Subscriber access provided by UNIV OF NEBRASKA - LINCOLN
Article
Regional effects of agricultural conservation practices on nutrient transport in the Upper Mississippi River Basin Ana Maria Garcia, Richard Brown Alexander, Jeff Arnold, Lee Norfleet, Michael J. White, Dale Robertson, and Gregory E. Schwarz Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.5b03543 • Publication Date (Web): 31 May 2016 Downloaded from http://pubs.acs.org on May 31, 2016
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 34
Environmental Science & Technology
TOC Art 47x26mm (600 x 600 DPI)
ACS Paragon Plus Environment
Environmental Science & Technology
Page 2 of 34
1
Regional effects of agricultural conservation
2
practices on nutrient transport in the Upper
3
Mississippi River Basin
4
Ana María García*†, Richard B. Alexander‡, Jeffrey G. Arnold§, Lee Norfleet¦, Michael J.
5
White§, Dale M. Robertson±, Gregory Schwarz‡.
6
AUTHOR ADDRESS
7
† U.S. Geological Survey, 3916 Sunset Ridge Rd., Raleigh, North Carolina 02906, United States
8
‡ U.S. Geological Survey, 432 National Center, Reston, Virginia 20192, United States
9
§ U. S. Department of Agriculture, Agricultural Research Service, Grassland Soil and Water
10
Research Laboratory, 808 East Blackland Rd. Temple, Texas 76502, United States
11
¦ U. S. Department of Agriculture, Natural Resources and Conservation Service, 101 East
12
Blackland Rd. Temple, Texas 76502, United States
13
± U.S. Geological Survey, Wisconsin Water Science Center, 8505 Research Way, Middleton,
14
Wisconsin 53562, United States
15
ACS Paragon Plus Environment
1
Page 3 of 34
Environmental Science & Technology
16
ABSTRACT
17
Despite progress in the implementation of conservation practices, related improvements in water
18
quality have been challenging to measure in larger river systems. In this paper we quantify these
19
downstream effects by applying the empirical USGS water-quality model SPARROW to
20
investigate whether spatial differences in conservation intensity were statistically correlated with
21
variations in nutrient loads. In contrast to other forms of water quality data analysis, the
22
application of SPARROW controls for confounding factors such as hydrologic variability,
23
multiple sources and environmental processes. A measure of conservation intensity was derived
24
from the USDA-CEAP regional assessment of the Upper Mississippi River and used as an
25
explanatory variable in a model of the Upper Midwest. The spatial pattern of conservation
26
intensity was negatively correlated (p = 0.003) with the total nitrogen loads in streams in the
27
basin. Total phosphorus loads were weakly negatively correlated with conservation (p = 0.25).
28
Regional nitrogen reductions were estimated to range from 5 to 34 percent and phosphorus
29
reductions from 1 to 10 percent in major river basins of the Upper Mississippi region. The
30
statistical associations between conservation and nutrient loads are consistent with hydrological
31
and biogeochemical processes such as denitrification. The results provide empirical evidence at
32
the regional scale that conservation practices have had a larger statistically detectable effect on
33
nitrogen than on phosphorus loadings in streams and rivers of the Upper Mississippi Basin.
34
INTRODUCTION
35
Quantifying the environmental benefit of agricultural conservation practices has been a priority
36
to stakeholders and stewards of rivers, lakes and estuaries that have been deteriorated by the
37
intensification of agricultural production in the United States. Agricultural conservation
ACS Paragon Plus Environment
2
Environmental Science & Technology
Page 4 of 34
38
programs, which range from voluntary technical assistance only to payment-based voluntary and
39
cross-compliance programs, have been implemented since the Food Security Act of 19851 with
40
an early focus on the viability of agricultural production through soil conservation. The Farm
41
Security and Rural Investment Act of 20022 substantially increased the level of public funding
42
for conservation and initiated the goal of maximizing environmental benefit 3,4. Subsequently,
43
the Conservation Effects Assessment Project (CEAP) was established to provide science-based
44
guidance on the best use of funding for conservation and to facilitate the alignment of
45
conservation programs with national environmental protection priorities such as the restoration
46
of the Gulf of Mexico.
47
Much of the experimental research in conservation documents local (field and farm scale)
48
benefits of conservation practices5,6 but broader off-farm effects have been more difficult to
49
observe7. Reviews of 14 Agricultural Research Service (ARS) benchmark studies8 and the 13
50
National Institute of Food and Agriculture watershed studies9 document the findings of a few
51
watershed (1 – 2500 km2) experimental studies that demonstrate improvements in stream water
52
quality attributable to conservation practices. Examples include a paired watershed experimental
53
approach that found that nutrient management reduced nitrate concentrations in tile drained
54
watersheds by 30% in small watersheds (~4-8 km2) in Iowa10. Kuhnle et al.11 documented a 60%
55
reduction in sediment loads over a 17-yr time period in an experimental catchment in northern
56
Mississippi (~21 km2) where 20% of cropland had been converted to permanent cover by
57
enrollment in the Conservation Reserve Program. Literature reviews have documented decreases
58
in particulate phosphorus losses with erosion control and increases in soluble phosphorus
59
availability with some structural practices, such as conservation tillage6,12. Overall, few
60
watershed experimental or monitoring studies have demonstrated improvements in water quality
ACS Paragon Plus Environment
3
Page 5 of 34
Environmental Science & Technology
61
that could be attributed directly to the implementation of conservation practices9. At regional
62
scales (> 10,000 km2) trend analyses of water quality records at monitored outlets of large,
63
predominantly agricultural watersheds have shown both increases and decreases in nutrients
64
during time periods when conservation practices were adopted. For example, Sprague et al13
65
found that flow-normalized stream nitrate concentrations had increased at most sites in the Upper
66
Mississippi during a time span that included some of the largest increases in conservation
67
funding (1980 – 2008); however, this time period also included increased fertilizer
68
consumption14,15 and significant hydrologic modification by artificial drainage16. By contrast,
69
Murphy et al.17 found recent (2000-2010) decreases in in flow-normalized stream nitrate
70
concentrations for the Iowa and Illinois Rivers. A key limiting factor of most of these
71
environmental and observational studies is the inability to measure and control for simultaneous
72
processes such as multiple nutrient sources, transport processes and hydrological variability8.
73
To address limitations of observational studies and trend analysis we present an alternative
74
empirical methodology to infer whether conservation practices have had a realized and
75
statistically detectable effect on nutrient loads in large, predominantly agricultural watersheds of
76
the Upper Mississippi River Basin. We apply the USGS SPARROW (SPAtially Referenced
77
Regression On Watershed attributes) model18,19 to investigate the correlations between water
78
quality changes across multiple (~700 for the Upper Midwest) monitoring stations and adopted
79
conservation practices. The SPARROW model was developed to extend existing regression-
80
based data analysis by introducing a spatial-referencing methodology that allowed a space for
81
time substitution, and therefore contrast trend studies which rely on inferences regarding changes
82
over time. The methodology improves the correlation of explanatory variables with water quality
ACS Paragon Plus Environment
4
Environmental Science & Technology
83
measurements over non-spatially referenced data analysis, thereby improving the ability to
84
discern causal influences.
85
Page 6 of 34
We employ data and farm-scale process modeling that was previously developed for the
86
regional Upper Mississippi River Basin CEAP assessment20. This includes the NRI-CEAP
87
Cropland Survey21, an unprecedented documentation of the location, extent, and types of
88
conservation practices that have been implemented. As part of the regional CEAP assessment,
89
the survey data were represented in field-scale simulation models using the Agricultural
90
Productivity and Extension (APEX) model22 incorporating the best-available process knowledge
91
of the expected effects on nutrient edge-of-field delivery. This approach allowed for multiple
92
conservation practices to be summarized and aggregated into nutrient-specific, integrated indices
93
of conservation activities on farms while ensuring that specific farm-level practices were not
94
disclosed, in compliance with disclosure clauses, specifically Section 1619 of the Food,
95
Conservation, and Energy Act, 2008. The indices of conservation intensity derived from CEAP
96
conservation modeling were incorporated into a SPARROW model of the Upper Mississippi
97
River Basin to statistically test for the downstream effects of established practices on nitrogen
98
and phosphorus transport in streams.
99
The Agricultural Research Service (ARS) benchmark studies8 and the National Institute
100
of Food and Agriculture watershed studies9 also documented the findings of watershed modeling
101
that forecasts improvements in stream water quality. Most of these studies applied a priori
102
scientific understanding of the physical system to project response and provide technologically
103
feasible conservation effects. Yet, the level of complexity afforded by process models can lead to
104
uncertainties in the linkages established between conservation practices and in-stream water
105
quality response given 8,23. The CEAP assessment19 for the Upper Mississippi River Basin
ACS Paragon Plus Environment
5
Page 7 of 34
Environmental Science & Technology
106
coupled the APEX-predicted nutrient loads from cultivated cropland with the Soil Water
107
Assessment Tool (SWAT) to physically represent how practices function in agronomic systems
108
and would impact downstream water quality in large river basins. However, it is not currently
109
known whether the downstream benefits forecasted by process models have in fact been realized
110
in the watersheds of the Upper Mississippi River Basin. Studies that further integrate
111
observational data and modeling have been advocated to advance conservation science and
112
policy8,24,25.
113
With the complimentary application of process understanding in the APEX modeling,
114
conservation data derived from the CEAP cropland survey and the empirical assessment
115
framework in SPARROW we investigate the measured effect – if any – of agricultural
116
conservation practices, including those were not strategically designed for environmental benefit.
117
With this, we intend to fill an important gap in conservation research and provide a regional-
118
scale empirical assessment of conservation effects and an estimate of the magnitude of
119
downstream impacts.
120
METHODS
121
We perform an inference analysis by evaluating the empirical associations that result from the
122
inclusion of conservation information into regression-based SPARROW analyses of nutrient
123
transport in the Upper Mississippi River Basin. The SPARROW model is a nonlinear least-
124
squares multiple regression on catchments of a hydrologic framework to solve a mathematical
125
expression of constituent mass. Mean annual in-stream constituent load at the outlet of
126
catchment, i, is expressed as a function of landscape and instream characteristics such that,
127
= ∑ ∑∈ + ′∑ ∑ exp∑
ACS Paragon Plus Environment
(1)
6
Environmental Science & Technology
Page 8 of 34
128
where is an explanatory variable representative of direct nutrient source, indexed by n,
129
having source coefficient , is a source-specific upland or land-to-water delivery
130
explanatory variable, indexed by m, mediated by source/delivery-variable-specific coefficient
131
, and the functional term, A(⋅), accounts for in-channel and reservoir processing, depending
132
on a set of k-indexed attenuation variables, , and associated coefficients, . The functional
133
term, A’, represents attenuation applied to load entering the reach network at catchment i, and is
134
evaluated as √ if reach i is a stream, and equal to if reach i is a reservoir. The term ∑∈
135
corresponds to constituent load leaving the set of reaches , directly upstream of reach i. In
136
model estimation, if any of the contributing upstream reaches are monitored, those reaches use
137
the monitored value of load to represent the load from that reach; otherwise, for predictions or
138
for unmonitored upstream reaches, the contributing load is estimated using the modeled load
139
given by equation (1). If reach i is monitored, a residual in logarithm space, , may be
140
determined from monitored load, , and predicted load, , given by equation (1) such that
141
= ln − ln . The model is estimated using nonlinear optimization methods to determine the
142
values of the coefficients , and that minimize the sum of squared residuals across all
143
monitored sites. The residuals are assumed to be independent, identically distributed, and have
144
zero mean. Further details on the theoretical development of the SPARROW model are provided
145
by Smith et al.18 and Schwarz et al.19
146
The starting framework for this study was based on previously published SPARROW models24
147
developed for the U.S. portion of the Great Lakes, Upper Mississippi, Ohio, and Red River
148
Basins, a 1,379,100 km2 area in the Upper Midwest, represented by approximately 12,000
149
catchments of a 1 to 500,000 scale hydrologic framework. The models used nonlinear least-
150
squares regression methods to obtain coefficient estimates that minimize the squared residuals
ACS Paragon Plus Environment
7
Page 9 of 34
Environmental Science & Technology
151
implied by the model in equation 1. The response variables in the regressions were mean annual
152
loads at 708 water quality stations for nitrogen and 810 for phosphorus. Robertson and Saad26
153
compiled data collected from 1970-2006 (with most water-quality records spanning the period
154
1980 to 2004) at a spatial density of approximately 18 water quality monitoring sites per 4-digit
155
HUC (figure 1). Nutrient loads were estimated using rating curve estimation methods and de-
156
trended to represent long term mean annual conditions centered at 2002, removing the potentially
157
confounding effects of intra- and inter- annual variations in climate and hydrology25. Further
158
information on the previously published Upper Midwest models including the spatial data
159
sources used as explanatory variables in the models, model robustness discussion and comments
160
on the use of rating curve estimation methods are provided in the Supporting Information (SI).
161
In order to move the descriptive exercise represented by the published models into the
162
experimental framework needed for this study, modifications were made to explanatory variables
163
to accommodate information on agricultural conservation practices. The STATSGO erodibility
164
factor (K-factor) was added as an explicit representation of erosion, an important feature given
165
that many conservation practices have been developed to support erosion control. The
166
explanatory variable for tile drainage flow was removed from the published Upper Midwest; the
167
estimated coefficient associated lower phosphorus delivery with higher tile-drainage intensity as
168
limited surrogate for erosion processes. This change did not necessarily remove the
169
representation of tile-flow from the models. In contrast to mechanistic models, where explicit
170
representation is needed to fully ‘account’ for a particular transport mechanism, the empirical
171
nature of SPARROW implies that estimated coefficients implicitly aggregate the effect of
172
multiple transport processes that are spatially correlated with stream nutrient loads. The effects
173
of processes that are not explicitly described by the explanatory variables of the model and are
ACS Paragon Plus Environment
8
Environmental Science & Technology
Page 10 of 34
174
uncorrelated with nutrient loads are relegated to the spatially explicit model error term and
175
quantified as part of the prediction uncertainties.
176
As detailed later in this section, information about conservation practices were obtained for
177
crop agriculture in the Upper Mississippi River Basin, which is 35% of the Upper Midwest study
178
area (figure 1). Conservation practices are incorporated into the model as a land-to-water
179
delivery variable. To accommodate this it is necessary to develop a regionally stratified
180
SPARROW model specification. This allows for different agricultural source coefficients
181
between the UMRB and remaining portions of the UM basin. The estimates of the coefficients is
182
statistically informed by a subset of the observed loads from monitoring stations located in the
183
UMRB (252 for total nitrogen and 324 for total phosphorus). The remaining model coefficients
184
are constrained to have the same values in both the UMRB and remaining areas of the UM basin.
185
The modified nitrogen model contains four crop agriculture-related variables: cropland (UM),
186
cropland (UMRB), manure from confined sources (UM) and manure from confined sources
187
(UMRB). To limit correlated terms we retained cropland as an explanatory variable for fertilizer
188
loss and nitrogen fixation by legumes in crop rotations.
189
ACS Paragon Plus Environment
9
Page 11 of 34
Environmental Science & Technology
Upper Mississippi R. Study Area Upper Midwest Study Area
UPPER MISSSIPPI MAJOR (HUC4) RIVER BASINS
Mississippi Headwaters (0701) Minnesota River Basin (0702) St. Croix River Basin (0703) Upper Mississippi-Black-Root Rivers (0704) Chippewa River Basin (0705) Upper Mississippi-Maquoketa-Plum Rivers (0706) Wisconsin River Basin (0707) Upper Mississippi-Iowa-Skunk-Wapsipinicon Rivers (0708) Rock River Basin (0709) Des Moines River Basin (0710) Upper Mississippi-Salt Rivers (0711) Upper Illinois River Basin only (0712) Lower Illinois River Basin (0713) Upper Mississippi exclusive of Missouri (0714)
Base from USDA-NRCS, USGS and EPA. Watershed Boundary Dataset 1:24,000, Accessed 05/26/2016.
190 191
Figure 1. Map of major river basins (4-digit HUCs) in the Upper Mississippi River Basin
192
(UMRB) and distribution of water quality monitoring sites used for SPARROW. Inset: Spatial
193
extent of the Upper Midwest SPARROW models.
ACS Paragon Plus Environment
10
Environmental Science & Technology
Page 12 of 34
194
We term the modified Upper Midwest models, ‘SPARROW models without conservation
195
variable’ (columns 1-3, table 1 and table 2) although we hypothesized that conservation effects –
196
if any, were contained in the aggregate of estimated coefficients for cropland agriculture. The
197
models are the comparison basis for the empirical test of correlation of conservation practices
198
with stream nutrient loads. To perform this test, conservation intensity variable was added as a
199
land-to-water delivery term in the SPARROW model to be treated as intensive transport
200
properties in SPARROW, similar to a soil characteristic. Land-to-water delivery explanatory
201
variables and estimated coefficients (Znmi and in equation 1) establish associations to
202
climatic, natural or anthropogenic landscape processes that affect contaminant transport to
203
streams in the reach network. A positive coefficient indicates that catchments with large values
204
of the associated land-to-water variable have an increased rate of delivery of a nutrient source
205
( ) to the stream, as expressed by the exp∑ term in equation 1. Therefore, the
206
magnitude of any effect of a conservation variable on nutrient load was not presumed but was
207
instead inferred from the estimation process.
208
We used a conservation variable that was a spatially explicit representation of information
209
obtained through the NRCS-CEAP regional assessment strategy20,27,28. The NRCS-CEAP
210
regional assessment strategy involved developing predictive scenarios that represented the
211
farmer practice information obtained through the NRI-CEAP Cropland Survey21. The survey
212
results were modeled with the Agricultural Productivity and Extension (APEX) model22 which
213
simulated a ‘baseline’ of edge-of-field loss of phosphorus, nitrogen, pesticides and sediment that
214
from the documented farming activities. A second APEX modeling scenario (i.e., no practice)
215
simulated farming activities without the voluntary conservation practices and incentives that are
216
included in the baseline scenario. For example, for the no-practice scenario (figures S1-b and S1-
ACS Paragon Plus Environment
11
Page 13 of 34
Environmental Science & Technology
217
e in SI), structural best management practices such as reduced tillage, grass terraces, grass
218
waterways, riparian buffers were removed and formerly cultivated land that is currently retired
219
was brought back into agricultural production. The results were aggregated to the HUC-8 level
220
and shared with USGS for this analysis. Further details on the CEAP modeling framework and
221
the HUC-8 APEX model results are provided in the SI.
222
The conservation variable was computed as the relative difference between the APEX loads for
223
the baseline and the no-practice scenario (figures S1-c and S1-f) such that,
224
∆
=
∗
−
/
(2)
225
where Pk* is the nutrient load for the kth HUC-8 for the no practice scenario in mass units (figures
226
S2-b and S2-e, Table S7, columns 5 and 6); Pk is the load for the baseline scenario, also in mass
227
units (figure S2-a and S2-d; Table S7, columns 2 and 3). To allocate the data to the Upper
228
Midwest SPARROW hydrologic framework, we calculated a nutrient-specific conservation
229
intensity variable, #$, defined as &crop
#$, = &
230
crop+
∆
(3)
231
where crop ⁄crop+ is the ratio of cropland the ith SPARROW catchment to area in cropland
232
within the kth HUC-8.
233
The conservation intensity variable, as defined by equation 3 (figures 2b and 2c) is dependent
234
on both process-model predictions of conservation effectiveness and the synthesis of
235
conservation implementation from gathered survey data. Therefore higher values are associated
236
with both higher expected effectiveness and higher levels of overall implementation. Using the
237
relative difference between modeled scenarios minimizes the dependence on assumptions and
ACS Paragon Plus Environment
12
Environmental Science & Technology
Page 14 of 34
238
uncertainties related to the APEX model development and calibration. To illustrate this, we can
239
conceptualize the APEX scenarios as simple loading models where load P = SLC, with S are
240
nutrient inputs, L represent transport processes that embed process understanding, such as
241
algorithms that account for tile-drainage, and C an impact factor related to conservation-specific
242
activities. The relative difference between APEX edge-of-field loadings for the baseline
243
scenario, Po and the no-practice scenario, P* (figures S1-c and S1-f) becomes
244
∆ = ∗ - ∗ − - / - = ∗ - ∗ − - / -
245
The actual specification of APEX loading is more complicated than this simple description yet
246
we can expect that some multiplicative effect of process understanding, assuming they are equal
247
in both scenarios, ‘drop out’ (L drops out in equation 2) and the quantity depends primarily on
248
relative source inputs and conservation practice.
(4)
249
Because of the empirical nature of SPARROW, correlations between the conservation variable
250
and other explanatory variables with similar spatial geography could limit the interpretability of
251
the empirical experiment. The correlation to cropland agriculture is small (Pearson correlation =
252
0.1 p < 0.0001 for TN and 0.2, p < 0.0001 for TP). We further assessed these issues by
253
performing a multivariate standard regression on all transport variables present in SPARROW
254
models for the central USA: (1) the published Upper Midwest Model, (2) the Mississippi and
255
Atchafalaya River Basin model29. The residuals of the multivariate regression, which we here
256
term ‘conservation variable residuals’ were used as explanatory variables in SPARROW models
257
(tables S.5 and S.6 in the supporting information). The statistical significance of the correlation
258
between the conservation variable residuals and in-stream loads was essentially the same as the
259
results in table 1: p = 0.003 for both nitrogen SPARROW models and residual of the phosphorus
260
conservation intensity variable was found to remained statistically insignificant (p = 0.25 to p =
ACS Paragon Plus Environment
13
Page 15 of 34
Environmental Science & Technology
261
0.51). We conclude that multi-collinearity does not impede process interpretations of results
262
obtained in the estimation: model statistics and estimated coefficients are essentially the same for
263
nitrogen and phosphorus models when correlations were removed.
264
ACS Paragon Plus Environment
14
Environmental Science & Technology
(a) Extent of cropland
(b) Conservation variable, TN
Percent area
kg/kg
Page 16 of 34
(c) Conservation variable, TP
kg/kg
0% - 10.9%
0 - 0.0008
11% - 32.9%
0.00081 - 0.00419
33% - 52%
0.0042 - 0.0101
52.1% - 74.4%
0.0102 - 0.0232
74.5% - 100%
0.0233 - 0.381
0 - 0.0015 0.00151 - 0.0072 0.00721 - 0.0179 0.018 - 0.0402 0.0403 - 0.931
Base from USDA-NRCS, USGS and EPA. Watershed Boundary Dataset 1:24,000, Accessed 05/26/2016.
265 266 267
Figure 2. Spatial distribution of (A) cropland, the explanatory conservation intensity variable for (B) total nitrogen (TN) and (C) total phosphorus (TP) for the Upper Mississippi River basin. Interval bounds are quantiles of the distribution.
268
ACS Paragon Plus Environment
15
Page 17 of 34
269
Environmental Science & Technology
RESULTS AND DISCUSSION
270
The conservation intensity explanatory variable was found to be inversely correlated with total
271
nitrogen delivery to streams as indicated by a statistically significant (p = 0.003) and negative
272
land-to-water delivery coefficient, indicating that higher levels of conservation intensity are
273
associated with lower levels of nitrogen delivery to nearby streams (columns 4-6, table 1). For
274
total phosphorus the inclusion of the conservation variable also resulted in a negative land-to-
275
water coefficient, but the statistical significance is weak by conventional statistical measures (p=
276
0.25) (columns 4-6, table 2). In fact, the confidence interval includes positive numbers (-4.09 –
277
3.55) implying that the mean coefficient is statistically indistinguishable from zero. Explanatory
278
variables in tables 1 and 2 differ in units and the estimated coefficients are not standardized and
279
not directly comparable. The magnitude of the conservation effects associated with the estimated
280
coefficients were evaluated at a regional-scale and are presented in the next section.
281
The SPARROW with conservation variable re-interprets the observed data set with an explicit
282
representation of conservation effects. Adding the conservation intensity variables to
283
SPARROW did not fundamentally change model interpretation: explanatory variables account
284
for 86% of the variance in the logarithm of observed nitrogen yield and 74% of the phosphorus
285
yield variability. Explanatory variables representing non-agricultural sources and transport
286
properties, unrelated to conservation practices remained unchanged. The eigenvalue spread of
287
each model increased slightly but remained well under 100 indicating that multi-colinearity
288
would not limit interpretation.
ACS Paragon Plus Environment
16
Environmental Science & Technology
289 290
Page 18 of 34
Table 1. Estimated source (αn) land-to-water (./0 ) and in-channel (12 .) coefficients for SPARROW nitrogen models with and without conservation intensity variable*. SPARROW without conservation variable †
Explanatory variable
Coefficient‡
Standard Error
p§
0.112
*
SPARROW with conservation variable Coefficient
‡
[95% CI]
Standard Error
p
Source Point source (kg/yr)
0.801
0.799
0.111
*
[0.6 – 1.02] Atmospheric deposition (kg/yr)
0.545
0.0391
*
0.552
0.0391
*
1.530
*
2.784
*
0.0580
* *
[0.5 – 0.65] 2
Cropland, UM (km )
12.10
1.540
*
12.1 [8.78 – 15.3]
Cropland, UMRB (km2)
**
2.630
19.30
20.1 [13.5 – 27.1]
Manure from confined sources, UM 0.251 (kg/yr) Manure from confined source, UMRB ** (kg/yr)
*
0.0583
*
0.251 [0.173 – 0.35]
0.340
0.108
*
0.417 [0.07 – 0.75]
0.119
--
--
-6.4
2.134
Land to water delivery Conservation intensity (kg/km2)
--
0.003
[-9.67 – -0.0619] Fraction of catchments with tiles
1.250
0.132
*
1.22
0.131
.
[0.977 – 1.49] Drainage density (km/km2)
0.107
0.0561
0.0567
0.104
0.0557
0.0632
0.0186
0.0743
[-0.00092 – 0.216] Temperature (deg. C)
-0.033
0.0187
0.0801
-0.033 [-0.0745 – 0.00942]
ACS Paragon Plus Environment
17
Page 19 of 34
Environmental Science & Technology
SPARROW without conservation variable †
Explanatory variable
Clay content (%)
Coefficient‡ 0.019
Standard Error
p§
0.00387
*
SPARROW with conservation variable Coefficient
‡
[95% CI] 0.02
Standard Error
p
0.00383
*
0.00026
*
0.0989
*
[0.01 – 0.03] Precipitation (mm)
0.002
0.00027
*
0.00173 [0.00126 – 0.00219]
In-channel processing Loss in small streams (m3/s)
0.432
0.0988
*
0.445 [0.315 – 0.675]
Loss in medium streams (m3/s)
0.212
0.0910
0.0203
0.221
0.0915
0.0161
[0.047 – 0.57] Loss in reservoirs (yr./m)
5.320
1.260
*
5.52
1.280
*
[0.0333 – 10.9] Model diagnostics Number of sites, UM
708
708
Number of sites, UMRB
252
252
0.400
0.400
R of nutrient yield
0.850
0.860
Eigenvalue spread
42.64
43.66
RMSE of residuals 2
*
Variables significant at the 1 percent significance level; UMRB, Upper Mississippi River Basin; UM, Upper Midwest drainage area exclusive of the Upper Mississippi River; CI, confidence interval. † Descriptions of the datasets used to derive the explanatory variables are provided as supporting information (table S.1.) ‡ Coefficient units are inverse of explanatory variable units. § Reported p-values are for a single-tailed t-test for source, channel transport, and reservoir-loss coefficients and a two-tailed test for land-to-water coefficients. ** Source terms that were associated with conservation
ACS Paragon Plus Environment
18
Environmental Science & Technology
291 292
Page 20 of 34
Table 2. Estimated source (αn) land-to-water (./0 ) and in-channel (12 .) coefficients for SPARROW phosphorus models with and without conservation intensity variable*. SPARROW without conservation variable
SPARROW with conservation variable
†
Explanatory variable
‡
Coefficient
Standard Error
Coefficientǂ §
p
[95% CI]
Standard Error
p
Source Point source (kg/yr)
0.998
0.137
*
1.00
0.137
*
14.5
*
0.00313
*
0.00305
*
0.00795
0.0501
0.0114
*
0.0107
*
2.24
*
1.306
0.2
0.0966
*
[0.54 – 1.49] 2
Urban non-point (km )
63.19
14.47
*
64.2 [-1.83 – 128]
Fertilizer,UM (kg/yr)
0.019
0.00309
*
0.019 [0.01 – 0.03]
Fertilizer,UMRB (kg/yr)
**
0.013
0.00286
*
0.0133 [0.0053 – 0.02]
Manure from confined sources, UM (kg/yr)
0.015
0.00787
0.0524
0.0156 [0.00 – 0.03]
Manure from confined sources, UMRB (kg/yr)**
0.065
Manure from unconfined sources (kg/yr)
0.062
0.0103
*
0.0693 [0.0414 – 0.0924]
0.0107
*
0.061 [0.035 – 0.08]
2
Forest and wetlands (km )
20.06
2.251
*
19.8 [13.9 – 25.5]
Land to water delivery Conservation intensity (kg/km2)
-1.49 [-4.09 – 3.55]
Soil permeability
-0.306
0.0956
*
-0.323 [-0.508 – -0.0885]
ACS Paragon Plus Environment
19
Page 21 of 34
Environmental Science & Technology
SPARROW without conservation variable
SPARROW with conservation variable
†
Explanatory variable
Coefficient‡ K factor
Standard Error
6.550
1.160
Coefficientǂ p§ *
[95% CI]
Standard Error
6.26
p
1.178
*
0.0684
0.0316
0.0967
.
0.948
.
[3.33 – 9.04] In-channel processing Loss in small streams (m3/s)
0.149
0.0686
0.0296
0.147 [0.0381 – 0.262]
Loss in medium streams (m3/s)
0.254
0.0959
*
0.262 [0.126 – 0.387]
Loss in reservoirs (m/yr)
3.860
0.944
*
3.89 [-1.29 – 6.37]
Model diagnostics Number of sites, UM
810
810
Number of sites, UMRB
324
324
RMSE of log-transformed residual
0.49
0.49
R2 of nutrient yield, UM Model
0.73
0.74
Eigenvalue spread
40.13
43.05
*
Variables significant at the 1 percent significance level; UMRB, Upper Mississippi River Basin; UM, Upper Midwest drainage area exclusive of the Upper Mississippi River; CI, confidence interval. † Descriptions of the datasets used to derive the explanatory variables are provided as supporting information (table S.1.). ‡ Coefficient units are inverse of explanatory variable units. §§ Reported p-values are for a single-tailed t-test for source, channel transport, and reservoir-loss coefficients and a two-tailed test for land-to-water coefficients. ** Source terms that were associated with conservation.
ACS Paragon Plus Environment
20
Environmental Science & Technology
293
Page 22 of 34
Our finding of a statistically significant negative correlation between stream nitrogen loads and
294
conservation intensity is consistent with the hydrological and biogeochemical processes on land
295
surfaces and in the subsurface that can potentially interact with conservation practices to yield
296
nitrogen reductions in streams. The structural and erosion control practices that have been
297
implemented in the Upper Mississippi River Basin have been shown to reduce runoff and peak
298
flows, increasing water infiltration and soil water holding capacity6. Reactive forms of nitrogen
299
(ammonia, nitrate) are readily available and highly mobile in nitrogen-enriched agricultural soils
300
where high rates of nitrogen mineralization and nitrification favor dissolved inorganic over
301
organic, sediment-bound nitrogen30. The routing of larger quantities of water to the subsurface
302
by conservation practices contributes to increased hydraulic storage that can lead to higher
303
denitrification rates31,32 and reductions in nitrogen delivery to streams when compared to
304
equivalent areas without conservation practices. In areas where extensive tile drainage short
305
circuits natural flow paths and contributes to larger nitrate loads in streams33,34, structural
306
conservation practices, such as stream riparian buffers, can also increase hydraulic storage and
307
reduce nitrogen delivery to streams35–37. Soil conditions conducive to denitrification have been
308
observed in the shallow sub-surface throughout the Upper Mississippi region38,39 but substantial
309
leaching to ground water has also been noted, which can delay the delivery of nitrogen to
310
streams by several decades (10 – 40 years)40,41. Therefore, the effectiveness of conservation
311
practices in reducing nitrogen delivery to streams is highly dependent on subsurface hydrological
312
and biogeochemical conditions that favor the permanent removal of nitrogen via denitrification
313
and/or appreciably delay the transport and delivery of nitrogen to streams by leaching to deep
314
ground waters.
ACS Paragon Plus Environment
21
Page 23 of 34
315
Environmental Science & Technology
The weak negative correlation between conservation intensity and stream phosphorus loads
316
may be explained by the effects of several key factors. First, the downstream response to
317
conservation-related reductions in particulate phosphorus6 is subject to long time lags: sediment-
318
bound phosphorus can be stored on farm fields and can take decades (~25 years) to move
319
downstream as particles are repeatedly deposited and re-suspended, which can delay measurable
320
in-stream responses to changes in upland particulate phosphorus delivery42. In our analysis,
321
observing the effects of conservation practices on particulate phosphorus may be limited by the
322
use of stream water quality data detrended to a baseyear (2002) that is approximately coincident
323
with the time period represented by the conservation intensity measure. Additional years of
324
observations that can support the use of a later base year for detrending (post 2005) in
325
SPARROW could provide a more complete evaluation of a lagged conservation response in
326
stream particulate phosphorus loads.
327
Second, some erosion control practices, specifically, no-till and reduced tillage have been
328
shown to increase soluble phosphorus levels in farm runoff12,43, which can potentially offset
329
expected downstream benefits derived from conservation-related reductions in particulate
330
phosphorus. In soils with no-till and other forms of reduced tillage where there is less vertical
331
mixing of the soil and integration of phosphorus fertilizers, phosphorus can accumulate near the
332
surface, where desorption processes can lead to elevated levels of soluble phosphorus in overland
333
runoff12. Reduced tillage can also facilitate the development of soil macropores with connections
334
to tile drainage systems that speed the delivery of soluble phosphorus to streams43. The
335
availability of soluble phosphorus is also enhanced in farm soils where the phosphorus sorption
336
capacity has been saturated by legacy applications of phosphorus-enriched manure and
337
fertilizers44. With a substantial fraction of total phosphorus delivered in soluble form, and
ACS Paragon Plus Environment
22
Environmental Science & Technology
338
increasingly via sub-surface pathways, many conservation practices designed to remediate
339
erosion loss have limited effectiveness at controlling dissolved phosphorus losses. This
340
collection of confounding processes may explain the lack of statistical evidence in the
341
SPARROW analysis of the effects of conservation practices on total phosphorus in streams of
342
the Upper Mississippi.
343
Page 24 of 34
Predictions of regional-scale conservation effects.
344
To quantify the magnitude of possibly realized conservation effects in major rivers in the
345
UMRB, nutrient loads were predicted by using the estimated coefficients from the ‘SPARROW
346
with conservation’ columns (tables 1 and 2) and equation S1. Conservation effects were
347
subtracted from the SPARROW stream loads based on an assumed “no practice” condition
348
where the conservation intensity variable was set to reflect the absence of conservation practices
349
(Zcp, i = 0). A parametric bootstrap that presumes normality for all model coefficients was
350
performed to provide 90 percent confidence intervals for the SPARROW predictions. The results
351
were then summarized to the 4-digit HUC level by including the effects of nutrient processing
352
during instream transport and computing loads delivered to the outlet of the basin (figure 3). The
353
SPARROW predictions are presented alongside HUC-4 conservation effects predictions
354
documented in the Upper Mississippi CEAP20 assessment. The CEAP results were developed
355
using the Soil and Water Assessment Tool45 to route APEX-derived cropland loads downstream
356
and simulate nonpoint source loadings from land uses other than cropland.
357
While the comparison presented in figure 3 is of two differing assessment frameworks, it
358
allows us to make several conclusions. Our process inferences regarding the conservation impact
359
of nitrogen are explicitly represented in the mechanistic models. Most of the reduction presented
ACS Paragon Plus Environment
23
Page 25 of 34
Environmental Science & Technology
360
by the CEAP assessment is the result of representing hydrologic modifications, such as changes
361
in soil moisture content and related effects on nitrogen cycling processes including enhanced
362
denitrification processes. The close agreement in predictions of nitrogen load reductions with the
363
two approaches is likely related to the high mobility of reactive nitrogen that is favored in
364
agricultural systems, especially those with tile drainage, which leads to a rapid downstream
365
water-quality response. Thus, the results can be considered to provide a limited validation of
366
existing process understanding and we isolate a possible benefit of conservation: an overall
367
increase in hydraulic storage, which when coupled with denitrification could lead to permanent
368
nitrogen losses that reduce nitrogen delivery to streams.
369
Although the predictions for total phosphorus reductions differed substantially (figure 3b) they
370
illustrate the difference in assessment approaches. The weak empirical association obtained with
371
SPARROW led to the minor reductions: on average 4 percent reduction at major river basin
372
outlets with some lower bounds for the 90% confidence interval below 0. The average reduction
373
for the CEAP approach was 30 percent, with a significant portion attributed to nutrient
374
management. The Upper Mississippi cropland CEAP survey documented substantially more
375
phosphorus management than for nitrogen20. The process models simulate nutrient reductions as
376
immediate and sustained which leads a technologically-feasible forecast of water-quality
377
improvements. With the SPARROW analysis -- an assessment of already realized benefits
378
during the 90s and early 2000s, it is not possible to forecast a response that is not represented in
379
the observational dataset. Nutrient management was fully incentivized in Farm Security and
380
Rural Investment Act of 2002 through the Environmental Quality Incentives Program. In
381
comparison to well established soil management practices that have been landscape features for
382
years, if not decades, nutrient management have dynamic implementation pattern that would lead
ACS Paragon Plus Environment
24
Environmental Science & Technology
Page 26 of 34
383
to lags in measured effects. Therefore it is possible that additional water quality records, beyond
384
the early 2000s, may be needed to actually observe the downstream benefits of the managed
385
phosphorus reductions.
386
Our combined process-based and empirical assessments compliment temporal analyses of
387
aggregated water-quality monitoring records by providing an across the landscape exercise
388
assessment that leverages space over time and provides some degree of experimental control.
389
The process knowledge contained in the field-scale APEX loads facilitated the development of
390
nutrient-specific, integrated measures of multiple conservation activities on farms that allowed
391
an empirical regional-scale assessment of both our process-based understanding conservation
392
effects and documented levels of adoption and implementation. Applications of this
393
methodology in other environmental settings are needed to further evaluate the impacts of
394
conservation practices. With this study we identify a mitigating effect of conservation practices
395
for the delivery of total nitrogen streams. However, these relatively short-term benefits are
396
potentially offset in future years by the influence of deep leaching that introduces delays in the
397
delivery of nitrogen to streams and postpones the impacts of agriculture on water quality. For
398
phosphorus, the chemistry of agricultural soils and more recent research on soluble phosphorus
399
mobility and transport are leading to shifts in understanding that are either not fully incorporated
400
in process modeling or modeled with uncertainties. Obtaining an improved understanding of the
401
relative importance of these different processes in US watersheds is needed to optimize the
402
management of nutrients for environmental benefits.
ACS Paragon Plus Environment
25
Page 27 of 34
Environmental Science & Technology
(a)
403
(b)
404
ACS Paragon Plus Environment
26
Environmental Science & Technology
Page 28 of 34
405 4-Digit HUC
HUC Name
0701
Mississippi Headwaters
0702
Minnesota River Basin
0703
St. Croix River Basin
0704
Upper Mississippi-Black-Root Rivers
0705
Chippewa River Basin
0706
Upper Mississippi-Maquoketa-Plum Rivers
0707
Wisconsin River Basin
0708
Upper Mississippi-Iowa-Skunk-Wapsipinicon Rivers
0709
Rock River Basin
0710
Des Moines River Basin
0711
Upper Mississippi-Salt Rivers
0712
Upper Illinois River Basin only
0713
Lower Illinois River Basin
0714
Upper Mississippi, Kaskaskia and Meramec Rivers
406 407
Figure 3. Comparison for 4-digit HUC of nutrient reductions attributed to conservation.
408
ASSOCIATED CONTENT
409
Supporting Information. Additional details on modifications to previously published
410
SPARROW models, comments on the use of mean annual nutrient loads, description of CEAP
411
conservation effects datasets and nutrient loads from APEX models for scenarios. This material
412
is available free of charge via the Internet at http://pubs.acs.org.
413
AUTHOR INFORMATION
414
Corresponding Author
415
*U.S. Geological Survey, National Water Quality Program, 3916 Sunset Ridge Rd., Raleigh,
416
North Carolina 02906, United States. Phone: 919-571-4058 Fax: 919-571-4041Email:
417
[email protected] ACS Paragon Plus Environment
27
Page 29 of 34
Environmental Science & Technology
418
Author Contributions
419
The manuscript was written through contributions of all authors. All authors have given approval
420
to the final version of the manuscript.
421
REFERENCES
422
(1)
U. S. Congress. Food security act of 1985; 1985; p 198.
423
(2)
U. S. Congress. Farm security and rural investment act of 2002; 2002.
424
(3)
Becker, G. S. The 2002 farm law at a glance. In CRS Report for Congress RS21233.
425
Washington DC: Congressional Research Service; 2002.
426
(4)
427
our knowledge; Schnepf, M., Cox, C., Eds.; Soil and Water Conservation Society: Ankeny, IA,
428
2006.
429
(5)
430
and Research Needs. A Conservation Effects Assessment Bibliography; Makuch, J. R., Gagnon,
431
S. R., Sherman, T. J., Eds.; Special Reference Briefs; National Agricultural Library: Beltsville,
432
Maryland, 2004.
433
(6)
434
Conservation on Cropland. The Status of Our Knowledge; Schnepf, M., Cox, C., Eds.; Soil and
435
Water Conservation Society: Ankeny, IA, 2006.
436
(7)
437
watersheds; National Academies Press, 1999.
438
(8)
439
conservation practices: a review of USDA-ARS’s conservation effects assessment project
440
watershed studies. Water Sci. Technol. 2011, 64 (1), 300.
Cox, C. Foreword. In Environmental benefits of conservation on cropland: the status of
Agricultural Conservation Practices and Related Issues: Reviews of the State of the Art
Reeder, R.; Westermann, D. Soil Management Practices. In Environmental Benefits of
National Research Council (US) Committee on Watershed. New strategies for America’s
Tomer, M. D.; Locke, M. A. The challenge of documenting water quality benefits of
ACS Paragon Plus Environment
28
Environmental Science & Technology
Page 30 of 34
441
(9)
How to Build Better Agricultural Conservation Programs to Protect Water Quality: The
442
National Institute of Food and Agriculture–Conservation Effects Assessment Project Experience;
443
Osmond, D. L., Meals, D. W., Hoag, D. L., Arabi, M., Eds.; Soil and Water Conservation
444
Society: Ankeny, IA, 2012.
445
(10)
446
S. Using the late spring nitrate test to reduce nitrate loss within a watershed. J. Environ. Qual.
447
2004, 33 (2), 669–677.
448
(11)
449
practice effects on sediment load in the Goodwin Creek Experimental Watershed. J. Soil Water
450
Conserv. 2008, 63 (6), 496–503.
451
(12)
452
Modeling phosphorus transport in agricultural watersheds: Processes and possibilities. J. Soil
453
Water Conserv. 2002, 57 (6), 425–439.
454
(13)
455
Tributaries, 1980 to 2008: Are We Making Progress? Environ. Sci. Technol. 2011, 45 (17),
456
7209–7216.
457
(14)
458
River Basin for 200 Years. BioScience 2003, 53 (6), 563–572.
459
(15)
460
record of natural and anthropogenic induced low-oxygen conditions from Louisiana continental
461
shelf sediments. Geology 2005, 33 (4), 329–332.
Jaynes, D. B.; Dinnes, D. L.; Meek, D. W.; Karlen, D. L.; Cambardella, C. A.; Colvin, T.
Kuhnle, R. A.; Bingner, R. L.; Alonso, C. V.; Wilson, C. G.; Simon, A. Conservation
Sharpley, A. N.; Kleinman, P. J. A.; McDowell, R. W.; Gitau, M.; Bryant, R. B.
Sprague, L. A.; Hirsch, R. M.; Aulenbach, B. T. Nitrate in the Mississippi River and Its
Turner, R. E.; Rabalais, N. N. Linking Landscape and Water Quality in the Mississippi
Osterman, L. E.; Poore, R. Z.; Swarzenski, P. W.; Turner, R. E. Reconstructing a 180 yr
ACS Paragon Plus Environment
29
Page 31 of 34
Environmental Science & Technology
462
(16)
Magner, J. A.; Alexander, S. C. Geochemical and isotopic tracing of water in nested
463
southern Minnesota corn-belt watersheds. Water Sci. Technol. J. Int. Assoc. Water Pollut. Res.
464
2002, 45 (9), 37–42.
465
(17)
466
tributaries, 1980-2010: An update. US Geol. Surv. Sci. Investig. Rep. 2013, 5169.
467
(18)
468
monitoring data. Water Resour. Res. 1997, 33 (12), 2781–2798.
469
(19)
470
Surface Water-Quality Model: Theory, Application and User Documentation; Techniques and
471
Methods; Report 6-B3; 2006; p 256.
472
(20)
473
Conservation Practices on Cultivated Cropland in the Upper Mississippi River Basin; United
474
States Department of Agriculture (USDA), Natural Resources Conservation Service (NRCS),
475
2012; p 187.
476
(21)
Goebel, J. J. Statistical methodology for the NRI-CEAP cropland survey; 2012.
477
(22)
Williams, J. W.; Izaurralde, R. C.; Steglich, E. M. Agricultural policy/environmental
478
extender model; BREC 2008-17; Texas AgriLIFE Research, Texas A&M University, Blackland
479
Research and Extension Center: Temple, TX, 2008.
480
(23)
481
National Institute of Food and Agriculture–Conservation Effects Assessment Project. In How to
482
Build Better Agricultural Conservation Programs to Protect Water Quality: The National
483
Institute of Food and Agriculture–Conservation Effects Assessment Project Experience; Deana
Murphy, J. C.; Hirsch, R. M.; Sprague, L. A. Nitrate in the Mississippi River and its
Smith, R. A.; Schwarz, G. E.; Alexander, R. B. Regional interpretation of water-quality
Schwarz, G. E.; Hoos, A. B.; Alexander, R. B.; Smith, R. A. Section 3. The SPARROW
Conservation Effects Assessment Project (CEAP). Assessment of the Effects of
Mazdak Arabi; Donald W. Meals; Dana LK. Hoag. Chapter 5: Watershed Modeling:
ACS Paragon Plus Environment
30
Environmental Science & Technology
Page 32 of 34
484
Osmond, Donald W. Meals, Dana LK. Hoag, Mazdak Arabi, Eds.; Soil and Water Conservation
485
Society: Ankeny, IA, 2012.
486
(24)
487
Conducting an External Review of the US Department of Agriculture Conservation Effects
488
Assessment Project; Soil and Water Conservation Society: Ankeny, IA, 2006.
489
(25)
490
payment programs: U.S. experience in theory and practice. Ecol. Econ. 2008, 65 (4), 737–752.
491
(26)
492
and Watershed Estimated Using SPARROW Watershed Models1: Nutrient Inputs to the
493
Laurentian Great Lakes by Source and Watershed Estimated Using SPARROW Watershed
494
Models. JAWRA J. Am. Water Resour. Assoc. 2011, 47 (5), 1011–1033.
495
(27)
496
DiLuzio, M.; Wang, X.; Atwood, J.; et al. Nutrient delivery from the Mississippi River to the
497
Gulf of Mexico and effects of cropland conservation. J. Soil Water Conserv. 2014, 69 (1), 26–40.
498
(28)
499
APEX output for cultivated cropland with SWAT simulation for regional modeling. Trans
500
ASABE 2011, 54 (4), 1281–1298.
501
(29)
502
in the Mississippi/Atchafalaya River Basin. J. Environ. Qual. 2013, 42 (5), 1422.
503
(30)
504
W.; Schlesinger, W. H.; Tilman, D. G. Human alteration of the global nitrogen cycle: sources
505
and consequences. Ecol. Appl. 1997, 7 (3), 737–750.
Soil and Water Conservation Society. Final Report from the Blue Ribbon Panel
Claassen, R.; Cattaneo, A.; Johansson, R. Cost-effective design of agri-environmental
Robertson, D. M.; Saad, D. A. Nutrient Inputs to the Laurentian Great Lakes by Source
White, M. J.; Santhi, C.; Kannan, N.; Arnold, J. G.; Harmel, D.; Norfleet, L.; Allen, P.;
Wang, X.; Kannan, N.; Santhi, C.; Potter, S. R.; Williams, J. R.; Arnold, J. G. Integrating
Robertson, D. M.; Saad, D. A. SPARROW Models Used to Understand Nutrient Sources
Vitousek, P. M.; Aber, J. D.; Howarth, R. W.; Likens, G. E.; Matson, P. A.; Schindler, D.
ACS Paragon Plus Environment
31
Page 33 of 34
Environmental Science & Technology
506
(31)
Maag, M.; Vinther, F. P. Nitrous oxide emission by nitrification and denitrification in
507
different soil types and at different soil moisture contents and temperatures. Appl. Soil Ecol.
508
1996, 4 (1), 5–14.
509
(32)
510
N2O, N2 and CO2 from soil fertilized with nitrate: effect of compaction, soil moisture and
511
rewetting. Soil Biol. Biochem. 2006, 38 (2), 263–274.
512
(33)
513
small central Indiana watershed. J. Environ. Qual. 1998, 27 (4), 884–894.
514
(34)
515
nutrient movement into subsurface tile drains on a silt loam soil in Indiana. J. Environ. Qual.
516
1991, 20 (1), 264–270.
517
(35)
518
wetland: southern Minnesota, USA. Environ. Geol. 2008, 54 (7), 1367–1376.
519
(36)
520
conservation buffers. J. Soil Water Conserv. 2002, 57 (2), 36A – 43A.
521
(37)
522
cycling? Simulating the effects of flood alterations on floodplain denitrification. Glob. Change
523
Biol. 2005, 11 (8), 1352–1367.
524
(38)
525
groundwater in agricultural areas. Water Resour. Res. 2012, 48.
526
(39)
527
Sources and Denitrification in the Mississippi River, Illinois. J. Environ. Qual. 2006, 35 (2), 495.
Ruser, R.; Flessa, H.; Russow, R.; Schmidt, G.; Buegger, F.; Munch, J. C. Emission of
Fenelon, J. M.; Moore, R. C. Transport of agrichemicals to ground and surface water in a
Kladivko, E. J.; Van Scoyoc, G. E.; Monke, E. J.; Oates, K. M.; Pask, W. Pesticide and
Magner, J.; Alexander, S. Drainage and nutrient attenuation in a riparian interception-
Lowrance, R.; Dabney, S.; Schultz, R. Improving water and soil quality with
Gergel, S. E.; Carpenter, S. R.; Stanley, E. H. Do dams and levees impact nitrogen
Liao, L.; Green, C. T.; Bekins, B. A.; Böhlke, J. K. Factors controlling nitrate fluxes in
Panno, S. V.; Hackley, K. C.; Kelly, W. R.; Hwang, H.-H. Isotopic Evidence of Nitrate
ACS Paragon Plus Environment
32
Environmental Science & Technology
Page 34 of 34
528
(40)
Tesoriero, A. J.; Duff, J. H.; Saad, D. A.; Spahr, N. E.; Wolock, D. M. Vulnerability of
529
Streams to Legacy Nitrate Sources. Environ. Sci. Technol. 2013, 47 (8), 3623–3629.
530
(41)
531
water nitrate in two small watersheds. J. Environ. Qual. 2003, 32 (6), 2158–2171.
532
(42)
533
Best Management Practices: A Review. J. Environ. Qual. 2010, 39 (1), 85.
534
(43)
535
L.; Bergstrom, L.; Zhu, Q. Managing agricultural phosphorus for water quality protection:
536
principles for progress. Plant Soil 2011, 349 (1-2), 169–182.
537
(44)
538
indexing for cropland: Overview and basic concepts of the Iowa phosphorus index. J. Soil Water
539
Conserv. 2002, 57 (6), 440–447.
540
(45)
541
Assessment Tool: Historical Development, Applications, and Future Research Directions. Trans.
542
ASABE 2007, 50 (4), 1211–1250.
Tomer, M. D.; Burkart, M. R. Long-term effects of nitrogen fertilizer use on ground
Meals, D. W.; Dressing, S. A.; Davenport, T. E. Lag Time in Water Quality Response to
Kleinman, P. J. A.; Sharpley, A. N.; McDowell, R. W.; Flaten, D. N.; Buda, A. R.; Tao,
Mallarino, A. P.; Stewart, B. M.; Baker, J. L.; Downing, J. D.; Sawyer, J. E. Phosphorus
Gassman, P. W.; Reyes, M. R.; Green, C. H.; Arnold, J. G. The Soil and Water
543
ACS Paragon Plus Environment
33