TECHNICAL REPORTS:TECHNICAL LANDSCAPEREPORTS AND WATERSHED PROCESSES

Combined Monitoring and Modeling Indicate the Most Effective Agricultural Best Management Practices Zachary M. Easton,* M. Todd Walter, and Tammo S. Steenhuis Cornell University Although water quality problems associated with agricultural nonpoint source (NPS) pollution have prompted the rapid and widespread adoption of a variety of so called “best management practices” (BMPs), there have been few realistic efforts to assess their combined effectiveness in reducing NPS pollution. This study used the Variable Source Loading Function (VSLF) model, a distributed watershed model, to simulate phosphorus (P) loading from an upstate New York dairy farm before and after the implementation of a suite of BMPs. With minimal calibration, the model calculates the dissolved P (DP) losses from impervious surfaces (e.g., barnyards), the plant/soil complex, field-applied manure, and loads associated with baseflow conditions. The simulated DP loads agreed well with measured loads for both the pre-BMP and post-BMP periods. More importantly, results showed that BMPs reduced DP loads by 35%, which is over half of the expected reduction if all manure was removed from the watershed, i.e., ~50% reduction. The model results indicate that had no BMPs been installed DP loads would be ~37% greater than observed at the watershed outlet. The most effective BMPs were those that disassociated pollutant loading areas from areas prone to generating runoff, i.e., hydrologically sensitive areas. By contrast, attempts to reduce P content in manure were somewhat less effective. This study demonstrates that a combination of distributed, mechanistic modeling and long-term monitoring provides better insights into the effectiveness of water quality protection efforts than either individually.

Copyright © 2008 by the American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America. All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Published in J. Environ. Qual. 37:1798–1809 (2008). doi:10.2134/jeq2007.0522 Received 3 Oct. 2007. *Corresponding author ([email protected]). © ASA, CSSA, SSSA 677 S. Segoe Rd., Madison, WI 53711 USA

A

gricultural activity clearly contributes nonpoint source (NPS) contaminants to streams, lakes, and estuaries (Puckett, 1995; Ekholm et al., 2000; Sharpley et al., 2001; Andraski and Bundy, 2003; DeLaune et al., 2004), which has prompted the development and implementation of a wide variety of so called “best management practices” (BMPs). Although some BMPs have been developed based on good applied research and credible science, many BMPs, in practice, fail to perform as anticipated (Walter et al., 1979; Dillaha et al., 1988, 1989; Schwer and Clausen, 1989; Lammers-Helps and Robinson, 1991; Schellinger and Clausen, 1992). In some cases this may be because BMPs represent a compromise between theory and practicality. Thus, the effectiveness of many BMPs to protect water quality remains unclear. For example, no-till agriculture, which was originally proposed as a soil conservation practice, may actually increase soluble phosphorus (P) transport (Walter et al., 1979; Novotny, 2003). This is alarming given soluble P is usually the primary nutrient driving eutrophication in freshwater aquatic ecosystems (Daniel et al., 1994, 1998; Parry, 1998). Similarly, buffer or vegetative filter strips, widely promoted as one of the most effective BMPs for a wide range of potential pollutants have been observed to be sources, rather than sinks, of some nutrients (Dillaha et al., 1988, 1989). While there are studies that show specific BMPs to be effective in individual situations (Magette et al., 1989; Lee et al., 2000; Inamdar et al., 2001; Brannan et al., 2000; Novotny, 2003; Gitau et al., 2004), it is often difficult to use these results to predict how effective the same BMPs will perform in different situations. One reason why BMPs do not always work as planned is that many are simply re-packaged soil conservation practices, which are not appropriate for many pollutants and situations (Walter et al., 1979, 2003; Novotny, 2003). Additionally, BMP effectiveness is often based on the ratio of output to input pollutant concentrations; thus, research that uses very large input concentrations will observe better efficiencies than are perhaps representative of most realistic conditions (Dittrich et al., 2003). Even without these problems, it is difficult to ascertain the cumulative effectiveness of combinations of BMPs. Ideally we would like to carry out a suite of controlled experiments at field and watershed scales using single BMPs and combinations of BMPs to directly measure the impacts. This is clearly not practical for many Dep. of Biological and Environmental Engineering, Cornell Univ., Ithaca, NY 14853. Abbreviations: BMP, best management practices; DP, dissolved phosphorus; E, NashSutcliffe (1970) Efficiency; HRUs, hydrologic response units; NPS, nonpoint source; NMP, nutrient management plan; SCS-CN, Soil Conservation Service Curve Number; STI, soil topographic index; STP, soil test phosphorus; VSA, variable source area; VSLF, Variable Source Loading Function model.

1798

reasons, not the least of which is a lack of control landscapes. Thus, we need to find ways to use the “real-world” landscapes as “living” experiments to evaluate the effectiveness of BMPs as implemented. Paired watershed studies offer a proven evaluation method of the cumulative effects of BMPs in this “living experiment” context, particularly under varying hydro-climatic conditions (Bishop et al., 2005) but it is not easy or entirely feasible with this kind of approach to see which BMPs are the most effective. Similarly, longterm monitoring may provide insights into system-wide changes that may be difficult to associate with specific BMPs or combinations of BMPs, and trends may be difficult to interpret in light of the impact of exogenous variables such as weather variability or climate change. Many studies that have supposedly demonstrated the cumulative effectiveness of BMPs are almost entirely based on model results (e.g., Sorrano et al., 1996; Walter et al., 2001; Santhi et al., 2003) and are thus subjected to problems with incomplete understanding of the relevant fate-transport systems, especially those models that require calibration of pollutant export coefficients associated with BMPs. Another problem with some models is that they simulate processes at scales much larger than those at which BMPs operate, especially targeted management practices (see Walter et al., 2007), and it is virtually impossible to objectively manipulate the model to meaningfully capture the impacts of the BMPs (Garen and Moore, 2005). We propose that the best way to gain insights into the effectiveness of water quality management practices is to combine watershed-scale water quality monitoring and distributed hydrological-water quality modeling. For this approach to work, the model needs to be based on physical processes as much as possible to correctly simulate the local hydrology and to predict water quality changes arising from BMP adoption. The model must also be parsimonious to minimize problems of equifinality that may arise for models with many input parameters that cannot be easily and independently quantified. In fact, there should be little or, ideally, no model calibration. The model must also be spatially distributed to a fine enough resolution to capture the impacts of individual BMPs (Garen and Moore, 2005). Ideally, the distributed model should be “validated,” i.e., tested, against both distributed and watershed-scale (stream) observations. For evaluating the BMPs, stream monitoring needs to be long enough to verify that the model correctly captures the watershed hydrology and water chemistry and accounts for weather variability. Ideally, monitoring will span periods with and without implemented BMPs. Following these guidelines, we attempted to evaluate BMPs designed to reduce NPS dissolved phosphorus (DP) loads in a northeastern U.S. watershed dominated by dairy production. Dissolved P pollution of surface water is a considerable problem in many areas of the country (Allan, 1995; DeLaune et al., 2004; Hamilton et al., 2004; Sharpley et al., 2004), and a variety of BMPs have been offered as a means to reduce P loss from agricultural watersheds. Our study site is in the Catskill Mountain region of New York State, for which the most rapid hydrologic flow paths have long been associated with variable source area hydrology (e.g., Dunne and Black, 1970; Dunne and Leopold, 1978; Srinivasan et al., 2002; Walter et al., 2003; Needelman

et al., 2004). Several researchers have been able to convincingly capture these VSA systems with distributed hydrological models (e.g., Frankenberger et al., 1999; Lyon et al., 2004; Mehta et al., 2004; Lyon et al., 2006a, 2006b; Schneiderman et al., 2007).

Model Description The Variable Source Loading Function (VSLF) model (Schneiderman et al., 2007) is a re-conceptualized version of the Generalized Watershed Loading Function (GWLF) model (Haith and Shoemaker, 1987) that distributes runoff based on variable source area hydrology and associated, source specific DP transport (Easton et al., 2008). We briefly describe VSLF here.

Runoff Model The VSLF model uses the Soil Conservation Curve Number (SCS-CN) (USDA-SCS, 1972) equation to predict runoff but incorporates novel re-conceptualizations of the method to account for variable source area (VSA) hydrology (for the full evolution of this model see Steenhuis et al., 1995; Lyon et al., 2004; Schneiderman et al., 2007). Briefly, watershed runoff response is characterized by a basin-wide effective storage parameter, Se, which has been traditionally determined qualitatively via Curve Number tables (e.g., USDA-SCS, 1972). In VSLF, Se is determined directly from the relationship between measured effective rainfall, Pe, and storm runoff, Q. To distribute runoff generating areas, the watershed is divided into equal areas, Ai, of wetness index, as defined by a soil topographic index, λ (Beven and Kirkby, 1979): ⎛ a ⎞⎟ λ = ln ⎜⎜ ⎟ ⎜⎝ T tan β ⎠⎟

[1]

where a is upslope contributing area per unit of contour line (m), T is the transmissivity (soil depth x saturated hydraulic conductivity), and β is the local topographic slope. The soil water deficit, σi, for each Ai is related to the basinwide effective storage, Se: σi =

2S e ( 1− Ai − 1− Ai+1 ) ( Ai+1 − Ai )

− Se

[2]

where Ai+1 is the cumulative area between a watershed’s highest soil topographic index, i.e., λ at the watershed outlet, and a particular topographic index, λi; recall from above, each Ai is an equal fraction of the watershed area. The basinwide effective storage, Se, and, thus, the propensity for runoff generation, increases and decreases as soil moisture decreases and increases, respectively, using functions from the Soil Plant Air Water (SPAW) model (Saxton, 1982). Runoff from each area, Ai, is simply the amount of precipitation in excess of the local soil water deficit: Qi = Pe–σi for Pe > σi,

[3a]

Qi = 0 for Pe < = σi

[3b]

Through this approach, the total watershed runoff quantity is calculated by a traditional runoff curve number approach while

Easton et al.: Combined Monitoring & Modeling to Indicate Effective Agricultural BMPs

1799

runoff generation is distributed preferentially to areas of surface saturation. For a complete description of the VSLF hydrology model and a full corroboration of model predictions against a variety of integrated (stream response) and distributed soil moisture measures see Schneiderman et al. (2007).

Phosphorus Model Despite the numerous processes affecting DP loss from a watershed, VSLF (Easton et al., 2008) explicitly accounts for DP loads from field-spread manures, chemical fertilizers, areas of plant/soil complexes, and impervious areas (barnyards), and loads associated with base-flow conditions (i.e., no storm flow from the landscape).

Dissolved Phosphorus Loss from Manured (Fertilized) Areas There are two pathways for DP from manure, loss in runoff or conversion of soluble P into less soluble forms by interaction with the plant/soil complex. Gérard-Marchant et al. (2005) showed that available manure DP declined exponentially with time: ⎛ Δt ⎞ M F,t = M F,t −Δt exp ⎜⎜ − ⎟⎟⎟ − DF,t −Δt ⎝ τ ⎠

[4]

where τ is the P immobilization rate (d), MF, t, is the available water-extractable P at time t per unit area (kg m−2 d−1), MF,t-Δt and DF,t-Δt are the water-extractable P (kg m−2 d−1) and the manure DP loss in runoff (kg m−2 d−1), respectively, from the previous time step. Following manure application, the initial water-extractable P in the manure (a fraction, ω, of the total manure P, MF, [kg m−2] in the manure) is added to the amount already present on the soil surface. A time step, Δt, of 1 d is used in the model. Based on Gérard-Marchant et al. (2005), Easton et al. (2007, 2008) developed an expression for the DP load as a function of cumulative runoff, Qt, and MF,t: ⎡ ⎛ k M Q ⎞⎤ DF,t = ⎢⎢ M F,t ⎜⎜⎜ F F,t t ⎟⎟⎟⎥⎥ ⎝⎜ 1 + kF M F,t Q t ⎠⎟⎦⎥ ⎣⎢ [5] where kF (m3 kg−1) is a reaction constant. Equations [4] and [5] can also be used to simulate DP from chemical fertilizers (Easton et al., 2007).

Phosphorus Loss from Plant/Soil Complex

DS,t = μ S M S Q t

[6]

where DS,t is the daily DP load in runoff (kg m−2 d−1), MS is the STP in the surface soil (kg m−3) at time t, Qt is the VSLF predicted runoff (m3 m−2 d−1) from Ai at time t, and μS is the soil-specific coefficient determined from sampled runoff events adjusted for temperature with an Arrhenius equation (Zheng et al., 2003): ⎛⎜ T −TS ⎞⎟ ⎜ 10 ⎠⎟⎟

μ S = μ T,S Q 10S ⎝

[7]

where μT,S is the reference export coefficient, Q10S is a factor change for a 10°C change in temperature, T is the average temperature at the soil surface (C°) at time t, and TS is the reference temperature (C°) at which μT,S was estimated. MS may vary with updated STP values from the land.

Dissolved Phosphorus Loss from the Barnyard An accumulation/wash-off equation is used to model the contribution of DP from the barnyard. The contribution of the barnyard to P stream loads varies temporally (e.g., higher contribution in the winter when cows are confined and more P is deposited and lower in the summer when they are pastured). The barnyard contribution might also vary due to changes in flow processes (summer storms vs. winter snowmelt). Thus, the rate of P buildup was allowed to vary based on the presence of snowfall in the watershed. This allows more rapid accumulation of manure P in the barnyard when cows are confined. Accumulation is modeled with an exponential build up equation: M I,t = M I,Max − (M I,Max − M I,(t −Δt ) )⎡⎣ exp (− kIΔt )⎤⎦

[8]

where MI,t is the DP load (kg m−2) from the barnyard, MI,(t-Δt) is the DP load (kg m−2) from the previous event, MI, Max is the maximum dissolved P load (kg m−2) in the barnyard, and kI is the exponential buildup factor (d−1). Wash-off of DP from the barnyard is estimated using a first-order relationship: DI,t = M I ⎡⎢1− exp (−kI,Q Q t )⎤⎥ ⎣ ⎦

[9]

where DI,t is the DP load (kg m−2) in runoff, kI,Q is the washoff coefficient (m2 m−3), and Qt is the VSLF predicted runoff (m3 m−2). DI,t is subtracted from MI,t before the next day is calculated with Eq. [8].

Phosphorus release from the plant/soil complex reflects both abiotic (sorption/desorption kinetics, soil moisture, temperature, precipitation) and biotic (organic matter mineralization, plant uptake) factors varying with climate and season (Hansen et al., 2002), which makes process-based modeling difficult. VSLF uses an export coefficient model that aggregates the relevant processes affecting P loss by using the mean expected P concentration in runoff from a homogeneous area (Torrent and Delgado, 2001; Sharpley et al., 2002) to predict P loss from the plant/soil complex. A simple linear relationship between the soil test P (STP) and DP in runoff was used to approximate P loss from soil via runoff as (Hively et al., 2005, 2006):

where LB,t is the DP load (kg d−1) in baseflow, Bt is the VSLF predicted baseflow volume at the watershed outlet (m3 d−1), and μB is the baseflow DP export coefficient adjusted for temperature

1800

Journal of Environmental Quality • Volume 37 • September–October 2008

Dissolved Phosphorus Loss Associated with Baseflow Baseflow, although it does not generally contain high P concentrations, can act as a constant background source of P (McDowell et al., 2001), and contribute substantially to cumulative P loads (Hooda et al., 1999; Caruso, 2000; Maguire and Sims, 2002; Hively et al., 2006). A lumped export coefficient approach is used to model baseflow P loss: LB,t = μ B Bt [10]

with an Arrhenius equation similarly to soils (Eq. [7]); again μB may vary with updated measured stream P concentrations: ⎛⎜ T −TB ⎞⎟ ⎟ ⎜ 10 ⎠⎟

μ B = μ T,B Q 10B ⎝

[11]

where μT,B is the reference baseflow export coefficient, Q10B is a factor change rate for a 10°C change in temperature, T is the average soil temperature at the mean watershed soil depth (C°) at time t, and TB is the reference temperature (C°) at which μT,B was measured. It is unclear what mechanisms actually control the P loads associated with baseflow, but it seems logical that, whether associated with ground water contributions or near/in-stream processes, the temperature effects should be dampened due to the moderating effects of relatively stable ground water temperature. Therefore, the TB parameter should reflect temperature damping observed at soil depth, Zt (m). Thus, we vary temperature at a depth ZT in the soil using a sine function (Brutsaert, 1982; Hively et al., 2006) as: − ZT ⎡ 2π ⎤ Z T (t , Z t )= TAVG +ΔT Ze sin ⎢ t − t φ )− t Z ⎥ ( e ⎦⎥ ⎣⎢ 365 [12] where TAVG (C˚) is the annual average temperature at the soil surface, ΔT (C˚) is the temperature deviation from the average, tφ is the time lag (d), and Ze (m) is the equivalent dampening depth. Since the baseflow coefficient, μB, integrates the effects of the entire watershed on DP loss it should be obtained from observed baseflow DP concentrations.

Stream Dissolved Phosphorus Loss (LT,t) (kg d−1) is the sum of contributions from manure (and/or chemical fertilizers) (DF,t), from the plant/soil complex (DS,t), from barnyards (DI,t), and those associated with baseflow (LB,t): N

LT,t = LB,t + ∑ A j (DF,t,j + DS,t,j + DI,t,j ) j =1

average precipitation (NCDC, 2005). The topography is steep and winter snow accumulation and spring snowmelt dominate the hydrology (Bishop et al., 2005). Long-term monitoring consisted of precipitation, stream flow, and stream water quality from 1993 to 1995, before BMPs were implemented on the farm, and again from 1997 to present, after BMP implementation; here we only use data through 2004. BMPs were installed mainly during 1995 to 1996, although some were installed in 2001 and 2002. Detailed descriptions of the study watershed are given in Bishop et al. (2003, 2005), Gérard-Marchant et al. (2006), Hively et al. (2005, 2006), and Hively (2004). Daily stream flows were recorded on a 10-min basis by a gauge at the watershed outlet, and integrated over each day. Observed DP concentrations were derived from flow-weighted sampling at the watershed outlet, as described in Bishop et al. (2003, 2005, 2006). Dissolved P is defined as molybdate reactive orthophosphate in filtered (45 um) Kjeldahl digested water samples. Daily minimum and maximum temperatures were obtained from a weather station located in Delhi, New York, about 20 km SW of the watershed (NCDC, 2005).

Pre-Best Management Practice Management Land use on the farm before BMP implementation included deciduous forest (53.4%), brush/shrub (2.9%), pasture (9.8%), permanent hay (24.1%), cropland (7.6%), barnyard (0.1%), rural road (1.4%), water (0.6%), and barnyard-impervious surfaces (0.3%) (Fig. 1a). Based on herd size and average manure production, one load of manure was spread approximately every 1.5 d (4.3 kg P load−1) and spread evenly (21 kg P ha−1) over the pastures during the growing season, hay fields following harvesting operations, and row crops in the early spring and late fall, pre-plant and post-harvest, respectively. During the winter, manure was generally spread near the barn because of difficulty accessing fields in the upper reaches of the watershed. Before BMP implementation, the farm discharged its milkhouse waste directly to the stream.

Post-Best Management Practice Management [13]

where Aj is the area of each modeled process.

Watershed Description, Best Management Practices, and Long-Term Monitoring Site Description and Monitoring To demonstrate our hypothesis that a combination of longterm monitoring and distributed water quality modeling provides valuable insights into the effectiveness of BMPs, we focus on a 164-ha watershed that consists of a single dairy farm. At the beginning of the monitoring period (1993) the herd size was 70 milking cows and 40 heifers, which increased 15 to 20% over the course of the monitoring (USEPA, 2007). The farm watershed is located in the Cannonsville basin in Delaware County, NY, and drains into the West Branch of the Delaware River and ultimately into New York City’s Cannonsville reservoir. The climate is humid continental, with an average temperature of 8°C, and 1123 mm yr−1

Best management practices were developed as part of a comprehensive Nutrient Management Plan (NMP) and included precision feeding to reduce manure P content, riparian buffers and associated cattle exclusion from the riparian areas, barnyard improvements, targeted manure spreading schedule, and interceptor drainage ditches to reduce flow to frequently saturated areas. Exclusionary fencing to keep cattle from the near-stream areas mandated the installation of animal trail/stream crossing in 2002. A targeted manure spreading schedule prioritized spreading to fields/pastures that were deemed least likely to generate runoff, i.e., to avoid spreading in hydrologically sensitive areas (Walter et al., 2000, 2001) as described in the New York Phosphorus Index (Geohring et al., 2002; Czymmek et al., 2003). The manure spreading schedule also prioritized spreading to fields with low STP levels (Czymmek et al., 2001, 2002). The post-BMP manure spreading schedule required the installation of a manure storage lagoon to allow the farmer to store manure during times of the year when runoff risk is high. Thus, manure was generally stored during the winter and early spring and heavily spread in late spring and late fall to empty the lagoon. The barnyard improvements

Easton et al.: Combined Monitoring & Modeling to Indicate Effective Agricultural BMPs

1801

Fig. 1. Watershed land use for (a) the pre-best management practice (BMP) period (1993–1995), and (b) the post-BMP period (1997–2004).

consisted of relocating the stream channel away from the barnyard, and re-grading and shaping the area to reduce runoff followed by more frequent cleaning of accumulated manure. After BMPs, the milkhouse waste was directed to the manure storage lagoon and was applied to fields during manure spreading. The distribution of land use and, subsequently, the hydrologic response units (HRUs) were altered because of BMP implementation, primarily through small reductions in fields and pastures as a result of fenced-off riparian buffer areas, which accounted for 3.2% of the watershed, and the addition of the drainage ditches which altered flow paths (Fig. 1b). A precision feeding program was implemented as a BMP in January 2001 (Cerosaletti et al., 2004). Finally, a cattle stream crossing was installed in 2002 near the barnyard.

Model Application Determining Model Parameters The VSLF hydrology model has three primary calibration parameters: the watershed storage parameter for pervious areas, Se (or Curve Number, CN), a flow recession coefficient, and a snowmelt factor. We calibrated these watershed-scale parameters to maximize the Nash-Sutcliffe (Nash and Sutcliffe, 1970) efficiency (E) and minimize the bias for stream flow during the pre-BMP period (1993–1995): Se = 14 (cm), CN = 64.5, recession coefficient = 0.07 (d−1), snow melt factor = 0.12 (cm °C d−1). Detailed descriptions of the calibration procedure are in Schneiderman et al. (2002, 2007) and Easton et al. (2008). For the pre-BMP situations, parameters for the wetness indices (Eq. [1]), a and β are taken directly from a 10 m digital elevation model (DEM) of watershed and T is extracted from the SSURGO soil database (USDA-NRCS, 2000) for each 10 m cell (Fig. 2a). These were divided into “bins” of average wetness index over 10 equal areas, Ai. For the postBMP situation, the DEM was modified to account for new

1802

drainage ditches, stream channel modifications, and barnyard improvements (Fig. 2b). Pre- and post-BMP distributions of land use were overlain on the wetness index distributions to create maps of HRUs that delineate areas of unique runoff susceptibility and land management (Fig. 2a, 2b). There are twelve parameters in the DP model, all of which were independently quantified a priori (Table 1). Most were estimated from published values and/or direct measurements. The plant/soil complex export coefficients, μT,S, were based on direct measures of Morgan STP (McIntosh, 1969) from the topsoil for each field on the farm as directed by the NMP (Sharpley et al., 2002; Maguire and Sims, 2002; McDowell and Sharpley, 2003; DeLaune et al., 2004) and rainfall simulations performed in the watershed by Hively (2004) and Hively et al. (2005, 2006). The contribution of the barnyard was estimated using the simulated runoff data from Hively (2004) and Hively et al. (2005), and from qualitative observations. As mentioned earlier, our understanding of the processes controlling DP loads under baseflow conditions are incomplete but by considering summer baseflow data we were able to determine the parameters for Eq. [10] and [11] a priori or at least independently from the rest of the VSLF model. To do this, we set TB to the average summer temperature at the mean watershed soil depth (60 cm) and fit a linear regression to (T−TB)/10 as a function DP concentrations measured during summer baseflow conditions; the y-intercept of this regression is the log of μT,B and the regression slope is the log of Q10B. To develop this regression we used ~50 measurements of μB taken during summer, low flow (< 1 mm d−1) periods. The Q10B and μT,B parameters were calculated for pre- and post-BMP periods independently to account for changes in processes controlling baseflow DP resulting from BMPs. Again, we do not know explicitly what these processes were. Only, μT,B changed between pre- and post-BMP situations (Table 1). The same method was used to determine Q10S (Eq. [7]); recall that we have direct measures of μT,S from Hively (2004) and

Journal of Environmental Quality • Volume 37 • September–October 2008

Fig. 2. Schematic of hydrologic response unit (HRU) delineation for (a) the pre-best management practice (BMP) period (1993–1995) and (b) the post-BMP period (1997–2004). HRUs are defined as the intersection of the areally weighted soil topographic index (STI) and land use.

Hively et al. (2005) (Table 1). This approach represents an example of where strategic monitoring is essential to gain realistic insights into what aspects of our conceptual understanding of the system (i.e., our model) are influenced by alterations that we do not fully understand. Indeed, this approach of using models to help identify knowledge gaps in our understanding of environmental systems is championed by Grayson et al. (1992).

Simulating Best Management Practices We consider four categories of BMPs: (i) manure P reduction (e.g., precision feeding), (ii) hydrologic alteration (e.g., drainage tiles, stream channel movement), (iii) land use changes (e.g., conversion of near stream fields to riparian buffers), and (iv) redistribution of P application to the landscape (manure spreading schedule). Manure P reduction by precision feeding was represented in the model by altering the MF parameter (Eq. [4] and [5]) based on actual measurements of manure P before and after precision feed-

ing was initiated. Hydrologic alterations were primarily achieved by recalculating the wetness indices based on the modified surface flow paths due to, for example, drainage ditches and stream channel modifications (Fig. 2b). The re-grading of the barnyard was similarly implemented in the model by adjusting the slope and aspect of the part of the DEM corresponding to barnyard to reflect this BMP, which directed flow away from the stream. The addition of near stream exclusionary fencing effectively reduced crop and pasture land uses, and subsequently fostered the growth of a riparian buffer, which was simulated in the model by a land use change (i.e., compare Fig. 1a and 1b). Redistribution of the P applications to the landscape was primarily achieved by the NMP’s manure spreading schedule, which prioritized spreading to areas with low STP levels and low hydrological sensitivity. We used the farmerkept records of where, when, and at what rate manure was spread to direct simulated P applications to the watershed. Based on these records, the average P application rate was 78x10−4 kg m−2 yr−1

Table 1. Model parameter values estimated a priori and a posteriori. Parameter Initial water-extractable phosphorus in manure ωMF τ Dissolved P immobilization rate Reaction constant kF Plant/soil complex export coefficient μT,S

Eq. 4, 5 5 5 7

Value 6.5/6.4x10–4† 9 1.7x10−2 1–385

Unit kg m−2 d m3 kg−1 mg kg−1

TS Q10S MI Max kI kI,Q TB μT, B

Base temperature, soil Q10 base coefficient, soil Maximum dissolved P accumulation in barnyard Dissolved P exponential build up factor Dissolved P wash-off coefficient Base temperature, baseflow Reference DP baseflow export coefficient

7 7 8 8 9 11 11

20.5 1.70 12.0x10−4 1–3 0.02 17 210/55x10–5‡

°C – kg m−2 d−1 m2 m−3 °C –

Q10B

Q10 base coefficient, baseflow

11

2.50

–

Source Total measured manure P and Gérard-Marchant et al. (2005) Gérard-Marchant et al. (2005) Derived from Gérard-Marchant et al. (2005) From Morgan’s soil test P and expt’s by Hively et al. (2005) and Hively (2004) Avg. monthly air temp. during Hively et al. (2005) expt’s Rearranging and solving Eq. [7] regression Estimated from runoff expt’s by Hively et al. (2005) Estimated from runoff expt’s by Hively et al. (2005) Estimated from runoff expt’s by Hively et al. (2005) Average summer soil temperature at 60 cm (Eq. [12]) Measured baseflow DP concentration for the pre/post BMP period Rearranging and solving Eq. [11] regression

† Phosphorus contents in manure P were based on total measured manure P and Gérard-Marchant et al. (2005) for the pre-BMP situation (6.5x10−4 kg m−2) and on direct measures from the farm for post-BMPs (6.4x10−4 kg m−2). ‡ Reference baseflow export coefficient was adjusted to observed P levels in stream flows < 1 mm d−1 during the pre-BMP (210 x10−5) and post-BMP (55 x10−5) periods.

Easton et al.: Combined Monitoring & Modeling to Indicate Effective Agricultural BMPs

1803

Fig. 3. Measured and modeled stream flow at the watershed outlet (measurements were not taken from June 1995 to November 1996 while best management practices [BMP] were installed).

but varied between 57 and 115x10−4 kg m−2 yr−1. In the pre-BMP period average application rates were 3.2x10−4 kg m−2, applied on average 30 times per year following the protocols discussed earlier, e.g., manure spread near the barn in the winter but distributed more evenly on agricultural fields during the growing season.

Fig. 4. Average modeled runoff loss for (a) the pre-best management practice (BMP) period (1993–1995) and (b) the post-BMP period (1997–2004).

(r2 = 0.81, E = 0.83) and the post-BMP (1997–2004) periods (r2 = 0.78, E = 0.84) (Fig. 3). The changes in surface topology introduced by adding the drainage ditches and the re-grading of the barnyard did not appear to substantially alter the stream response at the outlet, i.e., the same Se parameter provides similarly good predictions of stream flow at the outlet for both pre- and postBMP periods. One might expect the drainage ditches to alter the outlet flow substantially; however, in this small watershed, runoff or interflow intercepted by the ditches would likely reach the outlet within the same 1-d time step used by the model as with no ditches. However the drainage ditches, the relocation of the stream channel, and to some extent the re-grading of the barnyard did change the internal hydrologic response of the watershed (Fig. 4a and 4b). In particular, the addition of the drainage ditches effectively reduced the surface runoff losses in the down gradient areas by preventing upslope interflow from saturating the soil (Fig. 4b). No other BMPs implemented in the model appeared to affect the distributed or integrated hydrology of the watershed. Similarly, the VSLF simulated DP load agreed well with the measured loads for both the pre- and post-BMP situations (Fig. 5, Table 2). However, it is not apparent from Fig. 5 that the BMPs had any impact on water quality and quantified differences in Table 2 could be arguably attributed to natural inter-annual variability in weather. In Fig. 6 and 7 we compare the alternative scenarios, no BMPs and BMPs with manure P exported for the post-BMP period, 1997–2004. Figures 6a and 7c show that, had BMPs not been implemented, DP loads would have been ~37% higher than actually observed during the post-BMP period (Table 2). Additionally, the implementation of BMPs reduced the DP load by > 70% of the maximum potential reduction, i.e., if all manure is exported (Fig. 6b and 7d, Table 2). Not surprisingly, those parts of the landscape that received no manure showed no changes in their absolute contributions of DP to the stream, regardless of whether or not BMPs were implemented (Fig. 7, Table 3). Interestingly, the DP load associated with baseflow constituted the largest fraction of the entire DP load in the absence of BMPs, generally > 40% (Tables 2 and 3), and is nearly 25% lower with BMPs. Indeed, the winter reduction in baseflow DP constitutes the largest absolute rate reduction, i.e., > 0.083 kg d−1 of any component, uniquely exceeding the overall reduction of 0.067 kg d−1, possibly a result of the reduced manure spreading and subsequent storage during the winter period. In general, implementation of BMPs resulted in reduction of P loads from manure, manured soil, and those associated with baseflows (Table 3). Figures 7a-d show the

1804

Journal of Environmental Quality • Volume 37 • September–October 2008

Testing Best Management Practices We simulated the pre- and post-BMP periods, 1993 to 1995 and 1997 to 2004, respectively, using the appropriate management conditions to assure that the model was correctly predicting stream DP loads under both sets of conditions. Then we ran the model for the post-BMP period (1997–2004) using pre-BMP management to compare pre- and post-BMP water quality under the same weather conditions. Each modeling period (pre and post-BMP) were divided into summer (1 May–31 October) and winter (1 November–30 April) for analysis and comparison. To put these results in context, we also simulated stream water quality under the hypothetical condition in which all manure is exported from the watershed to estimate the maximum possible water quality improvements over the postBMP period. Under this “all manure exported” scenario, we maintained the hydrological modification and baseflow P levels associated with the post-BMP conditions. We also still simulate manure additions from grazing cows to the pasture land use, but no manure is applied on hay or row crops.

Results The VSLF modeled event based stream flow agreed well with measured stream flow in both the pre-BMP (1993–1995)

Fig. 5. Measured and modeled event and cumulative dissolved P export at the watershed outlet for the pre-best management practice (BMP) period (1993–1995) and for the post-BMP period (1997–2004).

differences in the distributed DP loads from across the landscape under the different simulated management conditions.

Discussion Long-term monitoring in this watershed (Bishop et al., 2006) documented a 53% reduction in overall DP loading by the end of October 2005 which is very close to the 58% (± 3%) reductions predicted by the model. The model reductions are based on differences between the no- and post-BMP situations for the period 1997 to 2004 (Fig. 4 and 5, Tables 2 and 3) and the 3% uncertainty is the difference between observed and predicted loads during the post-BMP period. The real power of a combined modeling-monitoring approach as described here is the ability to identify what practices are working and which contribute little. For example, barnyard BMPs appeared to have had little relative impact, at least at the outlet (Table 3). Although these types of BMPs may have large localized impacts, apparently the barnyard, in this situation, is simply too small a contribution to the overall DP load to have a substantial impact on the overall watershed load. The barnyard contributed 0.005 kg d−1 under the worst conditions to 0.001 kg d−1 under the best, i.e., all manure removed from the watershed (Table 3), relative to total watershed loads of over two orders of magnitudes larger than these. That being said, the reduction associated with the latter scenario represents a fivefold decrease in P export from the barnyard, which results in the largest relative

reduction in P export of any model component except removing surface applied manure from the watershed (Table 3). Interestingly, this model suggested that the redistribution of manure spreading resulted in a ~35% reduction in DP contributions from manure but only ~18% reduction from manure-spread soils (Table 3– NoBMPs vs. With BMPs, 1997–2004). This result is expected, given that soil accumulation of labile P is widely observed in systems with P applied in excess of plant needs (Compton and Boone, 2000; Scott et al., 2001), and re-assuringly captured in VSLF. The model results indicate that substantial impacts are found from protecting the riparian areas and excluding cows from stream and near-stream areas (Fig. 7b and 7c), and is reflected in the “baseflow” reductions in DP loading. It is unfortunate that VSLF Table 2. Measured and modeled stream P export by time period (pre and post-BMP) and scenario (baseline, no-BMPs, and baseline BMPs plus manure P export). Measured Modeled r2‡ E§ ––––––kg d−1–––––– Pre-BMP(1993–1995) 0.220 0.217 0.81 0.77 With BMPs (1997–2004)† 0.119 0.123 0.84 0.71 No BMPs (1997–2004) 0.119 0.190 0.61 −0.06 All Manure Export (1997–2004) 0.119 0.099 0.59 0.46 † Represents baseline scenario, BMPs as implemented on the farm. ‡ Coefficient of determination. § Nash-Sutcliffe Efficiency. Period†

Fig. 6. Measured and modeled event and cumulative dissolved P export at the watershed outlet during the 1997–2004 period. Modeled results are for (a) the scenario with no best management practices (BMPs) and (b) the scenario with the BMPs and all manure P exported from the watershed.

Easton et al.: Combined Monitoring & Modeling to Indicate Effective Agricultural BMPs

1805

Fig. 7. Average landscape dissolved P loss for: (a) the pre-best management practice (BMP) period (1993–1995), (b) the postBMP (baseline) period (1997–2004) with BMPs as implemented on the farm, (c) scenario with no BMPs simulated, and (d) the scenario with the BMPs simulated as in the baseline scenario, plus all manure P exported from the watershed.

was not able to explicitly predict this reduction, i.e., re-adjusting μT,B, to reflect observed pre- and post-BMP P loads during lowflow conditions. On the other hand, because VSLF was able to

account for many of the spatially distributed loadings, it facilitated our ability to isolate the low flow contributions. We speculate that the primary benefit is due to keeping the cows out of the stream rather than the “filtering” effects of the buffer. However, preventing the direct dumping of high P content milkhouse waste into the stream (essentially a point source) would clearly further reduce the baseflow P level (Bishop et al., 2006). Unfortunately, our current understanding is too rudimentary to isolate the various potential benefits of protecting riparian areas. On the positive side, although these results suggest that more research is needed to better identify the mechanisms leading to reduced DP loads through riparian area protection, this project suggests that protecting these areas may provide the best protection of water quality, particularly in landscapes dominated by VSA hydrology. The largest discrepancies between modeled and observed results were generally due to anomalous conditions, such as those observed by Mehta et al. (2004) in an adjacent watershed. In this study, one of the most obvious occurred during the July 2002 event in which there was no modeled or measured runoff in the watershed (0.264 mm stream flow) (Fig. 3), yet 5.8 kg of DP were measured at the outlet. The observed DP load was the result of a faulty valve in the manure storage lagoon that allowed approximately 273,000 L of manure to enter the stream. Bishop et al. (2005) estimate that 6.7 kg of DP were associated with the spill and delivered to the stream over several days. These types of BMPs are particularly prone to catastrophic results when they are compromised (e.g., Mallin et al. [1999] note several instances in which swine waste lagoons were breached or inundated, discharging large quantities of waste into the Cape Fear River in North Carolina). During situations like this, BMPs may pose in-

Table 3. Phosphorus contribution from model components by period (pre and post-BMP) and scenario (baseline, no BMPs, and baseline BMPs plus manure P export), as well as the percentage of total P loss by component, and the percentage of the watershed represented by the component. Land use for the pre-BMP and no-BMP scenarios is shown in Fig. 1a. Land use for the with BMPs and with BMPs and manure P export is shown in Fig. 1b. Period Pre-BMP (1993–1995) Summer Winter % of total DP loss % of watershed With BMPs (1997–2004)† Summer Winter % of total DP loss % of watershed No BMPs (1997–2004) Summer Winter % of total DP loss % of watershed All manure exported (1997–2004) Summer Winter % of total DP loss % of watershed † Represents baseline scenario, BMPs as implemented on the farm. ‡ Represents the contribution from soil under manured landscapes.

1806

Baseflow Non-manured soil Manure Manured soil‡ Barnyard/impervious –––––––––––––––––––––––––––kg d−1––––––––––––––––––––––––––– 0.101 0.005 0.027 0.070 0.004 0.072 0.003 0.007 0.036 0.003 0.128 0.007 0.045 0.104 0.004 48.2 2.2 12.9 34.8 1.8 – 56.7 41.6 1.7 0.024 0.006 0.020 0.069 0.004 0.021 0.004 0.015 0.057 0.005 0.027 0.007 0.025 0.080 0.004 20.0 4.6 16.1 55.7 3.7 – 58.4 39.9 1.7 0.093 0.006 0.031 0.084 0.004 0.080 0.004 0.026 0.070 0.005 0.110 0.007 0.050 0.098 0.004 43.9 1.7 14.8 38.0 1.7 – 56.7 41.6 1.7 0.024 0.006 0.001 0.045 0.001 0.021 0.004 0.001 0.037 0.001 0.027 0.007 0.001 0.053 0.001 20.0 9.5 0.7 70.5 1.7 – 58.4 39.9 1.7

Journal of Environmental Quality • Volume 37 • September–October 2008

creased water quality risks. This emphasizes that models are only as good as their input information. An obvious result from this investigation is that snowmelt periods during the winter and spring, i.e., highly hydrologically active periods, contribute more than twice the summer periods. Thus, in as much as we can reduce P availability during periods when there are high risks of runoff production, we are likely to make the most substantial improvement in water quality protection. Although this study does not report data beyond a seasonal level, we speculate that developing BMPs that incorporate finer-scale predictions of runoff risk will contribute important advances in protecting water quality. Indeed, the results found here are similar to those found by Walter et al. (2001), who suggested that avoiding so-called hydrologically sensitive areas could reduce DP by ~50% of the maximum possible reduction resulting from exporting all excess manure. Another interesting result from this study is that precision feeding (implemented in 2001) appeared to have a comparatively small effect (compared to the other BMPs implemented) on improved water quality, some of which might have been due to the later implementation (2001 vs. 1995–1996 for most other BMPs). However, this was surprising given the obvious concept that only by reducing P inputs into a watershed can long-term water quality improvements be realized. In this case, reduced P inputs in the cattle feed only lead to a 4% reduction in available DP. However, this reduction may prove significant in the long term; as less P is ultimately applied, the watershed P mass balance will approach steady state, which will ultimately result in less P export. Indeed, Bishop et al. (2006) saw a continued reduction in P load from the watershed (32% reduction in summer P loads and 14% overall reduction) when compared to results from Bishop et al. (2005), and speculate that precision feeding may have contributed substantially to the reduction, as no other BMPs had been implemented since 2002. However, reduced soil P levels in areas receiving lowered, or no, manure could also contribute to this observed reduction. One final note, while NMPs consist of a suite of management practices, and most are relatively easy to follow or require little maintenance once implemented (i.e., precision feeding, riparian buffers, cattle crossings, etc.), the prescribed manure spreading can be somewhat more difficult to adhere to in as much as it requires consideration of variables such as weather, soil moisture, and land use. While the NMP, as written by a farm planner, is suitably detailed to direct spreading on certain fields at certain times, actual spreading may deviate significantly from the plan due to many factors such as time constraints, planting schedules, manure storage capacity, environmental conditions, or unpredicted wet periods. Indeed, in this case the actual manure spreading deviated as much as 300% from applications prescribed by the NPN at certain times. However, it is clear that the NMP was highly successful in reducing P loss from the farm landscape. The point being, improvements are achievable even with substantial deviation from the NMP; i.e., moving toward a better strategy can get us a long way toward our water quality objectives.

Conclusions Using the VSLF model, we demonstrated that BMPs can lead to pronounced decreases in the loss of DP from agricultural landscapes. VSLF predictions of DP loads were generally within 3% of observed results and showed a 35% improvement in water quality after BMP implementation. More importantly, a combined modeling long-term water quality monitoring approach helps give insight into what practices will have the most substantial impacts. In this study, we found that BMPs that disassociate P loading from areas prone to generating storm runoff resulted in the most profound reduction in P loading to streams. By contrast, efforts to reduce manure P content had less obvious short-term impacts. Another important contribution of modeling-based efforts is the ability to put results in context, for example, by demonstrating the magnitudes of water quality improvements in relation to some “ideal” situation. In the case of this investigation, we showed that BMPs lead to > 50% of the water quality improvements that could be expected if all excess manure was exported from the watershed, at least over the time scale considered here. This study concluded that BMPs, implemented in concert, that protect riparian areas and streams from direct pollutant loading provide the most substantial water quality protection per land taken out of production. We acknowledge that watershed scale water quality monitoring is probably the most cost effective means of determining the net BMP impact but unfortunately this provides little insight into distributed processes. Combining these data with logical hypotheses allows us to build models that highlight what we think are the most effective practices and what parts of the system we still do not understand sufficiently. Specifically, this study demonstrates how combined modeling and long-term or strategic monitoring can work together to help us understand what management options are likely to result in real and immediate water quality improvements.

Acknowledgments We would like to acknowledge helpful comments made by the reviewers, and thank our collaborators at the NYS-DEC, NYC-DEP, USGS, and USDA-ARS (Penn State) for data collection and analysis on this project.

References Allan, J.D. 1995. Stream ecology: Structure and function of running waters. Chapman and Hall, London, United Kingdom. Andraski, T.W., and L.G. Bundy. 2003. Relationships between phosphorus levels in soil and in runoff from corn production systems. J. Environ. Qual. 32:310–316. Beven, K.J., and M.J. Kirkby. 1979. A physically-based, variable contributing area model of basin hydrology. Hydrol. Sci. Bull. 24:43–69. Bishop, P.L., W.D. Hively, J.R. Stedinger, M.R. Rafferty, J.L. Lojpersberger, and J.A. Bloomfield. 2005. Multivariate analysis of paired watershed data to evaluate agricultural best management practice effects on stream water phosphorus. J. Environ. Qual. 34:1087–1101. Bishop, P.L., M.R. Rafferty, and J.L. Lojpersberger. 2003. Event based water quality monitoring to determine effectiveness of agricultural BMPs. In Proceedings of the American Water Resources Assoc., International Congress on Watershed Management for Water Supply Systems, 29 June–2 July 2003. Bishop, P.L., M.R. Rafferty, and J.L. Lojpersberger. 2006. Effectiveness of whole farm planning and implementation in achieving water quality improvement and protection of New York City water supplies: Final Report to the Watershed Agricultural Council. New York State Dep. of

Easton et al.: Combined Monitoring & Modeling to Indicate Effective Agricultural BMPs

1807

Environmental Conservation, Albany, NY. 35 p. Brannan, K.M., S. Mostaghimi, P.W. McClellan, and S. Inamdar. 2000. Animal waste BMP impacts on sediment and nutrient losses in runoff from the Owl Run watershed. Trans. ASAE 43(5):1155–1166. Brutsaert, W.H. 1982. Evaporation in the atmosphere. D. Reidel Kluwer, Boston, MA. Caruso, B.S. 2000. Spatial and temporal variability of stream phosphorus in a New Zealand high-country agricultural catchment. N. Z. J. Agric. Res. 43:235–249. Cerosaletti, P.E., D.G. Fox, and L.E. Chase. 2004. Phosphorus reduction through precision feeding of dairy cattle. J. Dairy Sci. 87:2314–2323. Compton, J.E., and R.D. Boone. 2000. Long-term impacts of agriculture on soil carbon and nitrogen in New England forests. Ecology 81(8):2314–2330. Czymmek, K.J., Q.M. Ketterings, and L.D. Geohring. 2001. Phosphorus and agriculture: VIII. The new phosphorus index for New York State. What’s Cropping Up 11(4):1–3. Czymmek, K.J., Q.M. Ketterings, and L.D. Geohring. 2002. Phosphorus index: A risk assessment tool to help evaluate farm field management and promote surface water quality. Pending, March 2002 NEDPA Conf. Proc. Czymmek, K.J., Q.M. Ketterings, L.D. Geohring, and G.L. Albrecht. 2003. The New York phosphorus runoff index: User’s manual and documentation. Crop and Soil Science Ext. Publ. E03-13. Cornell Univ., Ithaca, NY. 64 p. Daniel, T.C., A.N. Sharpley, D.R. Edwards, R. Wedepohl, and J.L. Lemunyon. 1994. Minimizing surface water eutrophication from agriculture by phosphorous management. J. Soil Water Conserv. 49:30–38. Daniel, T.C., A.N. Sharpley, and J.L. Lemunyon. 1998. Agricultural phosphorus and eutrophication: A symposium overview. J. Environ. Qual. 27:251–257. DeLaune, P.B., P.A. Moore, D.K. Carman, A.N. Sharpley, B.E. Haggard, and T.C. Daniel. 2004. Evaluation of the phosphorus source component in the phosphorus index for pastures. J. Environ. Qual. 33:2192–2200. Dillaha, T.A., R.B. Reneau, S. Mostaghimi, and D. Lee. 1989. Vegetative filter strips for agricultural nonpoint source pollution control. Trans. ASAE 32:491–496. Dillaha, T.A., J.H. Sherrard, D. Lee, S. Mostaghimi, and V.O. Shanholtz. 1988. Evaluation of vegetative filter strips as a best management practice for feedlots. J. Water Pollut. Control Fed. 60:1231–1238. Dittrich, T.M., L.D. Geohring, M.T. Walter, and T.S. Steenhuis. 2003. Revisiting buffer strip design standards for removing dissolved and particulate phosphorus. p. 527–534. In Total maximum daily load environmental regulations II conference proceedings, 8–12 Nov. 2003, Albuquerque, NM. ASAE Publ. 701P1503. ASAE, St. Joseph, MI. Dunne, T., and R.D. Black. 1970. Partial area contributions to storm runoff in a small New England watershed. Water Resour. Res. 6:1296–1311. Dunne, T., and L. Leopold. 1978. Water in environmental planning. W.H. Freeman & Co., New York. 818 p. Easton, Z.M., P. Gérard-Marchant, M.T. Walter, A.M. Petrovic, and T.S. Steenhuis. 2007. Identifying dissolved phosphorus source areas and predicting transport from an urban watershed using distributed hydrologic modeling. Water Resour. Res. 43:W11414, doi:10.1029/2006WR005697. Easton, Z.M., M.T. Walter, E.M. Schneiderman, and T.S. Steenhuis. 2008. Including source specific phosphorus mobility in a non-point source pollution model for agricultural watersheds. Environmental Engineering- ASCE (in press). Ekholm, P., K. Kallio, S. Salo, O.P. Pietilainen, S. Rekolainen, Y. Laine, and M. Joukola. 2000. Relationship between catchment characteristics and nutrient concentrations in an agricultural river system. Water Res. 34:3709–3716. Frankenberger, J.R., E.S. Brooks, M.T. Walter, M.F. Walter, and T.S. Steenhuis. 1999. A GIS-based variable source area hydrology model. Hydrol. Processes 13:805–822. Garen, D.C., and D.S. Moore. 2005. Curve number hydrology in water quality modeling: Uses, abuses, and future directions. J. Am. Water Resour. Assoc. 41:377–388. Geohring, L.D., T.S. Steenhuis, M.T. Walter, M.F. Walter, Q.M. Ketterings, and K.J. Czymmek. 2002. Phosphorus risk assessment tools for New York State. ASAE-CIGR Paper 0022071, ASAE Annual International Meeting/CIGR XVth World Congress, 28–31 July, Chicago, IL. Gérard-Marchant, P., W.D. Hively, and T.S. Steenhuis. 2006. Distributed hydrological modeling of total dissolved phosphorus transport in an agricultural landscape: Part I. Distributed runoff generation. Hydrol. Earth Syst. Sci. 10(2):245–261.

Gérard-Marchant, P., M.T. Walter, and T.S. Steenhuis. 2005. Simple models for phosphorus loss from manure in runoff. J. Environ. Qual. 34:872–876. Gitau, M.W., T.L. Veith, and W.J. Gburek. 2004. Farm-level optimization of BMP placement for cost-effective pollution reduction. Trans. ASAE 47(6):1923–1931. Grayson, R.B., I.D. Moore, and T.A. McMahon. 1992. Physically based hydrologic modeling: II. Is the concept realistic? Water Resour. Res. 28(10):2659–2666. Haith, D.A., and L.L. Shoemaker. 1987. Generalized watershed loading functions for stream flow nutrients. Water Resour. Bull. 23(3):471–478. Hamilton, P.A., T.L. Miller, and D.N. Myers. 2004. Water quality in the nation’s streams and aquifers: Overview of selected findings. USGS, Washington, DC. 1265. p. 1–19. Hansen, N.C., T.C. Daniel, A.N. Sharpley, and J.L. Lemunyon. 2002. The fate and transport of phosphorus in agricultural systems. J. Soil Water Conserv. 57:408–416. Hively, W.D. 2004. Phosphorus loading from a monitored dairy farm landscape. Ph.D. diss. Cornell Univ., Ithaca, NY. Hively, W.D., R.B. Bryant, and T.J. Fahey. 2005. Phosphorus concentrations in overland flow from diverse locations on a New York dairy farm. J. Environ. Qual. 34:1224–1233. Hively, W.D., P. Gérard-Marchant, and T.S. Steenhuis. 2006. Distributed hydrological modeling of total dissolved phosphorus transport in an agricultural landscape: Part II. Dissolved phosphorus transport. Hydrol. Earth Syst. Sci. 10:263–276. Hooda, P.S., M. Moynagh, I.F. Svoboda, A.C. Edwards, H.A. Anderson, and G. Sym. 1999. Phosphorus loss in drainflow from intensively managed grassland soils. J. Environ. Qual. 28:1235–1242. Inamdar, S.P., S. Mostaghimi, P.W. McClellan, and K.M. Brannan. 2001. BMP impacts on nutrient and sediment yields from an agricultural watershed in the coastal plain region. Trans. ASAE 44(5):1191–1200. Lammers-Helps, H., and D. Robinson. 1991. Literature review and summary pertaining to buffer strips, research sub-program of the National Soil and Water Conservation Program; final report. p. 25. Univ. of Guelph, Guelph, ON, Canada. Lee, K.Y., T.R. Fisher, T.E. Jordan, D.L. Correll, and D.E. Weller. 2000. Modeling the hydrochemistry of the Choptank River Basin using GWLF and Arc/Info: I. Model calibration and validation. Biogeochemistry 49(2):143–173. Lyon, S.W., M.R. McHale, M.T. Walter, and T.S. Steenhuis. 2006a. The impact of runoff generation mechanisms on the location of critical source areas. J. Am. Water Resour. Assoc. 42:793–804. Lyon, S., J. Seibert, A.J. Lembo, M.T. Walter, and T.S. Steenhuis. 2006b. Geostatistical investigation into the temporal evolution of spatial structure in a shallow water table. Hydrol. Earth Syst. Sci. Dis. 10:113–125. Lyon, S.W., M.T. Walter, P. Gérard-Marchant, and T.S. Steenhuis. 2004. Using a topographic index to distribute variable source area runoff predicted with the SCS curve-number equation. Hydrol. Processes 18:2757–2771. Magette, W.I., R.B. Brinsfield, R.F. Plamer, and J.D. Wood. 1989. Nutrient and sediment removal by vegetated filter strips. Trans. ASAE 32(2):663–667. Maguire, R.O., and J.T. Sims. 2002. Soil testing to predict phosphorus leaching. J. Environ. Qual. 31:1601–1609. Mallin, M.A., M.H. Posey, G.C. Shank, M.R. McIver, S.H. Ensign, and T.D. Alphin. 1999. Hurricane effects on water quality and benthos in the Cape Fear watershed: Natural and anthropogenic impacts. Ecol. Appl. 9:350–362. McDowell, R.W., and A.N. Sharpley. 2003. Phosphorus solubility and release kinetics as a function of soil test P concentrations. Geoderma 112:143–154. McDowell, R., A.N. Sharpley, and G. Folmar. 2001. Phosphorus export from an agricultural watershed: Linking source and transport mechanisms. J. Environ. Qual. 30:1587–1595. McIntosh, J.J. 1969. Bray and Morgan soil extractants modified for testing acid soils from different parent materials. Agron. J. 61:259–265. Mehta, V.K., M.T. Walter, E.S. Brooks, T.S. Steenhuis, M.F. Walter, M. Johnson, J. Boll, and D. Thongs. 2004. Application of SMR to modeling watersheds in the Catskill Mountains. Environ. Model. Assess. 9:77–89. Nash, J.E., and J.V. Sutcliffe. 1970. River flow forecasting through conceptual models: Part I. A discussion of principles. J. Hydrol. 10:282–290. National Climatic Data Center. 2005. Weather/climate events. Available at http:// www.ncdc.noaa.gov/oa/climateresearch.html (verified 12 May 2008). National Oceanic and Atmospheric Administration, Washington, DC.

1808

Journal of Environmental Quality • Volume 37 • September–October 2008

Needelman, B.A., W.J. Gburek, G.W. Petersen, A.N. Sharpley, and P.J. Kleinman. 2004. Surface runoff along two agricultural hillslopes with contrasting soils. Soil Sci. Soc. Am. J. 68:914–923. Novotny, V. 2003. Water quality: Diffuse pollution and watershed management. 2nd ed. John Wiley & Sons, New York, NY. Parry, R. 1998. Agricultural phosphorus and water quality: A U.S. Environmental Protection Agency perspective. J. Environ. Qual. 27:258–261. Puckett, L.J. 1995. Identifying the major sources of nutrient water-pollution. Environ. Sci. Technol. 29:A408–A414. Santhi, C., J.G. Arnold, R. Srinivasan, and J.R. Williams. 2003. A modeling approach to evaluate the impacts of water quality management plans implemented in the Big Cypress Creek watershed. Total maximum daily load (TMDL) Environmental Regulations-II, Proc. of the November 2003 Conf. of the ASAE, Albuquerque, NM. Saxton, K.E. 1982. SPAW model predicts crop stress. Agric. Res. 30:16–26. Schellinger, G.R., and J.C. Clausen. 1992. Vegetative filter treatment of dairy barnyard runoff in cold regions. J. Environ. Qual. 21:40–45. Schneiderman, E.M., D.C. Pierson, D.G. Lounsbury, and M.S. Zion. 2002. Modeling the hydrochemistry of the Cannonsville Watershed with Generalized Watershed Loading Functions (GWLF). J. Am. Water Resour. Assoc. 38(5):1323–1347. Schneiderman, E.M., T.S. Steenhuis, D.J. Thongs, Z.M. Easton, M.S. Zion, G.F. Mendoza, M.T. Walter, and A.L. Neal. 2007. Incorporating variable source area hydrology into the curve number based Generalized Watershed Loading Function model. Hydrol. Processes 21:3420–3430. Schwer, C.B., and J.C. Clausen. 1989. Vegetative filter treatment of dairy milkhouse waste-water. J. Environ. Qual. 18:446–451. Scott, C.A., M.F. Walter, G.N. Nagle, M.T. Walter, N.V. Sierra, and E.S. Brooks. 2001. Residual phosphorus in runoff from successional forest on abandoned agricultural land: I. Biogeochemical and hydrological processes. Biogeochemistry 55:293–309. Sharpley, A.N., P.J. Kleinman, R. McDowell, M. Gitau, and R.B. Bryant. 2002. Modeling phosphorus transport in agricultural watersheds: Process and possibilities. J. Soil Water Conserv. 57:425–439. Sharpley, A.N., P. Kleinman, and J. Weld. 2004. Assessment of best management practices to minimize the runoff of manure-borne phosphorus in the United States. N. Z. J. Agric. Res. 47:461–477. Sharpley, A.N., R.W. McDowell, J.L. Weld, and P.J.A. Kleinman. 2001. Assessing site vulnerability to phosphorus loss in an agricultural watershed. J. Environ. Qual. 30:2026–2036. Sorrano, P.A., S.L. Hubler, S.R. Carpenter, and R.C. Lathrop. 1996.

Phosphorus loads to surface waters: A simple model to account for spatial pattern of land use. Ecol. Appl. 6(3):865–878. Srinivasan, M.S., W.J. Gburek, and J.M. Hamlett. 2002. Dynamics of stormflow generation: A field study in east-central Pennsylvania, USA. Hydrol. Processes 16(3):649–665. Steenhuis, T.S., M. Winchell, J. Rossing, J.A. Zollweg, and M.F. Walter. 1995. SCS runoff equation revisited for variable-source runoff areas. J. Irrig. Drain. Eng. 121:234–238. Torrent, J., and A. Delgado. 2001. Using phosphorus concentrations in the soil solution to predict phosphorus desorption to water. J. Environ. Qual. 30:1829–1835. USDA-NRCS. 2000. Soil survey geographic (SSURGO) database for Delaware County, New York. Available at http://soildatamart.nrcs.usda. gov (verified 12 June 2008). USDA-NRCS, Washington, DC. USDA-SCS. 1972. National Engineering Handbook, Part 630 Hydrology, Section 4, Chapter 10. USDA-SCS, Washington, DC. USEPA. 2007. Summary report: National monitoring program projects: New York City watershed section 319 monitoring program project. National Watershed Studies. USEPA, Washington, DC. Walter, M.T., M. Dosskey, M. Khanna, J. Miller, M. Tomer, and J. Wiens. 2007. The science of targeting within landscapes and watersheds to improve conservation effectiveness. p. 63–91. In M. Schnepf and C. Cox (ed.) Managing agricultural landscapes for environmental quality, strengthening the scientific base. Soil and Water Conservation Society, Ankeny, IA. Walter, M.T., V.K. Mehta, A.M. Marrone, J. Boll, P. Gérard-Marchant, T.S. Steenhuis, and M.F. Walter. 2003. Simple estimation of prevalence of Hortonian flow in New York City watersheds. J. Hydrol. Eng. 8:214–218. Walter, M.T., J.Y. Parlange, M.F. Walter, X. Xin, and C.A. Scott. 2001. Modeling pollutant release from a surface source during rainfall runoff. J. Environ. Qual. 30:151–159. Walter, M.F., T.S. Steenhuis, and D.A. Haith. 1979. Nonpoint source pollution control by soil and water conservation practices. Trans. ASAE 22:834–840. Walter, M.T., M.F. Walter, E.S. Brooks, T.S. Steenhuis, J. Boll, and K.R. Weiler. 2000. Hydrologically sensitive areas: Variable source area hydrology implications for water quality risk assessment. J. Soil Water Conserv. 55(3):277–284. Zheng, X.H., Y. Huang, Y.S. Wang, M.X. Wang, J.S. Jin, and L.T. Li. 2003. Effects of soil temperature on nitric oxide emission from a typical Chinese rice-wheat rotation during the non-waterlogged period. Glob. Change Biol. 9:601–611.

Easton et al.: Combined Monitoring & Modeling to Indicate Effective Agricultural BMPs

1809