EPA/620/R-94/014
May 1994
Environmental Monitoring and Assessment Program
Agroecosystem Pilot Field Program Report - 1992
by
C Lee Campbell, Technical Director
Jeff M. Bay
Anne S. Hellkamp
George R. Hess
Michael J. Munster
Karen E. Nauman
Deborah A. Neher
Gail L. Olson
Steven L. Peck
Brian A. Schumacher
Kurex Sidik
Mark B. Tooley
David W. Turner
This study was conducted in cooperation with
U.S. Department of Agriculture
Agricultural Research Service
Raleigh, NC 27711
U.S. Environmental Protection Agency
Office of Research and Development
Washington, D.C. 20460
Environmental Monitoring Systems Laboratory
Las Vegas, NV 89193
Printed on Recycled Paper
-------
Notice
This research has been funded by the United States Environmental Protection
Agency (EPA) through its Office of Research and Development (ORD) under Interagency
Agreements #DW12934170 with the U.S. Department of Agriculture (USDA),
Agricultural Research Service (ARS) and #DW12934747 with the USDA National
Agricultural Statistics Service (USDA NASS) and by the USDA ARS. It was conducted
with our research partners under the management of the Environmental Monitoring
Systems Laboratory - Las Vegas in support of the Environmental Monitoring and
Assessment Program (EMAP). Neither U.S. EPA nor USDA ARS endorses or
recommends any trade name or commercial product mentioned in this document to the
exclusion of others. They are mentioned solely for the purpose of description or
clarification.
Proper citation of this document is: -
Campbell, C. L., J. M. Bay, A. S. Hellkamp, G. R. Hess,.M. J. Munster, K. E. Nauman, D.
A. Neher, G. L. Olson, S. L. Peck, B. A. Schumacher, K. Sidik, M. B. Tooley, and D. M.
Turner. 1994. Environmental Monitoring and Assessment Program - Agroecosystem Pilot
Field Program Report -1992. EPA/620/R-94/014. U.S. Environmental Protection Agency,
Washington, D.C.
11
-------
Table of Contents
Notice . ii
List of Figures . . v
List of Tables vii
Glossary of Acronyms viii
Acknowledgements ix
1. Introduction 1
1.1 Mission and goals of the EMAP-Agroecosystems program 1
1.2 Conceptual approach 1
1.3 Purpose and Objectives . . . ; 6
1.4 Planning and Implementation 8
1.5 Statistical Methods 15
2. Results 17
2.1 Extent and Geographic Distribution of Annually Harvested
Herbaceous Crops 17
2.2 Crop Productivity . 27
2.3 Soil Quality - Physical and Chemical 32
2.3.1 Soil Physical Condition 32
2.3.2 Soil Chemical Properties 37
2.4 Soil Biotic Diversity 47
2.5 Water Quality 52
2.6 Agrichemical Use 55
3. Evaluation of Designs, Indicators, and Activities ,. . . 58
3.1 Preliminary Design Comparison 58
3.2 Successes and Challenges in Indicator Development and Evaluation 63
3.2.1 Crop Productivity 63
3.2.2 Successes and Challenges in Assessing Soil Quality . 69
3.2.3. Soil Biotic Diversity 79
3.2.4. Water Quality: Method Development 83
3.2.5. Extent and Geographic Distribution of Annually Harvested
Herbaceous Crops 84
3.2.6 Landscape Structure 90
3.2.7 Agrichemical Use 93
3.3 Evaluation of Pilot Activities 94
3.3.1 Interactions with USDA-NASS 94
3.3.2 Logistics 95
3.3.3 Quality Assurance 97
111
-------
4.
3.3.4 Information Management
Literature Cited
103
108
IV
-------
List of Figures
Figure 1.1 Conceptual model of an agroecosystem 4
Figure 1.2 An agroecosystem in its surroundings 5
.Figure 1.3 Ecoregions of the Southeastern U.S 7
Figure 1.4 Approximate sample locations for the 1992 Pilot 9
Figure 1.5 Cumulative distribution function for clay content of AP horizon at surface soil ... 15
Figure 2.1 Location of agricultural lands in North Carolina 18
Figure 2.2 Cumulative distribution functions for percent of land area with annually
harvested herbaceous crops per 260 hectare tract 21
Figure 2.3 Cumulative distribution function for number of annually harvested crops per
260 hectare tract 22
Figure 2.4 Cumulative distribution functions for percent of area covered by most common
annually harvested herbaceous crop per 260 hectare tract ' 23
Figure 2.5 Cumulative distribution functions for the proportion of fields of size S or
smaller . 25
Figure 2.6 Cumulative distribution functions for the proportion of total area of annually
harvested herbaceous cropland accounted for by fields of size S or smaller 26
Figure 2.7 Cumulative distribution functions for the nitrogen efficiency index for land
in field corn (maize) and wheat 28
Figure 2.8 Cumulative distribution functions for the nitrogen efficiency index for land in
soybean and cotton 29
Figure 2.9 Cumulative distribution functions for the nitrogen efficiency index for land
cropped to eight seed crops 30
Figure 2.10 Cumulative distribution functions for percent clay in AP horizon 35
Figure 2.11 Cumulative distribution functions for percent organic matter in AP horizon 36
Figure 2.12 Percent clay in the AP horizon versus percent organic matter content in the AP
horizon in each soil sample from North Carolina 38
Figure 2.13 Cumulative distribution functions for pH (in water) for AP horizon 41
Figure 2.14 Cumulative distribution functions for cation exchange capacity (CEC) in AP horizon . . 42
Figure 2.15 Cumulative distribution functions for available phosphorus in AP horizon 44
Figure 2.16 Cumulative distribution functions for total lead in AP horizon 45
Figure 2.17 Cumulative distribution functions for total cadmium in AP horizon . . 46
Figure 2.18 Cumulative distribution functions for maturity index for free-living
nematodes in AP horizon of soil 49
Figure 2.19 Cumulative distribution functions for maturity index for plant parasitic nematodes
in AP horizon of soil 50
Figure 2.20 Cumulative distribution functions for Shannon's trophic diversity index for
nematodes in AP horizon of soil 51
Figure 2.21 Cumulative distribution function for nitrate-N concentrations in wells on
North Carolina farms, Fall 1992 52
Figure 2.22 Cumulative distribution function for nitrate-N concentrations in samples
from ponds on North Carolina farms, Fall 1992 53
Figure 3.1 Cumulative distribution functions for the observed/expected yield index,
composite for land area cropped to barley, field corn (maize), cotton, hay,
oat, peanut, grain sorghum, soybean, sweet potato, tobacco, and wheat 64
Figure 3.2 Cumulative distribution functions for the land area cropped to field corn
(maize), hay, and soybean, respectively 65
Figure 3.3 Biplot for all regions combined 76
V
-------
Figure 3.4 Biplot for Piedmont region
Figure 3.5 Biplot for Coastal Plain region
Figure 3.6 Data flow for the 1992 Agroecoststem Pilot
77
78
105
VI
-------
List of Tables
Table 1.1 Societal values and related assessment questions 3
Table 1.2 Number of sample segments in each of the NASS strata. . . 10
Table 1.3 Land use or cover for the 1991-92 crop year on the sample unit in the Pilot
Field Program in North Carolina 11
Table 1.4 Occurrence of multiple sample units in the same field 13
Table 1.5 Number of sample units for soil chemistry, nematodes, ponds, and wells for
the 1992 Pilot in North Carolina 13
Table 2.1 Annually harvested herbaceous crops included in the Pilot 17
Table 2.2 Extent of annually harvested herbaceous cropland in North Carolina 19
Table 2.3 Sample sizes from the June Enumerative Survey for each landscape type ' 20
Table 2.4 Organic matter content for each soil textural type 37
Table 2.5 Percentages of samples occupied by each compartment of the graph from organic
matter and clay content in the AP horizon for each soil sample in Figure 2.12 ... 39
Table 2.6 Estimated use of fertilizer nitrogen and selected pesticides on annually harvested
herbaceous cropland in North Carolina, 1992 56
Table 3.1 Relative efficiencies of the Hexagon design to the Rotational Panel design 59
Table 3.2 Costs of the Hexagon and Rotational Panel sampling plans 60
Table 3.3 Relative efficiency of the Hexagon design to Rotational Panel design with respect
to estimation of extent of annually harvested herbaceous crops . 62
Table 3.4 Comparisons of coefficients of variation (=100*(a/u)) for the nitrogen efficiency
index and simple yields 66
Table 3.5 Comparisons of coefficients of variation (=100*(a/u)) for simple yields and for the .
observed/expected yield index 66
Table 3.6 Variance components for related soil chemical and physical measures 73
Table 3.7 Reliability ratios for soil chemical and physical measures for the Piedmont and
Central Plain regions and for regions of NC combined. 75
Table 3.8 Coefficients from a principal components analysis of variables of soil properties
and nematode community indices . , . 79
Table 3.9 Spearman correlations between indices and trophic groups (n=185) of soilborne
nematodes 80
Table 3.10 Variance components for three indices associated with nematode community
structures in soils 81
Table 3.11 Reliability ratios for several indices of nematode community structures for
various sampling plans 82
vu
-------
Glossary of Acronyms
AHHC annually harvested herbaceous crop
AIC Agroecosytem Information Center
ARG Agroecosystem Resource Group
ARS Agricultural Research Service
cdf cumulative distribution function
cv coefficient of variation
EMAP Environmental Monitoring and Assessment Program
EPA Environmental Protection Agency
ERL Environmental Research Laboratory
GIS geographic information systems
IM information management
JES June Enumerative Survey
LAN local area network
MCL maximum contaminant level
MI maturity index (for free-living nematodes)
MOU Memorandum of Understanding
NASDA National Association of State Departments of Agriculture
NASS National Agricultural Statistic Service
NCSU North Carolina State University
OM organic matter
PPI maturity index (for Plant-Parasitic Nematodes)
PSU Primary Sampling Unit
QAO quality assurance officer
QA/QC quality assurance/ quality control
SHAN Shannon index of trophic diversity
SRPG , soil rating for plant growth
USDA United States Department of Agriculture
via
-------
Acknowledgements
The Agroecosystem Resource Group (ARG) would like to thank all those who
contributed to the planning, execution, and analysis of the 1992 Pilot in North Carolina
and all those who shared their expertise with us.
The involvement of the USDA National Agricultural Statistics Service (NASS) has
become integral to the ARG. We especially thank our past and present administrative
contacts: Robert Bass, Ray Halley, Steve Manheimer, and Jim Gibson in Washington,
and Craig Hayes in Raleigh. Special thanks also to those who helped us "sweat the
details" of data collection and analysis: Sarah Hoffman, Becky Cross, Tom Sabel, and
David Luckenbach. Of course, the data would never have been collected without the
conscientious efforts of a group of 26 enumerators with the National Association of State
Departments of Agriculture, working through the North Carolina NASS office.
The statistical expertise, and guidance of Len Stefanski, Department of Statistics,
North Carolina State University, kept us on track during the preparation of this report.
The programming and assistance of Cathy Barrows, Department of Statistics, allowed
us to produce our cumulative distribution functions and is gratefully acknowledged.
Our increasing cooperation with the USDA Soil Conservation Service (SCS) is also
acknowledged, with thanks to our SCS liasons, Bob Smith and Bill Roth.
Thanks also to various U.S. Environmental Protection Agency (EPA) personnel
including Ann Pitchford and Sue Franson at EMSL-Las Vegas for their excellent
management support, and Charlie Smith, Bill Payne, and Andy Paeng at the Athens-
ERL, who assisted in the development of the water sampling protocols and analyzed
the water samples.
Gratitude is also expressed to those who have helped in the production of this
project report: Tracey Diggs for getting the drafts into shape, Carla Tutor for
photocopying and mailing, Phyllis Garris in the library, and Charles Harper for general
technical support. Also, our thanks to Clara Edwards for all the things she does to
make our program possible.
Deep admiration and gratitude are due to four emeritus members of the
Agroecosystem Resource Group. Dr. Walt Heck (USDA-ARS) was Technical Director
of the ARG during the planning and field phases of this pilot field program. His
capable and visionary leadership continues in his role as Associate Director for the
EMAP terrestrial groups. Dr. John Rawlings, Department of Statistics, North Carolina
State University, and Dr. Alva Pinker, Emeritus Statistician, helped lay a foundation for
the Agroecosystem program in both statistics and common sense. Finally, Dr. Julie
Meyer, now of the University of Wisconsin, was a pioneering member of the ARG, who
brought a special knowledge and goodwill to the initiation of soil quality indicators,
development of the survey questionnaire, and birth of the program.
IX
-------
-------
1. INTRODUCTION
1.1 MISSION AND GOALS OF THE EMAP-AGROECOSYSTEMS PROGRAM
The mission of the Agroecosystem Resource Group (ARG) of the Environmental
Monitoring and Assessment Program (EMAP) is to develop and implement a program
that will, in the long term, monitor and assess the condition and extent of the nation's
agroecological resources from an ecological perspective through an interagency process.
The specific objectives of the agroecosystem program are to:
o Estimate the status, trends, and changes in selected indicators of the condition of
the nation's agroecological resources on a regional basis with known confidence.
o Estimate the geographic coverage and extent of the nation's agroecological
resources with known confidence.
o Seek associations between selected indicators. of natural and anthropogenic
stresses and indicators of the condition of agroecological resources.
o Provide annual statistical summaries and periodic assessments of the nation's
agroecological resources.
1.2 CONCEPTUAL APPROACH
The EMAP-Agroecosystems monitoring effort is based on assessment questions
related to three societal values, which in turn are related to sustainability. Biotic and
abiotic condition indicators are being developed to address each assessment question.
1.2.1 Sustainability
The sustainability of agroecosystems is of primary importance to the people of the
United States and the world. Although there are several aspects of sustainability, the
ARG is interested in the ecological sustainability of agroecosystems.
An agroecosystem is ecologically sustainable if it maintains or enhances
its own long-term productivity and biodiversity, the biodiversity of surrounding
ecosystems, and the quality of air, water, and soil.
Two other facets of sustainability, economic and social, are addressed by other
federal agencies such as the U.S. Department of Agriculture (USDA) Economics Research
Service and the USDA National Agricultural Statistics Service (NASS).
A farm is economically sustainable if it is economically viable over the long
term.
-------
An agricultural system is socially sustainable if it meets the basic food and
fiber needs of society and maintains or enhances the quality of life for farmers
and rural communities.
1.2.2 Societal values
The three societal values for agroecosystems are the components of ecological
sustainability: quality of air, water, and soil; productivity; and biodiversity.
Agroecosystem performance depends on the quality of the air, water, and soil
entering and within the agroecosystem. In addition, agricultural practices can affect the
air, water, and soil of the agroecosystem and surrounding ecosystems.
In the traditional ecological sense, productivity is the rate at which a given trophic
level captures energy; net primary productivity is the net accumulation of plant biomass
per unit area per unit time. Society has a compelling interest in the biomass that is
harvested for food and fiber. However, in a monitoring context, other aspects of
productivity are of interest, for instance, biomass production in pastures and in ecotonal
areas such as windbreaks, and the efficiency with which resources are used in the
agroecosystem.
Biodiversity in a field and in the surrounding landscape affects agroecosystem
function and is affected by agricultural practices. Abundance and diversity of some
species, including pollinators and insect predators, can positively affect plant
productivity; diversity of others, such as parasites, can have negative effects on both
plants and animals. Genetic diversity is important as the raw material for adaptation
and in the prevention of devastating epidemics.
1.2.3 Assessment Questions and Indicators
Assessment questions are general, enduring, biologically oriented questions that
drive the program. Assessment questions are answered relative to some specified value
with information from indicators. Indicators are measures that reflect the condition of
an ecological resource or its exposure to stress. Condition is judged as acceptable
(nominal), marginal or unacceptable (submnominal) relative to standards established for
the specified value.
Although assessment questions are not actually answered in the 1992 Pilot, they
are the driving force behind it. Examples of the kinds of questions that EMAP-
Agroecosystems will answer in the future are in Table 1.1. Indicators currently under
consideration by the ARG are listed below; those in italics were explored in the Pilot and
are discussed in Chapters 2 and 3.
-------
Crop Productivity
Soil Biotic Diversity (nematode indices)
Land Use and Cover
Landscape Structure
Insect Diversity
Symptoms of Foliar Injury
Livestock Productivity
Soil Quality: physical/chemical
Water Quality: ponds and existing wells
Pest Management (including agrichemical use)
Biological Ozone Indicator (clones of white clover)
Habitat Quality for Wildlife
Genetic Diversity
Table 1.1 Societal values and related assessment questions.
Societal Value
Assessment Questions
Quality of Air, Water, and
Soil
oWhat proportion of agroecosystems has soil quality sufficient
to sustain both crop and non-crop productivity?
oWhat proportion of agroecosystems has acceptable diversity
of soil microbes and invertebrates?
Productivity
oWhat proportion of agroecosystems is attaining their
productive capacity (as determined by soil map unit, climate,
historic levels of production, etc.)?
Biodiversity
oWhat proportion of agroecosystems has acceptable diversity
of insects?
oWhat percent of agroecosystems has associated noncrop
areas with habitat suitable for wildlife species of interest?
oln what proportion of agroecosystems is diversity of wildlife
declining?
1.2.4 Conceptual Model
The ARG has developed a conceptual model of agroecosystems to assist in
formulating assessment questions, identifying appropriate measures for each indicator,
and identifying possible relationships among measurements in the development of
indices. It is also a valuable aid in communicating our ideas to a wide audience.
An agroecosystem (Fig. 1.1) is a dynamic association of crops, pasture, livestock,
other plants and animals, atmosphere, soils, and water. The agroecosystem includes not
only the field, but also the associated border areas such as windbreaks, fence rows,
ditchbanks, and farm ponds. The agroecosystem boundary .depends on, and varies with,
the process being considered.
Agroecosystems interact with larger landscapes (Fig. 1.2), which include
uncultivated land, drainage networks, human communities, and wildlife. The landscape
is the area that directly affects the ecology of the agroecosystem and is
-------
MANAGEMENT
PRACTICES
CLIMATE
\
BORDER AREA
WEATHER
CROPS
'S
'FORAGESd
'LIVESTOCK! ( LIFE
BIOTIC
.FIELD/ORCHARD/PASTURE
AGROECOSYSTEM
-f AIR, WATER, SOIL
+ ORGANISMS
-f CHEMICALS
-f AGRICULTURAL COMMODITIES
LANDSCAPE
REST OF WORLD
Figure 1.1. Conceptual model of an agroecosystem.
-------
—=3*5>
§
o
REST OF THE WORLD
KEV
•^•INFORMATION AND POLICY
—*=>EMAP INFORMATION
MATERIAL AND ENERGY FLOW
FLOW MANAGED BY FARMER
Figure 1.2. An agroecosystem in its surroundings;
-------
directly affected by agroecosystem processes. The landscape boundary depends on, and
varies with, the process being considered.
Materials and energy are brought to the agroecosystem from afar and exported
to distant points (rest of the world, Fig. 1.2).
Agroecosystems also belong to farms (Fig. 1.2). A farm is an economic entity in
which an operator uses information about market conditions and available resources to
manage agroecosystems for economic gain.
Government policy, programs, and regulations and socioeconomics influence
decisions made at all levels. In particular, policies are intended to change people's
actions to realize the interests of society.
1.3 PURPOSE AND OBJECTIVES
The 1992 Pilot was designed to evaluate aspects of the EMAP Agroecosystem
monitoring program critical to the implementation of a regional and national program.
It was designed to address these program aspects over an area large enough to provide
reliable answers to questions concerning the operation of the monitoring program but
small enough to be physically and fiscally manageable. There were three major
objectives for the 1992 Pilot:
1. Compare the relative efficiency, in terms of cost and precision, of two sampling
frames.
2. Evaluate an initial suite of indicators to: • L t.- t
o Assess the ability of each indicator to address the assessment questions of
interest; ,
o Establish an initial range of values for each indicator across the diverse
physiographic regions in the state;
o Assess spatial variability of indicator values within and among sample
units; .
o Identify the usefulness and sensitivity of each indicator m determining
ecological condition; and
o Determine cost-effectiveness for each indicator.
, 3. Develop and refine plans for:
o Sampling;
o Logistics;
o Quality Assurance;
o Data analysis, summarization, and reporting;
o Information management; and
o Ecological health indices and their interpretation.
-------
The 1992 Pilot was also conducted to establish a cooperative, working relationship
with USDA-NASS at both the state and national level. NASS has a well established,
nationwide network of enumerators and administrators experienced in conducting
national surveys. Also, growers throughout the U.S. are familiar with NASS personnel
and have confidence in NASS because of NASS's data confidentiality requirements.
North Carolina was selected for the 1992 Pilot because:
1. The physiographic diversity of the state is representative of the ecoregions of the
Southeastern region of the United States (Fig. 1.3).
2. NASS is organized by state. Limiting the Pilot to one state simplified problem
resolution.
3 Most of the ARG staff is located in Raleigh, NC, which facilitated logistic
activities.
Mid-Atlantic
Coastal Plain
Southern Coastal
Plain
Southern Florida
Coastal Plain
Figure 1.3 Ecoregions of the Southeastern U.S.
7
-------
1.4 PLANNING AND IMPLEMENTATION
A brief summary of activities and methods associated with the planning and
implementation of the 1992 Pilot are presented here. The Agroecosystem 1992 Pilot
Project Plan (Heck et alv 1992) should be consulted for a full description of activities and
methods.
1.4.1 Design and Statistical Considerations
The ARG considered two sampling plans in the 1992 Pilot: the NASS Rotational
Panel Plan and the EMAP Hexagon Plan. Each used the NASS Area Frame segments as
the basic sampling units. NASS Area Frame segments were defined by stratifying the
State of North Carolina based on intensity of agriculture, dividing each stratum into
Primary Sampling Units (PSUs) and then dividing a random sample of PSUs into six to
eight sample segments. Segment size depends on strata, but is approximately 1 square
mile (2.6 km2) for agricultural strata.
The 203 hexagons (40 km2 each) with their centroids in North Carolina were
divided into four subsamples (Overton et al., 1991). Of the 54 hexagons in the one
randomly selected subsample, three were over water. Thus, 51 hexagons distributed over
49 counties were used (Fig. 1.4).
1.4.2 Hexagon Sampling Plan
NASS identified the PSU that contained the centroid of each selected hexagon and
divided the PSU into segments according to standard NASS criteria. The segment
containing the centroid was identified and included as a sample segment. Area,
cultivated acreage, and number of fields were estimated within each segment.
8
-------
Distribution of Sample Points
A / County boundaries
^ Chosen with hexagon plan
.. Chosen with rotational panel plan
Figure 1.4 Approximate sample locations for the 1992 Pilot. Rotational panel sampling
points are stratified by intensity of agricultural land use (Table 1.2). Hexagon sampling
points occur on a regular grid with random initial placement of the grid.
-------
Strata
>50% cultivated
15-50% cultivated
<15% cultivated
Ag-Urban (>20 homes/mi2)
Commercial
Resort (>20 homes/mi2)
Non-Agricultural
Sample
Points
6
28
16
10
3
1
1
1.4.3 Rotational Panel Sampling Plan
Table 1.2 Number of sample segments in
each of the NASS strata. The Rotational Panel sample for
the Pilot consisted of (approximately) a
20% subsample of the June
Ehumerative Survey (JES). The
complete 1992 NASS sample for their
JES in North Carolina had 321
segments contained in 96 (of 100)
counties. Sample segments were
selected from the most recent
replication to enter the plan within
each stratum. The 65 segments selected
fell into 55 counties (Table 1.2).
1.4.4 Sample Unit Selections
During the JES, NASS
enumerators obtained land use data on
all areas of each sample segment. The
location of each field in each sample
segment was mapped on an aerial
photograph and its identification number and acreage recorded. Fields eligible for the
full survey were those that contained an annually harvested herbaceous crop (AHHC).
Prior to drawing the samples for the Fall Survey, field areas for both the hexagon
and rotational sampling plans were expanded using NASS expansion factors in order to
obtain lists of acres of AHHC in North Carolina for 1992. Multiplying an acre by its
NASS expansion factor converts that acre to the number of acres in North Carolina that
it represents. By expanding the field areas, each list represented an estimate of the
population of AHHC acres in North Carolina. The two lists of areas were then sampled
systematically to select every k* acre with random starts to identify the fields to be
visited under each of the two plans for the Fall Survey. The sampling rate (k) was set
to select an average of three selected acres per segment for a total of 348 selected
sampling sites. Each randomly chosen acre identified a field to be visited but was not
necessarily the sampled acre. A random start point for sampling site in each field was
identified upon arrival at the field. The fields selected were identified and marked on
the aerial photographs for use by NASS enumerators in collecting field data during the
Fall Survey.
For the 65 segments of the Rotational Panel Plan, a total of 195 sample units were
selected for the 51 segments of the Hexagon Plan, 153 sample units were selected (Table
1.3). (For 15 cases for the Rotational Plan and 20 cases) for the Hexagon Plan a field was
selected more than once for sampling. (Table 1.4). When this occurred, the appropriate
10
-------
Table 1.3 Land use or cover for the 1991-92 crop year on the sample unit in the Pilot
Field Program in North Carolina. Cover crops were not included. Samples were
selected using JES data, and some samples were chosen incorrectly and did not belong
to the desired resource class (annually harvested herbaceous crops). Data presented are
from the fall questionnaire. Questionnaire data are unavailable for some units because
of farmer refusal or inaccessibility.
Crop or land use
winter small grains (wheat, oat, barley, and
rye)
small grain - soybean double crop
soybean (not double cropped)
corn for grain
corn for silage
popcorn
cotton
hay (all)
peanut
grain sorghum
sweet potato
tobacco
cucumber
pasture, clover, and grasses
set-aside or mostly set-aside
idle cropland
mixture of crops or land uses
woodland or wetland
refusal or inaccessible
TOTAL
Number of
sample units,
NASS design
8
16
40
34
10
2
11
15
3
1
3
14
1
4
8
6
7
5
7
195
Number of
sample
units,
hexagon
design
5
10
29
34
6
0
9
17
8
0
0
7
0
7
0
2
3
1
15
153
11
-------
numbers of random start points were selected for actual sampling. In some cases, a
sample was not taken from the designated sample unit because it was refused by the
owner/operator, inaccessible, not in the designated resource class, or there was an
enumerator error (Table 1.5).
For ponds and wells, a number of ponds and wells per tract (i.e., per ownership
unit that had a portion of that unit in the selected segment) were identified from the
Rotational Panel plan only in the JES. Only those ponds and wells within the actual
NASS segment were included in the final list for sample selection. If a segment
contained 1 or 2 ponds, then each pond was included in the sample. If a segment
contained more than 2 ponds, the selection probability for each pond was 2/n, where
n is the number of ponds in that segment. If a segment contained 1, 2, or 3 wells, then
each was included in the sample. If a segment contained more than 3 wells, the selection
probability for each well was 3/m where m in the number of wells in the segment. A
total of 51 ponds and 81 wells was identified for sampling. In some cases (Table 1.5) a
sample was not taken from a pond or well because it was refused by the
owner/operator, inaccessible, or not properly identified originally.
12
-------
Table 1.4 Occurrence of multiple sample units in the same
field. Includes fields that were later lost from title sample because
of refusal, inaccessibility, etc.
Number of
times a field
was selected
1
2
3
5
7
Number of
NASS fields
159
9
6
0
0
Number of
hexagon fields
102
15
3
1
1
Table 1.5 Number of sample units for soil chemistry, nematodes, ponds, and wells for
the 1992 Pilot in North Carolina.
Soil
chemistry
Nematodes
in soil
Ponds
Wells
Number of Sample Units
NASS Rotational Panel Design
Sampled
185
185
40
61
Not Sampled8
10
10,
11
20
Hexagon Design
Sampled
136
0
0
0
Not Sampled3
17
0
0
0
a Sample not taken because it was refused, inaccessible, or not in the designated resource
class, or there was an enumerator error.
1.4.5 Sampling fields, ponds, and wells
Information on crops and land use for each field within a segment and the
presence or absence of farm ponds or wells was obtained from the JES. For each field
13
-------
identified after the JES as part of the pilot sample, a NASS enumerator contacted the
owner/operator of the field in the fall (November-December) and completed a
questionnaire designed specifically for EMAP Agroecosystems. The questionnaire
provided information on crop yields, land use and tillage history, fertilizer usage history,
pest management practices, fuel usage, and irrigation practices.
Soil samples were taken along an approximately 90 m transect for each sample
unit for the current season to determine soil physical and chemical properties and to
quantify nematode community trophic structure. The transect was located in a
pseudorandom manner and represented a 0.4 ha (=1 acre) portion of the field. Twenty
soil cores (2.5 cm diam x 20 cm deep) were taken at equally spaced intervals along the
transect. Soil was placed in a plastic bucket and mixed by hand. Portions were placed
into plastic bags, one bag for physical/chemical analysis and one for nematode analysis.
Bags were sealed, labelled, and shipped via Federal Express to commercial laboratories
for analysis (Agrico, Agronomic Service Laboratory in Ohio for physical and chemical
analysis; N&A Nematode Identification Services in California for nematode analysis). In
every sixth field two independent transects were used to estimate within-field variation.
In every 12th field, in addition to the additional or duplicate transect, twice as much soil
was taken and the composite was divided into two samples for each of the
physical/chemical and nematode analyses. The split samples were used to estimate
within-sample variability.
Two methods were used to obtain water samples from ponds. In the first method,
an enumerator took a boat out onto the' pond. Water samples were collected from three
points and three depths per point with a Kemmerer sampler and mixed to give one
sample for each pond. In the second method, an enumerator used a stoppered teflon
bottle at the end of a 5-m pole to obtain water samples from a shallow depth (0.3 m)
from four points along the bank of each pond. Water from the four bank locations was
mixed to form a composite sample. Each mixed sample was divided into two portions
which were placed in separate glass bottles. Bottles were placed in styrofoam cases and
shipped via Federal Express to the EPA Environmental Research Laboratory (ERL) in
Athens, GA, for analysis of nitrate concentration and concentration of selected pesticides.
For well water samples, wells were purged by allowing water to flow from the
water outlet. The purge time was 15 minutes if there was a tank between the well and
the spigot, 5 minutes otherwise. Samples were then collected and placed into two glass
bottles, one for nitrate analysis and one for pesticide analysis. Bottles were placed in
styrofoam cases and shipped via Federal Express to the EPA ERL in Athens, GA.
14
-------
1.5 STATISTICAL METHODS
The primary statistical tool used to summarize indicator data from the 1992 Pilot
is the cumulative distribution function (CDF). The estimated CDF for a particular
indicator gives the estimated proportion of the population that has values of the
indicator less than or equal to any specified value -S. Ninety percent confidence bands
are added to the CDFs in order to indicate the precision of the estimated CDF.
Typically, in this report, the population is the area of land in North Carolina, the
Piedmont region of North Carolina, or the Coastal Plain region of North Carolina
cultivated with annually harvested herbaceous crops. Marginal maps indicate the region
represented by the CDF. If no marginal maps are presented, the CDF represents all
possible samples for that measurement or index.
As an example consider the estimated CDF for clay content of soil (0-20 cm depth) in
the Coastal Plain of North Carolina (Figure 1.5).
Clay
Figure 1.5 Cumulative distribution function for clay content of AP horizon at surface
soil (0-20 cm deep).
The CDF shows that the proportion of land (in acres or hectares) cultivated with
annually harvested herbaceous crops in the Coastal Plain with a clay content of 15% or
less, for example, is estimated to be 0.80 with a 90% confidence interval of approximately
(0.72, 0.88). Similarly, an estimate of the median clay content of annually harvested
herbaceous cropland in the Coastal Plain is approximately 10%.
For each CDF, values of the measured or calculated quantity can be given for
specified quantilies on portions of the CDF. For example, a value of 10% clay at a
quantile of 50% would mean that 50% of the population of land cultivated in annually'
harvested herbaceous crops has 10% or less clay content in the AP horizon.
15
-------
A CDF, by definition, will range from 0 to 1 on the vertical axis. An interpolation
procedure is used to fit the estimated continuous CDF to the empirical step-function
CDF. The procedure connects the midpoints of the steps for each observed value and
is incorporated to remove bias. Hence, the estimated CDFs will not begin at 0 or end
at 1. For many indicators, this is not noticeable.
In the 1992 Pilot, most data were collected using the Hexagon sampling design
and the Rotational Panel sampling design (Section. 1.4). Nematodes were only sampled
for sample units of the Rotational Panel design. For indicators using fall questionnaire
or soil sample data and both designs, CDFs computed from each design were combined
to form the CDFs reported in Chapter 2. An average variance was computed for each
design using the variances at the observed values for the combined dataset. The
combined CDF was defined as:
_ _
3) , $=VAR(Fhex) /VAR(Frp)
The above weighting provides an approximately optimal combination of the two CDFs
in terms of overall variability.
Regardless of which sampling design is ultimately selected by the Agroecosystem
Resource Group, land-use and extent indicators will continue to be computed from
NASS's June Enumerative Survey in order to take advantage of this resource. Thus,
although land-use and extent data were collected in June using the Hexagon design,
CDFs for these indicators represent estimates based only on the complete June
Enumerative Survey performed using the Rotational Panel design. For June Enumerative
Survey and fall survey data, variances computed from the Rotational Panel design
account for the stratified nature of this design.
16
-------
2. RESULTS
Results are presented for the indicator categories land use, crop productivity, soil
quality-physical and chemical, soil biotic diversity, water quality and agricultural
chemical use. The results for each indicator are presented in a question and answer
form similar to what might be used in the future for annual statistical summaries. In
a few cases, boundaries for acceptable (nominal), marginal, and unacceptable
(subnominal) conditions are presented. In most cases, however, their boundaries are not
proposed and will be developed for future reports.
2.1 EXTENT AND GEOGRAPHIC DISTRIBUTION OF ANNUALLY HARVESTED
HERBACEOUS CROPS
Agricultural lands in North Carolina are scattered across the state in several
different types of landscape, largely intermixed with forest (Figure 2.1). For this Pilot,
we examined only a portion of North Carolina's agriculture — the land planted in
annually harvested herbaceous crops. Annually harvested herbaceous crops are herbaceous
plants that are harvested every year, regardless of whether the plant itself is annual or
perennial (Table 2.1). These crops are planted on about 1.68 million hectares in North
Carolina, covering some 13% of the total land area of the state (Table 2.2). Soybean,
accounting for one third of the annually harvested herbaceous cropland, is the most
common crop.
Table 2.1. Annually harvested herbaceous crops included in the Pilot.
Barley
Corn
Cotton
Hay
Alfalfa hay
Grain hay
Other hay
Oats
Peanuts
Potatoes
Rye
Soybeans
Sweet potatoes
Sorghum
Sunflowers
Tobacco
Burley
Flue-cured
Vegetables (all)
Winter wheat
17
-------
-------
Table 2.2 Extent of annually harvested herbaceous cropland in North Carolina.
Landscape Type
Forest-Agriculture Mosaic
Forested
Suburban-Agriculture Mosaic
Non-Agricultural
State Total
Extent of AHHCa
(hectares)
(with 95% confidence)
1,401,021 ± 135,249
217,259 ± 87,611
61,816 ± 35,353
0
1,680,096 ± 163,444
Proportion of
Total AHHCa
(percent)
83
13
4
0
100
a AHHC is annually harvested herbaceous crop.
Information about the extent and distribution of annually harvested herbaceous
crops was obtained by analyzing data from the complete National Agricultural Statistics
Service's June Enumerative Survey. The Survey design is based on stratification of land
by intensity of agricultural use (see Section 1.5). Each stratum is subdivided into
segments, approximately 260 hectares (one square mile) in size, a sample of which is
surveyed during the June Enumerative Survey. In North Carolina, most segments
contain more than one field, each of which is identified during the Survey.
What follows is a description of the extent and distribution of agricultural lands,
particularly land devoted to annually harvested herbaceous crops, in North Carolina.
Information is presented at the scale of broad landscapes (based on the strata); smaller
areas within similar landscapes (based on segments within strata); and individual fields
(based on fields within segments). As you examine this information, you may notice
that the sample size is not constant. Changes in sample size reflect differences in both
the questions and the scale of observation. Table 2.3 summarizes the different sample
sizes seen in this section of the report. No further information is provided about the
suburban-agriculture' mosaic because of the limited extent of annually harvested
herbaceous crops in that type of landscape and the small number of samples.
You may also notice that the other indicators summarized in this section are
reported for regions that differ from the landscapes shown in Figure 2.1 and Tables 2.2
and 2.3. The reasons for this inconsistency, and several potential solutions, are discussed
under the heading Reporting on Consistent Regions in Section 3.2.5.2.
19
-------
Table 2.3 Sample sizes from the June Enumerative Survey for each landscape type.
Landscape Type (Stratum)
Forest-Agriculture Mosaic
Forested
Suburban-Agriculture Mosaic
Non-Agricultural
State Total
Number of
Segments
170
80
53
18
321
Number of
Segments
with AHHC"
160
33
18
0
211
Number of
Fields
2671 -
266
49
0
2986
aAHHC is annually harvested herbaceous crops
20
-------
How much of the land is cropped?
If you look at a random 260 hectare tract of land in North Carolina, how much of that
area would you expect to be planted in annually harvested herbaceous crops? Because.crops are
unevenly distributed, the answer depends upon which part of the state you are in.
Figure 2.2 shows the proportion of 260 hectare tracts in which P% or less of the area is
cropped. There are separate graphs for parts of the state that are a forest-agriculture
mosaic and those that are forested. As you might expect, land in the forest-agriculture
mosaic landscape is more heavily cropped than land in forested landscapes.
SO 4O SO ^ O
= Area. Cropped
QUANTILES
5% 0.11
25% 9.09
50% 17.1
75% 32.0
95% 53.5
Sample Size: 170
o _ ot^
so ISO -41.O so
F» — Area Cropped
SO ~7 O 8 O
QUANTILES
5% 0
25% 0
50% 0.24
75% 5.16
95% 26.6
Sample Size: 80
Figure 2.2 Cumulative distribution functions for percent of land area with annually
harvested herbaceous crops per 260 hectare tract.
21
-------
:
How diverse are the state's croplands?
A more diverse agriculture may be less vulnerable to disease, pest infestation, and
changing economic conditions. The diversity of a collection of items — in this case, crops
— is usually measured in two ways: the number of different items, and their relative
abundance. When comparing the diversity of crops in several areas, each area should be
of equal size; diversity can change with area in surprising ways. First, let's look at the
number of different items. In a random 260 hectare tract of land, how many different
annually harvested herbaceous crops would you be likely to find? Again, the answer depends
on which part of the state you are in. Figures 2.3 shows the proportion of 260 hectare
tracts containing S or fewer crops. As you might expect, tracts in the forest-agriculture
mosaic landscape have greater diversity of crops than those in the forested landscape.
o o -
QUANTILES
5% 0.26
25% 2.49
50% 3.99
75% 5.30
95% 6.98
Sample Size: 170
Different Crops
QUANTILES
5% 0
25% 0
50% 0.57
75% 2
95% 4.75
Sample Size: 80
Number of Different Crops
Figure 2.3 Cumulative distribution function for number of annually harvested crops per
260 hectare tract.
22
-------
Do these cropped areas tend to be dominated by a single crop?
The answer to this question provides a measure of relative abundance, the second
aspect of diversity. For example, two tracts of equal size may each contain corn and
soybeans. In tract 1, corn may account for 90% of the cropped land, with 10% in
soybeans; in tract 2, corn may account for 60% of the land. Corn is said to dominate tract
1; in tract 2 the crops are said to have a more even diversity. Large areas dominated by
a single crop, particularly if crop rotation is not practiced, may be more vulnerable to
disease and pest infestations. For the 260 hectare tracts in which annually harvested
herbaceous crops occur, is there a single crop that dominates the cropped area? Figure 2.4
shows the proportion of tracts in which P% of the annually harvested herbaceous
cropland is accounted for by a single crop - the one with the largest area.
o - s -
o - -a. -
Covorod By iviost Common Crop
QUANTILES
5% 29.5
25% 42.2
50% 51.3
75% 64.4
95% 98.8
Sample Size: 160
<=» r~* ~i C3 C3
CZIovereci By IVIost C^ommon Crop
Figure 2.4 Cumulative distribution functions for percent of area covered by most
common annually harvested herbaceous crop per 260 hectare tract.
23
-------
How large are the crop fields?
If you were to measure the size of a random field of an annually harvested herbaceous
crop how large would you expect the field to be? The distributions in Figure 2.5 show the
proportion of fields of size S or smaller. There are separate distributions for the entire
state and for the forested and forest-agriculture mosaic landscapes. North Carolina
fields tend to be small, with a median field size of just over 2 hectares regardless of the
type of landscape.
How do different sized fields contribute to the total area of annually harvested
herbaceous cropland?
What proportion of the total area of annually harvested herbaceous cropland occurs in
fields of a given size or smaller? The distributions in Figure 2.6 show the proportion of the
total area of annually harvested herbaceous cropland accounted for by fields of size S
or smaller. For the state, half the total area of annually harvested herbaceous cropland
occurs rn: fields smaller than 6.09 hectares.
24
-------
— Agriculture ft/loasUc
10 IS 20 25 30 OS 4O 45 SO 55 SO
S — Field Size (Hectares)
QUANTILES
5% 0.40
25% 1.21
50% 2.07
75% 4.08
95% 11.8
Sample Size: 2671
O 5 10 15 20 25 00 OS 4O 45 SO 55 BO
S = f=leld Slze
QUANTILES
5% 0.35
25% 0.81
50% 2.02
75% 4.29
95% 10,7
Sample Size: 266
L-and wltti
•"» **• • - - - •
O S 10 IS 20 2S 30 35 4O 4S SO SS SO OS
S — «=iold Size (Hectares)
QUA^f^LES
5% 0.40
25% 1.09
50% 2.05
75% 4.17
95% 11.8
Sample Size: 2986
Figure 2.5 Cumulative distribution functions for the proportion of fields of size S or
smaller.
25
-------
QUANTILES
5% 1.21
25% 3.13
50% 6.07
75% 12.1
95% 27.1
Sample Size: 2671
O S 10
ts zo zs 30 as 40 -«s so
S — Field Size (Hectares)
eo es
S — l=le»Ic» Size (Hectares)
3O 3S 4O -«B SO S5 SO BS
S — i=iold Size (Hectares)
QUANTILES
5% 0.97
25% 3.24
50% 7.09
75% 11.9
95% 44.1
Sample Size: 266
QUANTILES
5% 1.21
25% 3.23
50% 6.09
75% 12.1
95% 26.9
Sample Size: 2986
Figure 2.6 Cumulative distribution functions for the proportion of total area of annually
harvested herbaceous cropland accounted for by fields of size S or smaller.
26
-------
2.2 CROP PRODUCTIVITY
Productivity of agroecosystems is a broad concept with many possible definitions.
Two indices of productivity are summarized here: nitrogen use efficiency and
standardized yield. So far, indicator development in this area has focused on annually
harvested herbaceous crops (AHHCs).
How efficiently is nitrogen being used to produce annually harvested herbaceous
crops?
Nitrogen is a key plant nutrient that is often applied in large quantities to
agricultural fields, either in the form of animal wastes or as commercial fertilizer. The
latter requires large amounts of energy to produce (Pimentel et al., 1973, Southwell and
Rothwell, 1977). If sustainability is a goal, these nitrogen subsidies should be used
efficiently. Efficiency is also important because nitrogen, regardless of its source, is
considered a contaminant if it reaches ground or surface water (Spalding and Exner,
1993, Angle et al., 1993).
A nitrogen efficiency index was calculated from the yield and management
information obtained from the 1992 Pilot questionnaire. The index is the ratio of the
amount of nitrogen applied on a field to the harvested yield from that field. In-double-
crop situations (fields with two harvested crops in a single season), the total weight of
applied nitrogen was divided by the total weight, of harvested material for both crops.
All yield values used in the nitrogen efficiency index as reported here were expressed
On a dry matter basis, using conversions that were obtained from university publications,
extension personnel, and MASS. Both the numerator and the denominator' of the index
are expressed in the same units, so the ratio is dimensionless. This approach gives an
indication of the efficiency of the system, not of individual plants. Also, note that
smaller index values represent greater efficiency. This "reverse" interpretation was
configured so that fields with no applied nitrogen could be included (otherwise division
by zero excludes them). Fields with reported yields of zero were excluded from this
calculation (see Section 3.2.1 for details).
Nitrogen inputs from commercial fertilizer were provided by NASS. For
estimating the nitrogen content of animal manures, conversion factors were used from
The 1991 North Carolina Agricultural Chemicals Manual (Barker et al., 1991). Manure
nitrogen could be estimated for only 19 sample units, out of a total of 33 on which
manure was applied to an AHHC (see Section 3.2.1). The remaining 14 sample units
were dropped from the analysis.
27
-------
The cumulative distribution functions of the nitrogen efficiency index for two
important cereal crops, field corn and wheat, are depicted in Figures 2.7. The two
curves are quite similar. Contrast these with the graphs for soybean and cotton in
Figures 2.8. The association of the soybean plant with N2-fixing bacteria increases its
efficiency (lowers the index).. Cotton's low efficiencies (large index values) probably
come from the relatively low weight of lint compared to grain.
QUANTILES
5% 0.004
25% 0.024
50% 0.031
75% 0.045
95% 0.079
o.oo 0.02 0.04 o.oe o.oa o.-to 0.12 0.14 o.-ie o.to 0.20
IM efficiency Index for Field Corn
QUANTILES
5% 0.000
25% 0.027
50% 0.035
75% 0.039
95% 0.074
"o.oo 0.02 0.04 o.oe o.os o.io 0.12 o.i4 o.-ie o.io 0.20
IM offlctoncy index for Wheat
Figure 2.7 Cumulative distributions functions for the nitrogen efficiency index for land
in field corn (maize) and wheat. Larger index values indicate lower efficiency. The 90%
confidence interval is shown by dashed lines.
28
-------
QUANTILES
5% 0.000
25% 0.004
50% 0.011
75% 0.017
95% 0.040
o.oo 0.02 0.04 o.oe o.os o.io 0.12 o.i 4 o.ie o.ia 0.20
rsl efficiency Index -for Soybean
QUANTILES
5% 0.000
25% 0.096
50% 0.141
75% 0.493
95% 0.493
O.I O.2 O.O O.4 O.I
IM efficiency index for Cotton
Figure 2.8 Cumulative distribution functions for the nitrogen efficiency index for land
in soybean arid cotton. Larger values indicate lower efficiency. Dashed lines show the
90% confidence interval. Note difference in scale on the horizontal axes.
In order to show the overall condition of cropland in a region, it is our goal to
present one composite index for nitrogen efficiency. It would be misleading to do so
when the harvested plant parts are so different; however, it seems acceptable to pool the
indices across the following seed crops for which we have both yield data and an
approximate conversion to dry matter: wheat, oat, barley, rye, sorghum, field corn
(maize), soybean, and peanut. Together, these crops represent 49% of the AHHC sample
units from the fall survey. The CDFs of the nitrogen efficiency index, combined across
seed crops, are summarized in Figure 2.9. Note that the distribution for the piedmont
has a long tail and a large 95th percentile, indicating that the most inefficient conversion
of nitrogen into harvested crop occurred there.
29
-------
long Oomtolnoci. n — 1
o.oi
IM effici
o.oa o.iz o.ie 0.20
y Indeoc -for Seed Crops
QUANTILES
5% 0.000
25% 0.011
50% 0.022
75% 0.034
95% 0.068
F*|qdmont Region, n—
QUANTILES
5% 0.000
25% 0.011
50% 0.020
75% 0.035
95% 0.087
O.O8 O.1S! O.ie O.S2O
IM offlcloncy Index for Seed Crops
o.c
o.
Flnln Region, n — SI
0.04 o'.os O.IE o.ie o.ao
IM efficiency index for Seed Crops
QUANTILES
5% 0.001
25% 0.010
50% 0.023
75% 0.033
95% 0.055
Figure 2.9 Cumulative distribution functions for the nitrogen efficiency index for land
area cropped to eight seed crops. Crops included are wheat, oat, barley, rye, sorghum,
field corn (maize), soybean, and 'peanut. Larger index values indicate lower efficiency.
30
-------
Are yields meeting expectations based on soil and climate?
The Pilot Plan for 1992 proposed an index comparing each reported yield to the
yield expectations for the soils on the sample field. These expected yields for soil map
units within a county were not available for a fixed point in time, so the calculation of
this index was postponed. In its place, an observed/expected index based on historical
county average yields is presented in Section 3.2.1.1, illustrating a method for combining
data across many crops.
31
-------
2.3 SOIL QUALITY - PHYSICAL AND CHEMICAL
Physical and chemical properties of soils largely determine and reflect the
productive potential of land. For the 1992 Pilot, we analyzed composite samples for:
particle size, organic matter, cation exchange capacity, select micro and macro nutrients,
and some contaminants related to sludge application. In general, geographical
differences were found between the individual parameters, which are likely related to
the innate characteristics of the soils. As discussed in Section 3, we are evaluating
whether it is necessary to consider innate soil properties while trying to determine soil
"condition."
Three cumulative distribution functions (CDFs) are presented for each parameter:
all North Carolina samples combined, Coastal Plain samples, and Piedmont samples.
Samples collected in the mountain region are not presented in a separate CDF because
there were only eight samples; however, these data are included in the CDFs for the
state summaries. For the CDFs presented, the sample size is n= 233 for the Piedmont
region and n= 80 for the Coastal Plain region.
In the following discussions, we refer to proportions,or percentages of samples,
e.g., "About 5% of the Piedmont samples had < 35% clay." Because of the statistical
design, these percentages or proportions also represent "the proportion of land area
planted in annually harvested herbaceous crops." In this context, "5% of the Piedmont
samples" is a simplified but analogous way of saying that "5% of the Piedmont area that
is planted in annually harvested herbaceous crops".
Acceptable (nominal) and unacceptable (subnominal) ranges are proposed which
may be related to whether a region has sufficient soil quality to sustain crop and non-
crop productivity. The proposed limits are tentative and are presented in this report
primarily as illustrations. The rationale for the limits are described in the following
sections.
2.3.1 Soil Physical Condition
The physical condition of soil was determined by analyzing composited samples
for particle size and organic matter content. Each of these parameters is presented
separately in the first two subsections. A third subsection combines the results to
provide a more integrated evaluation of soil physical condition. CDFs accompany the
first two subsections but not the third.
32
-------
2.3.1.1 Particle size
Do the soils have acceptable amounts of clay to sustain crop production?
Particle size analysis (percent sand, silt, and clay) reflects an innate soil property
that is related primarily to geologic parent materials, transport processes, and soil age.
An ideal soil for most plant growth and for soil management contains between 18-35%
clay. Soils with >35% clay are difficult to till, conduct water too slowly, and are not
porous enough to allow for adequate supply of oxygen to the plants. Soils with <18%
clay tend to be droughty, do not retain nutrients very well, and compact easily.
CDF results for percent clay are presented in Figure 2.10. About 5% of the
sampled Piedmont soils and none of the Coastal Plain samples had "too much" clay (i.e.,
> 35%). About 55% of the Piedmont samples and 90% of the Coastal Plain samples were
subnominal because they had "too little" clay (i.e., < 18%). About 60% of the Piedmont
samples had nominal amounts of clay (i.e., 18-35%), compared to about 10% of the
Coastal Plain samples.
2.3.1.2 Soil organic matter in AP horizon
Do the soils have acceptable levels of organic matter in order to provide aeration to
the roots and retain nutrients?
Soil organic matter is an essential component of a healthy soil. Organic matter
influences: 1) the capacity of a soil to supply nutrients and trace metals to plants; 2)
infiltration and retention of water; 3) aggregation and structure that affect air and water
relationships; 4) cation exchange capacity; and 5) soil color, which impacts temperature
relationships (Nelson and Sommers, 1982). Soils with low organic matter (<1%) content
tend to be dense and less able to retain and provide nutrients to plants. In North
Carolina, soils with organic matter (>20%) are organic soils that have been drained.
These soils may have drainage problems or other limitations to the productivity of
annually harvested herbaceous crops, although they could support non-crop, wetland
vegetation. Organic matter content in the AP horizon is a more responsive monitoring
parameter than soil texture, because OM content is highly dependent on management
factors.
CDFs for percent organic matter are presented in Figure 2.11. Soil with less than
1% organic matter is tentatively propososed as being in unacceptable condition for plant
growth, and those with more than 1% organic matter as acceptable. None of the soil
samples had more than 5% organic matter except for eight samples from the Coastal
Plain. All of these samples came from one segment within the Coastal Plain. This area
is mapped by the USDA Soil
33
-------
Conservation Service as a conglomeration of different soil types, all of which are mucks
(i.e., organic soils).
About 30% of the samples in NC had amounts of organic matter <1%. More than
30% of the Piedmont soils had 1% organic matter compared to about 20% of Coastal
Plain soils. No samples in the Piedmont had > 5% of organic matter; the eight muck
soils from the Coastal Plain accounted for 10% of the area with > 5% organic matter.
34
-------
QUANTILES
5% 3.43
25% 7.17
50% 11.96
75% 22.25
95% 34.13
20 . 30
Clay
QUANTILES
5% 3.50
25% 7.33
50% 13.68
75% 26.23
95% 35.98
QUANTILES
5% 3.46
25% 6.86
50% 10.18
75% 14.13
95% 23.88
Figure 2.10 Cumulative distribution functions for percent clay in AP horizon.
Cumulative proportions refer to the land area cropped with annually harvested
herbaceous crops.
35
-------
QUANTILES
5% 0.40
25% 0.89
50% 1.58
75% 2.31
95% 3.47
Plodmont RoQlon |
Organic Ivlatter
7/
Coatatctl Rlaln R«£|[on
3 -4- B 0
Organic IVIatter
QUANTILES
5% 0.36
25% 0.83
50% 1.40
75% 2.26
95% 3.04
QUANTILES
5% 0.57
25% 1.09
50% 150
75% 2.06
95% 9.11
Figure 2.11 Cumulative distribution functions for percent organic matter in AP
horizon. Cumulative proportions refer to the land area cropped with annually
harvested herbaceous crops.
36
-------
2.3.1.3 Combined assessment of organic matter and clay
Organic matter data cannot be interpreted apart from clay content data because
there is a correlation between the two properties. Results from the 1992 Pilot indicate
higher organic matter content in finer-textured soils (Table 2.4). The exception to this
is the organic soils from the Coastal Plain, which have high organic matter content and
low clay content.
Table 2.4. Organic matter content for each soil textural type. Clay percentage classes
are established by the USDA Soil Conservation Service.
% Clay
0-18
18-35
>35
Number of
Samples
212
98
11
O.M. Content
Range
0.1-18.7
0.6-4.7
1.9-3.4
Average
1.74
2.17
2.46
Median
1.2
2.2
2.3
The results of the two properties were graphed together to provide a snapshot of
the proportion of soils that have acceptable amounts of both clay and organic matter to
sustain plant growth (Figure 2.12). For example, 27% of the Piedmont samples with
acceptable amounts of organic matter had unacceptable subnominal amounts of clay.
The information from the various sections of the graphs is summarized in Table 2.5.
2.3.2 Soil Chemical Properties
Soils with acceptable chemical properties allow nutrients to become available to
plants and are not laden with toxic substances that threaten biological health of the
agroecosystem. To assess soil chemical condition, we measured: pH (in water), cation
exchange capacity (CEC), available phosphorus, total cadmium, and total lead. Cation
exchange capacity and pH are largely responsible for the availability and supply of
nutrients to the plants. Phosphorus is an essential plant nutrient. Cadmium and lead
are micronutrients that are toxic to biota at certain levels.
37
-------
2O
18
16
14
CD
3=J 12
as
riCf
o
I"
S>
O
Q
6
4
2
r>
a
a
a
a a
r D
f
: ^ •
f • Aa •
: A A _ A -dm
". a H AA23A
LnA|||Slgl@ABA
.
A
.^ ;A -
.1
•
, '
i • A
i A A
v A A • A A
i i , , 1 , , i i i . ...i i_L
O
10
2O 3 O
% Clay
40
50
Figure 2.12 Percent clay in the AP horizon versus percent organic matter content
in the AP horizon in each soil sample from North Carolina. Soils with 1% or
more organic matter and between 18-35% clay have acceptable physical condition.
See Table 2.5 for a summary of the percentages of samples in each compartment
of this graph.
38
-------
Table 2.5 Percentages of samples occupied by each compartment of the graph from
organic matter and clay content in the AP horizon for each soil sample in Figure 2.12. t
Shaded areas show the percentages of soil samples with nominal amounts of both clay
and organic matter content for the various regions and for the entire state.
Piedmont
n=233
< 18% clay
18-35% clay
>35% clay
O.M.
29
0
1-5% O.M.
27
>5% O.M.
0
0
Coastal Plain,
n=80
OM.
15
1-5% O.M.
64
0
>5% O.M.
10
0
Mountain, n=8
O.M.
0
0
0
1-5% O.M.
63
0
>5% O.M.
0
0
State, n=321
O.M.
25
0
1-5% O.M.
37
>5% O.M.
0
39
-------
2.3.2.1 Soil pH
Do the soils have pH levels that facilitate the availability of essential plant nutrients?
Soil pH is an indicator of possible chemical constraints to the growth of roots and
other biological communities. Chemical constraints usually associated with pH include
the availability of inhibitory compounds (e.g., aluminum, salts), or a nutrient deficiency
(e.g., phosphorus fixation) (Pierce et al., 1991). At a soil pH <4.5, root growth is often
limited due to toxicity of soluble trace metals and H+ ions. At pH values of 4.5-5.8,
aluminum is in a soluble form, which limits growth of sensitive plants and
microorganisms. Soil pH 6.5 is non-limiting to root growth. For these reasons,
unacceptable pH values are considered to be <4.5, marginal values between 4.5-5.8, and
acceptable values >5.8 for North Carolina soils.
None of the soil samples from the Piedmont had unacceptable pH values, and
only one sample from the Coastal Plain was below pH 4.5 (Figure 2.13). This same
sample had 18% organic matter and was one of eight samples from one segment.
2.3.2.2 Cation exchange capacity
Do the soils have cation exchange capacities that enable nutrient storage and supply?
Cation exchange capacity (CEC) is defined as the sum total of exchangeable
cations that a soil can adsorb. Soils with low CEC are susceptible to leaching and are
unable to retain nutrients and slowly release them over the growing season. Organic
matter and clay content account for most of a soil's CEC. Soils with low CEC may
reflect poor management practices.
As shown in the CDFs (Figure 2.14), a higher proportion of Coastal Plain samples
had higher CECs than the Piedmont samples. These data need to be examined to
identify if the high CECs in the Coastal region are the organic soils (i.e., the mucks),
which are present in the wetland areas. .
40
-------
QUANTILES
5% 4.93
25% 5.37
50% 5.69
75% 6.00
95% 6.44
QUANTILES
5% 4.97
25% 5.41
50% 5:73
75% 6.07
95% 6.48
Coasted Plain Ftealon
QUANTILES
5% 4.85
25% 5.30
50% 5.58
75% 5.80
95% 6.21
Figure 2.13 Cumulative distribution functions for pH (in water) for AP horizon.
Vertical lines on the CDF represent boundaries for unacceptable (<4.5)and
acceptable (>6.5) condition.
41
-------
OEC (meci/lOO a)
QUANTILES
5% 1.48
25% 3.71
50% 5.88
75% ,8.70
95% 14.15
FModmorit
10 S>0 30
OEG e|/100 Q)
QUANTILES
5% 1.35
25% .3.17
50% 5.12
75% 7.35
95% 11.22
QUANTILES
5% 2i87
25% 6.02
50% 8.15
75% 12.70
95% 29.81
Figure 2.14 Cumulative distribution functions for cation exchange capacity (CEC)
in AP horizon. Cumulative proportions refer to the land area cropped with
annually harvested herbaceous .crops.
42
-------
2.3.2.3 Available phosphorus
Do the soils have adequate plant available phosphorus to sustain plant growth?
Phosphorus is a macronutrient that directly affects plant growth and health and
is thus a common component in fertilization programs on any given field. In North
Carolina, the soil phosphorus levels for Coastal Plain soils (using the Bray method with
dilute hydrochloric acid and sulfuric acid extraction method) are classified as follows:
<10 ppm, low; 11-31 ppm, medium; 31-56 ppm, high; > 56 ppm, very high (Olsen and
Sommers, 1982). According to this scheme, more than 50% of the samples had very high
phosphorus levels (Figure 2.15). Data are being re-evaluated to identify if the
classification scheme cited by Olsen and Sommers is appropriate for the Bray analysis
used on the Pilot samples.
2.3.2.4 Lead and cadmium
Do the soils have lead or cadmium levels that pose health risks to the ecosystem?
Lead and cadmium occur naturally in soils, but in excess quantities may threaten
the health of the ecosystem. Soils that receive applications of municipal sludge are
candidates for excess lead and cadmium levels, and soils near roads or some factories
may contain high lead levels due to automobile traffic and emissions.
Average concentrations for lead in soils range from 15 to 25 ppm (Burau, 1982).
A small proportion of the Piedmont samples contained relatively high levels of lead (>
40 ppm, Figure 2.16). Some of the Piedmont soils also contained relatively high levels
of cadmium (Figure 2.17).
43
-------
| Ftoglono ComtoinociJ
QUANTILES
5% 13.72
25% 34.58
50% 74.40
75% 152.28
95% 280.53
QUANTILES
5% 13.03
25% 33.10
50% 72.79
75% 150.84
95% 236.87
QUANTILES
5% 25.91
25% 42.51
50% 82.15
75% 154.08
95% 308.35
Figure 2.15 Cumulative distribution functions for available phosphorus in AP
horizon. Cumulative proportions refer to the land area cropped with annually
harvested herbaceous crops.
44
-------
Combinoc*
IS 20 2S
Load (f>p>m>
30 35
QUANTILES
5% 3.47
25% 6.48
50% 8.56
75% 12.24
95% 18.30
20 2S
m)
QUANTILES
5% 3.36
25% 5.85
50% 9.02
75% 12.84
95% 18.82
Gonatal F*la!n
is 20 2e
Lead CF>F>rn)
QUANTILES
5% 3.85
25% 6.87
50% 7.98
75% 9.33
95% 14.21
Figure 2.16 Cumulative distribution functions for total lead in AP horizon. Vertical
lines on the CDFs represent a range of lead concentrations in soils (Burau, 1982).
45
-------
QUANTILES
5% 20.20
25% 37.88
50% 53.14
75% 79.69
95% 149.01
FModmont Ftoplor* |
•t oo saoo
Cadmium, (p>p>t>)
F*l«Un Fta
QUANTILES
5% 20.11
25% 35.44
50% 49.77
75% 76.88
95% 141.72
QUANTILES
5% 20.60
25% 41.87
50% 59.53
75% 76.14
95% 178.60
Figure 2.17 Cumulative distribution functions for total cadmium in AP horizon.
Cumulative proportions refer to the land area cropped with annually harvested
herbaceous crops.
46
-------
2.4 SOIL BIOTIC DIVERSITY
Biotic communities within soil are responsible for the decomposition of organic
matter, involved in many aspects of nutrient cycling, provide mechanisms for the
development and maintenance of pore spaces through which gases and water flow, and
interact in beneficial and detrimental ways with plant roots. Although soil biotic
communities are influenced in many ways by natural elements of the soil physical
structure and chemical composition, anthropogenic influences, such as tillage practices,
cultural practices, and contaminants, also have profound effects on soil biotic
communities. The status of the soil biota is, thus, vital to the overall characterization of
soil condition or health.
Nematodes are found in most soils throughout the world. They are diverse in
their feeding habits and occur as central members of the soil food web. Nematodes can
be classified as bacterivores, fungivores, omnivores, plant-parasites, and predators.
Predaceous nematodes occur higher in the trophic hierarchy than bacterivores and
fungivores and are more sensitive to disturbance than the bacterivores and fungivores.
Because of the range of feeding habits, trophic structure of nematode communites
reflects the relative degree of stability or disturbance of the general biota in soils.
What is the degree of stability or maturity in nematode communities in soil?
A highly stable or mature community of nematodes would indicate that minimal
disturbance or contamination has occured in the soil. Maturity indices for nematode
community structure depend on whether particular nematode families are colonists
(generalists or r-strategists) or persisters (specialists or ^-strategists) (Bongers, 1990;
Ricklefs, 1990), and therefore, reflect relative disturbance. Disturbances to the soil biota
include cultivation, applications of agricultural chemicals, and soil compaction. A lower
value for the maturity index for free-living or plant-parasitic nematodes indicates more
disturbance, i.e., mostly colonizers and few persisters (Bongers, 1990; Neher and
Campbell, 1994; Yeates, 1994).
For land cultivated in AHHC in the Piedmont and Coastal Plains regions of North
Carolina (Figure 2.18 and 2.19), only a small proportion of communities of soil
nematodes had a relatively high degree of stability or maturity. The proportion of land
area cultivated with AHHC in North Carolina having values for the maturity index for
free-living nematodes of less than 3.2 on a 1-5 scale (where 5 is the most mature or
stable community) was estimated to be 0.95 (Figure 2.18). Although the median value
for the maturity index for free-living nematodes was slightly greater for the land area
cultivated with AHHC in the Coastal Plain (2.47) than for Piedmont (2.28), the reverse
was true for the maturity index for plant-parasitic nematodes (2.48 vs. 2.74; Figure 2.19).
47
-------
A higher proportion of larger values (>3.0) was found in the Piedmont -than in the
Coastal Plains region for both the free-living and plant-parasitic nematodes. Variability
of maturity index values was much greater for land area cultivated with AHHC in the
Coastal Plain than in the Piedmont region due to smaller sample size.
What is the degree of diversity of nematode communites in soil?
A diversity index such as the Shannon index of trophic diversity (Shannon and
Weaver, 1949) describes the relative abundance and evenness of the distribution of
nematodes across trophic or food-preference groups. In agricultural soils, higher
diversity of trophic groups is correlated with an increase in the less abundant trophic
groups, i.e. fungivores, omnivores, and predators, relative to the more generally
abundant trophic groups, i.e., bacterivores and plant-parasitic nematodes.
Trophic diversity (Figure 2.20) values had a greater range for land area cultivated
with AHHC in the Piedmont than in the Coastal Plain with more higher index values
occurring in the Piedmont. Variability of the values for the trophic diversity index was
also greater (i.e., wider 90% confidence band about the CDF) for the land area cultivated
with AHHC in the Coastal Plain region due to smaller sample size than that of the
Piedmont region.
48
-------
QUANTILES
5% 1.53
25% 1.96
50% 2.28
75% 2.57
95% 3.19
O -I 234
Maturity Index for Free — living IMemettodes
QUANTILES
5% 1.53
25% 1.95
50% 2.28
75% 2.56
95% 3.20
O 1 2 3 4
Maturity Index for Free — living Nemeitodes
Coastal Plain Rocjlon. n — 32
QUANTILES
5% 1.38
25% 1.97
50% 2.47
75% 2.74
95% 3.05
Maturity Index for Free — living rMematodes
Figure 2.18 Cumulative distribution functions for maturity index for free-living
nematodes in AP horizon of soil. Cumulative proportion refers to land area in annually
harvested herbaceous crops.
49
-------
Maturity Index -for f=lant Parasitic Nematodc
N/latLjrlty Index -for F=lant Parasitic Menniatodejs
IVisiturity Index -for FMant F"ara.sitio ISIemsitodes
QUANTILES
5% 2.19
25% 2.44
50% 2.70
75% 2.89
95% 3.22
QUANTILES
5% 2.20
25% 2.53
50% 2.74
75% 2.91
95% 3.23
QUANTILES
5% 2.17
25% 2.27
50% 2.48
75% 2.83
95% 2.96
Figure 2.19 Cumulative distribution functions for maturity index for plant
parasitic nematodes in AP horizon of soil. Cumulative proportion refers to land
area cropped with annually harvested herbaceous crops.
50
-------
O.OH- .
O
QUANTILES
5% 2.09
25% 2.81
50% 3.38
75% 3.72
95% 4.24
Shannon's Trophic: Diversity Index
1234.
Shannon's Trophic Diversity Index
Shannon's Trophies Diversity Index
QUANTILES
5% 2.09
25% 2.75
50% 3.35
75% 3.71
95% 4.24
QUANTILES
5% 2.32
25% 3.19
50% 3.49
75% 3.77
95% 4.06
Figure 2.20 Cumulative distribution functions for Shannon's trophic diversity
index for nematodes in AP horizon of soil. Cumulative proportion refers to land
area cropped with annually harvested herbaceous crops.
51
-------
2.5 WATER QUALITY
Water is essential to any agroecosystem, and it carries materials from the
agroecosystem to the larger landscape. Two key places where water may be found are
in ponds and underground. The 1992 Pilot Field Program sampled these water sources
for two types of contaminants: nitrate and pesticides.
What is the distribution of nitrate concentrations in wells on farms in North Carolina?
Specifically, what percentage of these wells have nitrate levels above the U.S. EPA
maximum contaminant level (MCL) of 10 ppm nitrate-N?
Manure and fertilizer applications can result in leaching of nitrate into
groundwater, where it is a potential health hazard. For example, nitrate can cause
methemoglobinemia in infants, which is why the 10 ppm nitrate-N MCL was established
(Bouchard et al., 1992; Fedkiw, 1991). Sixty-one wells were sampled across the state
(Table 1.5). Although it was not a survey requirement that the wells be used for
drinking water, about 90% of them were in that category. Samples were not taken of
any treated well water. The detection limit was 0.20 parts per million of nitrate-N.
Samples with detectable nitrate concentrations less than that value were assigned values .
of 0.20 ppm. .
From the distribution of nitrate-N concentrations in the sampled wells (Figure
2.21) an estimated 3.5% of wells had greater than 10 ppm nitrate-N. " Using 90%
confidence limits, the upper bound of that estimate is 7.2% of wells and the lower bound
is 0%. This result agrees with an unpublished study finding 3.2% of 9000 wells in North
Carolina to have such nitrate concentrations (cited in Spalding and Exner, 1993.) The
maximum concentration detected was 19 ppm nitrate-N. One of the three wells found
to have greater than IQppm nitrate-N was a hand-dug, butket-type well.
" ~ *' ' QUANTILES
5% = 0.0000
25% = 0.2369
50% = .0.9462
75% = 2.6424
95% = 7.1025
2 4 6 8 1O 12 14 IB 18 2O
nitrate — N concentration CP>P»rn) In wells
Figure 2.21 Cumulative distribution function for nitrate-N concentrations in wells on
North Carolina farms, Fall 1992. Vertical line shows the EPA maximum contaminant
level for drinking water.
52
-------
What is the distribution of nitrate concentrations in farm ponds in North Carolina?
Nitrogen can be a nutrient that favors growth of green algae, but it is only one
of the important elements (phosphorus is another). The nitrate concentration values by
themselves have little ecological significance (Burkholder, personal communication),
though levels of nitrate-N over 40-100 ppm may be harmful to livestock (Fedkiw, 1991).
Nitrate-N concentrations in pond samples (n=40) are illustrated by the CDF in
Figure 2.22. Only the samples taken from the boat are included. Where replicate
samples were taken in a pond, only the first was used. Concentrations were generally
less than those found in wells (note the difference in the scales on the horizontal axes)
with a median concentration in ponds of Q.2 ppm, about five times lower than the
median value for wells.
QUANTILES
5% = 0.0000
25% = 0.0666
50% = 0.2005
75% = 0.2822
95% = 2.2381
oontreitlon (ppm) Iri ponds
nltrato — M
Figure 2.22 Cumulative distribution function for nitrate-N concentrations in samples
from ponds on North Carolina farms, Fall 1992.
How extensive is pesticide contamination of farm ponds and wells in North Carolina?
There is concern about field-applied pesticides reaching ground and surface
waters, mostly because of their possible impact on human health. Water samples were
tested for the following pesticides, and none was detected at concentrations above one
part per billion (1 ppb). Carbamates (aldicarb, carbofuran, methomyl, oxamyl) were
analyzed by high performance liquid chromatography, the others by gas
chromatography.
° Insecticides/nematicides:
aldicarb, carbofuran, chlorpyrifos, endosulfan, methomyl, oxamyl
53
-------
o Herbicides
alachlor, atrazine, cyanazine, linuron, propachlor, simazine
Samples were taken in the fall, and it is possible that pesticides were present
earlier in the year, closer to the time when such chemicals are widely applied to
agricultural fields.
Chromatograms from a few of the samples showed the presence of unidentified
compounds not on the above list, but this was not pursued further.
54
-------
2.6 AGRICHEMICAL USE
How much stress are we putting on the ecosystems of North Carolina by applying
chemical fertilizers and pesticides?
Next to human health, this question is at the heart of people's concerns about
agricultural chemical use. Unfortunately, it is too broad to be addressed by simply
finding out how much of which chemicals were used on North Carolina fields, though
such data may be useful from four perspectives:
(1) As good or bad in themselves. According to one school of thought, sustainable
agriculture must rely as little as possible on off-farm inputs.
(2) As stressors to agroecosystems. Data on fertilizer and pesticide use can be used in
association with other indicators measured on agroecosystems. For example, nematicide
use will eventually be an important variable for association with the nematode-based
indices of soil biological condition.
(3) As variables against which to normalize. Fertilizer inputs will affect crop growth,
and should be taken into consideration when designing crop productivity indices. An
example of this is the nitrogen efficiency index shown in Section 2.2 of this report.
(4) As stressors to adjoining ecosystems and to ground water. There is a strong interest
in identifying agriculture as a stressor to other systems. Such concerns may be justified,
but that sort of assessment is beyond the scope of this report. Knowing what chemicals
were applied is only a first step.
Recognizing these limitations, the use of four materials is summarized in Table 2.6.
o atrazine:
o Bt
carbofuran:
an old and popular herbicide that has received recent publicity
because of concerns about human health effects. It has become a
water quality issue, and resistance to atrazine has developed in a
number of weed species.
Bacillus thuringiensis, a bacterium used as a biological control agent
for insect pests. Its use may be an indication that the farmer is
trying to rely less on synthetic pesticides.
a highly toxic insecticide/nematicide
55
-------
° nitrogen:
an important plant nutrient, applied to fields in large quantities, that
can leach to ground water. Only nitrogen from commercial
fertilizers is shown in the table.
Table 2.6 Estimated use of fertilizer nitrogen and selected pesticides on annually
harvested herbaceous cropland in North Carolina, 1992. Standard error is in
parentheses. Units are kg/ha, except for Bt which is in billion international units of
potency per hectare.
Material
atrazine
Bacillus thuringiensis
(Bt)
carbofuran
fertilizer nitrogen
Area (ha)
of AHHCs
treated at
least once
299,000
(41,000)
13,000
(7,000)
36,000
(14,000)
1,269,000
(76,000)
Proportion
of area of
AHHCs
treated at
least once
0.178
(0.022)
0.008
(0.004)
0.021
(0.008)
0.755
(0.024)
Average
number of
treatments
per year
1.06
(0.03)
1.00
1.00
1.72
(0.06)
Average
rate3
1.65
(0.12)
9.88
1.18
(0.08)
84.3
(4.3)
Sample
Numberb
54
3
7
242
"Average rates are active ingredient per treatment, except in the case of nitrogen which is per crop
year.
bNumber of sample units on which there were applications of the material.
TSTo standard error exists because all data values used to compute the mean were equal.
All estimates in the above table were derived solely from the Fall 1992
questionnaire, except that the areas treated were calculated by multiplying the
proportion of area treated by the estimate of the total area of annually harvested
herbaceous crops. This estimate was made using the June Enumerative Survey data.
Interestingly, the pesticides were each found on a specific crop in our sample. All
applications of atrazine on sample fields were to corn (grain or silage) or popcorn. All
of tine fields with recorded carbofuran applications were either corn for grain or corn for
silage. All of the Bt applications were to tobacco.
56
-------
As expected, nitrogen applications dominate the picture, both in terms of area
treated and the amount applied. This lends further support to the importance of
nitrogen efficiency (Section 2.2).
Atrazine was used on a fairly substantial portion (almost 18%) of the area under
annual crops. The area to which carbofuran and Bt were applied is much smaller, so
small that the estimates associated with these materials have rather large standard errors.
57
-------
3. EVALUATION OF DESIGNS, INDICATORS, AND ACTIVITIES
3.1 PRELIMINARY DESIGN COMPARISON
Design comparisons were performed primarily for fall questionnaire and soil
sample data. Land-use indicators took advantage of the full June Enumerative Survey
dataset in order to get the most precise estimates possible. Other indicators, including
soil quality-physical/chemical, soil biotic diversity, and crop productivity indices, were
only calculated from the fall questionnaire and soil sample data. The design comparison,
then, is most relevant with respect to the these indicators.
Only statewide estimates were used and the comparison was done both
accounting for, and ignoring, the stratification (Cotter and Nealon, 1987) of the
Rotational Panel plan. Cost and precision have been considered in the evaluation of the
relative efficiencies of the two sampling plans.
Typically, the efficiency of one design relative to another is defined as the ratio
of the variances for the statistic of interest, adjusted for different sample sizes when
necessary. For example, the relative efficiency of design one to design two is defined
as the variance under design two divided by the variance under design one with respect
to the parameter of interest. In EMAP the parameter of interest is the cumulative
distribution function (CDF). A method which evaluates the overall precision of the CDF
was used for comparing the two sampling plans for chemical measures of soil quality
and crop productivity indicators. Nematode indices could not be used because
nematode data were collected only with the Rotational Panel plan. Agrichemical indices
are not included since they are not summarized with CDFs in this report. Crop specific
indices of crop productivity are not presented because of relatively small sample sizes.
For the relevant fall questionnaire and soil sample indicators, a CDF was computed
under each plan. Bu was defined as:
where Varh(t) is the estimated variance at t for the CDF under the Hexagon plan and
Var,p(t) is the estimated variance under the Rotational Panel plan. The sums are over
equally spaced increments ranging from
tj =max(min(Hexagon dataset), min(Rotational Panel dataset)),
to
t40=rmn(max(Hexagon dataset), max(Rotational Panel dataset)).
58
-------
The limits of the sum were defined as such to prevent the relative efficiencies from being
dominated by variances in the tails of the CDFs. The method was also tried for a 20-
point 't' vector and for the combined data 't' vector; results were very similar. To adjust
for differences in sample sizes, Bu was multiplied by (n^/nj, i.e., the ratio of the sample
sizes. This adjusted value is a measure of the relative efficiency of the Hexagon
sampling plan to the Rotational Panel sampling plan, ignoring costs (Table 3.1).
Table 3.1 Relative efficiencies3 of the Hexagon design to the Rotational Panel design.
Indicator
Ignoring Stratification
Relative
Relative Efficiency w/
Efficiency Cost Factored in
Accounting for Stratification
Relative
Relative Efficiency w/
Efficiency Cost Factored in
Soil Quality-
Physical/Chemical
Phosphorus 0.81
Cadmium 0.80
CEC 0.72
Clay 1.00
Lead 1.13
Organic Matter 0.78
Sand 1.04
pH 0.90
Crop Productivity
0.58
0.57
0.51
0.71
0.81
0.56
0.74
0.64
0.80
0.78
0.69
0.97
1.11
0.77
1.02
0.89
0.57
0.56
0.49
0.69
0.79
0.55
0.73
0.63
Nitrogen Efficiency (using nitrogen from commercial fertilizer and manure)
All Seed Crops 0.80 0.57 0.77 0.55
Observed/Expected Yield
All Crops 0.95 0.68
(all crops for which expected yields were available)
0.92
0.66
aA relative efficiency less than one implies that the Rotational Panel design is more
efficient, a relative efficiency greater than one implies that the Hexagon design is more
efficient. ,
Relative efficiencies greater than one indicate that the Hexagon design is more
efficient, whereas relative efficiencies less than one indicate that the Rotational Panel
design is more efficient. For example, consider soil chemistry
59
-------
when stratification is ignored in the variance calculations tinder the Rotational Panel
design. For estimating phosphorus content in soil, the Hexagon design was estimated
to be 81% as efficient as the Rotational Panel design. When costs were factored in (to
be described below), the Hexagon design was estimated to be only 58% as efficient as
the Rotational Panel design. For estimating lead content in soil, on the other hand, the
Hexagon design was estimated to be 13% more efficient than the Rotational Panel design
when costs were ignored but only about 81% as efficient when costs were accounted for.
Results are nearly the same when variance calculations accounted for stratification.
The adjusted relative efficiency can be thought of as the information per sample
unit of the Hexagon plan relative to the Rotational Panel plan, where information is
defined as the inverse of the average variance. To factor in costs, total costs, excluding
salaries of the Agroecosystem Resource Group staff, were documented for each plan
using records kept by NASS (Table 3.2). The Hexagon sampling plan required NASS
enumerators to visit segments in June that they normally would not visit and, hence,
costs were more per sample unit.
Table 3.2. Costs of the Hexagon and Rotational Panel sampling plans.3'b • . .
Item
Constructing segments
(Prorated over 20 years)
June Enumerative Survey
SUBTOTAL
Number of segments
Cost per segment
Fall survey costs0
NASS labor & administration
Equipment, shipping & lab analyses
TOTAL
Number fall questionnaires listed
as complete
Cost per fall survey sample unit
•ir-ii ""'" ' . I ' ' l ' "i"' " " :' 1 ...—~
$
Hexagon
$ 586
$ 17,289
$ 17,875
51
$ 350
$ 13,005
$ 14,388
$ 45,268
138
328
Rotational Panel
0
$ 8,025
$ 8,025
65
$ 123
$ 16,575
$ 19,323
$ 43,923
188
$ 234
"Since nematode data were only collected from the Rotational Panel plan;
are not included.
'Table entries are rounded.
°Cost differences for the Fall Survey reflect only differences in sample sizes.
nematode costs
60
-------
The relative cost per sample unit of the Hexagon Plan ($328.00) to the Rotational Panel
plan ($234) was calculated as 328/234=1.40. The relative information per unit was
divided by the relative cost per unit to obtain the information per dollar of the Hexagon
Plan relative to the Rotational Panel Plan. This index is a measure of the relative
efficiency accounting for differences in costs (Table 3.1).
Accounting for stratification had little effect on the relative efficiencies. This is
not surprising because NASS's stratification is designed to improve estimates of extent
from segment level data and not specifically designed to improve estimates of field level
data, e.g., fall questionnaire and soil sample indices. When costs were ignored for soil
quality, the estimated relative efficiencies seemed to favor the Rotational Panel design
for five of the eight variables and seemed to favor the Hexagon design for only lead
content. The soil chemistry measures were, however, not all independent. The relative
efficiencies, ignoring costs, for the crop productivity indicators nitrogen efficiency and
observed/expected yield indices indicated that the Rotational Panel design was at least
as efficient as the Hexagon design for both indices. When costs are factored in, the
Rotational Panel design appears to be more efficient for all fall questionnaire and soil
sample indices examined.
Entries in Table 3.1 are estimated relative efficiencies and, therefore are subject to
sampling variability. Assessing the variability in the estimated relative efficiencies is
difficult and we made no attempt to do so. Two conclusions can be reliably drawn from
Table 3.1. First, the Rotational Panel design is no less efficient than the Hexagon design.
Second, the Rotational Panel design is the more cost-effective design.
In addition to the above comparisons for fall data, the two sampling plans were
compared with respect to the estimate of extent of annually harvested herbaceous crops
from the June data (Table 3.3). Relative efficiencies in Table 3.3 were derived from the
estimated variances of the estimates. NASS samples high-agriculture strata with greater
intensity than they sample low-agriculture strata and, thus, segments in the Rotational
Panel sample tend to be more agricultural than do segments in the Hexagon sample.
The impact of NASS's stratification is evident in Table 3.3 where the Rotational Panel
Plan appears to be considerably more efficient than the Hexagon Plan in the estimation
of extent of annually harvested herbaceous crops.
61
-------
Table 3.3 Relative efficiency3 of the Hexagon design to Rotational Panel design with
respect to estimation of extent of annually harvested herbaceous crops.
Indicator
Ignoring Stratificationb
Relative
Relative Efficiency w/
Efficiencv Cost Factored in
Accounting
Relative
Efficiency
for Stratificationb
Relative
Efficiency w/
Cost Factored in
Extent of AHHC 0.63
0.22
0.43
0.15
"A relative efficiency less than one implies that the Rotational Panel design is more
efficient, a relative efficiency greater than one implies that the Hexagon design is more
efficient.
''Refers to whether stratification was accounted for in variance computations.
Both designs provided good spatial coverage of North Carolina with the 51
Hexagon segments falling in 49 counties and the 65 Rotational Panel segments
distributed over 55 counties. Prior to drawing the sample for the fall survey, the lists
of acres were ordered first by crop and then by segment number. The ordering by crops
diminished the spatial coverage of the two sampling plans for the fall questionnaire and
soil sample. Because segments were numbered according to location, and the lists are
sampled systematically, ordering first by segment number should help maintain the
spatial coverage. For the 1993 Nebraska Pilot the annually harvested herbaceous
cropland acres will be ordered only by segment number and the spatial coverage will
be preserved better.
In summary, when differences in costs were considered, the Rotational Panel
design appeared to be more efficient than the Hexagon design. If costs are ignored, the
Rotational Panel design tends to perform as well or better than the Hexagon design for
most fall questionnaire and soil sample indices, with the only exception being lead
content of soil. For estimation of extent from the June Survey, the Rotational Panel
design was more efficient than the Hexagon design.
62
-------
3.2 SUCCESSES AND CHALLENGES IN INDICATOR DEVELOPMENT AND
EVALUATION
3.2.1 Crop productivity
At least three important questions surround the development of a crop productivity
indicator. These relate to combining data across crops, indicator performance, and
conceptual and operational challenges. Something about each of these was learned in
the 1992 Pilot.
How can data from different crops be brought together in a single productivity index?
In order to make broad regional statements, it would be useful to be able'to "add
apples and oranges" into one overall index of productivity. A simple method of
standardizing yield was attempted with the data from the 1992 Pilot.
Ideally, standardization of any index should rely on a long-term baseline of data.
As a test, county average yields were used. These were obtained from NASS for the
period 1980-1990. The index summarized in Figures 3.1 and 3.2 is calculated by dividing
the reported yield for each sample unit by this 11 year county average. The composite
graphs in Figure 3.1 include most but not all AHHCs: barley, field corn (maize), cotton,
hay, oat, peanut, grain sorghum, soybean, sweet potato, tobacco, and wheat. Data were
available to calculate this index for 77% of the valid sample units, much more than^the
49% for the combined nitrogen index for seed crops (Section 2.2).
The median values of all the illustrated observed/expected indices are greater
than 1.0, indicating that technology or perhaps weather has caused the yields as a whole
to be greater than they were during the 11-year reference period. This index presented
for a single year has little meaning in itself, except to allow many species to be plotted
on the same graph. Over the long term, however, this type of index should reflect
changes and trends in overall crop productivity.
It is interesting to look at the pieces that go into the overall index. The CDFs for
the observed/expected index for three individual crops (Figure 3.2) are quite different
from each other. The median index value for land in soybeans is 1.04, indicating a near-
average situation. Land in corn showed greater productivity than during .the baseline
years, and almost all of the distribution for hay lies to the right of the 1.0 line.
63
-------
[ Regions Comtolnqd, r> •—
O.OO O.25 O.SO O.75 1.OO 1.25 1.5O 1.75 2.OO 2.25 2.5O
Observed/Expected 'Yield — Composite
QUANTILES
5% 0.59
25% 0.99
50% 1.17
75% 1.45
95% 1.99
QUANTILES
5% 0.56
25% 0.96
50% 1.14
75% 1.49
95% 2.30
O.OO O.25 O.5O O.75 1.OO 1.2S 1.SO 1.75 2.OO 2.25 2.5O
Otosorved/Expected "Vleld — Composite
O.OO O.2S O.SO O.76 1.OO 1.25 1.5O 1.7S 2.OO 2.25 2.6O
Yield — Composite
QUANTILES
5% 0.65
25% 1.01
50% 1.18
75% 1.44
95% 1.86
Figure 3.1 Cumulative distribution functions for the observed/expected yield
index, composite for land area cropped to barley, field corn (maize), cotton, hay,
oat, peanut, grain sorghum, soybean, sweet potato, tobacco, and wheat.
64
-------
Observed/Expected Yield
QUANTILES
5% 0.51
25% 0.97
50% 1.35
75% 1.50
95% 1.88
e= O.B
'O.O O.S 1.0 1.S &.0 2.S 3.0 3.S -t.O •O.-S B.O
Observed/Expected Yield — Hay
QUANTILES
5% 0.91
25% 1.07
50% 1.61
75% 2.46
95% 4.72
Observed/Expected Yield — Soybeans
QUANTILES
5% 0.52
25% 0.84
50% 1.04
75% 1.19
95% 1.61
Figures 3.2 Cumulative distribution functions for the land area cropped to field
corn (maize), hay, and soybean, respectively. A vertical reference line shows
where the index value is 1.0.
65
-------
How did the two crop productivity indices perform?
Variability is the only performance criterion that has been examined so far. In
order to detect trends it is desirable that within-year variance and random year-to-year
fluctuation be small compared to the trend "signal". Low noise would also be helpful
for making comparisons among regions. Tables 3.4 and 3.5 show the coefficients of
variation (CV) for the nitrogen efficiency index and the observed/expected ratios, as
compared to the CVs for the corresponding unadjusted yields. Number of sample units
differ between the two tables; see "Data completeness and quality" below.
Table 3.4 Comparisons of coefficients of variation (=100*(a/u)) for the nitrogen
efficiency index and unadjusted yields (weight per unit area). Data for the entire
state of North Carolina. Nitrogen calculations based on manure plus commercial
sources.
Category
eight seed crops, combined data
corn for grain
wheat
soybean
cotton
CVfor
unadjusted yield
58
37
21
34
22
CV for nitrogen
efficiency index
100
80
62
139
89
No. of
sample
units
146
60
31
57
15
Table 3.5 Comparisons of coefficients of variation (=100*(a/u)) for unadjusted
yields and for the observed/expected yield index. Data presented are for the
entire state of North Carolina.
Category
all crops with available data
corn for grain
soybean
hay
CV for
unadjusted yield
83
36
34
72
CVfor
observed/
expected index
46
33
31
63
No. of
sample
units
231
67
58
32
66
-------
In every case, the nitrogen index shows a much greater CV than the
corresponding unadjusted yield. This is disappointing, and further work is needed to
find the best indicators of efficiency. The observed/expected ratio gives slight to
moderate reductions in the coefficient of variability. The expected yields used in the
denominator were county average yields for each crop, so the resulting index essentially
corrects for crop species (a substantial effect) and location within the state.
What challenges were met, and what obstacles remain?
Data completeness and quality
The index values, especially the combined index for nitrogen efficiency, include
data from only a portion of the sample. Sometimes sample units were unusable because
questionnaire data were simply missing, for example when crops were not harvested by
the time of the fall visit. In other cases there was a lack of conversion factors or
reference yields, leading to the exclusion of otherwise usable data. Further research is
needed in this area. The inclusion of nitrogen from manures is a special challenge (see
below).
Many issues surround the data that go into the indices. The factors used to
convert yields to dry weights and to estimate the nitrogen content of manures are of
unknown accuracy. The survey data themselves are subject to a number of nonsampling
errors.
Where yield=0, data could not be used in the nitrogen efficiency index (no
division by zero). This may skew the distribution downward slightly. For the nitrogen
use efficiency of seed crops, 24 such sample units were excluded. It is difficult to
determine why those yields were 0, but in 14 of the 24 samples, the crop seemed not to
have been intended for harvest (usually a cover crop). Crop failure is the apparent
explanation for the other 10 sample units. To include those units in the calculation
would be to ignore the fact that biological production was not actually zero. This index
is not strictly biological, of course, since only the harvested portion of biomass is
considered. When yield data were available for only one of two crops in a single year,
that sample unit was included in the nitrogen efficiency index for the individual crop
but excluded from the composite. Many other such judgements had to be made during
index development.
Estimating the amount of nitrogen contributed by application of manure
The estimated nitrogen content of manures was assigned to the crop to which the
material was applied, though realistically only about half will be available for plant
uptake during the first year. No conversion factors have been obtained for cases where
the animal waste was in the form of a slurry (n=4 sample units). Sample units that
67
-------
received sewage sludge applications in the last five years (n=3) were excluded from the
calculation, since no details were known on the rate or time of the sludge application.
If a cover crop received manure or commercial fertilizer, the nitrogen will be credited
to the next harvested crop, though this situation did not occur for the seed crops
illustrated in this report.
An alternative nitrogen efficiency index can be calculated that includes only
commercial sources of nitrogen, which depend heavily on nonrenewable resources.
Fewer conversion factors are necessary to calculate this index. The resulting CDF is not
illustrated, but it is nearly identical to Figure 2.9 for the state, because so few sample
units received manure applications.
Classifying index values as good or bad
A major limitation at this stage is that no acceptable/unacceptable interpretation
can be given for the crop productivity indices. Such cutoff points must either be defined
or the index will have to be used only for the difficult job of trend detection.
Future directions for indices of crop productivity
Nitrogen use efficiency and standardized yield are only two of many possibilities
for evaluating productivity in agroecosystems. It is suggested by Lai (1991) that the
most limiting or nonrenewable input be used as a part of a larger index to quantify
sustainability. In this spirit, water use efficiency may be included when the ARG begins
work in the western states. Regression approaches also may be tried.
The use of a simple long-term average in the observed/expected index is only for
testing purposes. Variants of it may be used in associative studies. It is desirable to
have expectations that are somehow tied to soil map unit as well as location, but this has
not yet been accomplished because the expected yield data available from the Soil
Conservation Service are not all from the same time period. An adjustment for time
trends will be needed before the data are useful.
In addition to soil and management, the key variable driving productivity is
weather. Neither of the indices shown here accounts for variability due to temperature,
precipitation, etc., and efforts are underway to use crop simulation models to do so.
68
-------
3.2.2 Successes and Challenges in Assessing Soil Quality
We have met with a number of challenges in attempting to define and assess soil
quality. Issues emerged related to societal values, integrating soil assessments with
other indicator assessments, relating the measured values to agroecosystem condition,
and identifying whether or not our values represented sustainable conditions or not. But
our first task was to evaluate the soils data. In many ways, we haven't graduated
beyond this step, although after accomplishing this, we have a much clearer direction
of where we need to go.
Our first step was to make the task of data evaluation manageable, and we
accomplished this by examining each of the distributions of the various parameters and
developing CDFs for each. Then we assigned ranges of acceptable and unacceptable to
each of the parameters in relation to crop production. These ranges were related to
generic plant growth; for example, we identified the "ideal" ranges of clay content, pH,
etc. We then grouped the parameters according to whether they related to soil physical,
chemical, or biological condition. Soil physical condition was estimated by evaluating
organic matter and clay content; soil chemical condition was determined by evaluating
pH, cation exchange capacity, available phosphorus, and some trace elements; and soil
biological condition was evaluated by evaluating the nematode communities. Framing
our data in this way enabled us to pose fairly specific assessment questions, which in
turn facilitated our data summaries.
We found this approach to be very useful for summarizing data about individual
soil properties, but after all our summaries were complete, we noticed that we still could
not draw very meaningful conclusions about soil quality as related to agroecosystem
condition. For the soil physical properties, then, we combined the clay and organic
matter data and redefined the proportion of agroecosystems in North Carolina that had
"nominal" soil physical condition. While this approach was useful, it does not go far
enough. We recognize that additional data integration needs to proceed. Data need to
be integrated within indicators (such as clay and organic matter) but also need to be
integrated across categories. We identified three deficiencies in our approach to the 1992
Pilot data. These are described below, along with potential approaches to addressing
these issues in forthcoming studies.
3.2.2.1 The approach did not account for different plant requirements.
Data evaluation did not incorporate information about which plants were being
grown in the soil or the specific requirements of the plant. Implicit in this approach is
that there is an "ideal" soil for all types of annually harvested herbaceous plants. This
approach required that the ranges for soil properties be sufficiently large to include the
span of requirements for a wide range of crops, which included crops grown extensively
such as wheat, corn, cotton and tobacco, and crops grown in relatively small areas such
69
-------
as watermelons and cucumbers. The data ranges could be narrowed if we took into
account the specific soil requirements for each crop.
We possess a wide range of crop and management information for each of the
fields from which our soil samples were drawn. We should identify the specific crop
that is grown in each soil and assign a suitability value for each parameter that relates
the crop requirements to the soil conditions. Using this approach would then enable us
to evaluate whether the soil was being used for the type of crop for which it is best
suited. Such information may be useful for linking soil parameters to sustainability or
agroecosystem condition.
3.2.2.2 Agronomic potential and management information are not considered.
Our approach to data analysis does not consider soil potential or land
management practices. Consequently, there is no way to determine whether a soil is in
good or bad condition relative to its potential. Management practices are an integral
factor of sustainability in agroecosystems.
Some soils have naturally high potential for crop and non-crop productivity, while
others do not. It is not enough to merely measure soil properties and proceed with an
assessment of sustainabilty. The link between soil properties and sustainability requires
information about management factors. Examination of the management factors alone
does not give us information about whether the management practices are producing
desired results. Examination of the soil properties alone does not provide information
about sustainability. Soil property evaluation may show a "good" quality soil in the
midwest and a "marginal" soil in North Carolina. The reality may be that the deep,
midwestern loessial soil may be in the process of degrading due to rill erosion or
continuous cropping, whereas the North Carolina soil may be holding steady or
improving due to responsible management. Future data analyses will integrate
information about management practices and soil type to arrive at an indicator of
sustainability.
3.2.2.3 Societal values were not integrated into the assessment.
The integrated assessment of clay and organic matter contents revealed some low
clay content soils with unexpectedly high amounts of organic matter. A closer
evaluation of these outliers revealed that these supposedly "nominal" or "optimal" soils
were actually organic (muck) soils, probably drained wetlands. This finding challenged
our soil raring system by touching the nerve of wetlands protection. In turn, this
challenged our agricultural approach of evaluating soils and confirmed that we were
really evaluating "crop-ability" of the soils, and not tying this evaluation to other societal
values and issues. In order to resolve the issue of how to evaluate cropped wetlands,
we may need to frame our questions in the context of the much-debated societal values.
Since the societal values, as well as the indicators, are still evolving, we do not have a
70
-------
recommended procedure for addressing this elusive issue. At this point, we simply note
that it is an issue and will continue to be aware of potential conflict of evaluating the
"crop-ability" of soils without considering the broader issues of societal values.
3.2.2.4 Some successes were realized in simultaneously analyzing two soil
properties.
For soil physical properties, we have met with some successes in jointly analyzing
the clay and organic matter data. The integrated evalulation of clay and organic matter
contents revealed that the two properties were inversely related. This leads us to
conclude that it may be necessary to establish organic matter ranges for each range of
clay class (0-18%, 18-35%, and over 35% clay). The deterministic relationship between
these properties requires that we develop indices of soil quality that integrate a myriad
of properties. The integration of clay and organic matter contents is an attempt to
integrate factors to make statements about physical condition of the soil; now we need
to attempt to integrate information about physical, chemical, and biological condition,
as well as soil potential and management factors to make statements about the overall
soil condition and sustainability. Our first step towards such a holistic assessment is
framingxthe data in the context of the soil type. This will free us from the concept of an
ideal soil, and will preclude us from concluding our investigations with a map of
agroecosystem condition that looks remarkably like a Major Land Resource Areas map.
3.2.2.5 Suggestions for future efforts.
While we met with some successes in analyzing organic matter and clay contents
together, we recognize that we still fall short in incorporating all of the factors that we
need for a truly integrated assessment. Several soils with nominal organic matter, for
example, had undesireable pH levels. It is clear that several factors need to be
considered simulataneously, and it may not be appropriate to arrive at an assessment
of soil physical condition in addition to soil chemical condition and soil biological
condition. We may need to integrate all three of these into one finding. We will
approach this integrated effort by utilizing the Soil Conservation Service's Soil Rating
for Plant Growth (SRPG) model, developed by Joyce Scheyer at the National Cooperative
Soil Survey. This model incorporates information about the surface properties of soil
(which we could measure) and some profile information pertinent to the soil series
(which we could obtain from the SCS databases). Results from this model will get us
closer to a truly integrated assessment for plant productivity, yet we will still need to
incorporate information about management practices and biological indices to approach
an assessment of sustainabilty.
In addition to using the SRPG model, we will design our next sample collection
around the use of some intact soil samples. In the 1992 Pilot, we utilized composited
samples collected by NASS enumerators. These samples were comprosed of 10
subsamples that were homogenized in the field then ground up in the laboratory prior
71
-------
to analysis. This approach to sample analysis precludes any evaluation of the soil tilth.
While tilth is largely a qualitative and intuitive measure, it can be semi-quantified by
measuring bulk density, pore size distribution, aggregate size and stability, mechanical
impedance, organic matter content, and some measure of water movement (e.g.,
infiltration, permeability, or hydraulic conductivity). Most of these measurements
require intact core samples rather than ground samples. We plan to use John Doran's
(USDA-ARS) kit for collecting and analyzing field samples, which he developed
specifically for assessing soil quality.
Other ideas that are being considered include additional measurements and
analyses and different ways of analyzing the data, e.g., with the use of indices. These
are briefly outlined below.
Measurements. We are currently looking at comparing bulk density of soil cores
with reconstituted bulk density, which is determined in the laboratory by grinding,
packing, and wetting the soil to create a controlled sample for comparison purposes. A
ratio of reconstituted to field bulk density could be compared across all fields and soil
types.
Ratios of cation exchange capacity to clay may also be evaluated in the different
soil types to identify the theoretical maximum clay fraction of the soil material, and
compare it with the measured clay fraction. Since CEC is also related to organic matter,
we would expect to see relationships between the theoretical versus measured clay
content and the organic matter content of the soil.
Further studies could evaluate the relationship between soil pH and macro and
micronutrient levels in soils. Consideration should be given to native soil conditions
(e.g., jarasite and sodium-affected soils) and to management practices (e.g., lime
application, irrigation). A ratio of lime application to pH could be attempted to examine
the efficiency of lime application, which would be related to sustainability of
management practices.
Analyses. Subsequent analyses are planned to examine variance components
resulting from grouping the soils by textural class and climatic regime, and possibly by
soil order (e.g., forest soil, grassland soil, wetland). Such an approach will attempt to
facilitate comparisons among and between soil types. Future efforts will also focus on
the type of crops that are supported and the resulting yield from those crops. This
approach will facilitate the comparison of the expected to the observed yield based on
soil type.
Indices. Several possible approaches to providing an index of soil quality are
possible. To answer the question of whether the conditions are suitable for plant
growth, we will evaluate the SPRG model developed the SCS National Cooperative Soil
Survey. This model provides incorporative, implementive measures of soil physical and
72
-------
chemical quality and a rating for potential biomass production. The model does not
currently incorporate any measures relating to the soil biota; however, attempts will be
made to add this component to the model. The model has also been tested only in
South Dakota and Indiana, and thus we intend to validate the model for other regions.
Calibration may be based on soil type or climatic data.
3.2.2.6 Components of variance and reliability ratios for pilot data
Analysis of the 1992 Pilot data revealed that the current sampling design of one
transect per field and one composite sample per transect is sufficient to detect differences
among segments and fields. Except for cadmium, the ranking of variance components
from high to low was among segments, among fields, among transects, and within
samples (Table 3.6). For cadmium, the ranking from high to low variance was among
transects, among fields, among segments, and within samples. Similar variance
component analyses will be applied to the SPRG model results to identify if differences
can still be detected when all the data are combined.
Table 3.6 Variance components for related soil chemical and physical measures.
Among
Parameter Segmentsfa2.)
Sand(%) 354.72
Clay(%) 94.12
pH (1:1 water) 0.10
Organic Matter (%) 1.75
Phosphorus (ug/ml) 3920.
Cation exchange 13.43
capacity (meq/lOOg)
Lead (ug/ml) 15.73
Cadmium (ug/ml) 775.9
Among
Fields(cF2f)
64.41
16.68
0.07
0.65
3164.
4.58
10.03
823.4
Among
Transects(g2f)
20.98
5.93
0.04
0.60
1139.
3.13
2.58
940.3
Within
Samples(q2^)
4.00
3.56
0.02
0.04
419.
0.83 .
1.08
124.1
Reliability ratios were calculated to determine the amount of signal versus noise in
the data. Reliability ratio is defined as variance (signal)/(variance(signal) + variance(noise))
where the "signal" refers to the sampling or among segment and among field variability
and "noise" refers to the measurement or within field and among samples variability
(Table 3.7). As a reference point, a reliability ratio of 0.5 implies that the variance due to
noise is as large as the variance of the signal. Conversely, a reliability ratio near 1.0
indicates that the variability from noise is a very small part of the overall variability.
Ratios greater than 0.9 were calculated for sand content (%) and clay content (%) for all
regions combined and separate regions with a sampling design that includes one transect
per field and one laboratory determination per transect or composite soil sample (Table
3.7). Reliability ratios were greater than 0.70 for soil pH, organic matter (%), extractable
phosphorus, and total lead for separate regions with a sampling designed with one
73
-------
transect per field and one laboratory determination per transect (Table 3.7). For cation
exchange capacity, reliability of estimates in the Coastal Plain region of NC exceeds 0.70
with a design of one transect and one laboratory determination per transect, but a design
of two transects and one laboratory determination per transect is necessary to exceed a
reliability ratio of 0.70 in the Piedmont region (Table 3.7). For total zinc, estimates of
reliability equal 0.9 in the Piedmont region of NC with the current sampling design, but
require one transect with two laboratory determinations per transect to exceed a reliability
ratio of 0.7 in the Coastal Plain region of NC (Table 3.7). For cadmium, estimates of
reliability equal 0.81 in the Coastal Plain of NC with the current sampling design but
require two transects with one laboratory determination each per field to exceed a
reliability ratio of 0.70 in the Piedmont region of NC (Table 3.7).
Gabriel's biplot is an informative plot derived from principal component analysis
that visually shows (1) the relationship among the independent variables, (2) the relative
similarities of the individual data points, and (3) the relative values of the observations for
each independent variable (Rawlings, 1988). It is called a "biplot" because both row
(observation) and column (variable) information are displayed in one plot. The vectors
in the plot represent the variables. If a vector length is close to unity, then it is well
represented by the plot. If any vectors are not close to unity, then they are poorly
represented by the plot and determining relationships among these variables should be
avoided. The angle between two well represented variable vectors reflects their pairwise
correlation. The correlation is the cosine of the angle because the variables were centered
before the eigenanalysis was done (Rawlings, 1988), Therefore, angles of 0, 90, or 180
degrees indicates correlations of 1.0,0.0, or -1.0, respectively. The numbers on the bottom
and left axes are the lengths of the projected vectors in that dimension. The numbered
points (+) on the graph represent the observations. Points close together have similar
values with respect to the set of independent variables, and vice versa. The numbers on
the upper and right axes are scale values for the numbered observations for the
corresponding principal components.
The biplots illustrate associations among soil chemistry variables for all regions
combined (Figure 3.3), Piedmont region (Figure 3.4), and Coastal Plains region (Figure 3.5).
As expected, Bray 1 and Bray 2 methods for measuring extractable phosphorus were
highly correlated, and sand and clay content were inversely correlated (Figures 3.3-3.5).
Amount of calcium and magnesium found in soil samples was moderately correlated, and
percent organic matter and clay content were positively correlated in the Piedmont region
but not in the Coastal Plain region (Figures 3.4 and 3.5).
74
-------
Table 3.7 Reliability ratios for soil chemical and physical measures3 for the Piedmont and
Central Plain regions and for regions of NC combined.
Parameter tb
% Sand 1
1
2
% Clay 1
1
2
PH (1:1) 1
1
2
% O.M. 1
1
2
P (Bray II) 1
1
2
CEC 1
1
2
Zn 1
1
2
Pb 1
1
2
Cd 1
1
2
dc
1
2
1
1
2
1
1
2
1
1
2
1
1
2
1
1
2
1
1
2
1
1
2
1
1
2
1
Piedmont
Region
0.96
0.96
0.98
0.91
0.92
0.95
0.73
0.75
0.84
0.80
0.81
0.89
0.92
0.93
0.96
0.62
0.63
0.77
0.90
0.92
0.95
0.90
0.91
0.95
0.55
0.56
0.71
Coastal Plain
Region
0.92
0.93
0.96
0.94
0.96
0.97
0.79
0.86
0.88
0.81
0.82
0.90
0.72
0.75
0.84
0.88
0.90
0.94
0.62
0.71
0.77
0.75
0.79
0.86
0.81
0.83
0.89
Regions
Combined
0.94
0.95
0.97
0.92
0.93
0.96
0.76
0.79
0.86
0.79
0.79
0.88
0.82
0.84
0.90
0.82
0.84
0.90
0.89
0.91
0.94
0.88
0.89
0.93
0.60
0.61
0.75
calculated as signal/(signal + noise) which corresponds to:
a2,. + C52f / [c^g + o^f + (
-------
lios ' Z uojsuauuia
CO
CM .
o
o
OX)
I ..-
3
S'O
O'O
lios ' 3
in
o
1
~5
i
in
9
S'O-
Figure 3.3 Biplot for all regions combined.
76
-------
820
sd|diues I!OS ' Z
0-6
co
CM -|
p
d
a>
3-
I
—i—
9'0
O'O
9Z-Q-
q
d
o
S
in
9
8
5
—i—
S'O-
'2
Figure 3.4 Biplot for Piedmont region.
77
-------
Sfr'O
se|daiBS i;os ' Z
O'O
SZ'O-
Sfr'O-
1
in
5
co
2
W
O
O
CO
«S
9
in
o
o
o
at
-------
3.2.3 Soil Biotic Diversity
Three indices were computed for the nematode community in each soil sample:
o
o
o
Maturity index for free-living nematodes (MI) (Bongers, 1990)
Maturity index for plant parasitic-nematodes (PPI) (Bongers, 1990)
Shannon index of trophic diversity (Shannon and Weaver, 1949)
Diversity of nematodes by feeding preference were estimated using the Shannon
diversity index, Nl= exp [-ZP^lnP,)], where P, is the proportion of trophic group i in the
total nematode community (Ludwig and Reynolds, 1988).
The MI was calculated as the weighted mean of the values assigned constituent
nematode families (and the genera and species they contain) (Bongers, 1990): MI or PPI=
(Z v( * ft)/n where i?,=the colonizer-persister (c-p) value assigned to family i, /—the
frequency of family i in a sample, and n=total number of individuals in a sample. C-p
values range from 1-5; however, plant-parasitic taxa are assigned c-p values from 2-5
because there are no plant-parasitic colonizers designated as 1 (Bongers, 1990).
3.2.3.1. Associations among soil properties and with nematode communities
Nematode maturity indices were compared to various soil chemical and physical
properties thought to influence populations of nematodes i.e., clay content, sand content,
pH, organic matter, extractable phosphorus, exchangeable potassium, exchangeable
calcium, exchangeable magnesium, cation exchange capacity, total lead, and total cadmium
(Table 3.8).
Table 3.8 Coefficients from a principal components analysis of variables of soil properties
and nematode community indices. The first five principal components for MI, PPI,
Shannon's index of trophic diversity (SHAN), and related physical and chemical
parameters for soils are given.
1
2
3
4
5
MI
-0.01
0.07
0.68
0.11
-0.09
K
PPI
-0.10
-0.30
0.15
0.20'
0.66
Ca
SHAN
-0.03
0.00
0.66
-0.00
-0.29
Mg
Sand
-0.37
0.06
0.02
0.16
0.25
CEC
Clay
0.37
-0.13
0.03
-0.28
-0.08
Pb
EH
0.09
0.65
-0.02
0.01
0.13
Cd
OM
0.35
-0.18
0.07
0.09
0.22
-0.26
0.15
-0.12
0.49
-0.24
1
2
3
4
5
0.21
-0.01
-0.15
0.60
-0.33
0.32
0.44
0.03
0.12
0.20
0.37
0.30
0.05
-0.04
0.20
0.30
-0.25
0.07
0.41
0.13
0.28
-0.20
-0.13
-0.15
-0.28
0.27
-0.11
-0.01
0.19 ~
-0.01
79
-------
Principal component analysis involves the linear transformation of the original set
of variables to a set of orthogonal variables such that the first principal component
accounts for the largest possible amount of the total dispersion, the second principal
component accounts for the largest possible amount of the remaining dispersion, and so
on. The first five principal components accounted for about 74% of the dispersion in the
data. The first principal component (i.e., variables with the largest coefficients) was
primarily soil chemical and physical properties. Conversely, the third principal
component consists primarily of MI and trophic diversity (i.e., SHAN), and the fifth
principal component consists primarily of PPL Nematode indices tend to load on different
principal components than do the soil chemical and physical properties, thus implying that
the nematode indices provide information different from that provided by soil chemical
properties.
3.2.3.2 Relationship among nematode community indices
Among the nematode indices only the MI and Shannon's diversity index highly
correlated with each other (Table 3.9) Many nematode families that feed on bacteria are
colonists (Bongers, 1990; Neher et al., 1994) and the presence of such families would
decrease the Maturity Index for free-living nematodes (Table 3.9). Bacterivores are
generally abundant in agricultural soils. As a result the trophic group, bacterivores, is
inversely related to the maturity index of free-living nematodes and the Shannon trophic
diversity index.
Table 3.9 Spearman correlations between indices and trophic groups (n=185) of soilborne
nematodes.
PP SHAN BACP FUNGb
MI 0.05 0.43*** -0.64*** -0.15*
PPI 0.01 -0.04 0.07
SHAN -0.17* 0.26**
BACT ' 0.37***
FUNG
OMNI
PRED
* p<0.05, **p<0.01, ***p<0.001
"number of bacterivorous nematodes
bnumber of fungivorous nematodes
cnumber of omnivorous nematodes
dnumber of predaceous nematodes
cnumber of plant-parasitic nematodes
OMNF
0.25***
0.02
0.39***
0.28***
0.19*
PREDd
0.02
0.01
0.40***
0.25***
0.16*
0.27***
PLPAR6
-0.12
0.18*
-0.36***
0.24**
0.24***
0.06
0.18*
80
-------
Table 3.10 Variance components for three indices associated with nematode community
structures in soils. Actual variance values are presented.
. Indices
Maturity Index (MI)
Plant Parasitic Index (PPI)
Shannon Trophic Diversity
Among Among
Segments Fields
0.031
0.026
0.002
0.066
0.014
0.080
Within
Fields
0.017
0.032
0.173
Among
Samples
0.127
0,031
0.134
Various components among samples and within fields were generally numerically
greater than among fields and among segments (Table 3.10). These differences in
magnitude of variance components are probably due to biological variation in numbers
of nematodes within each composite sample and within fields. These differences in
magnitude and relative ranking of the variance components is a source of concern about
the acceptability of this indicator of condition.
The ability to differentiate ecological condition of soil among fields with statistical
confidence improves, if the variance among segments and among fields exceeds the
variance within fields and within samples (Table 3.11). A measure of this is the reliability
ratio and values >. 0.70 are desirable (see Table 3.7 for method of calculations for the
reliability ratio). The results suggest that sampling plans which have multiple
measurements per field may greatly improve the ability of these indices to differentiate
ecological condition of soil among fields with statistical confidence.
One of the biggest challenges with this indicator is to locate additional personnel
who have the expertise and time to identify nematodes to taxonomic family, especially
nematodes that are not parasites of higher plants.
81
-------
Table 3.11 Reliability ratios for several indices of nematode community structures for
various sampling plans'1
Indices
Maturity Index (MI)
Plant Parasitic Index (PPI)
Shannon Trophic Diversity
ib .
i
i
i
2
2
2
3
3
3
1
1
1
2
2
2
3
3
3
1
1
1
2
2
2
3
3
3
rfc
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
Reliability
Ratio
0.402
0.546
0.620
0.574
0.707
0.766
0.669
0.783
0.831
0.388
0.457.
0.486
0.559
0.627
0.654
0.656
0.716
0.739
0.211
0.255
0.274
0.348
0.406
0.430
0.445
0.506
0.531
"Calculated using the variance component estimates in Table 3.10.
bNumber of samples taken within a field.
•Number of subsamples of a composite soil sample analyzed by the enumeration
laboratory.
82
-------
3.2.4 Water Quality: Method Development
How do the two methods of sampling pond water compare?
The standard method of sampling from a body of water requires a boat and a
device to retrieve water samples from several depths. Water was collected from three
points per pond and mixed. At each point, water was taken at three depths with a
Kemmerer sampler. An alternate method was tested in which a stoppered teflon bottle
at the end of a 5-m pole was used to draw a shallow (0.3 m) water sample from four
points along the bank of each pond. Water from the four bank locations was mixed to
form a composite sample. There was no significant difference (P>0.6) in pond nitrate
concentration due to the sampling method, and the results of the methods were highly
correlated (r>0.99). However, precision was slightly greater with the boat sampling
method, based on variances between replicates within ponds (data not shown).
83
-------
3.2.5 Extent and Geographic Distribution of Annually Harvested Herbaceous Crops
3.2.5.1 Successes and lessons learned
Full June Enumerative Survey used for extent estimates
Extent estimates are based on data from the complete National Agricultural
Statistics Service (NASS) June Enumerative Survey, the largest sample available. During
data analysis, extent was also estimated using the rotational panel and hexagon-selected
samples. These samples, which are much smaller than the June Enumerative Survey
sample, provided less precise estimates of extent. We decided to use data from the full
June Enumerative Survey because they gave the most precise and accurate estimate of
extent.
Strata map
The digitized NASS strata map is a very useful presentation tool and served as the focal
point for describing the geographic distribution of the annually harvested herbaceous
cropland resource. It provides a clear picture of the distribution of the agricultural
resource throughout the state.
Strata maps for many of the other states will be unavailable for several years.
Creating these maps is presently a labor-intensive and inexact process, both for us and
NASS. Fortunately, NASS has developed a new system of area frame construction that
will both simplify the construction of strata maps and improve their accuracy. As new
frames are created for each state, these maps will become available. In the meantime, it's
possible that an Advanced Very High Resolution Radiometer (AVHRR)-derived land cover
map may provide similar information.
Diversity index selection
Measures of diversity are used to answer two questions:
(1) How many crops is one likely to find in an area?
(2) How even is their relative abundance in that area?
The first question is answered with a species richness measure. For each segment,
the observation is the number of different annually harvested herbaceous crops in the
segment. For this measure of diversity, segments with no crops must be included.
In the ecological literature, the Shannon index is the most commonly used measure
of relative abundance. However, explaining the Shannon index is always difficult, largely
because it does not have a direct biological interpretation. While it is relatively easy to
84
-------
explain a trend in the Shannon index, it is difficult to describe what a single value means.
For these reasons, the Shannon index will not be used here.
Another index - the Berger-Parker index - is both simpler to calculate and easier to
interpret. For each segment i, the Berger-Parker index is:
= MAX(n.p
where Nt is the acres of annually harvested herbaceous crop in the segment and MAX(nf//)
is the number of acres of the most common crop in the segment. The Berger-Parker index
is a measure of dominance (the inverse of evenness): the proportion of cropland occupied by the
crop with the largest acreage. The index varies 0 < d < 1, where a value near 1 means that
a single crop dominates the area, and smaller values mean the crops are more evenly
represented. For this measure of diversity, segments with no crops must be excluded
because the index is undefined. The Berger-Parker index was found to be closely,
negatively correlated to the Shannon index for our data (Spearman r = -0.91).
Sanity Test
The estimated total land area of North Carolina was calculated from NASS data by
summing the expanded area of all segments. This gave an area estimate of 77,525,091 ±
495,202 (at 95% confidence) hectares. According to the 1987 Census of Agriculture, the
total land area of North Carolina is 77,184,583 hectares. This is within the 1%
Measurement Quality Objective.
3.2.5.2 Challenges and issues
Most of the challenges result from differences between what we would like to get
from June Enumerative Survey data, and what NASS expects from those data. These
differences are, for the most part, minor. The order in which the issues are arranged
should not be taken to imply relative importance.
Measurement error
Each datum used to calculate resource extent is the area of a field. Although NASS
takes great pains to control the quality of their data these measurements are not 100%
accurate. We do not know the frequency and magnitude of measurement error, and it is not
accounted for in the estimates of variance in the Land Use and Cover report. The variance
calculations used here assume that each measurement is completely accurate. We don't
know how to deal with this problem yet, but our statisticians are thinking about it.
85
-------
"Other crop" identification
During the June Survey, NASS records the acreage of infrequently occurring crops in an
"Other crops category. Although the exact crop is recorded on the survey form, some crops are
aggregated during data entry and the dataset we receive does not allow us to determine what these
other crops" are. We ivould like to have a way of identifying the crops for which acreage is
recorded in "Other crops."
The diversity of annually harvested herbaceous crops is one of our indicators.
Calculation of diversity requires as input the number and area of each different crop
found in each segment. During the analysis of data from the 1992 North Carolina June
Enumerative Survey we found three "Other crop" variables: CROPOTHP, COPTL1PL, and
COPTL2PL. CROPOTHP and COPTL2PL were both used to record "undetermined crops";
COPTL1PL was used to record all vegetables, including melons. The estimated extent of
"Other crops" is larger than the estimated extent for some crops that are specifically
identified (e.g. COPTL1PL = 8535 hectares; sunflower = 383 hectares). This makes it
difficult to justify writing "other crops" off as insignificant. Also, for segments in which
multiple "other crops" are recorded, we have no way of determining how many of these
are different; and therefore no way of calculating the number of different crops in the
segment.
We are also aware that "other crops" include crops that are not annually harvested
herbaceous crops. These appeared in the CROPOTHP variable, which is one of the
variables used to select fields for EMAP sampling.
This issue has been brought to NASS's attention in the form of a Survey
specification for the 1994 North Carolina and Nebraska surveys. One solution which we
would find acceptable is to have the identity of the "other crop" coded into another
variable. For example, if acreage is recorded in CROPOTHP, code the crop name in
another variable called CROTHPID (or something like that). Other solutions might be to
assign unique variable names to all crops, or to assign individual "other crops" to the
COPTLnPL variables (instead of groups of crops).
Idle land
We found the recording of "idle land" to be very confusing. Although a separate field
was expected for idle land, it is sometimes combined with crop area. The reason for this
has to do with farming practices, particularly for cotton and tobacco. A large amount of
idle land can occur in a tobacco field as farmers plant their acreage allotment leaving
spaces throughout the field. Further, the number of idle acres could be reported in a
different column in the June Enumerative Survey table. For example, the tobacco acreage
could be reported in field 2, while its idle acres might be reported in field 15. BUT these
were not drawn off as separate fields on the photograph; rather, the two numbers (2 and
15 in this example) were written in the same field on the photo.
No action has been taken to rectify this issue because we have yet to determine
how idle land fits into our resource classes. Two possibilities are:
86
-------
(1) Include it with cropland or pasture, depending on its previous (and intended
future) use.
(2) Establish an idle land resource class.
Once we have determined how we wish to view idle land, we will work with NASS to
sort out the confusion.
For 1992, idle land was included for extent, but was not eligible for fall sampling.
Cropland pasture
NASS defines and records two kinds of pasture: permanent pasture and cropland
pasture. Permanent pasture is grass and rangeland that is not in a regular crop-pasture
rotation. Cropland pasture is land in a crop-pasture rotation, being used for pasture in
the current year. The area of permanent pasture and cropland pasture sometimes appear
in the same record as a crop. This was a surprise, as we expected pastures would be
drawn off separately. The acreage involved was usually small.
A decision must be made regarding whether cropland pasture is part of the cropland or
pasture resource. This decision will affect both extent estimates and indicator development.
If cropland pasture is included in cropland, it is unclear how the productivity indicator
would be measured on a field that is currently cropland pasture.
Cropland pasture was not eligible for fall sampling in 1992, nor was it included for
extent.
Hay
"Grains harvested for hay" duplicates acreage that appears as either winter wheat
planted, barley planted, oats planted, or rye planted. For extent, this problem was
accounted for by use of the subtractive method.
Geographic coordinates for June Enumerative Survey (JES) data
We would like to have the latitude and. longitude of the centroid of all segments that are
part of the JES. During the 1992 pilot, coordinates were included on each record of the
EMAP samples (both the rotational panel and hexagon-selected samples), but not for the
complete JES. Although the variables MLAT and MLONG appeared in the JES dataset,
the values were all set to missing. The solution we prefer would be an ASCII file
containing segment number, latitude, and longitude. This could be directly imported into
our geographic information system (GIS). Another solution would be to include the
coordinates on each JES record. A 1994 JES survey specification requesting this change has
been submitted and will be discussed with NASS.
87
-------
Reporting on consistent regions
The regions used to summarize extent and geographic distribution were different than those
used to summarize other indicators. Most indicators were regionalized by Major Land
Resource Areas (MLRA); extent and distribution were reported by NASS strata. We could
have reported extent within each MLRA by post-aggregating either rotational panel or
hexagon selected data by MLRA. However, the sample sizes were small and extent
estimates very imprecise; instead we chose to use the full June Enumerative Survey.
As part of our future analyses, and for the development of reports, we would like to post-
aggregate extent data to the same ecological and political regions used for other indicators (e.g.
hydrologic units, Major Land Resource Areas, ecoregions). Because we have access to these
regions in a GIS format, availability of segment centroids would allow us to post-
aggregate data to any and all of these regions. Analyses and reports could then be made
within consistent boundaries. An alternative to obtaining segment centroid coordinates
would be to have NASS provide a table of segments cross-referenced to specific regional
identifiers. In the long run, this alternative is likely to be more costly and time-consuming
than using segment centroids.
Communication of data needs
There was much confusion surrounding specific attributes of the survey data when we
received it. Not all of the data were included and the meaning of variables in the datasets
was unclear. The confusion was a result of our failure to communicate exactly what we
needed to NASS, which was, in part, a result of our lack of understanding of NASS
procedures.
Data presentation
Although cumulative distribution graphs provide a lot of information, they may not be the
best presentation tool for all audiences.. For example, histograms may be more interpretable
for discrete data like crop diversity (Figure 2.3). The appropriate presentation techniques
are very audience-dependent. Statisticians and scientists may be comfortable with
cumulative distribution functions; histograms, bar charts, and pie charts may work better
for general audiences.
Link to societal values
The few statements in the land use and cover section that attempt to relate the measures
to societal values are weak and point out the need for further work. Further analysis of our 1992
data for correlation among indicators may help, but field research may be required to
answer other questions. For now, use of these data is best limited to statements of the
extent and geographic distribution of the annually harvested herbaceous crop resource.
-------
At what scales should crop diversity over space be measured?
In this report it was measured over a 260 hectare area — the size of a NASS
segment. This may not be the (only) appropriate scale. If we're interested in making
statements about spatial diversity and vulnerability to disease, pest infestation, and erosion
we must first measure them at various scales and find out which (if any) are correlated.
Likewise, if we want to make statements about crop diversity and wildlife. Keep in mind
that these findings are all likely to be region-specific. Even if we don't want to make
correlations, if diversity is measured at all it should be done at several scales. (Perhaps the
crop diversity of each ecoregion should be calculated.)
We need a measure of crop diversity over time — rotation
This could be an important indicator of vulnerability to disease and pest
infestation, and of sustainability in general. We are working on a crop rotation index that
may be calculated using crop history data obtained during the Fall Survey.
89
-------
3.2.6 Landscape Structure
The work on landscape structure described in the 1991 Pilot Plan (Heck et al., 1992)
has been slowed by several obstacles, including the learning curve for ARC/INFO, data
management difficulties, and changing priorities. Although no results are presented,
much has been learned.
3.2.6.1 Successes and lessons learned
Appropriate data format: Raster, not vector
For broad geographic areas covered in fine spatial detail, the vector format is not
appropriate; raster data are strongly preferred.
Preliminary work was conducted using Thematic Mapper-based land cover data for
the Albemarle-Pamlico Drainage Basin, a 27,500 square-mile area that includes parts of
North Carolina and Virginia. Classification was at a spatial scale of (approximately) 28
meter x 28 meter pixels. These are the only extensive, classified data presently available
for North Carolina. The land cover data were delivered in ARC/INFO vector format,
separated into files representing the USGS 1:100,000 quad sheets. The classification system
included 19 categories of land cover. We wanted to aggregate land cover into eight classes,
in a seamless coverage of the region.
Working with this much data in vector format proved unacceptable. Software and
hardware limitations were encountered; even when things worked, analyses were very
slow. After the study began, a raster-based module (GRID) was released for ARC/INFO.
Portions of the data were converted from vector to raster format, which proved much
easier to work with.
Calculating of measures of landscape pattern
Once the data were available in a raster format, developing algorithms to calculate most
measures landscape pattern detailed in the Pilot Plan was straightforward. In most cases, the
procedure involved exporting information from ARC/INFO and performing the analysis
using SAS.
3.2.6.2 Challenges and issues
Thematic classification: How much detail?
The level of thematic classification detail required is unknown. To date, EMAP-
Landscape Characterization has attempted to produce a single classification system for
land use and cover. This approach results largely from the belief that one can define a
one-to-one association between land cover and EMAP Ecosystem Resource Groups.
90
-------
Combined land use and cover classifications confound the issues of use and cover, with
a resulting loss of information. Separate coverages of land use and cover are preferred, using
GIS to overlay them as needed. Ecosystem Resource Group membership could then be
decided using information from any number of thematic data layers (e.g., land use, land
cover, ecoregion, soil type, elevation).
Classification of broad geographic areas
Completing a consistent, and accurate, land cover map for the nation is a formidable task,
yet it is critical to the development of landscape indices. Vegetation over specific areas can be
highly variable, yet disparate vegetation can have similar reflectance; the same vegetation
can even have different reflectance at different times of the day or year. Even within the
relatively small area of the Albemarle-Pamlico basin, data for adjacent scenes were
acquired months apart, different parameters were used for different geographic regions,
and some cover classes were dropped for particular regions. Calculation of measures of
pattern across scene lines using these data is thus difficult, if not meaningless.
Appropriate scales of space and time
The appropriate spatial and temporal scale(s) for monitoring and research depends on the
question(s) being asked and the scale(s) at which pertinent processes occur. We need to gain a
better understanding of the scales at which agroecosystem processes occur and collect data
at those scales. We look to cooperative work among EMAP Resource Groups and the
EMAP Landscape Ecology and Landscape Characterization groups to help us address
these issues.
Ecosystem boundaries
The question of where one "ecosystem" ends and another begins is unanswerable without
reference to particular processes or organisms. Further, one cannot define the extent of an
ecosystem without reference to particular processes or organisms. While it is possible to
delineate and calculate the extent of land covers at various scales without reference to a
particular process or organism, a patch of a single cover does not necessarily bound an
ecosystem. Again, the belief that one can define a one-to-one association between land
cover and EMAP Ecosystem Resource Groups leads us astray. This is another argument
against combining classification themes into a single data layer; doing so hampers our
ability to redefine system boundaries for different processes and organisms.
Measurement error and confidence in measures of landscape pattern
Landscape ecologists have been using GIS representations of remotely-sensed data to
calculate measures of landscape pattern but have devoted little effort to quantifying the uncertainty
in these measures. Without statistical confidence in the measures used, scientists cannot
evaluate correlations between landscape pattern and ecological processes. Without
91
-------
statistical confidence one cannot use measures of pattern to detect differences in
landscapes over space or changes in a landscape over time. This problem should be of
great interest to all EMAP Resource Groups using measures of landscape pattern, and
should be on the research agendas of both the EMAP Landscape Characterization and
Landscape Ecology groups.
Classified land cover data in a CIS, usually derived from remotely-sensed data, are
frequently used to calculate measures of landscape pattern. Classification error is usually
reported in the form of an error matrix based on an accuracy assessment of the classified
data. An error matrix clearly describes the accuracy of each classification category, as well
as the nature of the confusion among categories. Further uncertainty is introduced when
multiple GIS maps are overlaid - a common procedure among GIS users. Although
awareness of these and other error-related issues is common in the remote sensing and
classification communities, the implications of these issues have yet to be addressed by
landscape ecologists. There are no procedures for using an error matrix to generate confidence
intervals for measures of landscape pattern. We are working to overcome this problem.
Interpretation of measures of landscape pattern
Theoretical work indicates that landscape pattern measures may reflect the ability of
organisms to inhabit and traverse a landscape, the potential for materials or disturbances to move
from one part of the landscape to another, or the types of processes that are shaping the landscape.
But these relations are hypothetical and difficult to test. We look to cooperative work among
landscape ecologists, the EMAP-Landscape Ecology Group, and the EMAP Resource
Groups to begin addressing this issue. This implies that work at the landscape scale must
be designed carefully to serve not only as a method to monitor patterns, but also as a
method to test hypotheses at multiple spatial scales.
Interactions with EMAP-Landscape Characterization and EMAP-Landscape Ecology
The operational details of the. interaction between Resource Groups and the EMAP-
Landscape Characterization Group need to be clearly defined. There have been major changes
within EMAP-Landscape Characterization over the last 4 years. During these changes, the
roles and responsibilities of the Landscape Characterization Group have not been clearly
communicated to the Resource Groups. Further discussions are essential to determine
how the Resource Groups and EMAP-Landscape Characterization can interact successfully.
One of the changes in the EMAP-Landscape Characterization group was the spinoff
of an new group, EMAP-Landscape Ecology. The manner in which the EMAP-Landscape
Ecology group and the Resource Groups will interact is also undefined.
92
-------
3.2.7 Agrichemical Use
How does EMAP-Agroecosystems pesticide use information compare to what other
groups do?
More extensive pesticide surveys are currently done by others for specific crops, for
example the Agricultural Chemical Usage reports published by NASS for selected crops
(e.g. Agricultural Chemical Usage: 1992 Field Crops Summary, USDA-NASS, Agricultural
Statistics Board, Washington, D.C., March 1993, Ag Ch 1 (93)).
One strength of the EMAP-Agroecosystems survey is that it included most field
crops and, in that respect, can provide more complete information about our resource than
the current NASS Agricultural Chemical Usage report. The NASS report is based on a
sample of only major crops and is only conducted in the states that account for most of
the area of each crop. In a very few cases, a comparison can be made between the EMAP
survey data and the Agricultural Chemical Usage findings. According to the latter, the
average rate of atrazine applied to corn in North Carolina was only 1.43 kg/ha/treatment
in 1993, considerably less than our value. Yet that same report would indicate that the
total area to which atrazine was applied was 363,000 ha for corn alone, much more than
our estimate of the total. These differences are approximately 1.5 to 2 standard errors in
magnitude. Attempts to reconcile the numbers were unsuccessful.
93
-------
3.3 EVALUATION OF PILOT ACTIVITIES
3.3.1 Interactions with USDA-NASS
The 1992 Pilot was a successful test of cooperation between EMAP-Agroecosystems
and NASS. NASS personnel were involved in four areas of the Pilot: (1) collecting data
during their regular JES, (2) drawing the sample of fields to be visited in the fall, (3)
designing and conducting the fall questionnaire and, simultaneously, (4) collecting fall soil
and water samples. Two primary contacts were provided by NASS: one in Washington,
D.C. and one in North Carolina. One member of the ARG served as liaison officer to
NASS; however, the day-to-day cooperation was very decentralized. The technical
director, indicator leads, statisticians, and information manager all worked with various
NASS personnel as needed.
One area of particular sensitivity has been the confidentiality of data from
individual NASS surveys. Each member of the ARG who participates in collection or use
of the data must sign a "certification and restrictions on use of unpublished data (off-site)"
(NASS Form ADM-043). This form states that the signee will abide by United States Code
Title 18, Sections 1902 and 1905, and Title 7, Section 2276. These codes deal with crop
market information divulgence and personal identity of the data source. Any violations
of these codes is punishable by fines and/or imprisonment. A memorandum of
understanding is being drafted between NASS and the USDA-Agricultural Research
Service (ARS) that will guide future handling of confidential data (see Section 3.4.4).
Both the interview and sampling for the 1992 Pilot were done by part-time
"enumerators" who work for NASS through the National Association of State Departments
of Agriculture (NASDA). They are accustomed to doing various NASS surveys and
adapted well to the new job of taking soil and water samples (see Section 3.3.3). The
Agroecosystem program is not restricted to using only enumerators. Some future
monitoring may require field crews with more technical training and expertise.
The ARG and NASS together developed the questionnaires and enumerator's
manual. Training was also a joint responsibility. All printed materials, including labels,
were produced by NASS. For the 1992 Pilot, the ARG purchased all equipment,
assembled the kits, and distributed them at the training school. In the future, more of
these logistical responsibilities will be handled by NASS. Surveys and sampling as well
as all other NASS activities were conducted on a reimbursable basis and were funded
through an interagency agreement between U.S. EPA and NASS.
Although everything eventually fell into place, there were some elements, such as
water sampling procedures, that were being revised up until the last minute. This is
something that is of concern to NASS, so the ARG is trying to adjust our timetables for
future programs.
Other aspects of cooperation with NASS are covered in Sections 3.4.4 (Information
Management) and 3.1 (Preliminary Evaluation of Designs).
94
-------
3.3.2 Logistics
The 1992 Pilot in North Carolina was generally a logistical success. The logistics
activities can be broken down into questionnaire performance and survey administration,
soil sampling, water sampling, shipping and sample tracking, laboratory relations, and
reporting back to the respondents. The survey and sampling period ran from the end of
October until mid-December.
Questionnaire performance and survey administration. Most aspects of the survey went
well. The questionnaire was completed for 326 (94%) of 348 sample units (Table 1.3).
There were 11 questionnaires not completed because of operator refusal and 11 for which
the operator was inaccessible. Of the 326 "completed" questionnaires, some had partial
data: two were missing fertilizer information, and 11 more were missing both fertilizer and
pesticide information. One drawback of the sampling scheme was that often fields were
sampled more than once (Table 1.4), so that failure to contact or receive cooperation from
one operator meant the loss of several samples. This problem will be corrected in the 1993
pilot by drawing the sample from a larger set of fields.
Areas for improving the questionnaire were identified. For example, the occurrence
of double crops and cover crops had to be deduced on a case-by-case basis, because no
specific code was provided to identify them. That code was added in 1993. The questions
designed to arrive at fuel use (not reported here) were also difficult to enumerate. This
section was changed for 1993 so that field operations, but not fuel use, were tallied. The
question about crop rotations as a pest control practice was too vague to give meaningful
data for 1992.
Soil sampling. Soil was not collected for 27 of the sample units (Table 1.5), usually
because the farmer refused or was inaccessible. On six of those 27, however, the field was
a woodland or wetland and should not have been selected for sampling. In two other
cases the enumerator made a mistake and failed to get the soil sample. Also, several
farmers allowed soil sampling but declined the interview. Only one farmer (representing
seven sampling units) gave the interview but refused to allow soil sampling.
The soil sampling procedure itself was done satisfactorily (see Section 3.3.3). Some
of the enumerators reported difficulty with laying out the transect, use of the soil sampler,
and mixing the soil.,
Water sampling. Water was not collected for 11 of 51 ponds and 20 of 61 wells
(Table 1.5) due to farmer refusals or inaccessibility. Ponds and wells were sampled only
from the Rotational Panel design.
Water sampling procedures and equipment purchases were not finalized until the
last minute. Considerable effort was made to borrow equipment for this part of the
project, but in the end most of it had to be purchased, including depth finders and several
95
-------
small jon boats. To prevent enumerators from getting steel splinters, nylon clothesline
cords were substituted for the cables that came with the Kemmerer samplers. However,
one of the cords broke during use. The sampler was recovered from the bottom of the
pond, and all enumerators were given the original steel cables to use.
Shipping and sample tracking. The use of Federal Express as a courier for shipping
samples was successful. At first the enumerators were calling to have the samples picked
up at their homes, but then the NASS office determined where the nearby Federal Express
offices were, so that samples could be dropped off later each day. The procedure went
smoothly and samples were delivered in a timely manner. Note that the water samples
were sent via "priority" overnight service, whereas the soil went under the regular
government contract service.
There were a few accounting difficulties with Federal Express. Without a weight
written on each "airbill" label, a large default weight was sometimes charged. It also
seemed that several of the charges on the final billing report may have been incorrect.
Sample tracking was done by the ARG logistics officer, who received database files
from NASS and faxed sheets from the laboratories. Identification numbers of the samples
sent were compared to those of the samples received. Although a rather tedious
procedure, it was useful in identifying, for example, transcription errors at the laboratories.
Federal Express was very efficient in providing information about packages that were
thought to be lost (using their airbill identification number).
Laboratory relations. Identification time for nematodes exceeded initial estimates, and
a few errors were found in the soil chemistry data, but in general the laboratories
performed well. Some delays were also experienced with receipt of data from the EPA
Athens-ERL, where the water samples were analyzed. Nitrate testing was done by a
federal scientist and pesticide testing by a contractor: Technology Applications, Inc.
Reporting back to the respondents. By mail, NASS reported some soil and water data
back to the individual farmers who participated in the 1992 Pilot. One unfortunate
incident was that there was a mix-up in the units in which the nitrate data were reported.
Thus, farmers were told that concentrations were about 4 times their actual values. NASS,
fearing that it would create confusion and negative public relations, did not send out the
corrected results. This is not the ideal situation, but at least the concentrations were not
understated, and some people had had their wells retested.
96
-------
3.3.3 Quality Assurance
Two separate quality assurance (QA) programs were used during data collection for
the 1992 Pilot. NASS has a well-established QA program, which was used during the
collection of the questionnaire data for its annual JES and for the EMAP-Agroecosystems
fall survey. All other organizations participating in sampling and analytical activities were
required to follow QA procedures developed by the ARC, in addition to their own quality
assurance activities. These organizations included:
o the ARC;
° NASS for the collection of soil and water samples;
o U.S. EPA Environmental Research Laboratory (ERL) in Athens, Georgia,
for water sample analysis;
o U.S. EPA Environmental Monitoring Systems Laboratory in Las Vegas, Nevada;
o Agrico Agronomic Services Laboratory in Washington Court House, Ohio,
for soil sample analysis; and
o N&A Nematode Identification Service in Davis, California,
for nematode identification and enumeration.
Overall, the quality assurance program, although immature, was successful in
obtaining data of high and known quality during the Pilot. Quality assurance activities
occurred at six specific steps detailed below.
3.3.3.1 Area frame development
During the development of the area frames used to determine sampling locations,
quality assurance activities were performed by NASS in accordance with the procedures
specified in Area Frame Design for Agricultural Surveys (Cotter and Nealon, 1987). The
NASS area frame was successfully converted to a geographic information system (GIS)
coverage in ARC (a commercial package). The GIS lead for the ARG verified that the
primary sampling units developed around EMAP hexagon centroids actually contained
the appropriate centroid.
3.3.3.2 Enumerator training
Training of enumerators for the collection of questionnaire data (June and fall) was the
responsibility of NASS. NASS conducted a national training school in April 1992 to train
the NASS supervisors. A state school was held in May 1992 to train enumerators for the
JES. Each enumerator was provided with a manual that included the standard operating
97
-------
procedures for the collection of the verbal survey data. The school for the Fall Survey was
conducted in October 1992. Enumerators were instructed on collection of soil and water
samples as well as questionnaire data. This later training school was conducted
cooperatively by NASS and the ARG, including the EMAP-Agroecosystems Quality
Assurance Officer (QAO). An enumerator's manual was again provided.
For administration of questionnaires, enumerator training consisted of a thorough
discussion of all questions and possible appropriate answers. Enumerators were also
instructed on how to deal with unusual or unexpected situations. For soil and water
sampling, all procedures were presented and discussed in a classroom setting, followed
by a field demonstration and practice.
3.3.3.3 Collection and handling of survey data
Survey data collection, processing, and output were the responsibility of NASS.
Complete NASS QA procedures were not used, but the following steps were taken. As
usual, enumerators each worked under the guidance of a supervisory enumerator.
Questionnaires were sent back to the NASS state office, where they were checked for
completeness and reasonableness in the first part of a two-stage edit process. If this
manual edit turned up incomplete or questionable data, the enumerator was contacted
again, or if necessary, the farmer was phoned directly from the state office. The step of
having a supervisory enumerator review two surveys from each enumerator's workload
was skipped because of the tight sampling schedule.
After the manual edit, a detailed computer edit was done to test that data were
within expected ranges and were internally consistent. Any problems at this stage were
brought to the attention of the statistician coordinating the survey in North Carolina, who
resolved them based on the original questionnaires and notes made by enumerators. A
statistician in Washington, D.C. also reviewed pesticide usage for outliers and
reasonableness. Any questions were relayed to the statistician in Raleigh who resolved
them. It was not necessary to re-contact any enumerators or farmers for this step. After
edits, a final summary was written.
3.3.3.4 System, and performance audits
Field technical system audits of soil and water sampling were performed. Two technical
systems audits were performed on two different NASS survey crews. These audits were
conducted by the EMAP-Agroecosystems QAO in cooperation with a member of the
North Carolina NASS office.
The NASS enumerators performed their soil and water sampling efficiently. Only a
few problems or concerns were identified, and they were minor and were not expected
to have any significant influence on the integrity of the samples.
98
-------
A technical systems audit was performed on the Agrico Agronomic Services Laboratory
in Washington Court House, Ohio, by the EMAP-Agroecosystems QAO. The purposes
of the audit were: to observe the operations of the analytical laboratory, to review existing
preliminary data as a pre-QA check (preliminary performance audit), and to answer
questions regarding the analyses to be performed.
In general, the Agrico laboratory was found to be generating acceptable data and was
following ARG protocols. Most areas of concern were relatively minor and could easily
be incorporated into laboratory operations. However, two areas of major concern were
noted during the audit. First, samples were processed through a 14-mesh sieve whereas
most soils are processed through a 10-mesh sieve to meet the standard USD A size
definition for soil (i.e., 2-mm or finer). The other area of concern was that sample aliquots
were obtained for analysis on a volumetric and not on a gravimetric (i.e. sample with
known weight) basis. Whereas this type of sample aliquoting may be appropriate for soil
fertility testing, the potential variability in the sample results can affect the long-term
monitoring efforts of the EMAP Agroecosystems program.
It was noted during the review that laboratory quality control (QC) samples were
prepared and analyzed at a higher frequency at Agrico than required for most U.S. EPA
programs indicating the laboratory's concern and effort to obtain high quality results.
Additional laboratory control samples were incorporated into each analytical batch. These
samples were collected and processed by Agrico in bulk and have established acceptance
windows to allow the technicians to rapidly assess, during ongoing sample analysis, the
functioning of the equipment and the accuracy of the measurement system.
During the performance audit phase of the laboratory audit, results of the other
QA/QC samples (to be discussed) were examined for both accuracy and precision. Where
comparable procedures were used, accuracy results were generally within the determined
acceptance windows for those given parameters. Precision among the analytical replicate
and field split samples was generally within a relative percent difference of 10% or less
for all parameters undergoing analysis indicating that the sample preparation techniques
in the field and laboratory were functioning well. Upon examination of the data from the
field replicate samples, larger relative percent differences were identified, indicating a high
variability within the field, in particular for phosphorus.
3.3.3.5 Use of QA/QC samples
Soil samples
Precision was assessed through the analysis of analytical duplicates created in the
laboratory, field split samples (i.e., two samples obtained from a single composite soil
sample), and field replicates (i.e., two different samples collected from two different
transects within the same field). Field split and replicate samples were sent blind to the
laboratory. Although no measurement quality objectives had been set for precision in the
EMAP-Agroecosystems program at the time of the audit, with few exceptions, the data
99
-------
generated for analytical duplicates were within a relative percent difference of 10% or less
for all parameters examined.
E^W* IVeJLMifcVBW*. JLV^Jk V**H«»*.T v*.»— VL*. »A »*•.f.r -fc.
for all parameters examined.
In addition to the precision samples, a performance evaluation sample was obtained
by the EMAP-Agroecosystems QAO from the EMAP-Forests program and incorporated
into the sample stream. Several replicates of this sample were submitted throughout the
time period soil sampling was being performed by the NASS enumerators. The laboratory
did not know which samples were the performance evaluation samples. Where
comparable determinative procedures were used, accuracy results were generally within
the pre-determined acceptance windows for those given parameters.
Water samples
Water samples were analyzed for nitrate during November and December 1992,
following the procedures outlined in Payne (1992). No technical systems audit was
performed on the Athens laboratory by EMAP. Aqueous samples were directly injected
into the ion chromatograph, so no fortified samples were needed beyond the nitrogen
standards used for calibration. The correlation coefficients indicated no significant day-to-
day variance in the standard curve. Forty-eight samples (29% of the total) were split for
replication. The average ratio between duplicate and original came out to be 99.98%, with
a standard deviation of 2.9. Six samples (3%) were considered outliers and analyzed in
triplicate.
Of the 12 pesticides for which the water was tested, none were detected. All
available evidence indicates that these zero values are legitimate. A second extraction and
chromatography were done on 10% of the samples, all of which showed no detectable
levels of the compounds. Fortified spike samples for each compound were analyzed at
1, 2, and 3 ppb (only 1 ppb for alachlor, aldicarb, and cyanazine). Average percent
recoveries for the replicated 1 ppb spikes ranged from 60% (aldicarb) to 285% (cyanazine).
This shows that the compounds would have been detectable at or above 1 ppb, though
the target accuracy of 70% and target precision of ± 10% would not have been met.
Laboratory blanks were included for each batch (10% of total). No field blanks were sent
to the laboratory, but field contamination was clearly not a problem.
Note that sample bottles were chilled on ice and then shipped overnight in
styrofoam cases within cardboard boxes. They arrived the next morning in good
condition. The temperature of the first few samples was tested and found to be 10-15 C.
This shows either that the samples stayed cool during transit or that they re-chilled en
route in the aircraft. Stability studies showed that the samples were extracted (24 hr) and
the extracts analyzed (1-2 mo), well within the periods during which the compounds were
stable in storage (7 d and 6 mo, respectively).
100
-------
3.3.3.6 Data validation
Upon receipt of the completed data sets from MASS, Agrico Agronomic Services
Laboratory, N&A Nematode Identification Service, and ERL-Athens, ARG statisticians and
indicator leads, in cooperation with the ARG information manager, validated the
assembled database. For example, an ARG statistician evaluated the minimum and
maximum values of the soil parameters to ensure that the values were within the range
of possible values. An ARG statistician and indicator lead also checked all soil parameter
values which were greater than five standard deviations from the sample mean to be sure
that values were correct. The ARG ensured that data were missing from the dataset only
when the soil samples had not been taken. Questions about specific values for soil or
water parameters that appeared to be anomalous were referred to the appropriate
personnel at the respective analytical laboratories.
Although NASS checks for reasonableness and consistency, the EMAP-
Agroecosystems logistics coordinator followed up on apparent problems with the
questionnaire data when they arose. Some of these problems indicate that data will need
to be collected differently in the future, while other problems show that extra edit checks
are needed. In most cases, the questionnaire data were not changed once they reached
the ARG.
No attempt has been made to quantify these nonsampling errors in the
questionnaire data, but several deserve mention. During the fall visit it sometimes
happened that a different crop was in the sample field than what was expected; sometimes
the field had no annually harvested herbaceous crops at all. Survey information and soil
samples were taken anyway, except when the site was found to be woods or wetland.
Some, but not all, "of the differences can be explained as changes that occurred between
June and fall. The sample was drawn using data from the June Enumerative Survey;
however, even field sizes were sometimes reported differently between June and fall,
indicating that a different field was visited on the two trips. The yield and management
information are surely subject to similar uncertainties. The fact that the "unit weight" was
not consistent within crop species gives another indication of possible survey problems,
although they were used as given.
Unexpectedly, there were a few cases where an enumerator "estimated" yield and
other data when a farmer was unavailable (out-of-state). This will be expressly forbidden
for the 1993 survey in Nebraska. When the respondent was available but harvest was not
complete, yield data were usually recorded by NASS as missing values; however, in some
of those sample units, an estimated yield and harvest date were recorded. Estimated data
were not excluded from the 1992 Pilot analyses.
QA checks were not performed on the determination of soil map units, but the
results did vary when some were done twice by different people. The procedure involves
comparing the NASS aerial photograph (showing the outline of the sample field) to SCS
101
-------
soil survey maps. In the recent 1993 Pilot, some map units will be determined on the
ground by SCS scientists as a check on the accuracy of the map-and-photo method.
The primary reporting tool for this report is the cumulative distribution function
(CDF). ARG statisticians submitted data sets that had been analyzed by hand to the CDF
computer programs to verify that the estimated distribution and its confidence bands were
being computed correctly.
102
-------
3.3.4. Information Management
3.3.4.1. Introduction
Information management activities were focused in three areas: 1) the development
of a working relationship with NASS, particularly in the area of data confidentiality and
data movement; 2) the development of an initial infrastructure for the Agroecosystems
Information Center (AIC) - items such as hardware, software, and networks were
acquired, installed, and configured to satisfy the requirements of the ARG for pilot
activities; and 3) data management. Because this was the first field season for the ARG,
much of the work was new and involved documenting activities and working closely with
indicator leads and statisticians.
3.3.4.2 NASS interactions
The development of a cooperative relationship with NASS was a primary goal of
the 1992 Pilot. The process of familiarization with each other's activities, reasoning,
terminology, methods, personnel, etc., was very successful. This success has already
shown benefits in the planning for the ARG 1993 Pilot Field Program.
Working within the confines of the NASS data confidentiality regulations was a
challenging activity for information management. Sample level identifiers were masked
to remove any systematic identification and replaced with unique/ "random" identifiers.
When sample level data were required with identifiers, ARG personnel traveled to the
NASS State Office in North Carolina to acquire data and_perform aggregation or masking
until that data would pass screening for removal from the premises. This proved to be
very cumbersome because the types of analyses required relied heavily on the sample
level identifiers.
To address the issues of data confidentiality and transfer for the future pilot
activities, a Memorandum of Understanding (MOU) between NASS and ARS is in the
draft stage, which will specify that the ARG can have the data with identifiers at the ARG
headquarters. The MOU will require that ARG personnel sign NASS Form ADM-043
(Certification and Restrictions on Use of Unpublished Data) and that the Information
Manager develop a security plan and procedures which are subject to periodic audit by
NASS. Restrictions on the data would still remain for any "public" use of the data that
does not adhere to the requirements of Form ADM-043 and pass a confidentiality
screening by NASS. The provisions of the MOU will allow the ARG to integrate and
analyze data in a much less restricted manner while protecting data confidentiality. Once
the EMAP Information Management System is on-line and the ARG is ready to become
a functioning node of that system, the question of "public" release of the data will be
addressed.
103
-------
The movement of pilot data from NASS to the ARC proved to be more difficult
than expected. Several data sets were required from NASS: JES for all of North Carolina;
the JES that was performed on the EMAP Hexagon segments; the Fall Survey data; and
several reference files. Most of the difficulties encountered were either related to
miscommunication or data confidentiality. A post-pilot meeting was held with NASS to
address these problems. As a result of that meeting, an e-mail link between NASS and
the ARC has been established, protocols and standards have been developed for data
interchange, and the MOU with NASS is under development.
3.3.4.3. The Agroecosystem Information Management System Infrastructure
Emphasis was given to the acquisition, installation, and learning of a technology
infrastructure for the ARC for the 1992 Pilot. The analysts used Sun workstations
configured with at least 24 megabytes of memory, with local and shared disks, and color
displays. The remaining staff each have an MS-DOS personal computer with at least 8
megabytes of memory.
A suite of software was available to the analysts. SAS and S-Plus for statistical
analysis; WordPerfect for word processing; ARC/INFO and Arc View for CIS; and Corel
Draw for graphics. Indicator leads, management, and support staff had WordPerfect,
Freelance, and various spreadsheets available for use, as well as access to the workstations.
All of the computers are attached to a local area network (LAN). The network has
a print, file, compute, and mail server attached to it that is accessible to all users. By
structuring the network in this way, expensive devices can be shared by all of the staff
and they can be configured in the most efficient manner. The network's most important
feature, file sharing, made the analysis of the 1992 Pilot data logistically simpler than if
files had not been shared. By providing access to current versions of common data files,
version control was easier to maintain,
The LAN is connected to the NCSU campus network, which is in turn connected
to the Internet. This allows members of the ARG housed in locations other than the ARC
headquarters to have access to all of the facilities described above through the NCSU
network and modem bank. Other important aspects of the Internet connection are
electronic mail to NASS and other EMAP personnel and the ability to become a node on
the EMAP Information System that is currently under development.
3.3.4.4. Data management
The movement and management of data (Fig. 3.6) was critical to the success of the
pilot. Planning for data management consisted of NASS/ARG interactions to determine
process, schedule, and flow of data from the field to the ARG. The processes were
monitored carefully and adjustments made as necessary. The flow of data, although
104
-------
Samples to Labs
Surveys to NASS
somewhat awkward, worked as planned. Once results were received from the analytical
laboratories, they were merged with the survey and sample design data.
From the perspective of the
ARG, NASS performed their IM
related duties well. When questions
or problems arose, they were solved
promptly. The ability to go back to
the source of the data (the NASS
enumerator) was tested and was
successful. NASS showed great
flexibility in adapting to changing
needs during the pilot, sometimes
transferring similar data sets to the
ARG several times. A person from
NASS to serve as the specific liaison
with the ARG would greatly
improve understanding and
coordination, decrease confusion,
and generally benefit the ARG effort.
This arrangement is being pursued.
As verification and validation
routines were performed on the data,
errors were encountered. For errors
in the laboratory sample data, direct
communication with the lab cleared
up all mistakes. For errors in the fall
survey data, there was not a clearcut
procedure for tracing back in the
data stream to find the source of the
error and to correct it. Most of the
time the ARG went to the NASS
State Office and consulted the
original paper form. This worked
reasonably well for a one state pilot, but it would have to be refined for larger field
projects.
When errors were discovered in the data, corrections were made only'by the
Information Manager. Corrected data sets were numbered consecutively and notes of the
corrections were attached to the data by means of documentation files. These files are
accessible to the ARG by means of a Gopher server on the ARG workstation server. The
Gopher entries were linked dynamically to the files so that changes would be available
•Subject lo NASS
confidentiality
Figure 3.6 Data flow for the 1992 Agroecosystem
Pilot.
105
-------
immediately to the staff. This method is also being used to document the ARG GIS
Library.
A challenge that confronted the analysts was to integrate the sample and survey
data into one data set and analyze duplicate and split-sample data without specific
identifiers attached to the data sets. Both of these tasks were difficult to accomplish
because the identifiers were withheld or changed to ensure data confidentiality. To solve
most of the problem, a cross-reference file was provided by the NASS that used a
sequential number (value from 1 to N, where N is the number of sample units) as the
identifier to link the samples and surveys together. This did not solve the problem of
analyzing the duplicate and split samples because the sequential number applied only to
the primary sample for that sample site. Eventually, enough information was included
in the cross-reference file to distinguish the specific type of sample. This problem was
encountered again when comparisons of the two sample designs required strata and sub-
strata identifiers for the NASS segments. NASS is understandably reluctant to release
segment numbers because when combined with a map that identifies them, they define
exact location of a 1 square mile area.
The ARG made some use of external data during the 1992 Pilot. The North
Carolina State Soil Survey Database was used for the soil quality and crop productivity
indicators, the Soil Conservation Service's Land Resource Region boundaries were used
as ecologically meaningful regions for data summarization, NASS strata were used for
the land use indicator, NASS aerial photography was used for soil map unit identification,
and climate data were obtained for use in indicator research.
3.3.4.5. Future Efforts
The focus for information management during the 1992 Pilot was not on systems
development, but ongoing through the procedure for the first time with close attention to
the individual processes. Most of the data management and analysis tools were created
de novo and many times performed manually, which allowed for a thorough
understanding of the details. This understanding will be used to review the requirements
for the Agroecosystem information system.
For future monitoring efforts, the ARG and NASS need to cooperate more closely
in the information management arena. Much remains to be learned about each other's
operations, equipment, software, and expertise that could be used to benefit the EMAP
program. NASS has recently finished an exhaustive search for a relational database
management system to use for their future enterprise data management system. Since the
EMAP Information Management System is being developed with a relational database
management system, it is to the benefit of our working relationship to be aware of the
parallel development efforts. The Soil Conservation Service is starting a similar process
to re-engineer their soils databases.
106
-------
Areas such as version control, data documentation, database management, security,
data modelling, metadata, verification/validation, and sample tracking need to. be
developed. The management of the data in 1992 was accomplished using SAS. Using SAS
for management of data was found to be unacceptable. It is imperative that the ARC
procure and began using a relational database management system as soon as possible.
107
-------
4. LITERATURE CITED
Angle, J.S., Gross, CM., Hill, R.L., and Mclntosh, M.S. 1993. Soil nitrate concentrations
under corn as affected by tillage, manure, and fertilizer application. /. Environ.
Qual. 22:141-147.
Barker, J.C., Zublena, J.P., and Campbell, C.R. 1991. Livestock manure production rates
and nutrient content, pp. 322-324 in The 1991 North Carolina Agricultural Chemicals
Manual. The College of Agriculture and Life Sciences, North Carolina State
University at Raleigh. January 1991.
Bongers, T. 1990. The maturity index: an ecological measure of environmental
disturbance based on nematode species composition. Oecologia 83: 14-19.
Bouchard, D.C., Williams, M.K., and Surampalli, R.Y. 1992. Nitrate contamination of
groundwater: sources and potential health effects. Journal AWMA. September
1992. pp. 85-90.
Burkholder, JoAnn (NCSU Dept. of Botany), personal communication re. significance of
nitrate in ponds. 10 September 1993
Cotter, J., and Nealon, J. 1987. Area frame design" for agricultural surveys. U.S.
Department of Agriculture, National Agricultural Statistics Service, Research and
Applications Division, Area Frame Section. Washington, DC.
EMAP. 1993. Environmental Monitoring and Assessment Program Guide. EPA/620/R-
93/012. Research Triangle Park, NC: U.S. Environmental Protection Agency,
Atmospheric Research and Exposure Assessment Laboratory.
Fedkiw, J. 1991. Nitrate Occurrence in U.S. Waters (and Related Questions): A Reference
Summary of Published Sources from an Agricultural Perspective, United States
Department of Agriculture. Washington, D.C. September 1991.
Heck, W.W., Campbell C.L., Finkner A.L., Hayes CM., Hess G.R., Meyer J.R., Munster
M.J., Neher D., Peck S.L., Rawlings, Smith C.N., and Tooley M.B. 1992.
Environmental Monitoring and Assessment Program - Agroecosystem 1992 Pilot
Project Plan. EPA/620/R-93/010. U.S. Environmental Protection Agency,
Washington, D.C.
Lai, R. 1991. Soil structure and sustainability. Journal of Sustainable Agriculture 1(4 ):67-92.
Ludwig, J.A., and Reynolds, J.F. 1988. Statistical Ecology: A Primer on Methods and
Computing. John Wiley, New York.
108
-------
Neher, D.A., and Campbell, C. L. 1994 Nematode communities and microbial biomass
in soils with annual and perennial crops. Appl. Soil Ecol. 1: (in press).
Neher, D.A., Peck, S.L., Rawlings, J.O., and Campbell C.L. 1994. Measures of nematode
community structure for an agroecosystem monitoring program and sources of
variability among and within agricultural fields. Plant and Soil (in press).
Nelson, D. W., and Sommers L. E. (1982). "Total carbon, organic carbon, and organic
matter." In Methods of Soil Analysis, Part 2. American Society of Agronomy and
Soil Science Society of America, Madison, WI.
Olsen, S. R., and Sommers L. E. (1982). "Phosphorus." In Methods of Soil Analysis, Part
2. American Society of Agronomy and Soil Science Society of America, Madison,
WI.
Overton, W.S. 1993. Probability Sampling and Population Inference in Monitoring
Programs. Department of Statistics, Oregon State University, Corvallis, OR.
Overton, W.S., White D., and Stevens D.L. 1991. Design Report for EMAP -
Environmental Monitoring and Assessment Program. EPA/600/3-91/053. U.S.
Environmental Protection Agency, Washington, D.C.
Payne, William R. (Project Officer). 1992. QAP for Agroecosystem EMAP North
Carolina Pilot including Sampson County Project: Nitrate Analysis. Task#: 416.
Internal Document of ERL-Athens. October 26, 1992.
Pimentel, D., Kurd, L.E., Bellotti, A.C., Forster, M.J., Oka, I.N., Sholes, O.D., and
Whitman, R.J. 1973. Food production and the energy crisis. Science 182: 443-449.
Rawlings, J.0.1988. Applied Regression Analysis. Wadsworth & Brooks/Cole Advanced
Book & Software. Pacific Grove, California.
Ricklefs, R. E. 1990. Ecology. Third Edition. Chiron Press, New York.
Shannon, C.E., and Weaver, W. 1949. The Mathematical Theory of Communication.
Univ. Illinois, Urbana.
Southwell, P.H., and Rothwell, T.M. 1977. Report on Analysis of Output/Input Energy
Ratios of Food Production in Ontario. March 31, 1977. School of Engineering,
University of Guelph, Ontario, Canada. Contract serial number OSW76-00048.
Spalding, R.F., and Exner, M.E. 1993. Occurrence of nitrate in groundwater — a review.
/. Environ. Qual. 22:392-402
109
-------
Stevens, D.L., Jr. 1993. Implementation of a National Environmental Monitoring
Program, submitted to the Journal of Environmental Management.
Yeates, G.W. 1994. Modification and qualification of the nematode maturity index.
Pedobiologia (in press).
USDA-National Agricultural Statistics Service. 1992. Agricultural Chemical Usage: 1992
Field Crops Summary, USDA-NASS, Agricultural Statistics Board, Washington,
D.C, March 1993, Ag Ch 1 (93)
110
•ftU.S. GOVERNMENT PRINTING OFFICE: 1994 - 550-001/80394
------- |