Recommendations for PQUAL Sensitivity to Inputs
Guido Yactayo, Richard Tian, Gary Shenk, Lewis Linker, Gopal Bhatt
Chesapeake Bay Program
June 2014
1. Introduction
The Phase 5.3.2 (P532) HSPF watershed model uses the Agricultural Chemical Model (AGCHEM)
and the Pervious Quality Model (PQUAL) modules for the simulation of nutrient export from
land sources. AGCHEM simulates biological and chemical processes, while PQUAL simulates
nutrient fluxes using a simpler coefficient-based method. Although AGCHEM is a robust module
from academic and scientific standpoints, its high level of complexity represents an obstacle for
comprehension within the management community and introduces variability in response to
inputs that is not readily attributed to measureable watershed characteristics. Using only the
PQUAL module in phase 6 will be more straightforward for comprehension and decision
implementation.
Unlike the AGCHEM module, the PQUAL module is not generally sensitive to inputs of nutrients
such as atmospheric deposition, fertilizer, and manure. However, variations in these inputs are
featured in management scenarios. The effect of changes in nutrient inputs on nutrient
outflows from the land simulation must be estimated and then applied to the PQUAL
simulation. These input-output relationships are referred to in this document as sensitivity.
This technique allows for the direct use of multiple models in the estimation of these
sensitivities. The use of multiple models has been recommended to the CBP partnership by the
CBP Scientific and Technical Advisory Committee on several occasions (e.g., Friedrichs, et al,
2012).
Inputs, outputs, and sensitivity to inputs of existing watershed models used in the Chesapeake
Bay Watershed have been evaluated. The sensitivities obtained from the analysis of all models
will be used to specify functional links between nutrient inputs and outputs. Sensitivities will be
presented and then implemented within the PQUAL module in the phase 6 version of the
Chesapeake Bay Watershed Model.
2. Objectives
a) To compare watershed models inputs, outputs, and sensitivity to inputs.
b) To build, present and implement PQUAL sensitivities in the next version of the Chesapeake
Bay Watershed model.
3. Methods
Five different watershed models previously used in the Chesapeake Bay Watershed are
included in this analysis: the Phase 5.3.2 (P532) HSPF watershed model, the Spatially
Referenced Regression on Watershed Attributes (SPARROW) model, the Agricultural Policy
Environmental Extender (APEX) model, Annual Phosphorus Loss Estimator (APLE) model, and
the Functional Linear Concurrent Model (FLCM).
The Phase 5.3.2 CBP Watershed Model
For the 2017 mid-point assessment, a shift is planned from a combined PQUAL and AGCHEM
simulation, which was systematically used in the previous phases of the Chesapeake Bay
watershed model, to a full version of PQUAL. Yet, the Chesapeake Bay Program must develop a
new version of PQUAL as robust as the ACHEM prior to application. To this end, The
Chesapeake Bay Program Modeling Team has conducted a sensitivity analysis of the P532
AGCEHM simulation between all nutrient inputs and outputs in the Chesapeake Bay watershed.
This analysis incorporated input loadings and predicted yields information from fourteen
different P532 scenarios, including simulations from 1985 to 2011. The scenario list includes
runs such as the “No Action” scenario, in which all management practices are removed, and the
“E3” simulation, a theoretical scenario where maximum levels of pollution management control
are applied to the fullest extent. All scenarios were simulated over a ten-year hydrologic period
from 1991 to 2000, and average annual data of both inputs and outputs were used for the
analyses.
Multi-variate regressions, slopes between nutrient export and input loading, and ratios
between nutrient export and input loading were calculated to identify P532 landuse
sensitivities. These sensitivities are presented in Tables 1-4.
Table 1. Model sensitivity slopes between nutrient export and total loading on different land uses. TN: Total
nitrogen, DIN; Dissolved inorganic nitrogen; ON: Organic nitrogen; TP: Total phosphorus; PO4: Phosphate; OP:
Organic phosphorus; W: with; w/o: without. Blank cells indicate not applicable.
Land use
TN
DIN
ON
TP
PO4
Forest (for)
0.05
0.04
0.003
High-tillage w manure (hwm)
0.21
0.15
High-tillage w/o manure (hom)
0.48
Low-tillage w manure (lwm)
0.19
0.07
0.12
0.11
0.015
0.46
0.02
0.1
0.1
0.002
0.14
0.06
0.1
0.09
0.01
0.08
0.08
0
0.1
0.1
0.001
Hay w nutrient management (hym)
Hay w/o nutrient management (hyo)
0.3
0.3
0.005
Alfalfa (alf)
0.03
0.03
-0.002
Pasture (pas)
0.06
0.05
0.013
Non-regulated pervious urban (npd)
0.12
0.1
0.012
OP
Table 2. Model sensitivity slopes of multi-variate regression between total nitrogen (TN) export and different
types of loading on different land uses. A: atmospheric deposition; M: Manure; F: Fertilizer; U: Uptake; L:
Legume nitrogen fixation; w: with; w/o: without. Blank cells indicate not applicable.
Land use
A
M
Forest (for)
0.05
High-tillage w manure (hwm)
0.35
High-tillage w/o manure (hom)
0.44
Low-tillage w manure (lwm)
0.33
F
0.19
0.16
U
L
0.29
-0.09
0.34
0.23
-0.004
0.43
0.31
-0.13
0.36
Hay w nutrient management (hym)
Hay w/o nutrient management (hyo)
0.27
0.02
Alfalfa (alf)
0.22
0.08
Pasture (pas)
0.15
0.06
Non-regulated pervious urban (npd)
0.15
0.1
0.07
Table 3. Model sensitivity slopes of multi-variate regression between total phosphorus (TP) export and different
types of loading on different land uses. A: atmospheric deposition; M: Manure; F: Fertilizer; U: Uptake; L:
Legume nitrogen fixation; w: with; w/o: without. Blank cells indicate not applicable.
Land use
A
M
F
U
L
Forest (for)
High-tillage w manure (hwm)
0.13
High-tillage w/o manure (hom)
0.11
0.1
Low-tillage w manure (lwm)
0.11
0.09
Hay w nutrient management (hym)
0.09
0.1
0.15
0.09
Hay w/o nutrient management (hyo)
Alfalfa (alf)
Pasture (pas)
Non-regulated pervious urban (npd)
Table 4. Ratio between nutrient exports and total loading on different land uses. TN: Total nitrogen, DIN;
Dissolved inorganic nitrogen; ON: Organic nitrogen; TP: Total phosphorus; PO4: Phosphate; OP: Organic
phosphorus; W: with; w/o: without. Blank cells indicate not applicable.
Landuse
Forest (for)
High-tillage w manure (hwm)
High-tillage w/o manure (hom)
Low-tillage w manure (lwm)
Hay w nutrient management
(hym)
Hay w/o nutrient management
(hyo)
Alfalfa (alf)
Pasture (pas)
Non-regulated pervious urban
(npd)
TN
0.29
0.28
0.23
0.26
DIN
ON
TP
PO4
OP
0.13
0.16
0.17
0.11
0.1
0.08
0.02
0.2
0.03
0.1
0.1
0.004
0.16
0.1
0.09
0.08
0.01
0.37
0.73
0.07
0.3
0.62
0.05
0.07
0.11
0.02
0.22
0.07
0.15
0.08
0.08
0.004
0.09
0.09
0.003
Spatially Referenced Regression on Watershed Attributes (SPARROW) model
SPARROW is a spatially explicit, mass-balance watershed model that uses nonlinear regressions
to quantify the relationship between observed nutrient fluxes in nontidal streams and inputs
and factors that affect their overland and in-stream fate and transport. The SPARROW model
has been used to provide empirical estimates of the source, fate, and transport of nutrients
across the United States including the Chesapeake Bay Watershed.
Ator et al. (2011) simulated the fate, transport, and flux of nitrogen and phosphorus in the
Chesapeake Bay Watershed using SPARROW, and the model was calibrated to mean annual
conditions in 2002 in the Chesapeake Bay watershed. The NHDPlus segmentation and attributes
developed by Wieczorek and LaMotte (2010a) were used in this study. The NHDPlus is a
geospatial dataset that incorporates features of the National Hydrography Dataset (NHD) and
the National Elevation Dataset (NED). The NHDPlus includes stream networks based on the
1:100,000-scale NHD. Inputs and outputs from this study were compared to a P532 scenario
run for a 10-year hydrology simulation period from 1991 to 2000 using a constant
representational input data set for the year 2002.
The NHDplus atmospheric deposition attributes include inorganic nitrogen for the year 2002
compiled for every catchment of NHDPlus (Figure 1), and the atmosphere was assumed to be a
negligible source of phosphorus. Both models have similar input rates and spatial distribution
(Figure 1). The wet atmospheric input load in 2002 by SPARROW and P532 were 195 and 183
millions of pounds, respectively. SPARROW only includes wet atmospheric deposition while
P532 includes both wet and dry inputs. The dry atmospheric deposition by P532 in the year
2002 was 276 millions of pounds.
Figure 1. P532 and SPARROW atmospheric deposition inputs for the year 2002.
Ruddy et al. (2006) created a data set that represents the estimated amount of nutrient inputs
to the land surface, compiled for every catchment of NHDPlus for the conterminous United
States. Inorganic fertilizer estimates were calculated from the Association of American Plant
Food Control Officials fertilizer sales data, Census of Agriculture fertilizer expenditures, and U.S.
Census Bureau county population. P532, on the other hand, used crop uptake and Census of
Agriculture yields to determine the inorganic fertilizer application rates. Both models have
similar nitrogen input rates; however, they did not show a similar spatial distribution (Figure
2).The nitrogen fertilizer input load in 2002 by SPARROW and P532 were 424 and 624 millions
of pounds, respectively.
Figure 2. P532 and SPARROW total nitrogen fertilizer inputs for the year 2002.
Both models have similar phosphorus input rates and similar spatial distribution (Figure 3). The
phosphorus fertilizer input load in 2002 by SPARROW and P532 were 88 and 82 millions of
pounds, respectively.
Figure 3. P532 and SPARROW total phosphorus fertilizer inputs for the year 2002.
The nitrogen and phosphorus content of livestock wastes was estimated using Census of
Agriculture. The procedures include: (1) estimating animal populations, (2) calculating nitrogen
and phosphorus content of the animal manure, and (3) estimating the component of nitrogen
and phosphorus from confined and unconfined livestock. SPARROW estimates of nitrogen input
from manure do not account for loss through volatilization. These estimates represent the total
nitrogen content in manure as excreted by each livestock group.
Both models have almost the same nitrogen and phosphorus input rates and spatial distribution
(Figure 4 and 5). The nitrogen manure input load in 2002 by SPARROW and P532 were 491 and
478 millions of pounds, respectively. The phosphorus manure input load in 2002 by SPARROW
and P532 were 134.14 and 134.12 millions of pounds, respectively.
Figure 4. P532 and SPARROW total nitrogen manure inputs for the year 2002.
Figure 5. P532 and SPARROW total phosphorus manure inputs for the year 2002.
Despite differences in atmospheric and fertilizer input mass, spatial application rates and
approach to assess water quality, the models showed almost the same output at 9 USGS
stations (Figure 6 and 7).
Figure 6. P532 and SPARROW nitrogen outputs at 9 USGS stations.
Figure 7. P532 and SPARROW phosphorus outputs at 9 USGS stations.
SPARROW uses an estimated source-specific coefficient to interpret the proportion of the
applied or deposited nutrient mass that is transported to the tributaries. For example, Ator et
al. (2011) estimated that approximately 24% of the total nitrogen from fertilizer was
transported to the Chesapeake Bay tributaries in 2002 (a source-specific coefficient of 0.237).
Similarly, less than 6 % of total nitrogen from manure and total phosphorus from both fertilizer
and manure reached watershed streams. Source-specific coefficients were 0.058, 0.038, and
0.025, respectively.
These rates can be interpreted as sensitivities to inputs. An additional pound of fertilizer input
results in 0.237 lbs delivered to an NHD+ stream as an average for the entire Chesapeake basin.
The strength of SPARROW is that it is a regression approach that is based on observed data. A
caveat is that these sensitivities are average sensitivities, not instantaneous sensitivities. The
effect of adding a single pound of fertilizer to an unfertilized crop is likely different from the
effect of adding an additional pound of fertilizer to an already overfertilized crop. An ideal
sensitivity varies over the range of application or is derived from model behavior near typical
application rates.
The Decision Support System (DSS) SPARROW is an online tool that allows mapping predictions
of long-term average water quality conditions and changing nutrients source contributions. We
conducted an exercise in the Susquehanna Basin to evaluate SPARROW sensitivity to inputs.
After adding 100% more fertilizer and manure, it was found that the DSS model is very sensitive
to nitrogen fertilizer input but not very sensitive to nitrogen manure, or to phosphorus fertilizer
or manure inputs. The mean and median sensitivity slopes (ratio between output and input) are
in agreement with SPARROW’s manure and fertilizer source specific coefficients found
previously in the literature (Figure 8 and 9).
Figure 8. DSS SPPAROW 2002 total nitrogen ratio between predicted yield and input load for the Susquehanna
Basin.
Figure 9. DSS SPPAROW 2002 total phosphorus ratio between predicted yield and input load for the
Susquehanna Basin.
The Agricultural Policy / Environmental eXtender (APEX) model
USGS-NRCS (2013) evaluated the impacts of conservation adoption on cropland in the
Chesapeake Bay watershed. The APEX model was used to estimate field-level effects of the crop
management and conservation practices. Concurrently, the SWAT model (Soil and Water
Assessment Tool) was used to simulate non-point source loadings from land uses other than
cropland and to route instream loads from one watershed to another.
The APEX model was developed to extend the Environmental Policy Integrated Climate (EPIC)
model capabilities of simulating land management impacts for small-medium watersheds and
heterogeneous farms. APEX simulates day-to-day farming activities, wind and water erosion,
loss or gain of soil organic carbon, and edge-of-field losses of water, soil, and nutrients.
HSPF and SWAT usually report a good performance when simulating streamflow and nutrients.
Im et al. (2003) found that HSPF streamflow and nutrient simulations more closely approximate
observed values than those predicted by SWAT in an urbanizing watershed in northeastern
Virginia. Moriasi et al. (2007) has reviewed ranges of daily and monthly model calibration
statistics for surface runoff, streamflow, sediment, and nutrients from several publications, and
both HSPF and SWAT reported satisfactory calibration statistics. Based on HSPF and SWAT
capabilities, we could anticipate that the models would produce similar results in the
Chesapeake Bay.
Very little information has been found to evaluate APEX and SWAT in the Chesapeake Bay
region, and no spatial information or tabular detailed information has been publicly available
for review. However, information from a study published by USGS-NRCS (2013), which included
total inputs and outputs for four subregions of the Chesapeake Bay watershed, was used to
calculate the ratio between predicted yields and input loads.
Functional Linear Concurrent Model (FLCM)
FLCM employs a functional linear concurrent approach to predict loads in forested regions by
using combinations of satellite-derived data such as vegetation and disturbance indexes (Figure
10), nitrogen atmospheric deposition, precipitation and physiographic and land cover
parameters.
Figure 10. FLCM monthly Normalized Difference Vegetation Index (NDVI) satellite-derived data for eight major
basins of the Chesapeake Bay watershed.
The model was calibrated for the Chesapeake Bay region and the simulation period was from
2001 to 2009. Parameters and inputs vary by month, enabling predictions to be made on a
monthly basis. Inputs and outputs are available for every major basin of the Chesapeake Bay
watershed (Figure 11).
Figure 11. FLCM monthly predicted nitrate in forest regions for eight major basins of the Chesapeake Bay
watershed.
Functional links between nutrient inputs and outputs were also calculated using this model. The
ratio between monthly yield and monthly input load was calculated for every major basin of the
Chesapeake Bay watershed from 2001 to 2009 (Figure12). The median ratio was 0.26 and the
average was 0.33.
PROBABILITY
Figure 12. FLCM monthly total nitrogen ratio between monthly yield and input load for eight major basins of the
Chesapeake Bay watershed.
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
0.8
1
1.2
RATIO
Figure 13. Cumulative distribution function of total nitrogen ratio between monthly yield and input load.
1.4
Annual Phosphorus Loss Estimator (APLE) Model
The APLE model is a Microsoft Excel spreadsheet model that runs on an annual time step. The
model simulates sediment bound and dissolved P loss in surface runoff. It does not consider
subsurface loss of P through leaching to groundwater or artificial drainage networks. It is
intended to simulate edge-of-field P loss for uniform fields of several hectares in size, or smaller
(Vadas, 2014).
APLE was used to simulate phosphorus loss in every county of the Chesapeake Bay watershed
and in five types of agricultural landuses from 1992 to 2005 (Mulkey, in prep). The landuses
simulated were high-till with manure (hwm), high-till with manure nutrient management (nhi),
low-till with manure (lwm), low-till with manure nutrient management (nlo), and pasture (pas).
APLE model’s phosphorus sensitivity to inputs was assessed by calculating the ratio between
annual predicted yield and input loads for every county of the Chesapeake Bay watershed.
Figures 14 - 18 show APLE’s sensitivity distribution for the entire simulation period.
The implementation of the APLE model will likely take one of two forms following further study.
APLE may be used directly to estimate phosphorus losses from cropland as part of each
scenario run. Though this model closely matches recent research, its implementation would
preclude the use of multiple models for estimating phosphorus. Alternatively, multiple runs of
APLE could be used to establish relationships between inputs and yield specific to each region’s
characteristics. This prediction could then be used with predictions from other models in an
ensemble framework.
Approximately 90% of hwm and nhi landuse ratios and 98% of lwm, nlo, and pas landuse ratios
were lower than one. Hwm and nhi landuse showed higher median and average ratios than the
rest of the landuses. The median and the average ratios are listed in Table 5.
1
0.9
0.8
PROBABILITY
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.0001
0.001
0.01
0.1
1
10
RATIO
Figure 14. Cumulative distribution function of total phosphorus ratio between yield and input load for high-till
with manure (hwm) landuse.
1
0.9
0.8
0.7
PROBABILITY
0.6
0.5
0.4
0.3
0.2
0.1
0
0.0001
0.001
0.01
0.1
1
10
100
RATIO
Figure 15. Cumulative distribution function of total phosphorus ratio between yield and input load for high-till
with manure nutrient management (nhi) landuse.
1
0.9
0.8
0.7
PROBABILITY
0.6
0.5
0.4
0.3
0.2
0.1
0
0.00001
0.0001
0.001
0.01
0.1
1
10
RATIO
Figure 16. Cumulative distribution function of total phosphorus ratio between yield and input load for low-till
with manure (lwm) landuse.
1
0.9
0.8
0.7
PROBABILITY
0.6
0.5
0.4
0.3
0.2
0.1
0
0.00001
0.0001
0.001
0.01
0.1
1
10
RATIO
Figure 17. Cumulative distribution function of total phosphorus ratio between yield and input load for low-till
with manure nutrient management (nlo) landuse.
1
0.9
0.8
PROBABILITY
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.00001
0.0001
0.001
0.01
0.1
1
10
RATIO
Figure 18. Cumulative distribution function of total phosphorus ratio between yield and input load for pasture
(pas) landuse.
Table 5. APLE median and average ratio between predicted yield and input loads.
Landuse
Median Ratio
Average Ratio
high-till with manure (hwm)
0.247
0.448
high-till with manure nutrient management (nhi)
0.234
0.466
low-till with manure (lwm)
0.103
0.186
low-till with manure nutrient management (nlo)
0.097
0.191
pasture (pas)
0.081
0.155
4. Results and Recommendations
The main similarities between P532 and SPARROW in their simulations of 2002 conditions were
that both models: 1) used similar methods to calculate manure inputs, 2) showed similar
manure and phosphorus fertilizer input rates and spatial distribution, 3) had almost the same
total manure input, and 4) showed almost the same output at 9 USGS stations and from the
entire Chesapeake Bay watershed.
The main differences found between SPARROW and P532 consisted of the following: 1)
SPARROW uses only wet atmospheric deposition while P532 includes both wet and dry, 2) the
models used different methods to calculate fertilizer inputs, 3) SPARROW total nitrogen
fertilizer input was significantly lower than the P532, and 4) the models did not show similar
nitrogen fertilizer input rates or spatial distribution.
Preston and Brakebill (1999), Ator et al. (2011), Moore et al. (2011), and Preston et al. (2011)
have published studies in the Chesapeake Bay Watershed region using the SPARROW model. All
SPARROW source-specific coefficients found in the literature were compiled and used in the
sensitivity to input comparison.
The sensitivities obtained from APEX, SPARROW, APLE, and P532 models that represent the
relationship between nutrient predicted yields and input loads for cropland are shown in
Figures 19, 20, 21, and 22. We recommend using the median of the distribution of all model
coefficients to build the sensitivity to inputs in PQUAL agricultural landuses (Figures 19, 20, 21,
and 22).
We recommend using sensitivity values of 0.255 and 0.225 to calculate fertilizer and manure
nitrogen losses, respectively. We can interpret it as for every pound of nitrogen fertilizer that is
entered into PQUAL, 0.255 pounds would make it to the tributaries.
Figure 19. APEX, P532, and SPARROW models sensitivity to fertilizer nitrogen inputs.
Figure 20. APEX, P532, and SPARROW models sensitivity to manure nitrogen inputs.
We recommend using a sensitivity value of 0.1 to calculate both fertilizer and manure
phosphorus losses. We can interpret it as for every pound of phosphorus fertilizer that is put
entered into PQUAL, 0.1 pounds would reach the tributaries. Future versions of this document
will highlight a dependence on storage rather than inputs.
Figure 21. APEX, APLE, and P532, and SPARROW models sensitivity to fertilizer phosphorus inputs.
Figure 22. APEX, APLE, P532, and SPARROW models sensitivity to manure phosphorus inputs.
In forested regions, the P532 ratio between yield and input load was 0.29, whereas the FLCM
median ratio was 0.26 and the average was 0.33. We recommend using a sensitivity value of
0.29 to calculate total nitrogen losses in forested regions. It may be interpreted as for every
pound of atmospheric deposition that is entered into PQUAL, 0.29 pounds would make it to the
tributaries.
In grasslands, the P532 ratio between yield and input load was 0.07, whereas the APLE median
ratio was 0.081 and the average was 0.155. We recommend using a sensitivity value of 0.081 to
calculate total nitrogen losses in grasslands. We can interpret it as for every pound of
atmospheric deposition that is entered into PQUAL, 0.08 pounds would reach the tributaries.
5. References
a) Ator, S., J. Brakebill, and J. Blomquist. 2011. Sources, Fate, and Transport of Nitrogen and
Phosphorus in the Chesapeake Bay Watershed: An Empirical Model. USGS. Available online at
http://pubs.usgs.gov/sir/2011/5167/
b) Friedrichs, M., K.G. Sellner, and M.A. Johnston. 2012. USING MULTIPLE MODELS FOR
MANAGEMENT IN THE CHESAPEAKE BAY: A SHALLOW WATER PILOT PROJECT. Chesapeake Bay
Program Scientific and Technical Advisory Committee Report. No. 12-003, Edgewater, MD. 2012
09.
c) Im, S., K. Brannan, S. Mostaghimi, and J. Cho. 2003. A Comparison of SWAT and HSPF
Models for Simulating Hydrologic and Water Quality Responses from an Urbanizing Watershed.
ASAE. Paper Number: 032175.
d) Moore, R., C. Johnston, R. Smith, and B. Milstead. 2011. Source and Delivery of Nutrients to
Receiving Waters in the Northeastern and Mid-Atlantic Regions of the United States. JAWRA.
47(5):965-990.
e) Moriasi, D., J. Arnold, M. Van Liew, R. Bingner, R. Harmel, and T. Veith. 2007. Model
Evaluation Guidelines for Systematic Quantification of Accuracy in Watershed Simulations.
ASABE. Vol. 50(3): 885−900.
f) Preston, S., R. Alexander, G. Schwarz, and C. Crawford. 2011. Factors Affecting Stream
Nutrient Loads: A Synthesis Of Regional Sparrow Model Results For The Continental United
States. JAWRA. 47(5):891-915.
g) Preston, S. and J. Brakebill. 1999. Application of Spatially Referenced Regression Modeling
for the Evaluation of Total Nitrogen Loading in the Chesapeake Bay Watershed. U.S. Geological
Survey. Available online at http://md.water.usgs.gov/publications/wrir-994054/pdf/SPARROW.pdf
h) Ruddy, B., D. Lorenz, and D. Mueller. 2006. County-Level Estimates of Nutrient Inputs to the
Land Surface of the Conterminous United States, 1982–2001. USGS. Available online at
http://pubs.usgs.gov/sir/2006/5012/
i) Smith, R., G. Schwarz, and R. Alexander, 1997, Regional interpretation of water-quality
monitoring data, Water Resources Research, v. 33, no. 12, pp. 2781-2798. Available online at
http://water.usgs.gov/nawqa/sparrow/intro/intro.html
j) Schwarz, G., A. Hoos, R. Alexander, and R. Smith. 2006. The SPARROW Surface WaterQuality Model—Theory, Applications and User Documentation: U.S. Geological Survey,
Techniques and Methods. Available online at http://water.usgs.gov/nawqa/sparrow/sparrowmod.html
k) USDA-NRCS. 2013. Impacts of Conservation Adoption on Cultivated Acres of Cropland in the
Chesapeake Bay Region, 2003-06 to 2011. 113p. Available online at
http://www.nrcs.usda.gov/wps/portal/nrcs/detail/national/technical/nra/ceap/?cid=stelprdb1
240074
l) Vadas, P. 2014. Annual Phosphorus Loss Estimator - Theoretical Documentation (Version
2.4). Available online at http://ars.usda.gov/Services/docs.htm?docid=21763
m) Wieczorek, M., and A. LaMotte. 2010a. Attributes for NHDPlus Catchments (version 1.1) for
the Conterminous United States: Average Atmospheric (Wet) Deposition of Inorganic Nitrogen,
2002. USGS. Available online at
http://water.usgs.gov/GIS/metadata/usgswrd/XML/nhd_atdep.xml
n) Wieczorek, M., and A. LaMotte. 2010b. Attributes for NHDPlus Catchments (version 1.1) for
the Conterminous Unites States: Nutrient Inputs from Fertilizer and Manure, Nitrogen and
Phosphorus (N&P), 2002. USGS. Available online at
http://water.usgs.gov/GIS/metadata/usgswrd/XML/nhd_nutrients.xml
o) Wieczorek, M., and A. LaMotte. 2010c. Attributes for NHDPlus Catchments (version 1.1) for
the Conterminous United States: Nutrient Application (phosphorus and nitrogen) for Fertilizer
and Manure Applied to Crops, 2002. USGS. Available online at
http://water.usgs.gov/GIS/metadata/usgswrd/XML/nhd_cropsplit02.xml