for:
Fishing                   	
Wading                    	
Picnicking                	
Sitting                   	
Playing ball, games       	
Walking                   	
Bird watching             	
Ice skating               	
Other	   	
b)
Do you or your family  actually
do any of these activities  there?
                                           (If none of the categories above
                                            are checked:)  Do you visit the
                                            stream at all?
                                                  yes  	
                                                  no   	
                                                  Why?	
                                     207

-------
 4.a)   Approximately  how many feet  from the  stream is your house, located?

                                 	 feet

   b)   How is  it  situated with respect  to  the stream?

        It  is right  next  to the stream           	*

        It  has  a view  of  the stream

        It  lies within a  five minute
         walk  from  the stream                   	

        It  is farther  than five minutes  away
         from  the stream
 5.
Would you rather  live nearer to the
(In a house like  your present one,
due to moving, etc.)  Check one.

I would rather stay where I am
  now, the house  is very nicely
  located with respect to the
  stream                        	

I would rather1 live in a house
  right next to the stream      	
•I would rather live in a house
  with a view of  the stream
       Comments
 stream or farther away -from  it?
disregarding inconveniences
                                              I would rather live with-
                                                in a five minute walk
                                                from the stream

                                              I would rather live even
                                                farther away:  the farther
                                                the better
                                              It doesn't make any
                                                difference to me.  The
                                                stream is not important
                                                to me one way or the other
       If you own your home, do you
         think the value of the house
         is affected by the presence
         of the creek?


       Value not affected by creek

       $	more because of creek

       $	less because of creek

       Present value of your house   $
                                       If you rent,  by how much do  you
                                         think the presence of  the creek
                                         increases or decreases your
                                         monthly rent from what it would
                                         be otherwise?

                                       Rent not affected by creek  	

                                       $	more because of creek

                                       $	less because of creek

                                       Present rent:  house       $ 	

                                                     apartment    $ 	
-7.     In general, do you think that houses with the stream bordering their
       back yards are worth more or less because of their closeness  to the
       stream?
                      a little                      a little
       much more        more  	      same	        less  	 much less_
                                     208

-------
 8.  Judging from your own observations, how polluted do you think the
     stream is?
       Very        somewhat      a little       not at all
     polluted	  polluted	  polluted	    polluted 	   don't know	

     If polluted:
     What arc the most noticeable signs of pollution in this particular stream?
 9.  If you had information that the stream was polluted and  could  identify
     the major polluters, what would you do?
     I would not do anything, since.I think adequate action would
       be taken by appropriate agencies                                 	
     I would not do anything since ray actions would have little effect   	
     I"would do one or more of the following:
       wri^e letters to government officials      	
       organize neighbors for legal action        	
       write letters to major polluters           	
       organize or take part in clean-ups         	
       join organizations of my neighbors         	
       picket major polluters                     	
10.  Would you be willing to pay an increase in your -.real estate  tax to help
     clean up the stream or do you think that is really not necessary?
     I am willing to pay more           	     How much more?  $	
     I would not be willing to pay more 	
       Why?_
11.  How would you rate yourself with respect to environmental problems
     and pollution?-
       very         somewhat       a little       not at all
     concerned	  concerned	  concerned	   concerned	don'tknov._
12.  How long have you lived in your present home or in the immediate
     neighborhood?
                       	years
                                       209

-------
Household  Information:
Number of  adults  in household
Number of. children in"1 household
Adults:
Sex
Age
Marital Status
 Single
 Married
 Divorced
 Widowed
.d



boys,
girls
yourself
M

F
_years
M

ages
, ages
2 -
F
vears


3 4
M F M F
years years
Education
 High School
 College
 Higher                        	years   	years   	years   	years
Occupation                     	   	   	   	
Yearly Income  (Before Taxes)  If family, put down joint income of husband
               and wife, plus contributions of other adults, in first column.
               If not a family, give individual incomes of each adult household
               member in appropriate column.
Under $ 5,000                  	        	        	        	
$ 5  -  10,000                  	        		        	
$10  -  15,000                  	        	        	        	
$15  -  20,000                  	        	        	        	
$20  -  25,000                  	        	       ...	        	
Over  $25,000                  	        	        	        	
Where did the  adult members of your household spend most of their lives
before they were 18 years of age?
in rural area                  _____        	        	        	
In suburban area               	        ___        	        	
in urban area                  	        	        i            	

      Please mail completed questionnaire in prepaid envelope to the:
                  Regional Science Research Institute
                            GPO Box 8776
                   Philadelphia, Pennsylvania  19101

                                    210

-------
                       VIII.  CONCLUSIONS
EVALUATION OF EFFECTS
The research carried out under this project has made it pos-
sible to make a rough evaluation of the effects summarized
in Table 1 in the Introduction to this report.  Such a rough
evaluation is given in Table 46, whose form resembles that
of Table 1.  The effects in question refer to a land use
plan such as the Brandywine Plan which would involve some
limitation on overall development but would pertain primar-
ily to restricting development in critical areas.

Several of the value-related effects listed in the table were
not the subject of original research, namely "ecological"
benefit, reduced flood damages, and reduced need for water
treatment.  A survey of the literature indicates that the
benefits of the plan would be rather small for water treat-
ment.  In the case of flood damages, the reduction in losses
due to lower peak flows should be moderate,in line with the
moderate reduction in peak flows resulting from the plan.
Flood damages might also be moderately or even strongly re-
duced by the plan in that development is prevented from
encroaching on areas subject to flooding.  Quantitative es-
timates of the ecological benefits are not possible but it
appears clear that greater ecological benefits would result
from better water quality and more open space than from
more stable streamflow.

One of the physical effects listed in Table 46 was not the
subject of original research in this study, namely the effect
of urbanization in reducing low or base flow of the stream.
It has been hypothesized that the rendering of land impervious
by preventing infiltration of water into the soil will cause
a decrease in base flow.  However , there is some evidence
that in typical urbanized areas this effect is compensated
for by importation of water into the watershed.  At any rate,
the effect appears to be small.

Our studies of stream channel enlargement, which is relevant
to peak flow characteristics (i.e., average annual flood),

                           211

-------
                                          Table 46
         ROUGH EVALUATION OF EFFECTS  OF A PLAN FOR STREAM VALLEY PRESERVATION
               Physical Changes
                   Potential Magnitude of Value-Related Effects
Type of Change
t stable streamflow
.gher base flow
iwer peak flows
Probable
Magnitude
of Change
small
moderate
More
Active
Use
small
small
More
Visual
En j oyment
small
small
More
Ecological
Benefit
small
small
Reduced
Flood
Damages
moderate
Reduced Need
for Water
Treatment
small
small
Better water quality
More open space
moderate
 large
moderate  moderate    large
                                  small
 large
large
large
moderate

-------
 indicates  rather  clearly  that  the  overall density of devel-
 opment  is  considerably more  important  to peak flow increase
 than the distance of  development from  the stream.  Therefore,
 the  effect of  a plan  such as the Brandywine Plan would be
 only a  moderate reduction in peak  flows relative to the
 flows which would occur with normal development.  Our studies
 of environmental  preferences do not indicate that the reduc-
tion  in  peak flows would have more  than a small effect on the
 value derived  from active use  and  visual enjoyment.

 Our  research with regard  to water  quality indicates that
 water quality  contaminants are generally positively related
 to amounts of  various types of urban development.  For devel-
 opment  involving  houses with on-site disposal of sewage the
 distance from  the stream  was found to  be a significant factor.
 These results  would indicate that  the  effect of a Brandywine-
 type plan  would be of moderate size.   With regard to the
 value-related  effects of  water quality, our studies of
 environmental  preference  have  found that persons seeing a
 stream  for the first  time are  able to  perceive water quality
 reasonably well but do not appear  to give this factor much
 weight  in  their stated preference  for  stream sites.  However,
 for  persons living near stream sites significant correlations
 have been  found between water  quality  and the amount of active
 use.  From this information we have concluded that better
 water quality, as would result from the plan, would have a
 moderate effect on active use  and  visual enjoyment, for the
 range of improvement  in water  quality  which might be expected
 to result  from the plan.

 With regard to open space, our studies have indicated that
 although somewhat less development tends to occur in stream
 valleys than elsewhere, even in the absence of strong planning
 controls,  the  restriction of development near streams, as
 envisioned in  the Brandywine Plan, would result in there
 being much more open  space than would  be preserved with typ-
 ical urbanization.  It is felt that this large open space
 effect  would have correspondingly  large effects on active
 use  and enjoyment.

 In general, the overall conclusion from Table 46 is that the
 value-related  effects associated with  open space may be as
 great or greater  than those  associated with streamflow and
 water quality  characteristics.
                           213

-------
The value-related effects listed in Table 46 may all be reflec-
ted in the value of land located near protected areas.  The
existence of such effects has been demonstrated by our study
of the area surrounding Pennypack Park in Philadelphia.  The
reflection of values in real estate is important since it
provides a way to measure value-related effects which them-
selves can be identified only through elaborate attitude
surveys.  Whether or not the real estate values generated by
a land use protection plan will exceed the land value lost
because of restriction of development must be established in
future research.
IMPLICATIONS FOR PROJECT PLANNING
The benefits listed in Table 46 should be considered potential
benefits, many of which will be fully realized only if people
who value the environment move into the planned area.

Our research results indicate that the appreciation of and
concern for the natural environment of the average person are
rather mild.  Although most persons seem to be able to identify
water pollution reasonably well, environmental quality does
not seem to make a drastic difference to their appreciation
of an area or the use which they are likely to make of it.
Furthermore, there are indications that considerations invol-
ving the natural environment may not be as important as other
considerations in choosing a residential location.  It is
possible that most people do not systematically enter envir-
onmental considerations in their calculations of residential
value, even though these considerations may actually be
important to them.  It is especially likely that the possibi-
lity of future deterioration, and the consequent benefits of
preservation are not given adequate weight.  The implications
of environmental change do not appear to be considered as
explicitly as, for example, the implications of change in the
"type of neighbors."

A conclusion indicated by our research would be that a typical
cross-section of people would not be willing to pay a large
price for the type of environment guaranteed by a Brandywine-
type plan.  There is a good possibility, however, that sub-
groups exist within the population that would pay such a cost.
This means, first, that for initial projects, areas must be
selected carefully to assure the support of existing residents.


                            214

-------
Second, there may have to be a merchandizing process by which
members of that sub-group of the population who especially
value environmental considerations are made aware of the
advantages of the plan and are attracted to the area.  Market
analyses, which focus on the specific sub-group rather than
on a general cross-section as in most of our research, will
be needed.  This is not unlike the approach to marketing
employed by most large-scale house builders.

This discussion has referred to demonstration projects which
involve only a small proportion of a metropolitan area.  The
goal is that eventually such plans might be enacted on a
much larger scale.  It should be noted that large-scale
programs differ from small-scale single projects in a signi-
ficant respect.  For a small-scale project such as the
Brandywine Plan the restriction of development in critical
areas of the watershed does not appreciably affect the over-
all supply of land in the metropolitan housing market.  There-
fore, there is no automatic tendency for land values to rise
in the unrestricted portions of the watershed because of
scarcity.  Increases in land values must be due entirely to
the environmental benefits of the plan.  For a whole metro-
politan area, however, the enaction of restrictions would
cause an effective reduction in the  overall supply of land
(i.e., the amount of land at various distances from the
metropolitan center).  Therefore there would be an automatic
tendency for land values to rise in unrestricted areas.  This
provides a virtual guarantee that property tax revenues can
be increased to pay for recompensing the owners of restricted
land.  Therefore, environmental planning is, at least in theory,
somewhat easier on a large scale.

Due to the above considerations, small-scale demonstration
projects may need proportionately more public funds than large-
scale proposals.  The fact that small-scale projects may have
to be rather heavily subsidized does not mean that larger-
scale programs will have to be.

A number of small-scale projects should be begun in the near
future.  These would be valuable in demonstrating the benefits
of environmental preservation and thereby increasing public
awareness of the possibilities of better patterns of urban
development and the methods of achieving it.


                            215

-------
ISSUES ARISING FROM THIS STUDY
Our results suggest two major areas in which future plans
and future research might be different in emphasis from the
Brandywine Plan.  The first relates to the physical effects
of urbanization on stream conditions.  The research summar-
ized here has focused on two factors:  the density of devel-
opment and the distance of development from the stream.
Density has been found to be critically important, but to a
certain extent must be taken as a given over a watershed.
Distance of development from the stream was^ found to be
considerably less important.  Therefore, although much more
should be learned about the effects of urban patterns on
streamflow and water quality, future efforts might well,
then, pay somewhat more attention to the design of develop-
ment than to density and location.  For example, in the
case of peak streamflow increase due to impervious develop-
ment, little is known concerning the  amount of improvement
which might be brought about by design of areas such as
parking lots so as to retard the flow of rainwater in its
flow to the stream.  Also, storm sewers on local streets
might be purposely underdesigned to provide temporary storage
of storm water.  It is suspected that the improvement that
might be brought about by such methods would be greater than
that from seeking an optimal location pattern for development.

The second possible change in emphasis relates to the value
of open space to the residents of an area.  The research
reported here has indicated that most persons place only a
moderate value on streams and stream quality.  Therefore,
the most important aspect of a plan such as the Brandywine
might be the preservation of open space, per se.  Although
there is a natural association between the preservation of
open space and the preservation of stream quality, it may be
advantageous in formulating future plans to attempt explicitly
to maximize the availability of open space itself to residents
of the area.  This may, of course, involve protection of open
space in areas other than stream valleys.

Considerable further research needs to be done regarding the
value of open space to the residents of an area, with the
object of determining what size, arrangement, and topographic
characteristics of open-space parcels provide the greatest
benefit.   It is important to continue to consider the preser-


                            216

-------
vation of water quality as a goal in open space planning.
The choice of land to be preserved should represent some
mix of the two strategies of maximizing open space value
and minimizing stream quality deterioration.
                           217

-------
                    ' BIBLIOGRAPHY
Alonso, W- 1964.  Location and Land Use:  Toward a General
     Theory of Land Rent.  Cambridge, Mass.: Harvard
     University Press.

American Public Works Association. 1969.  Water Pollution
     Aspects of Urban Runoff.  FWQA Research Report No.
     WP-20-15, Washington: Department of the Interior,
     January.

Anderson, D.G. 1968.  "Effects of Urban Development on
     Floods in Northern Virginia," U.S. Geological Open-
     File Report.

Babbitt and Doland. 1955.  Water Supply Engineering.  New
     York: McGraw-Hill Company.

Bringham, E.F. 1965.  "The Determinants of Residential
     Land Values," Land Economics, Vol. 31, pp. 325-334.

Brown, R.M., N.J. McClelland, R.A. Deininger and R.G.
     Tozer. 1970. "A Water Quality Index-- Do We Dare?"
     presented at National Symposium on Data and Instru-
     mentation for Water Quality Management, University
     of Wisconsin, Madision, Wisconsin, July 20-23.

Buckman, H.O. and N.C. Brady. 1960.  The Nature and Pro-
     perties of Soils.  New York: Macmillan Company, 6th
     Edition.

Bucks County Division of Natural Resources. 1969.  Water
     Quality Monitoring Program.  Year End Report.

Bureau of the Budget. 1967.  Standard Industrial Classi-
     fication Manual. Washington: U.S. Government Print-
     ing Office.
                             218

-------
Carrothers, G.A.P. 1956.  "An Historical Review of the
     Gravity and Potential Concepts of Human Interaction,"
     Journal of the American Institute of Planners,  Vol.
     22, pp. 94-102.~

Carter, R.W. 1961.  "Magnitude and Frequency of Floods
     in Suburban Areas," Short Papers in the Geologic
     and Hydrologic Sciences: U.S. Geological Professional
     Paper 424-B, p. B9-B11.

Chapin, F.S. and S.F. Weiss. 1962.  Factors Influencing
     Land Development.  Chapel Hill, North Carolina:
     Institute for Research in Social Science, University
     of North Carolina, August.

Clonts, H.A., Jr. 1970.  "Influence of Urbanization on
     Land Values at the Urban Periphery," Land Economics,
     Vol. 46, pp. 489-497.

Coughlin, R.E. and K.A. Goldstein. 1970.  "The Extent of
     Agreement among Observers on Environmental Attrac-
     tiveness," RSRI Discussion Paper Series: No.  37,
     February.

              and J. Fritz. 1971.  "Land Values and En-
     vironmental Characteristics in the Rural-Urban Fringe,"
     RSRI Discussion Paper Series: No. 45, May.

    	 S. Sheldon, and T.R. Hammer. 1971.   "The
     Intensity of Development Along Small and Medium-Sized
     Streams in Suburban Philadelphia," RSRI Discussion
     Paper Series: No. 50, September.

    	 F. Taieb, and B.H. Stevens. 1972.   Urban
     Analysis of Branch Library System Planning.   Westport,
     Conn.:  Greenwood Publishing Corporation.

    	 T.R. Hammer, T.G. Dickert and S. Sheldon. 1972.
     "Perception and Use of Streams in Suburban Areas: Effect
     of Distance from Residence to Stream," RSRI Discussion
     Paper Series, forthcoming.
                             219

-------
Craik, K.H. 1968.  "Human Responsiveness to Landscape:
     An Environmental Psychological Perspective," Student
     Publication of the School of Design, North Carolina
     State University, Raleigh, N.C.

	. 1969.  Forest Landscape Perception.  Final
     Report, Institute of Personality Assessment and
     Research Bulletin, University of California, Berkeley

Czamanski, S. 1966.  "Effects of Public Investment on
     Urban Land Values," Journal of the American Institute
     of Planners, Vol. 32, pp. 204-217.

David, E.L. 1968.  "Lakeshore Property Values: A Guide to
     Public Investment in Recreation," Water Resources
     Research, Vol. 4, pp. 697-707.

Donnelly, T.G., F.S. Chapin, and S.F. Weiss. 1964.  A
     Probabilistic Model for Residential Growth.  Chapel
     Hill, North Carolina: Institute for Research in
     Social Science, University of North Carolina, May.

Downing, P.B. 1969.  "Estimating Residential Land Value
     by Multivariate Analysis," paper presented at Con-
     ference of the Committee on Taxation, Resources,
     and Economic Development, University of Wisconsin-
     Mi Iwauke e, June.

Eckenfelder, W.W. (Ed.). 1969.  Manual of Treatment Pro-
     cesses .  Water Resource Management Series, Vol.  1,
     Environmental Science Services Corporation.

Economic Systems Corporation. 1970.  Storm Water Pollu-
     tion for Urban Land Activity, a report to the Federal
     Water Quality Commission, Washington: U.S. Govern-
     ment Printing Office, July.

Gloyna, E.F. and W.W. Eckenfelder (Eds.). 1968.  Advances
     in Water Quality Improvement.  Water Resources
     Symposium No. 1, University of Texas Press.

Grava, S. 1969.   Urban Planning Aspects of Water Pollution
     Control.  New York: Columbia University Press.
                             220

-------
Haig, R.M. 1962.  "Toward an Understanding of the Me-
     tropolis," Quarterly Journal of Economics, Vol.  40,
     pp. 179-208.

Hammer, T.R. 1970.  "Criteria for Measurement of Stream
     Channels as an Indicator of Peak Flow History,"
     RSRI Discussion Paper Series: No. 36, February.

	. 1971.  "The Effect of Urbanization on Stream
     Channel Enlargement," unpublished Ph.D. dissertation,
     University of Pennsylvania, August.

	 , E.T. Horn, IV, and R.E. Coughlin. 1971.   "The
     Effect of a Large Urban Park on Real Estate Value,"
     RSRI Discussion Paper Series: No. 51, September.

Hawkins, R.W. 1969.  "The Effect of Changes of Streamflow
     Regimen on Reservoir Yield," Water Resources Re-
     search, Vol. 5, pp. 1115-1119.

Hem, J.D. 1959.  Study and Interpretation of the Chemical
     Characteristics of Natural Water.  Geological Survey
     Water Supply Paper 1473.  Washington: U.S. Govern-
     ment Printing Office.

Institute for Environmental Studies, Regional Science
     Research Institute, and U.S. Geological Survey.  1968.
     The Plan and Program for the Brandywine.  Philadel-
     phia: Institute for Environmental Studies, University
     of Pennsylvania.

Isard, W. and E. Romanoff. 1967.  "Water Utilization
     Input-Output Coefficients," Technical Paper No.  5.
     Cambridge, Mass.: Regional Science Research In-
     stitute, August.

Jacobs, P. and D. Way. 1968.  Visual Analysis of Land-
     scape Development.  Harvard University, Graduate
     School of Design, Department of Landscape Archi-
     tecture.

Kawashima, T., T.R. Hammer, and R.E. Coughlin. 1970.
     "Preliminary Analysis of the Effects of Urbanization
     on Water Quality," RSRI Working Paper.


                             221

-------
Keene, J.C. and A.L. Strong. 1970.  "The Brandywine Plan,"
     Journal of the American Institute of Planners, Vol.
     36, pp. 50-58.

Klein, L. 1959.  River Pollution. 1; Chemical Analysis.
     New York: Academic Press, Inc.

Kneese, A.V. and B.T. Bower. 1968.  Managing Water Quality:
     Economics, Technology, Institutions. Baltimore: Johns
     Hopkins Press.

Knetsoh, J.L. 1964.  "The Influence of Reservoir Projects
     on Land Values," Journal of Farm Economics, Vol. 46,
     pp. 520-538.

Knos, D.S. 1962.  Distribution of Land Values in Topeka,
     Kansas.  Lawrence, Kansas: Center for Research in
     Business, University of Kansas, May.

Lansing, J.B. and R.W. Marcus. 1969.  "Evaluation of
     Neighborhood Quality," Journal of the American In-
     stitute of Planners, Vol. 35, pp. 195-199.

Leopold, L.B., M.G. Wolman, and J.P  Miller. 1964.
     Fluvial Processes in Geomorphology.: San Francisco
     W.H. Freeman and Company.

	. 1968.  "Hydrology for Urban Land Plann-
     ing," U.S. Geological Survery Circular 554, Washington.

	. 1970.  "Let's Sing 'Auld Lang Syne1
     for the Upper Brandywine," Natural History, June-
     July.

McGauhey, P.H. 1968.  Engineering Management of Water
     Quality.  New York: McGraw-Hill Company.

McKee, J.E. and H.W. Wolf. 1963.  Water Quality Criteria.
     The Resources Agency of California, State Water Re-
     sources Control Board.

Menchik, M.D. 1971.  "Residential Environmental Pre-
     ferences and Choice: Some Results Relevant to Urban
     Form," RSRI Discussion Paper Series; No. 46, March.


                            222

-------
Michelson, W. 1966.  "An Empirical Analysis of Urban
     Environmental Preferences," Journal of the American
     Institute of Planners, Vol. 32, pp. 355-360.

Nebieker, J.H. 1969.   The Technology of Pollution Abate-
     ment," The Massachusetts Heritage, Vol. VII, No. 3,

Park, R.E. 1925.  "The Growth of the City," in Park,
     Burgess, and McKenzie, The City.  Chicago: The Uni-
     versity of Chicago Press.

Peterson, G.L. and E.S. Neumann. 1969.  "Modeling and
     Predicting Human Response to the Visual Recreation
     Environment," Journal of Leisure Research, Vol. 1,
     pp. 219-237.

Rabinowitz, C.B. and R.E. Coughlin. 1970.  "Analysis of
     Landscape Characteristics Relevant to Preferences,"
     RSRI Discussion Paper Series: No. 38, March.

	 and 	. 1971.  "Some Experiments in Quan-
     titative Measurement of Landscape Quality," RSRI
     Discussion Paper Series: No. 43, March.

Rogers, C.E. and Associates. 1968.  Nature's Challenge
     to Man, Guidelines for Conservation and Development
     of Open Space in the Lower Brandywine Valley.  The
     Brandywine Valley Association.

Rummel, R.J. 1967.  "Understanding Factor Analysis," Con-
     flict Resolution, Vol. XL, pp. 444-480.

Scherer, U. and R.E. Coughlin. 1971.  "The Influence of
     Water Quality in the Evaluation of Stream Sites,"
     RSRI Working Paper, October.

Seymour, D.R. 1966.  "Analysis of the Effects of Zoning
     Change on Property Values in Two Pittsburgh Suburbs,"
     unpublished Ph.D. dissertation, University of
     Pennsylvania.
                            223

-------
Shafer, E.L., Jr., J.F. Hamilton, Jr. and E.A. Schmidt.  1969
     "Natural Landscape Preferences: A Predictive Model,"
     Journal of Leisure Research, Vol. 1, pp. 1-20.

Soil Conservation Service, U.S.D.A., in cooperation with
     Pennsylvania State University and Pennsylvania De-
     partment of Agriculture. 1963.  Soil Survey for
     Chester and Delaware Counties, Soil Survey for Bucks
     County, and Soil Survey for Montgomery County.
     Washington: U.S. Government Printing Office.

Soil Surveysfor Montgomery County and Chester and Dela-
     ware Counties, Pennsylvania, prepared by the U.S.
     Department of Agriculture.

Stevens, B.H. and R.E. Coughlin. 1966.  "An Outline for
     Economic Evaluation of Stream Valley Preservation,"
     RSRI Discussion Paper Series: No. 10, July.

Thompson, P. 1969.  "Brandywine Basin--Defeat of an Almost
     Perfect Plan," Science, Vol. 163, pp. 1180-82.

Ven Te Chow, (Ed.). 1964.  Handbook of Applied Hydrology--
     A Compendium of Water Resource Technology.  New
     York; McGraw-Hill, Inc.
                           224

-------
             PUBLICATIONS PRODUCED IN WHOLE
             IN MAJOR PART UNDER THIS GRANT
Coughlin, Robert E. and Karen A. Goldstein, THE EXTENT
     OF AGREEMENT AMONG OBSERVERS ON ENVIRONMENTAL
     ATTRACTIVENESS, RSRI Discussion Paper Series: No.
     37, February 1970.

       and James Fritz, LAND VALUES AND ENVIRONMENTAL
     CHARACTERISTICS IN THE RURAL-URBAN FRINGE, RSRI
     Discussion Paper Series: No. 45, May 1971.

	, Sallie Sheldon and Thomas R. Hammer, THE INTENSITY
     OF DEVELOPMENT ALONG SMALL AND MEDIUM SIZED STREAMS
     IN SUBURBAN PHILADELPHIA, RSRI Discussion Paper
     Series: No. 50, September 1971.

	, Thomas R. Hammer, Thomas G. Dickert and Sallie
     Sheldon, PERCEPTION AND USE OF STREAMS IN SUBURBAN
     AREAS:  EFFECTS OF WATER QUALITY AND OF DISTANCE
     FROM RESIDENCE TO STREAM, RSRI Discussion Paper
     Series: No. 53, March 1972.

Hammer, Thomas R.,THE EFFECT OF URBANIZATION ON STREAM
     CHANNEL ENLARGEMENT, Ph.D. doctoral dissertation,
     University of Pennsylvania, August 1971.

	, Edward T. Horn, IV, and Robert E. Coughlin, THE
     EFFECT OF A LARGE URBAN PARK ON REAL ESTATE VALUE,
     RSRI Discussion Paper Series: No. 51, December 1971.

	, CRITERIA FOR MEASUREMENT OF STREAM CHANNELS AS AN
     INDICATOR OF PEAK FLOW HISTORY, RSRI Discussion
     Paper Series:  No. 36, February 1970.

	9 STREAM CHANNEL ENLARGEMENT DUE TO URBANIZATION,
     RSRI Discussion Paper Series: No. 55, May 1972.

	9 PROCEDURES FOR ESTIMATING THE HYDROLOGIC IMPACT
     OF URBANIZATION, RSRI Occasional Paper, September
     1971.

                         225

-------
	, STREAM CHANNEL ENLARGEMENT DUE TO URBANIZATION,
     Water Resources Research, Vol. 8, No. 6, December
     1972.

Kawashima, Tatsuhiko, Thomas R. Hammer- and Rober E.
     Coughlin, PRELIMINARY ANALYSIS OF THE EFFECTS OF
     URBANIZATION ON WATER QUALITY, RSRI, 1970.

Menchik, Mark D., RESIDENTIAL ENVIRONMENTAL PREFERENCES
     AND CHOICE:  SOME PRELIMINARY EMPIRICAL RESULTS
     RELEVANT TO URBAN FORM, RSRI Discussion Paper
     Series: No. 46, March 1971.

Rabinowitz, Carla B. and Robert E. Coughlin, ANALYSIS
     OF LANDSCAPE CHARACTERISTICS RELEVANT TO PRE-
     FERENCE, RSRI Discussion Paper Series: No. 38,
     March 1970.

	 and Robert E. Coughlin, SOME EXPERIMENTS IN QUAN-
     TITATIVE MEASUREMENT OF LANDSCAPE QUALITY, RSRI
     Discussion Paper Series: No. 43, March 1971.

Scherer, Ursula and Robert E. Coughlin, THE INFLUENCE
     OF WATER QUALITY IN THE EVALUATION OF STREAM SITES,
     RSRI, 1971.
                            226

-------
                     Acknowledgments
The senior  authors  of this report, Robert E. Coughlin and
Thomas R. Hammer, wish to acknowledge their many associates
in the research,  all of whose names should appear on the
title page.  Major  contributions to the research reported
here were made  by Thomas G.  Dickert, James Fritz, Karen A.
Goldstein,  Edward T.  Horn, IV, Mark D. Menchik, Carla B.
Rabinowitz,  Ursula  Scherer,  Sallie Sheldon, Sonia Wordley,
and John R.  Welson.   The efficient and reliable supporting
services and patience of Jacqueline Harmon and Vartine
Berberian have  been invaluable.  Also vital to the research
were the untiring and painstaking efforts of Bonnie Allen,
Tatsuhiko Kawashiina,  Harriet Newburger, Julie Rose, and
Leslie Wilson.

We also gratefully  acknowledge the encouragement and general
direction provided  by Luna Leopold and Benjamin H. Stevens,
and the inspiration of all of our colleagues on the ill-
fated Upper East  Branch Brandywine Project, which provided
the initial impetus  for this research program.
*U.S. GOVERNMENT PRINTING OFFICE:1973 514-156/363 1-3
                            227

-------
SELECTED WATER
RESOURCES ABSTRACTS
INPUT TRANSACTION FORM
                                                                w
    Stream Quality Preservation Through Planned  Urban
    Development
    Coughlin,  R.  E., and T. R. Hammer
   Regional Science Research Institute
   P.  0.  Box 8776
   Philadelphia,  Pa.  19101
12. Sponsoring
                  r ™  Environmental Protection Agency
                                                                              May 1973
                                                                   R...,jrtN<.
                                                                    16110  DYX
                                                                   Same
                                                                              .n
-------