Water Resour Manage
DOI 10.1007/s11269-009-9517-z
Spatially Explicit Modeling of Land Use Specific
Phosphorus Transport Pathways to Improve TMDL
Load Estimates and Implementation Planning
Erica J. B. Gaddis · Alexey Voinov
Received: 23 June 2007 / Accepted: 6 October 2009
© Springer Science+Business Media B.V. 2009
Abstract Diffuse pollution from urban stormwater and agricultural runoff are
among the leading causes of water pollution in the USA. A process-oriented,
stakeholder-driven research approach was implemented in the small heterogeneous
watershed of St. Albans Bay, Vermont to model the relative load of phosphorus
from all sources, including diffuse transport pathways, and compared to goals and
assumptions outlined by a Total Maximum Daily Load (TMDL) developed for
phosphorus in Lake Champlain. Mass-balance and dynamic landscape simulation
models were used to describe the distribution of the average annual phosphorus
load to streams (10.57 t/year) in terms of space, time, and transport process. The
E. J. B. Gaddis · A. Voinov
Gund Institute for Ecological Economics, University of Vermont,
601 Main Street, Burlington, VT 05405, USA
E. J. B. Gaddis
Rubenstein School of Environment and Natural Resources, University of Vermont,
George D. Aiken Center, 81 Carrigan Drive, Burlington, VT 05405, USA
A. Voinov
Department of Computer Science, University of Vermont,
351 Votey Building, Burlington, VT 05405, USA
Present Address:
E. J. B. Gaddis (B)
SWCA Environmental Consultants, 257 East 200 South, Suite 200,
Salt Lake City, UT 84111, USA
e-mail: egaddis@swca.com
Present Address:
A. Voinov
International Institute for Geo-information Science and Earth Observation (ITC) P.O. Box 6,
7500 AA Enschede, The Netherlands
URL: http://www.itc.nl

E.J.B. Gaddis, A. Voinov
majority of the phosphorus load comes from two subwatersheds dominated by clay
soils, Stevens and Jewett Brooks. Dissolved phosphorus in surface runoff from the
agricultural landscape, driven by high soil phosphorus concentrations, accounts for
41% of the total load to watershed streams. Direct discharge from farmsteads and
stormwater loads, primarily from road sand wash-off, are also significant sources.
Results reported in this study could help target watershed interventions both in terms
of the types and locations of recommended best management practices (BMPs).
The study offers an approach to attaining TMDL diffuse pollution targets in a
cost-effective and participatory manner and could be replicated for other TMDL
processes around the country.
Keywords Diffuse pollution management · Non-point source pollution ·
Spatially explicit watershed model · Landscape model · TMDL ·
Phosphorus · Vermont
1 Introduction
Nutrient pollution accounted for approximately 10% of the water quality impair-
ments identified by the EPA in the USA in 2009. Diffuse (nonpoint source) pollution
is the remaining large untreated source of nutrient pollution in many parts of the
country. Diffuse pollution is typically not regulated, leaving governmental bodies to
implement change through education and incentives which are only sometimes suc-
cessful at achieving targeted reductions identified through the total maximum daily
load (TMDL) process. This is in part because the identification and quantification
of diffuse sources and pollutant transport mechanisms have proven difficult at the
watershed scale. Furthermore, although maximum pollutant loads can be calculated
for a waterbody, there is often considerable uncertainty in quantifying the relative
roles of specific phosphorus transport pathways from the landscape. Overcoming this
challenge would help watershed managers to better target actions that will lead to
cost-effective attainment of TMDL load allocations.
Lake Champlain, like many fresh water lakes, has received excess nutrient runoff
for the past 50 years (VTANR and NYDEC 2002) due to changes in agricultural
practices and rapid development of open space for residential uses (Hyde et al.
1994). The effect of excess nutrients on the health of Lake Champlain has been
most dramatically witnessed in bays such as St. Albans Bay (Fig. 1), which exhibit
eutrophic algal blooms every August (Hyde et al. 1994). The Vermont Agency of
Natural Resources and the New York Department of Environmental Conservation
(DEC) completed a TMDL for phosphorus in Lake Champlain that was approved by
the USEPA in 2002 (VTANR and NYDEC 2002). Under this TMDL the St. Albans
Bay watershed was allocated a total annual nonpoint source (diffuse) phosphorus
load of 4.2 metric tons per year (t/year), requiring an estimated reduction of 33%
from diffuse sources. Past efforts to reduce phosphorus loss from the watershed,
beginning in the early 1980s, have not resulted in significant reductions of loading to
the bay, nor have they yet been successful in reducing nuisance algal blooms during
summer months (Meals 1996). This is in part due to the relatively low mixing rate of
water in St. Albans Bay with the rest of the lake, but it also suggests that important
phosphorus sources or transport pathways may have been overlooked in the past.

Spatially Explicit Modeling of Land Use Specific Phosphorus Transport Pathways
Fig. 1 Lake Champlain Basin and detailed map of St. Albans Bay watershed hydrology
By 1991, St. Albans Bay watershed still had one of the highest phosphorus export
rates in the Lake Champlain Basin (VTANR and NYDEC 2002). Quantification of
the sources and pathways of phosphorus movement across the landscape would be
useful in prioritizing where and how phosphorus reduction efforts in the basin should
be focused in the future.
This paper represents one piece of a comprehensive study of the St. Albans Bay
watershed in Northern Vermont. The overall study was process-oriented in that it
involved a holistic accounting of key watershed processes, identified by stakeholders,
at appropriate temporal and spatial scales, requiring integration of mechanistic
models and watershed monitoring. Monitoring of water quality in streams in the
watershed and hydrologic modeling were used together to calculate the mean
annual total phosphorus loads to St. Albans Bay from each of five subwatersheds
(Gaddis 2007). Several modeling tools were used to apportion the total load for each
subwatershed to specific phosphorus sources and transport pathways. Identifying the
relative importance of each phosphorus source and its transport pathway to streams
in the watershed is the primary focus of this paper. The results of other components
of the study are reported elsewhere and include water quality monitoring (Gaddis
2007), participatory modeling with stakeholders in the watershed (Gaddis et al.
2009), estimation of total phosphorus load to St. Albans Bay during different
hydrologic flow regimes (Gaddis 2007), and optimization of phosphorus reduction
strategies across the watershed (Gaddis 2007). In this paper, we report on the relative
importance of specific phosphorus sources and transport pathways in the St. Albans
Bay watershed. These include diffuse sources separated into specific landscape runoff
processes (erosion, surface runoff, subsurface runoff, and road sand wash-off), point
sources, and semi-diffuse sources such as the impact of waterfowl and farmstead

E.J.B. Gaddis, A. Voinov
runoff. In addition, we discuss how the results could be used to identify more cost-
effective strategies to reduce phosphorus load to the bay in the future.
2 Watershed Description
The 130 km2 watershed feeding St. Albans Bay is dominated by agriculture (57%
of the watershed area), at the same time that the urban and suburban areas (14%
of the watershed area) are growing. The watershed is currently home to over 20,000
people primarily in the City of St. Albans and the Town of St. Albans. The upper
reaches of the watershed are steep and dominated by deciduous forests while the
lower portions are flat and dominated by agriculture. The City of St. Albans is located
in the middle of the watershed and is surrounded by suburban development. Other
developed areas of the watershed are concentrated along Highway 7, which runs
North–South through the middle of the watershed. Between 1850 and 1990 there was
an overall shift in the watershed, and throughout Vermont, from small integrated
farms to concentrated feeding operations with larger herd sizes. This has led to
increased feed and fertilizer imports to the state, resulting in an excess of phosphorus
in dairy farming areas (Hyde et al. 1994; Cassell et al. 2002; Magdoff et al. 1997).
The St. Albans Bay watershed is drained by five streams: Jewett Brook, Stevens
Brook, Rugg Brook, Guayland Brook, and Mill River. The area draining to Jewett
Brook, the northern most stream in the watershed, is dominated almost entirely by
agriculture with very little forest and/or suburban area. The headwaters for Rugg
and Stevens Brooks originate in the steep forested eastern portion of the watershed.
These streams run through suburban areas and the City of St. Albans where they
receive stormwater runoff before flowing through predominantly agricultural areas
in the low elevation, flat portion of the watershed. Guayland Brook originates in the
suburban area downstream of the City of St. Albans and runs through low density
residential and agricultural areas. Mill River, the largest and southern most drainage
in the watershed, drains agricultural, forest, and low-density residential lands. Mill
River joins with Rugg Brook before flowing into St. Albans Bay (Figs. 1 and 2).
The soils in the watershed are primarily loam (varying from silty to stony loams)
however a large portion of the lower parts of Jewett and Rugg Brook subwatersheds
have clay soils (Fig. 3). These very flat portions of the watershed were once char-
acterized by wetlands and are naturally poorly drained. Agriculture has only been
made possible in this part of the watershed through tiled field drains. Surface runoff
is substantially reduced in areas that have functioning subsurface drains because soil
saturation is rare.
3 Methodology
A distributed landscape model was used to formalize concepts of watershed
processes and as such explore the mechanisms and underlying driving forces of
phosphorus movement and transport in the landscape. Simpler spreadsheet-based
models were used to estimate point source and aggregate diffuse phosphorus loads in
the watershed. Together, the landscape model and spreadsheet models were used to

Spatially Explicit Modeling of Land Use Specific Phosphorus Transport Pathways
Fig. 2 Map of St. Albans Bay
watershed land use
0 1 2 3 4 5 60.5
Kilometers
Landuse
Water
Residential
Commercial
Industrial
Transportation
Other Urban
Forest
Wetland
Row Crop
Pasture
St. Albans Bay
Stevens
BrookGuayland
Brook
Rugg Brook
Mill River
Jewett
Brook
0.50 1 2 3 4 5 6
Kilometers
0.50 1 2 3 4 5 6
Kilometers
Soil Phosphorus (mg/kg)
0 - 3
4 - 6
7 - 10
11 - 15
16 - 20
21 - 25
26 - 30
31 - 35
35+
St. Albans Bay
Stevens
BrookGuayland
Brook
Rugg Brook
Mill River
Jewett
Brook
Soil Texture
Hydric
Clay
Loam
Silty loam
Fine sandy loam
Sandy loam
Slaty loam
Stony loam
Loamy fine sand
Loamy sand
Stony
Rock Outcrop
Muck
St. Albans Bay
Stevens
BrookGuayland
Brook
Rugg Brook
Mill River
Jewett
Brook
Fig. 3 Map of soil texture and soil phosphorus concentration in the St. Albans Bay watershed

E.J.B. Gaddis, A. Voinov
evaluate the relative importance of all sources and transport pathways of phosphorus
in the St. Albans Bay watershed.
3.1 Spatially Explicit Landscape Model to Simulate Diffuse Phosphorus
Pathways and Sources
A landscape model was developed to simulate the dynamics of phosphorus transport
from diffuse sources in the St. Albans Bay watershed. This model was developed
as six interconnected modules (Evapotranspiration, Snow hydrology, Hydrology,
Plant dynamics, Sediment, and Phosphorus) using Stella systems modeling software
(iseesystems.org). A seventh module was used as a central module for input data
(time-series and spatial data), conversion rates, and other universal model parame-
ters. The distributed landscape model is partitioned into a spatial grid of square
unit cells (30 × 30 m in this application) in which each of the process modules is
implemented. Stella modules are linked with time series inputs, spatial inputs, and
parameter table files into a complete landscape simulation model using the Spatial
Modeling Environment (SME) software (Maxwell and Costanza 1997a, b; Maxwell
1999; Costanza and Voinov 2004). SME automatically converts Stella generated
modules into a C++ driver allowing the user to run the modules as one complete
process model in each cell of the landscape at a given time step (one day for the St.
Albans Bay watershed model application). Horizontal fluxes of water and nutrients
are accounted for with cell–cell head differences of surface water and groundwater
(Voinov et al. 1999a).
The linkage of ecological and hydrologic modules using SME is described else-
where (Costanza and Voinov 2004) as the Landscape Modeling Framework (LMF).
Feedbacks among the biological, chemical and physical model components are
important structural attributes of this framework (Maxwell and Costanza 1995;
Maxwell 1999; Voinov et al. 2004). The integration of physical, biological, and social
factors into a dynamic model provides a platform for decision makers to explore the
dynamics of a watershed system (Madani and Mariño 2009). A simulation run using
SME gives a visual representation of the landscape as it evolves over time reflecting
changes in hydrology, water quality, and material flows between adjacent cells. The
approach differs from other watershed modeling suites in that the user is not limited
to pre-designed modules allowing users to alter existing modules and add additional
modules with the user-friendly Stella software.
Templates for the hydrology, plant dynamics, and evapotranspiration modules
came from the Library of Hydro-Ecological Modules, originally developed for the
Patuxent River watershed in Maryland (Voinov et al. 1999a, b). These modules were
modified for use in the St. Albans Bay watershed, Vermont, which is smaller, colder,
and more heterogeneous than the Patuxent River watershed. New modules, devel-
oped specifically for the St. Albans Bay watershed were erosion/sediment transport,
phosphorus dynamics, and snow hydrology modules. The full set of equations used
in the modules is available on-line at http://www.likbez.com/LHEM.
3.1.1 Hydrology Modules (Hydrology, Snow, and Evapotranspiration)
The hydrology module drives the erosion and phosphorus transport modules since
all movement of sediment and phosphorus is assumed to be transported in water.

Spatially Explicit Modeling of Land Use Specific Phosphorus Transport Pathways
Water is tracked, in units of meters/day (assumed to be homogenous across a
cell), through pooled surface water, flowing water in channels, and two layers of
unsaturated water in soil. Water moves within a cell through the processes of runoff,
infiltration, percolation, surface and soil evaporation, plant transpiration, and tile
drainage. Besides precipitation and snow melt, water also enters some cells through
groundwater recharge in tile drained areas or in perennial streams. Water is lost from
the cell to groundwater, the dynamics of which were not modeled.
Daily precipitation and snow melt drive the hydrologic model. The former was
obtained from the St. Albans NCDC climate station (COOP ID 437032) and the
latter is an output from the snow hydrology module described below. Surface runoff
is estimated using the curve number method developed by the Natural Resources
Conservation Service (USDA 1997) and incorporates simulated Antecedent Mois-
ture Condition (AMC). Interception is calculated using the leaf area index, an
output from the plant dynamics module. Soil specific infiltration rates, percolation
rates, and porosity were obtained from the STATSGO database available for
the watershed (soils.usda.gov/survey/geography/statsgo). Land use specific percent
impervious cover was obtained from a technical report, specific to Vermont, which
was completed by the Center for Watershed Protection in 1999. Soil evaporation and
plant transpiration were calculated using equations from Neitsch (2000).
We assumed that a field was tile drained if it was characterized as poorly drained
(hydrologic soil group D) on the US General Soil Map (soils.usda.gov/survey/
geography/statsgo), has a slope less than 3%, and is of agricultural land use. These
assumptions were confirmed reasonable by agricultural stakeholders in the water-
shed. Twelve percent of the entire St. Albans Bay watershed (22% of the agricultural
area in the watershed) is estimated to be drained. In areas where tile drains are
installed on agricultural fields, water in the unsaturated zone, down to the level of the
soil’s field capacity, is assumed to drain through a tile drain to a channel or ditch at
each time step. Tile drains are generally placed at a depth of 0.5 m, which determined
the default model depth of the middle layer of the unsaturated zone.
Water movement between cells is accounted for with modified versions of algo-
rithms already developed and described in the literature (Voinov et al. 1999a, b).
The only modification made to the algorithms is the separation of the surface water
and channel water stocks in the hydrology module such that water is allowed to
pond in the cell as surface water for several days while some water, calculated as
a function of curve number, is transported to channels and is available for transport
downstream. In Stevens Brook, stormwater and wastewater discharge were diverted
to outfall locations within the watershed.
A separate evapotranspiration module calculates potential evapotranspiration,
using the Penman–Monteith equation (Neitsch 2000; Shuttleworth 1993; Voinov
et al. 2004). Leaf surface resistance and aerodynamic resistance are calculated
using equations from Shuttleworth (1993). Stomatal leaf conductance, leaf area
index, were obtained from Breuer and Frede (2003). The major climatic input for
evapotranspiration is daily average wind speed, which was obtained from the St.
Albans NCDC climate station.
The snow module simulates snow accumulation, sublimation, and melt in units
of meters/day of snow water equivalent (SWE). Snow fall data was obtained from
the St. Albans climate station. Snow interception by plants is calculated using leaf
area index, an output from the plant dynamics module. Sublimation is modeled as

E.J.B. Gaddis, A. Voinov
a function of wind speed, air temperature, humidity, slope and aspect. Snow melt is
modeled separately for rain on snow and dry periods. The latter is a function of air
temperature and snow density, aspect, and slope (Gray and Prowse 1993; Male and
Gray 1981). Process-based distributed snow melt models have been found to improve
stream flow predictions at the catchment level when compared to temperature-index
models (Zeinivand and De Smedt 2009).
3.1.2 Sediment Module
The sediment module simulates overland surface erosion from a cell to channels
(intermittent rills or perennial streams). The basis of the sediment module was
derived from RUSLE 1.06 (Rendard et al. 1997) which was modified to a daily time
scale to calculate potential soil loss, in units of kilogram per day per cell. Climatic
input data for the sediment module include maximum storm intensity (mm/h) and
daily precipitation, both obtained from the St. Albans NCDC climate station. Soil
erosivity factors came from the SSURGO soil database compiled by the NRCS
(soildatamart.nrcs.usda.gov). The slope factor is based on percent slope calculated
using GIS for each cell (Rendard et al. 1997; McCool et al. 1987). Coefficients
for cover crop and land management practices were obtained from NRCS staff in
Colchester and St. Albans, Vermont and assumed to be representative of agricultural
practices in the watershed. Potential soil loss is modified by overland sediment
transport capacity, which is a function of surface runoff (Foster 1982). In this way tile
drainage and other runoff mediators are accounted for in calculating actual estimated
erosion. The minimum of potential soil loss and overland sediment transport capacity
is used as the final overland erosion value.
Road sand applied during winter storms washes off during spring melts and rain
storms and contributes a unique source of phosphorus to streams in the area. The
City of St. Albans reported that the Public Works Department applies 600 tons of
sand per year over the course of an average of 20 snow storms per year. This value
was divided by the area of road in the city to derive an application rate of 0.058 kg
sand per square meter applied per storm. This estimate is used as a constant in the
sediment module, trigged by actual snow events. In this module, a snow event is
considered large enough to receive sand if snow fall is greater than 0.025 m/day. Sand
wash off is assumed to occur when a rain storm is greater than 30 mm, otherwise,
a proportion of the sand is washed off depending on the severity of the storm or
melt event.
3.1.3 Phosphorus Module
The phosphorus module captures four distinct diffuse phosphorus transport
processes: surface erosion, tile drainage, dissolved phosphorus released from soil,
and road sand wash-off. Some of these processes are specific to particular land use
types such as tile drainage (agriculture) and road sand wash-off (roads). The phos-
phorus module incorporates processes that are modeled mechanistically and some
that could only be determined empirically. Dissolved and particulate phosphorus are
tracked separately in the model. No attempt is made to model in-stream phosphorus
processes, though we recognize that this is an important process in the movement of
phosphorus through stream channels and final load estimates to St. Albans Bay. All

Spatially Explicit Modeling of Land Use Specific Phosphorus Transport Pathways
processes are modeled as phosphorus load to streams (g/cell) rather than to the bay.
The initial condition for soil phosphorus was input as a map of soil phosphorus
concentrations specific to land use-soil type combinations throughout the watershed
(Fig. 3). The map was created based on soil phosphorus tests on agricultural and
urban soils in the watershed. Agricultural soil phosphorus data was obtained from
the NRCS with the permission of landowners in the watershed. The fields sampled
represent 5.2% of the agricultural fields in the watershed, with a range of soil, slope,
and crop characteristics. In addition, we collected soil phosphorus samples from over
70 residential, commercial, and park lawns in the city, representing 5.8% of the
properties in the city.
The inputs to soil phosphorus are atmospheric deposition, manure application,
and fertilizer. The average daily atmospheric deposition of phosphorus was calcu-
lated based on the annual AirMON data available from the National Atmospheric
Deposition Program (nadp.sws.uiuc.edu/AIRMoN), which measures atmospheric
deposition in Underhill, Vermont. The total annual phosphorus load in 2003 was
0.021 kg/ha/year. Agricultural manure and fertilizer application assumptions were
obtained from local farmers and confirmed as typical application rates for the area
by the local NRCS office. Application rates of fertilizer on the urban landscape were
estimated using survey results collected of homeowners in the City of St. Albans
(Homziak n.d.). Weatherization of rock is assumed to be a sufficiently slow process as
to not be captured in this daily time step model. Phosphorus leaves the soil through
plant uptake (Wood et al. 1984), dissolved phosphorus in surface runoff (Sharpley
et al. 2002), surface erosion, and tile drainage.
The total dissolved phosphorus (g/day) that leaves the soil in runoff is calcu-
lated by multiplying the surface water runoff, from the hydrology module, by dis-
solved phosphorus concentration in runoff (g/m3) determined by the concentration
of phosphorus in the soil. Soil phosphorus concentrations required to reduce dis-
solved phosphorus in surface runoff are often significantly less than concentrations
maintained for optimal crop growth (Heathwaite et al. 2000) due to the dynamics
of soil availability dictated by adsorption and desorption processes (McDowell and
Sharpley 2001).
The regression equations linking soil phosphorus to dissolved phosphorus concen-
tration come from Sharpley et al. (2002) and are different for corn and pasture land
use types. It was assumed that urban lawns behave in a manner similar to pasture.
Since only the water running off of the soil can pick up soil derived phosphorus, this
value is then multiplied by the proportion of the cell that is not impervious.
The phosphorus concentration in tile drainage was assumed to be a constant
concentration based on monitoring data collected for soil type and land use specific
combinations. We sampled eight tile drains in the watershed on different soil type-
land use combinations during each season, and validated our findings with tile drain
concentrations reported for drainage waters in Quebec in an unpublished manuscript
by Enright and Madramootoo in 2004. The phosphorus load from tile drainage was
calculated by multiplying the soil type-land use specific phosphorus concentration in
tile drainage by the modeled volume of water drained from each cell, as calculated
in the hydrology module, during each daily timestep. This method was selected
because a mechanistic model could not be validated due to the lack of a strong
relationship between soil phosphorus concentration and tile drainage concentrations
also reported in Quebec by Enright and Madramootoo in 2004.

E.J.B. Gaddis, A. Voinov
Particulate phosphorus removed from soil as erosion is calculated by multiplying
the kilograms of soil eroded, calculated in the sediment module, by the concentration
of phosphorus in the soil, a parameter tracked in the phosphorus module.
The volume of road sand that washes off of roads and other impervious surfaces,
an output of the sediment module, is multiplied by the concentration of phosphorus
in the sand which was determined empirically through samples we collected of the
road sand applied by the City of St. Albans (0.78 g/kg).
3.1.4 Execution of Complete Landscape Model
The landscape model was run separately, using SME, for each of the five subwa-
tersheds: Stevens Brook, Rugg Brook, Jewett Brook, Mill River, and Guayland
Brook. The model was driven with climatic data from 1984 to 1989. These years were
selected because they represent a continuous climatic data set for 5 years with both
wet and dry years and a variety of climatic patterns.
The model was configured to output results as both time series and spatial maps.
Total discharge values at stream monitoring locations in the watershed were output
as time series data. Maps were output as raster files at a 30 × 30 m resolution. All
map outputs represented cumulative loads over the entire model simulation and were
used to calculate total phosphorus movement within the watershed for each land use
type. Phosphorus movement was calculated separately for the four transport path-
ways of phosphorus modeled in the landscape: surface erosion, dissolved phosphorus
in surface runoff, tile drainage, and road sand wash-off.
Total phosphorus load to the streams was estimated using a post-processing
method (Tim et al. 1992) to account for phosphorus attenuation in the landscape.
The closer a cell was to a stream the more phosphorus from that cell is delivered to
the stream.
3.2 Point-Source Load to Streams
Wastewater treatment plants are the only regulated point sources of phosphorus
in the watershed (VTANR and NYDEC 2002). Although the maximum annual
phosphorus load from these sources had been previously estimated and utilized
in the Lake Champlain Phosphorus TMDL, a more detailed analysis of this load
was necessary to understand the dynamics of combined sewer overflow in the
watershed. Phosphorus discharged from the City of St. Albans wastewater treatment
plant was calculated using a combination of empirical data and model output from
the hydrology and phosphorus modules of the landscape model. Phosphorus in
wastewater discharge was estimated on a daily timestep, and summarized over 5-
years to calculate an annual average. Separate phosphorus loads were calculated for
treated sewage and stormwater overflow.
3.3 Phosphorus Load to Streams from Dairy Farmsteads
Storm runoff from farmsteads where animals are concentrated for feeding and/or
milking are considered semi-diffuse sources rather than diffuse sources and were cal-
culated in aggregate for the watershed with help from the NRCS office in St. Albans.
There are an estimated 20 to 30 farmsteads in the St. Albans Bay watershed. The
specific location of farmsteads and data about their operations is either unavailable

Spatially Explicit Modeling of Land Use Specific Phosphorus Transport Pathways
or proprietary. Farmstead phosphorus discharge was estimated separately for milk
house effluent, manure runoff, and silage leachate. Unless otherwise noted, all of
the assumptions in these calculations were derived in consultation with the Franklin
County NRCS office in St. Albans.
Phosphorus load from milk house effluent was calculated by multiplying the
volume of milk house effluent per day for each farmstead (1,135 to 4,542 l/year), by
the average percent in the watershed that is not treated (10%) and the concentration
of phosphorus in milk house effluent as described in Krider (1999).
The total phosphorus load originating from manure (t/year) that runs off from
barns and barnyards is calculated as the product of the volume of manure deposited
on barnyards untreated (1,195 m3/year), the average concentration of P2O5 in ma-
nure (1,321 g/m3 manure), and the percentage of manure deposited on the barnyard
that runs off to streams (3%). The volume of manure deposited on barnyards was
calculated as the product of the number of animals in barns or barnyards (6,399
animal units), the percentage of time animals spend in the barn (80%) or barnyard
(20%), the percentage of manure treated from the yards (70%), and the quantity of
manure produced by each animal unit (43.8 l/AU/day). The percent of manure in the
barnyard that actually runs off to streams is calculated as an average erosion rate for
barnyards in the area, using the sediment module.
Phosphorus contained in silage leachate is calculated as the product of the number
of silos in the watershed (10), the quantity of silage in each silo (4,535 t/silo), the
concentration of phosphorus in silage (0.08 kg/t silage; NRAES 1993), and the
amount of silage leachate that is transported to streams (40%).
3.4 Waterfowl
Although waterfowl do not provide a new source of phosphorus to the watershed,
their presence in large numbers during migration periods can mobilize phosphorus
from the land (crops and vegetation) to the water. In the case of the St. Albans Bay
watershed, waterfowl feed primarily on wetland plants, young crops, and waste grain
on fields. Estimates of the number of waterfowl, primarily geese, present in the St.
Albans watershed and the length of residence during each season were provided by
William Crenshaw, a wildlife biologist with the State of Vermont Fish and Wildlife
Department. Multiplying the total number of geese by the days of residence in the
watershed gives the number of geese days per season. Geese days estimated for the
St. Albans Bay watershed ranged from 4,500 to 9,000 during the summer season
(90 days), 2,000–5,000 during the spring migration period (14 days), and 5,000–10,000
during the fall migration period (14 days). Geese days were multiplied by average
phosphorus excretion per goose, 0.45 g/day (Post et al. 1998).
4 Model Calibration and Uncertainty
Model uncertainty is usually a major concern when results are delivered to model
users such as stakeholders and decision-makers. We have found that by opening
up the modeling process itself and by engaging stakeholders in model development
and analysis, the issue of uncertainty is reduced, since the group is well aware of
all the assumptions that went into the model construction and know that the best

E.J.B. Gaddis, A. Voinov
knowledge and data have been incorporated. The process creates trust in the model
and its output (Gaddis et al. 2009) and very much facilitates discussions about model
validity, limitations, and uncertainty.
The Landscape Modeling Framework provides a transparent and flexible platform
for model building, application, analysis and interpretation. Because the model
was developed in consultation with a group of stakeholders, this transparency was
critical to gaining the confidence of stakeholders in the model results as well as in
clearly explaining the model assumptions and limitations. The ability to modify the
model to incorporate newly identified processes and sources of phosphorus in the
watershed (e.g. adding the role of waterfowl) partially compensated for the lack of
data available for extensive model calibration and validation.
4.1 Hydrology
Initial calibration efforts focused on calibration of the hydrologic modules with
discharge data collected during the 1980s by the Rural Clean Water Program for
which discharge and water quality data are available at intervals of several days
over 8 years. The years 1984 through 1989 were selected for model calibration using
observed climatic data from that time period and a land use and curve number
map from 1992. Calibration parameters associated with groundwater recharge, spring
runoff, tile drainage, vegetation interception, maximum travel distance down stream
(per day), were used to calibrate the model to annual discharge data. However, daily
calibration of the hydrologic model was challenging because of suspected errors
in either the measured discharge data and/or the climatic data driving the model.
None of the subwatershed outputs correlated well with discharge data (R2 < 0.23).
Separating the uncertainty resulting from model error versus that associated with
data error was difficult and would require gauging the subwatersheds with more
exact monitoring equipment (Hession and Storm 2000). Examination of patterns of
discharge in the measured data and model output as well as precipitation, indicates
several conflicts between the discharge data and the precipitation data. In some
cases discharge was an order of magnitude greater than precipitation, according to
the measured data and occurred during periods when snow melt was not a factor.
However, in other cases records of significant precipitation did not correspond to a
runoff response in the discharge data. The St. Albans NCDC climate station sits at
the highest point in the watershed and may not reflect climatic conditions throughout.
Some improvements in hydrologic calibration were made when climatic data were
interpolated between the St. Albans station and the South Hero climatic station,
which is more representative of conditions along the bay. Additional discharge data
was collected in 2004 and 2005 as part of this project for the purpose of model
calibration. Unfortunately, calibration of the hydrologic model with this data was
not possible because NCDC stopped data collection at this site during this period.
The hydrology module was applied to small watersheds in Maryland, during a
previous study, where climatic and discharge data are more reliable with calibration
resulting in a correlation coefficient (R2) of 0.62 (Gaddis et al. 2006). The snow melt
module was the only new component to the hydrology module for the St. Albans
Watershed application. This module was calibrated for the pixel cell in which the
snow weather station resides in St. Albans to a high level of certainty (R2 = 0.82;
Fig. 4).

E.J.B. Gaddis, A. Voinov
temporal timing of phosphorus sources associated with watershed processes. In this
way the model is a decision support tool used to compare the relative importance of
phosphorus sources and transport pathways, similar to the Spatial Decision Support
System (SDSS) developed by Rao and Kumar (2004). The St. Albans Bay Watershed
model differs from the SDSS in that it runs on a daily time step and provides the user
with information about temporal patterns in the watershed. Stakeholders gain a good
understanding of the system and the processes involved which helps to avoid the
‘uncertainty deadlock’ that often results from black box models. Daily uncertainty
in model runs was deemed acceptable and discussed with the stakeholder working
group, who plan to use the model results to identify cost-effective implementation
projects.
4.2 Phosphorus and Sediment Loads
The watershed model was used to output total phosphorus and sediment movement
within each subwatershed as well as total load delivered to streams. The current total
phosphorus load to streams in the St. Albans Bay watershed was estimated in a
separate part of our research to be 10.6 t/year. The methodology and a discussion
of this estimate are available in Gaddis (2007). This value is taken as a starting point
for the findings presented in this paper, which focuses on the relative importance
of phosphorus sources and transport pathways in the watershed. These values were
then converted to area weighted loading rates and compared to other studies in the
Lake Champlain basin (Tables 2 and 3).
The total phosphorus load to St. Albans Bay calculated in the Lake Champlain
Phosphorus TMDL (7.7 t/year) is slightly higher (11%) than the estimated total
phosphorus load to the bay in this study (6.9 t/year; Gaddis 2007). The difference in
estimated loads represents model uncertainty in both studies. Our estimate of total
phosphorus load to streams in the St. Albans Bay watershed model is 10.6 t/year,
which is substantially higher than the load to the bay. The difference between
these two loads represents in-stream processing which is not accounted for in the
model. The area weighted loading rate of phosphorus to streams in the St. Albans
Bay watershed is slightly lower than loading rates modeled for Little Otter Creek
watershed, another predominantly agricultural watershed in the southern Lake
Champlain Basin (Meals et al. 2008). The area weighted loading rates modeled for
cultivated areas in the St. Albans Bay watershed (0.84 kg/ha/year) are higher than
Table 2 Annual area weighted phosphorus loading rates (kg/ha/year) for St. Albans Bay watershed
compared to other studies and literature
St. Albans Bay Lake Champlain Pike River Little Otter Creek
watershed TMDL watershed watershed
Phosphorus movement 1.20 – – –
in the landscape
Phosphorus load 0.84 – – 1.00
to streams
Phosphorus load to bay 0.55 0.61 0.70 –
Reference This study VTANR and Michaud et al. Meals et al.
(Gaddis 2007) NYDEC (2002) (2008) (2008)

Spatially Explicit Modeling of Land Use Specific Phosphorus Transport Pathways
Table 3 Annual area weighted phosphorus loading rates (kg/ha/year) for St. Albans Bay land uses
compared to other literature
St. Albans Bay Lake Champlain Pike River Literature range
watershed TMDL watershed (kg/ha)
Average of all cultivated 0.84 0.42 1.30 0.09–2.66
sources
Corn fields 1.23
Hay/Pasture fields 0.35
Developed diffuse 0.84 1.50 0.54–1.39
sources
Forest diffuse sources 0.26 0.04 0.09–0.44
Reference This study VTANR and Michaud et al. Hegman et al.
(Gaddis 2007) NYDEC (2002) (2008) (1999)
the loading rate used in the Lake Champlain TMDL (0.42 kg/ha/year) but lower
than modeled rates for the Pike River Watershed (1.3 kg/ha/year).
The total sediment load to the bay, estimated with water quality data and modeled
hydrologic output, was 970 t/year which equates to approximately 0.1 kg/m2/year.
This value is significantly lower than sediment loading rates in the Pike River
watershed to the north of St. Albans. The modeled sediment load to streams for
St. Albans Bay watershed was 1,250 t/year. Discrepancies between total sediment
estimates to the bay and total sediment estimates to the stream reflect in-stream
processes that are captured in the empirical data collected for discharge to the bay
but are not captured in the landscape model. It also reflects model uncertainty and
error. Comparing sediment load delivered to streams to that delivered to the bay
suggests that Jewett, Stevens, and Guayland Brooks contribute sediment load as a
result of in-stream erosion whereas Rugg Brook and Mill River accumulate sediment.
5 Results
5.1 Diffuse Sources and Pathways
In total, diffuse sources account for 8.1 t/year or 76% of the total phosphorus load
to watershed streams (Table 4). The distribution of phosphorus loss across the
watershed is relatively uneven with substantially more phosphorus loss per area from
Stevens and Jewett Brook, especially in the City of St. Albans and clay agricultural
soils (Fig. 5). The City of St. Albans accounts for a larger percentage (5%) of the
load than its area accounts for in the watershed (2%). The remainder of Stevens
Brook, also accounts for slightly more diffuse phosphorus (30%) than its relative
area (28%). These higher loads are balanced by lower loads coming from Rugg
Brook and Mill River. Guayland Brook and Jewett Brook account for the same
percentage of the load (4% and 25% respectively) as their areas account for in the
watershed (4% and 24% respectively). Corn land use accounts for the majority of
diffuse phosphorus to watershed streams especially considering the relatively small
percentage of the watershed occupied by this land use type, (Table 5). Developed
sources also supply a higher percentage of the total phosphorus diffuse load (19%)
than the area occupied by its land use in the watershed. However, hay and pasture

E.J.B. Gaddis, A. Voinov
Table 4 Summary of
phosphorus loads by transport
process or point source
Phosphorus source Phosphorus Percent of total
load (t/year) to streams
Ag sub-surface drainage (tile) 0.77 7.3%
Ag surface erosion 0.85 8.0%
Ag dissolved surface runoff 4.37 41.3%
Subtotal agriculture 5.99 56.6%
Non-city development road sand 0.98 9.2%
Non-city development dissolved P 0.12 1.1%
Non-city development erosion 0.02 0.19%
Subtotal non-city development 1.12 10.5%
City stormwater road sand 0.28 2.7%
City stormwater dissolved P 0.12 1.1%
City stormwater surface erosion 0.03 0.3%
Subtotal city stormwater 0.43 4.1%
Forest and wetlands 0.51 4.8%
Subtotal forest and wetlands 0.51 4.8%
Geese 0.01 0.1%
Subtotal geese 0.01 0.1%
Subtotal diffuse sources 8.06 76%
Milk house 0.13 1.2%
Manure runoff 0.75 7.1%
Silage leachate 0.70 6.6%
Subtotal farmstead discharge 1.58 14.9%
P in sewage/storm treated 0.84 8.0%
P in sewage/storm overflow 0.07 0.7%
NWCC discharge 0.003 0.03%
Subtotal wastewater 0.91 8.7%
Subtotal point sources 2.49 24%
Total 10.6 100%
lands supply a much smaller percentage of the total load than the area occupied
by this land use, suggesting that as land is converted from hay and pasture to
corn and/or development, phosphorus losses increase. Both trends are occurring
in the watershed. Forest and wetland land uses supply a much smaller proportion
of phosphorus than their area, which is expected as these land uses are relatively
undisturbed. The forest estimates could be high as they are driven by soil phosphorus
concentrations which were not directly sampled but interpolated from minimum
values of similar soil types in other land uses in the watershed.
Dissolved phosphorus in surface runoff from agricultural fields accounts for 41%
(4.37 t/year) of the total load to streams from the landscape and is the single most
important diffuse source of those modeled in the St. Albans Bay watershed (Table 4).
Clay soils, in Jewett Brook and Stevens Brook drainages, release more phosphorus
in this process per area than other soil types (Fig. 6). This results from the high soil
P concentrations found in many clay soils in the watershed compounded by high
surface runoff from clay soils.

Spatially Explicit Modeling of Land Use Specific Phosphorus Transport Pathways
Fig. 5 Map of total
phosphorus load delivered to
streams in the St. Albans Bay
watershed
0
0-1 kg/ha
1-2 kg/ha
2-3 kg/ha
3-4 kg/ha
4-5 kg/ha
Diffuse Phosphorus Load
The process of surface erosion accounts for 8% (0.85 t/year) of the total phospho-
rus load to streams. The process is relatively well distributed across the landscape
with areas adjacent to streams and with a high slope contributing more phosphorus
from surface erosion than other areas (Fig. 6). The erosion output from the spatial
model represents only surface erosion of sediment, and the phosphorus attached to
it. It does not include in-stream erosion which is an important component of the total
watershed load resulting from hydrologic changes such as peak flows associated with
stormwater runoff.
Tile drains in agricultural areas represent an important source of phosphorus to
surface waters, providing approximately 7% (0.77 t/year) of the total load to streams.
The highest tile drain loads come from the clay soils in Stevens Brook and Jewett
Brook drainages (Fig. 6), which results from the high concentration of phosphorus in
tile drainage water from clay soils compared to other soils.

E.J.B. Gaddis, A. Voinov
Table 5 Summary of all phosphorus loads by land use
Phosphorus Percent of total Percent land
load (t/year) P load to streams use in watershed
Corn diffuse sources 4.92 46% 22%
Hay/pasture diffuse sources 1.12 11% 35%
Total Ag diffuse sources 6.04 57% 57%
Ag point sources 1.58 15% 57%
Total agricultural 7.62 72% 57%
Developed diffuse sources 1.52 14% 14%
Developed point sources 0.91 9% 14%
Total developed 2.43 23% 14%
Forest diffuse sources 0.45 4% 23%
Other diffuse sources 0.06 1% 6%
Total 10.56 100% 100%
Road sand wash-off represents 12% (1.26 t/year) of the total load of phosphorus
to streams in the St. Albans Bay watershed (Fig. 6), which is significant considering
it is a source that had not been identified prior to this study and could be a relatively
easy source to reduce by the urban community. It is a recognized source in other
northern watershed studies (Oberts 1986) and as the St. Albans Bay watershed
becomes more developed, this could become a more significant source.
5.2 Semi-Diffuse and Point Sources of Phosphorus
Although the phosphorus problem in the St. Albans Bay watershed is primarily
a diffuse problem, there are several processes that cannot be captured with the
spatially explicit model. Sources associated with farmstead runoff are semi-diffuse
in that they represent concentrated diffuse sources that drain to a specific, though
unknown, point along a stream. Waterfowl are also semi-diffuse in that geese
concentrate in specific locations in the watershed. Wastewater treatment effluent
Particulate P from
Surface Drainage
Desolve P from
Surface Drainage
Phosphorus loss via
subsurface Drainage
Phosphorus loss via
road sand washoff
Phosphorus Load(kg/ha*year)
0
0 - 0.25
0.25 - .50
.50 - .75
.75 - 1
1+
Fig. 6 Maps of diffuse phosphorus load to streams for each phosphorus transport pathway

Spatially Explicit Modeling of Land Use Specific Phosphorus Transport Pathways
represents the only regulated point sources in the watershed. Together all of these
sources account for 2.5 t/year, or 24% of the total load to watershed streams.
The estimated total phosphorus load to streams from farmstead discharge is
1.58 t/year representing 15% of the total load to streams in the watershed. This load
can be further broken down into milk house effluent (0.13 t/year), manure runoff
(0.75 t/year), and silage leachate (0.7 t/year).
Geese are responsible, on average, for mobilizing 0.005 to 0.01 metric tons of
phosphorus per year throughout the St. Albans Bay watershed, less than 0.1% of
the total load to watershed streams.
The wastewater treatment plant discharges an estimated load of 0.91 t/year to
Stevens Brook, representing 8.6% of the total watershed load to streams. Of this
0.84 t/year is discharged as treated sewage and 0.07 t/year as wastewater overflow
during storms. This load is still well below the load allocation in the Lake Champlain
Phosphorus TMDL and permitted discharge for the wastewater treatment plant of
2.8 t/year (VTANR and NYDEC 2002).
A much smaller wastewater treatment plant for the Northwest Regional
Corrections Facility also discharges phosphorus to Jewett Brook. Average annual
phosphorus load from this facility is estimated to be 0.003 t/year which is approxi-
mately ten times less than the allocated load in the Lake Champlain TMDL for this
facility (0.028 t/year) but very close to the estimate provided for 2001 in the TMDL
(0.004 t/year).
6 Discussion
6.1 Importance of Identifying Phosphorus Transport Pathways to Achieve
TMDL Targets
Although diffuse pollution from particular land uses are typically lumped together
in TMDL documents, the actual transport processes that make up diffuse pollution
may vary across the landscape. For example, phosphorus may leave the landscape
through processes of erosion, subsurface drainage, or in a dissolved form in water
that runs off the landscape. Interventions required to address each of these processes
are very different. Net source reduction of pollution is a key component to any
long-term solution to phosphorus problems. Net reduction includes interventions
that either reduce the import of phosphorus to the landscape or increase the export
of phosphorus in the form of biomass, compost, etc. Process alterations include
temporal buffering and stabilization of a pollutant in the landscape. Whereas the
former refers to a mechanism by which an intervention might reduce pollution or
heavy flows temporarily, the latter refers to sequestration of a pollutant in soil, plant
matter, or other stable forms thus reducing the net movement of pollutants in the
landscape. Enhanced sinks or capture interventions aim to recapture phosphorus
that is moving across the landscape before it reaches water bodies or attenuate it
in biomass or soil.
The list of feasible watershed interventions for the processes and sources identi-
fied in this study is extensive and ranges from structural changes implemented on
a centralized scale, such as stormwater ponds and treatment, to behavioral changes

E.J.B. Gaddis, A. Voinov
at the level of individual homeowners and farmers, such as eliminating the use of
excessive fertilizer in home gardens and lawns. In order to attain TMDL diffuse
pollution reduction targets, phosphorus sources and transport processes must be
matched with appropriate watershed interventions (Fig. 7).
SOURCE REDUCTION PROCESS
ALTERATION
CAPTURE
------------------------Ag Dissolved Runoff (41%)------------------------->
Reduce fertilizer (NMP)
Convert manure to export product
Reduce phosphorus in feed (NMP)
Use exclusion
Tillage
Timing of fertilizer/manure
application (manure storage)
Tiling fields

“EAF steel slag barriers for surface
runoff P reduction”
------------------------Farmstead Discharge (15%)------------------------->
Pasture more animals Manure storage Farmstead treatment
------------------------Road Sand Washoff (12%)------------------------->
Reduce use of road sand
Road sweeper

Reduce impervious cover
“Better back roads”
Sediment traps
Centralized detention ponds and
treatment
------------------------Wastewater Discharge (8.6%)------------------------->
Storm sewer separation Expand wastewater treatment
plant
Improve treatment capability
(below 0.5 mg/l)
------------------------Ag Surface Erosion (8%)------------------------->
Reduce fertilizer (NMP)
Convert manure to export product
Reduce phosphorus in feed (NMP)
Cover crops
Tillage
Tiling fields
Buffers and/or filter strips

------------------------Subsurface Field Drainage (7.3%)------------------------->
Reduce Soil P (reduce fertilizer
and manure application)
Un-tile

“EAF steel slag barriers for surface
runoff P reduction”
------------------------In-stream Erosion (Unknown)------------------------->
Retain water in landscape
(centralized & local retention)
Reduce impervious cover
Detach roof drains from storm
pipes
Armoring stream banks
Stream buffering
Restore geomorphology

Dredging sediments
In-stream detention
------------------------Developed Dissolved Runoff (2.2%)------------------------>
Reduce commercial and residential
fertilizer usage

Local retention (i.e. rain
barrels, rain gardens,
infiltration trenches)
Reduce impervious cover
Centralized detention ponds and
treatment

------------------------Developed Surface Erosion (0.5%)------------------------>
Local retention (i.e. rain
barrels, rain gardens,
infiltration trenches)
Reduce impervious cover
“Better back roads” program
Sediment traps
Centralized detention ponds and
treatment

Fig. 7 Summary of watershed interventions categorized by process and reduction mechanism

Spatially Explicit Modeling of Land Use Specific Phosphorus Transport Pathways
6.2 Opportunities for Innovative Implementation Efforts for the St. Albans
Bay Watershed
Several significant sources of phosphorus identified in this study could be addressed
through watershed interventions that have not yet been implemented in the St.
Albans Bay watershed. Phosphorus in road sand wash off and field drainage water
are two important sources in the St. Albans Bay watershed that are not specifically
identified in the Lake Champlain Phosphorus TMDL, although they are imbedded
in loads associated with agricultural and developed land uses. Road sand applied
during winter storms that washes off streets during spring melt and storm events
contributes 12% of the phosphorus to streams in the watershed. This could be one of
the most cost-effective sources to reduce in the short-term either through improved
road sweepers or replacement of sand with other deicing mechanisms. Tile drainage
from fields represents 7.3% of the total phosphorus loss to streams and offers a good
opportunity for innovative implementation efforts. Drainage water is collected in a
series of small pipes that discharge directly to ditches and stream channels. These
discharges are effectively small point sources that could receive some sort of pre-
treatment prior to discharge.
Dissolved phosphorus leaving agricultural fields through surface runoff represents
41% of the total phosphorus loss to streams in the watershed. Leaching of dissolved
phosphorus into surface drainage requires correction of the net imbalance of phos-
phorus imported to the watershed (Gaddis 2007). In addition, new technologies are
being developed to help mitigate the impact of dissolved phosphorus runoff through
phosphorus sink enhancement (Drizo et al. 2006).
Phosphorus from farmsteads are an important source of phosphorus that is
currently imbedded into the agricultural diffuse phosphorus estimate in the Lake
Champlain Phosphorus TMDL. Farmstead discharge is a semi-diffuse source which
offers the opportunity for treatment and phosphorus load reduction using existing
technologies.
6.3 Spatial Targeting of Watershed Interventions
The majority of the diffuse phosphorus load comes from the Stevens Brook water-
shed, which accounts for more load than its relative area in the watershed. Much of
this imbalance comes from the City of St. Albans. Treatment of the concentrated
stormwater load in the city is a priority for achieving TMDL reduction targets for
developed land uses. Across the agricultural landscape the highest phosphorus loads
come from clay soils that are saturated with phosphorus and are found primarily in
the Jewett Brook and lower Stevens Brook subwatersheds. These areas should be
the priority for agricultural BMP implementation in the future (Fig. 7).
6.4 Conclusions
The results of this study indicate that the majority of phosphorus loss to streams in
the St. Albans Bay watershed is diffuse. Of the diffuse sources, the most important
source is dissolved phosphorus in agricultural surface runoff and tile drain water

E.J.B. Gaddis, A. Voinov
followed by road sand wash-off in the developed landscape. Direct discharge to
streams from farmsteads also represents a significant load to streams of which the
majority comes from barnyard manure runoff and silage leachate. Addressing these
four sources with available and innovative technology and management practices
would be the most effective means to move towards achievement of TMDL targets.
This study offers a mechanism to identify the relative importance of specific
transport processes of pollutants from the landscape to streams and could inform the
types and locations of BMPs recommended for pollutant reduction. The approach
has implications for setting realistic TMDL diffuse pollution allocations and could
help communities attain targets in a more focused manner by targeting efforts to the
transport processes that are most important, the land use and soil types that appear
to be particularly problematic, as well as areas of the watershed that are the most
impaired. This approach could be replicated for other TMDL processes around the
country in designing more effective Watershed Management and Implementation
Plans for TMDL attainment.
References
Breuer L, Frede H (2003) PlaPaDa: an online plant parameter database for eco-hydrological
modelling approaches. Institute for Landscape Ecology and Resources Management.
http://www.uni-giessen.de/∼gh1461/plapada/plapada.html
Cassell EA, Meals DW, Aschmann SG, Anderson DP, Rosen BH, Kort RL, Dorioz JM (2002) Use of
simulation mass balance modeling to estimate phosphorus and bacteria dynamics in watersheds.
Water Sci Technol 45:157–168
Costanza R, Voinov A (eds) (2004) Landscape simulation modeling: a spatially explicit dynamic
approach. Springer, New York
Drizo A, Forget C, Chapuis RP, Comeau Y (2006) Phosphorus removal by electric arc furnace steel
slag and serpentinite. Water Res 40:1547–1554
Foster GR (1982) Modeling the erosion process. In: Haan C, Johnson H, Brakensiek D (eds)
Hydrologic modeling of small watersheds. St. Joseph, Michigan
Gaddis EJB (2007) Landscape modeling and spatial optimization of watershed interventions
to reduce phosphorus load to surface waters using a process-oriented and participatory
research approach: a case study in the St. Albans Bay watershed, Vermont. Disserta-
tion, University of Vermont. http://www.uvm.edu/giee/?Page=about/students/Erica_Gaddis.
html&SM=about/about_menu.html
Gaddis EJB, Vladich H, Voinov A (2006) Participatory modeling and the dilemma of diffuse nitrogen
management in a residential watershed. Environ Model Softw 22:619–629
Gaddis EJB, Falk HH, Ginger C, Voinov A (2009) Effectiveness of a participatory modeling effort
to identify and advance community water resource goals in St. Albans, Vermont. Environ Model
Softw. doi:10.1016/j.envsoft.2009.06.004
Gray DM, Prowse TD (1993) Snow and floating ice. In: Maidment DR (ed) Handbook of hydrology.
McGraw-Hill, New York
Heathwaite L, Sharpley A, Gburek W (2000) A conceptual approach for integrating phosphorus and
nitrogen management at watershed scales. J Environ Qual 29:158–166
Hegman W, Want D, Borer C (1999) Estimation of Lake Champlain basinwide nonpoint source
phosphorus export. Lake Champlain Basin Program Technical Report
Hession WC, Storm DE (2000) Watershed-level uncertainties: implications for phosphorus manage-
ment and eutrophication. J Environ Qual 29:1172–1179
Hyde K, Kamman N, Smeltzer E (1994) History of phosphorus loading to St. Albans Bay, 1850-1990.
Technical report no. 7B, Lake Champlain Basin program, North Hero, Vermont

Spatially Explicit Modeling of Land Use Specific Phosphorus Transport Pathways
Krider JN (1999) Agricultural waste management field handbook. Part 651 national engineering
handbook. Natural Resources Conservation Service, United States Department of Agriculture,
Washington DC
Madani K, Mariño MA (2009) System dynamics analysis for managing Iran’s Zayandeh-Rud river
basin. Water Resour Manage 23:2163–2187
Magdoff F, Lanyon L, Liebhardt B (1997) Nutrient cycling, transformations, and flows: implications
for a more sustainable agriculture. Adv Agron 60:1–73
Male DH, Gray DM (1981) Snowcover ablation and runoff. In: Gray D, Male D (eds) Handbook of
snow: principles, processes, management and use. Pergamon, Toronto, pp 360–436
Maxwell T (1999) A parsi-model approach to modular simulation. Environ Model Softw 14:511–517
Maxwell T, Costanza R (1995) Distributed modular spatial ecosystem modeling. Int J Computer
Simul 5:247–262
Maxwell T, Costanza R (1997a) A language for modular spatio-temporal simulation. Ecolog Model
103:105–113
Maxwell T, Costanza R (1997b) An open geographic modelling environment. Simul J 68:175–185
McCool DK, Brown LC, Foster GR (1987) Revised slope steepness factor for the Universal Soil Loss
Equation. Trans Am Soc Agric Eng 30:1387–1396
McDowell RW, Sharpley AN (2001) Approximating phosphorus release from soils to surface runoff
and subsurface drainage. J Environ Qual 30:508–520
Meals DW (1996) Watershed-scale response to agricultural diffuse pollution control programs in
Vermont, USA. Water Sci Technol 33:197–204
Meals DW, Cassell EA, Hughell D, Wood L, Jokela WE, Parsons R (2008) Dynamic spatially explicit
mass-balance modeling for targeted watershed phosphorus management—I. Model develop-
ment. Agric Ecosyst Environ 127:189–200
Michaud AR, Beaudin I, Deslandes J, Bonn F, Madramootoo CA (2008) SWAT-predicted influence
of different landscape and cropping system alterations on phosphorus mobility within the Pike
River watershed of south-western Québec. Can J Soil Sci 87:329–344
Neitsch SL (2000) Soil and water assessment tool: theoretical documentation. Grassland, Soil and
Water Research Laboratory, Temple
NRAES (1993) Silage production from seed to animal. NRAES-67, Northeast Regional Agricultural
Engineering Service, USDA, Syracuse
Oberts GL (1986) Pollutants associated with sand and salt applied to roads in Minnesota. Water Res
Bull 22:479–483
Post DM, Taylor JP, Kitchell JF et al (1998) The role of migratory waterfowl as nutrient vectors in a
managed wetland. Conserv Biol 12:910–920
Rao KHVD, Kumar DS (2004) Spatial decision support systems for watershed management. Water
Resour Manage 18:407–423
Rendard KG, Foster GR, Weesies GA et al (1997) Predicting soil erosion by water: a guide to
conservation planning with the Revised Universal Soil Loss Equation (RUSLE). Agriculture
handbook number 703, Agricultural Research Service, USDA, Washington DC
Rykiel EJJ (1996) Testing ecological models: the meaning of validation. Ecol Model 90:229–244
Sharpley AN, Kleinman PJA, McDowell RW, Gitau M, Bryant RB (2002) Modeling phosphorus
transport in agricultural watersheds: processes and possibilities. J Soil Water Conserv 57:425–
439
Shuttleworth WJ (1993) Evaporation. In: Maidment DR (ed) Handbook of hydrology. McGraw-Hill,
New York
Tim US, Mostaghimi S, Shanholtz VO (1992) Identification of critical nonpoint pollution source
areas using geographic information systems and water quality modeling. Water Res Bull 28:877–
887
Thornthwaite C, Mather J (1957) Instructions and tables for computing potential evapotranspiration
and the water balance. Laboratory of Climatology Publication No. 10. Centerton, New Jersey
USDA (1997) Part 630—hydrology. In: National engineering handbook. US Dept of Agriculture—
Soil Conservation Service, Washington, DC National Technical Information Service, Springfield,
Virginia
Voinov A, Voinov H, Costanza R (1999a) Landscape modeling of surface water flow: 2nd Patuxent
watershed case study. Ecol Model 119:211–230
Voinov A, Costanza R, Wainger L et al (1999b) Patuxent landscape model: integrated ecological
economic modeling of a watershed. Environ Model Softw 14:473–491

E.J.B. Gaddis, A. Voinov
Voinov A, Fitz C, Boumans R et al (2004) Modular ecosystem modeling. Environ Model Softw
19:285–304
VTANR and NYDEC (2002) Lake Champlain Phosphorus Total Maximum Daily Load (TMDL).
Vermont Agency of Natural Resources, Department of Environmental Conservation and New
York State Department of Environmental Conservation, Waterbury, VT and Albany, NY
Wood T, Bormann FH, Voigt GK (1984) Phosphorus cycling in a northern hardwood forest—
biological and chemical control. Sci 223:391–393
Zeinivand H, De Smedt F (2009) Hydrological modeling of snow accumulation and melting on river
basin scale. Water Resour Manage 23:2271–2287