Quantifying regional effects of best management practices on nutrient losses from agricultural lands

Results and Discussion

While data sets representative of multiple specific BMPs were tested during model specification as described, each regional model was limited in its final form. For the Midwest N model and the Southeast P models, testing demonstrated that the BMP data sets were representative of other collinear soil conservation BMPs. For the Southeast N model and both Northeast models, the variable testing confirmed that distinct processes were represented with each of the BMPs. For purposes of this study in this discussion, we use the term BMP for broad practices that are surrogate for multiple activities and we reserve the name of the specific BMP (such as cover crops or no-till) to models that are representative of the specific activity.

Nitrogen. Increasing the intensity of BMPs in the Midwest by 50% from 2012 levels increased agricultural N loads by a range of 3.24% to 4.73% among HUC2s, and overall N loads by a range of 1.52% to 3.14% (figure 3). The Ohio River basin (HUC 05) had the largest percentage change in agricultural N loads and the Upper Mississippi River basin (HUC 07) had the largest overall change in overall N loads. The increase in BMPs resulted in an average increase in agricultural N yield of 402 kg km^–2. When BMP intensity was decreased by 50%, agricultural N loads decreased by a range of 4.74% to 7.31% among HUC2s, and overall N loads decreased by a range of 2.28% to 4.77%. The largest change in agricultural N loads and overall N loads occurred in the Missouri River basin (HUC 10). Base FWMC of N in Midwest HUC2s ranged from 1.243 to 4.466 mg L^–1 (table 3). A 50% increase in BMPS increased N concentrations by 0.019 to 0.140 mg L^–1, and a 50% decrease in BMPs decreased N concentrations by 0.028 to 0.213 mg L^–1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Regional Spatially Referenced Regression On Watershed-attributes (SPARROW) model results for nitrogen (N) loads from applied increases and decreases of best management practice (BMP) implementation. (a) N load percentage of change from agricultural sources with a 50% increase in BMP intensity; (b) N load percentage of change from all nutrient sources with a 50% increase in BMP intensity; (c) N load percentage of change from agricultural sources with a 50% decrease in BMP intensity; (d) N load percentage of change from all nutrient sources with a 50% decrease in BMP intensity. Shaded regions of the maps represent HUC2 basins (Seaber et al. 1987) used in this study.

Alternate model specifications could not be tested in the N model because it included all of the BMPs tested in this study. Because the Northeast N model has two BMPs with different coefficient signs, results from BMP changes reflect the net effect from changes in both BMP variables (figure 3). The two HUC2s in the Northeast region (New England–HUC 01 and Mid-Atlantic–HUC 02) had similar patterns of N-load change with BMP-intensity change, with slightly larger magnitude for New England. On average for the Northeast, increasing BMPs by 50% increased agricultural N loads from 1.78% to 1.89% and overall N loads from 0.11% to 0.67%, and increased agricultural N yield by 302 kg km^–2. When calculated for cover crops and conservation tillage + no-till individually, increasing cover crops by 50% decreased agricultural N yield by 790 kg km^–2, and increasing conservation tillage + no-till by 50% increased agricultural N yield by 532 kg km^–2. Decreasing BMPs by 50% decreased agricultural N load from 2.97% to 3.12%, and overall N load from 0.19% to 1.12%. Base FWMC of N in Northeast HUC2s were 0.663 mg L^–1 and 1.654 mg L^–1 (table 3). A 50% increase in both BMPs increased N concentrations by 0.001 to 0.011 mg L^–1, and a 50% decrease in BMPs decreased N concentrations by 0.001 to 0.019 mg L^–1.

Increasing cover crop intensity by 50% in the Southeast N model (the South Atlantic-Gulf Region-HUC 03) decreased agricultural N load by 3.5% and decreased overall N load by 0.89% (figure 3). A 50% increase in cover crops decreased agricultural N yields by 329 kg km^–2. Decreasing the amount of cover crops in the region resulted in a 6.51% increase in agricultural N loads and a 1.66% increase in overall N loads. The base FWMC of N in the Southeast was 0.319 mg L^–1 (table 3). A 50% increase in cover crops decreased the N concentration by 0.011 mg L^–1, and a 50% decrease in cover crops increased the N concentration by 0.021 mg L^–1.

The simulations predict decreased agricultural N yields with cover crops with some regional variation: the effect predicted for agricultural fields in the Northeast is twice that predicted for the Southeast. There are several possible explanations for the regional variation that include seasonal influences. First, in colder climates representative of the Northeast we expect a longer season for cover crop implementation. This would require additional fertilization to produce these crops, and depending on nutrient management a surplus of inputs could dominate the impact of the BMP. In addition, as average temperatures increase in the north-south direction, we could also see an increase in N uptake and denitrification, which favors lower N yields in the Southeast.

Phosphorus. In the Midwest P model, increasing BMPs by 50% from 2012 conditions resulted in a 3.97% to 8.77% decrease in agricultural P loads and a 2.63% to 4.38% decrease in overall P loads (figure 4). The agricultural load decreased most in the Great Lakes Region (HUC 04), and the overall P load decreased most in the Upper Mississippi Region (HUC 07). A 50% increase in BMP intensity resulted in a 706 kg km^–2 decrease in agricultural P yields. When BMP intensity was decreased by 50%, there was a 7.59% to 14.2% increase in agricultural P loads and a 3.4% to 6.44% increase in overall P loads. The base FWMC in Midwest HUC2s ranged from 0.135 to 0.525 mg L^–1 (table 3). A 50% increase in BMPs decreased P concentrations by 0.005 to 0.020 mg L^–1, and a 50% decrease in BMPs increased P concentrations by 0.009 to 0.028 mg L^–1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Regional Spatially Referenced Regression On Watershed-attributes (SPARROW) model results for phosphorus (P) loads from applied increases and decreases of best management practice (BMP) implementation. (a) P load percentages of change from agricultural sources with a 50% increase in BMP intensity; (b) P load percentages of change from all nutrient sources with a 50% increase in BMP intensity; (c) P load percentages of change from agricultural sources with a 50% decrease in BMP intensity; (d) P load percentages of change from all nutrient sources with a 50% decrease in BMP intensity. Shaded regions of the maps represent HUC2 basins (Seaber et al. 1987) used in this study.

In the Northeast P model, increasing conservation tillage + no-till intensity by 50% resulted in a decrease in the agricultural P load of about 14% for both Northeast HUC2s and by a range of 2.23% to 3.57% for the overall P load (figure 4). Increasing conservation tillage + no-till intensity from base 2012 levels resulted in a 544 kg km^–2 reduction in agricultural P yields. When conservation tillage + no-till intensity was decreased by 50% the agricultural P load increased by approximately 30% in both HUC2s, and the overall P load increased by a range of from 4.7% to 7.6% in the region (figure 4). Base FWMC in the Northeast HUC2s were 0.050 mg L^–1 and 0.141 mg L^–1 (table 3). A 50% increase in conservation tillage + no-till decreased P concentrations by 0.001 mg L^–1 and 0.005 mg L^–1, and a 50% decrease increased P concentrations by 0.002 mg L^–1 and 0.011 mg L^–1.

In the Southeast P model, increasing BMP intensity by 50% resulted in a decrease in the agricultural P load of 4.8% and 1.5% for the overall P load (figure 4). Increasing BMP intensity from base 2012 levels resulted in a 26 kg km^–2 decrease in agricultural P yields. When BMP intensity was decreased by 50%, the agricultural P load increased by 8.9% and the overall P load increased 2.8%. Base FWMC for P in the Southeast HUC2 was 0.048 mg L^–1 (table 3). A 50% increase in BMPs decreased P concentrations by 0.002 mg L^–1, and a 50% decrease in BMPs increased P concentrations by 0.004 mg L^–1.

The coefficients of the BMP variables in each model demonstrate the effects of enhancing (positive coefficient) or diminishing (negative coefficient) delivery of nutrients caused by practices applied in the watershed. Cover crops in the Southeast and Northeast N models had similar negative coefficient values, which indicated N loads decreased where cover crops were implemented. Based on these results, the changes in N loads were similar as the application rate of cover crops increased in the two regions. The diminishing effect of cover crops on N loads agreed with the findings of many investigators (Hanrahan et al. 2018; Lee et al. 2016; Yeo et al. 2014). The Northeast N model also included reduced tillage practices (conservation tillage + no-till) with a positive coefficient; the enhancing effects of reduced tillage practices on N loads agreed with findings of Baker (1993) and Alberts and Spomer (1985). The combined effect of increasing intensity of both cover crops and reduced tillage practices in the Northeast was to increase N loss to streams. The effects of reduced tillage on N loads were nearly twice as large in the Northeast as compared to no-till practices in the Midwest.

The coefficients of reduced tillage and cover crops were negative in P models for all areas, suggesting these BMPs effectively reduced P losses. Reduced tillage was a significant variable in the Southeast model; however, cover crops also contribute to reducing P loads. The coefficient for cover crops in the Midwest P model indicated that cover crops and reduced tillage practices result in decreased P loss to streams. This agrees with findings by Wang et al. (2015) and Sharpley et al. (2000).

Results from SPARROW simulations with varying BMP intensity indicate that both cover crops and reduced tillage practices reduced P losses from agricultural areas, and reduced tillage practices increased N losses. In all three regional P models, BMPs (both cover crops and reduced tillage) reduced P export from the watersheds. Both of these BMPs are designed to reduce soil erosion and therefore suggest that erosion control has a significant effect on the fraction of instream P loads associated with soil loss. Although particulate P comprises a major part of the P loads, reducing the dissolved fraction is also integral to reducing P losses. Nevertheless, the advantages of using soil conservation measures to reduce P losses as predicted in this study are consistent with findings from other studies (Stevens et al. 2009; Richardson and King 1995; Andraski et al. 1985). In the Northeast, the change in agricultural P load caused by the increase in BMPs was nearly twice what was predicted in the other regions. The larger effect of the BMPs may have been caused by a larger area of application rather than it being more effective in the Northeast, since the magnitude of the coefficient of the BMP was similar to that in the Midwest.

In mitigating the P loss from agricultural fields, careful design and implementation of conservation strategies must also account for dissolved forms of P. Studies conducted by many investigators have concluded that runoff and subsurface flow are major transport vectors of dissolved P (Heathwaite and Dils 2000; Andraski et al. 1985; Sharpley et al. 1987). Tools for assessing the potential for P loss from agricultural areas are critical to the development of effective BMPs (e.g., P-index [Sharpley et al. 2001]) because they help to identify the processes that pose the greatest risk to water quality. In runoff from agricultural fields where P fertilizers have been applied, many studies have demonstrated the increased risk of P loss (Duncan et al 2017; Vadas et al. 2008; Nash et al. 2004). Accumulation of P from historic fertilizer and manure use form pools of legacy P that further the risk to water quality. In a recent review of P dynamics in the Northeast, Kleinman et al. (2019) describe legacy P as generally being more labile than P associated with the soil because it can be easily dissolved. Cover crops and reduced tillage that control erosion to mitigate P losses may be rendered less effective if a substantial part of P losses from an area are from legacy P. Furthermore, we note that changes in hydrology play a critical role in P transport (Kleinman et al. 2011) and water management such as tile-drainage can change the impact of BMPs by altering preferential flow pathways. The Midwest model accounted for some degree of water management impact with the inclusion of a tile-drainage variable. Future studies could consider other similar water management activities such as irrigation and open channel drainage in the Southeast.

Over the past decade, cover crops have gained support as an effective management strategy to reduce nutrient and soil losses from agricultural fields (Arbuckle and Roesch-McNally 2015; Hively et al. 2009; Sainju and Singh 1997). Historically, the use of cover crops lagged behind the use of other conservation practices, e.g., reduced tillage, due to economic drivers and other considerations such as managing seeding and cover crop termination (Roesch-McNally et al. 2018; O’Connell et al. 2015). The results of this study provide evidence that the gains achieved with cover crop adoption can play an important role in mediating P and N loss to streams, which is in agreement with the findings of other investigators (Shelton et al. 2018; Kladivko et al. 2014; Sharpley et al. 1991). A survey of farmers across the United States (2012 Agricultural Census) showed conservation tillage and no-till were adopted on 90% more land than cover crops as of 2012 (USDA NASS 2014). However, based on the 2017 Agricultural Census (USDA NASS 2019), the use of cover crops has outpaced conservation tillage and no-till by 41%. Although the 2007 Agricultural Census reported cover crop area as a lumped value, which included idle cropland and soil-improvement land, that area also increased over the survey period (USDA NASS 2009). Growing adoption of cover crops can largely be attributed to scientific and technological advances as well as advances that have come from efforts to improve farmers’ understanding of the advantages and disadvantages associated with adopting cover crops: improved understanding, for example, that cover crops are more efficient at reducing N losses than conservation tillage in some areas, or that interactions between different conservation practices must be accounted for in nutrient management plans (Dunn et al. 2016; Arbuckle and Roesch-McNally 2015).

Quantifying Regional Effects of Best Management Practices on Nutrient Loading to Lake Erie. In 2011, the International Joint Commission (IJC) identified nutrient contamination as a leading factor in the development of nearshore eutrophication and the reemergence of harmful algal blooms (HABs) in Western Lake Erie (in the Midwest SPARROW model) (Reutter et al. 2011). Many studies have identified P as the limiting nutrient in Lake Erie (Wilhelm et al. 2003; DeBruyn et al. 2004; Guildford et al. 2005) and a key driver of lake eutrophication (Vollenweider 1968). Eutrophication in Lake Erie in the past two to three decades has raised concerns about the occurrence of HABs and how best to forecast their occurrence and size using estimates of P loading (Obenour et al. 2014). The 1972 Great Lakes Water Quality Agreement was implemented with the target of reducing point source P loading by 50% by 1982 (IJC 1986). In 2012, the IJC updated the P target to a 40% reduction (186 t) from 2008 levels (IJC 2012).

In the BMP simulation results presented in this paper for the Midwest, increasing BMP intensity increased N loads and decreased P loads delivered to Western Lake Erie. Specifically, if the amount of no-till in the Midwest N model were to increase by 50%, an additional 2.62 × 10⁶ kg of N could be delivered to Western Lake Erie, which represents a 3.61% increase (table 4). Adding 50% more cover crops in the Midwest P model reduces the amount of P delivered to the lake by 1.90 × 10⁵ kg, which represents a 6.00% reduction (table 4). These results corroborate the conclusions from other modeling approaches (Scavia et al. 2016) that although some reduction in agricultural P and N loads are achievable with conservation BMPs, like the ones tested in this model, these measures alone will not be successful in reaching the nutrient targets to restore the impacted ecosystems. In 2006, 94% of the agricultural land in the Great Lakes had some form of conservation or nutrient-management practice in place to reduce nutrient losses (USDA NRCS 2011). In an assessment of nutrient control scenarios using multiple models (Soil and Water Assessment Tool [SWAT] [Arnold et al. 2012] and SPARROW), the most effective management scenarios for the reduction of nutrient loads included similar BMPs to those tested in this study (the ones tested in this study are vegetative BMPs), land use conversion, and nutrient management (Scavia et al. 2016).

The USDA-CEAP (USDA NRCS 2011) conducted a model-based regional assessment of BMPs for the baseline period from 2003 to 2006 in the Lake Erie Basin. The authors of the study concluded that conservation practices accounted for N and P load reductions of 17% and 18%, respectively, relative to not having implemented any conservation practices. When the authors simulated N and P delivery to rivers and streams as a result of increasing conservation treatment to critically undertreated areas (1.15 × 10⁴ km²) and to all under-treated areas (3.20 × 10⁴ km²), N loads were reduced by an additional 4% to 14%, respectively, and P loads by an additional 3% to 20%, respectively (table 4). Because of differences in the methodologies used in the CEAP study and the study that is being presented here, comparisons of the two studies are only qualitative. For P, this study and the CEAP study agree that increasing the implementation of BMPs has a positive effect on water quality, albeit falling short of the proposed target reduction, but the studies do not agree on effects for N. We speculate that BMPs do not have a similar effect on N loads in the Midwest in large part due to the characterization of the BMPs on the landscape. The Midwest model domain was the largest of the three models used in this study and with respect to BMPs may have the most spatial variability in BMP implementation. A regionalized approach to the characterization of BMPs in the Midwest may present opportunities for improvements in the specification of BMPs in the model. The characterization of BMPs in the SPARROW models are the best representation relative to the available data used to construct the models, but they (along with other similar regional water quality models) are not able to predict the influence of all BMPs on the landscape.

Quantifying Regional Effects of Best Management Practices on Nutrient Loading to the Chesapeake Bay. In the Chesapeake Bay watershed (in the Northeast SPARROW model), the Total Maximum Daily Load (TMDL) established in 2010 targeted 25% and 24% reductions in annual N and P loads, respectively (USEPA 2010). This equated to annual load limits of 84.3 million kg y^–1 for N and 5.69 million kg y^–1 for P (Linker et al. 2013). Based on the results of this study, increased intensity of BMPs in the Northeast increases N loads and decreases P loads. If the amount of BMPs in the Northeast N model were to increase by 50%, the Chesapeake Bay could receive an additional 1.13 × 10⁶ kg of N, which represents an increase of approximately 0.91% (table 4). The Atlantic Ocean could receive an additional 1.63 × 10⁶ kg of N from the Northeast United States, which represents an increase of 0.56% (table 4). For the Northeast P model, the P load delivered to the Chesapeake Bay could decrease by 5.06 × 10⁵ kg, a 6.29% reduction, as a result of a 50% increase in BMPs (table 4). The P load delivered to the Atlantic Ocean could decrease by 8.10 × 10⁵ kg, which represents a 3.42% reduction (table 4).

We compare the results presented in this study with those from a CEAP regional watershed study of the Chesapeake Bay (USDA NRCS 2013). From 2003 to 2011, the percentage of agricultural area in the Chesapeake Bay watershed with conservation tillage increased 23%, no-till by 16%, and cover crops by 40% (USDA NRCS 2013). In the same study, SWAT model simulations were used to assess the impact of conservation practices on N and P loads delivered to the Chesapeake Bay. The study concluded that conservation practices in 2011 reduced N and P loading by 17% and 21% respectively, when compared to conditions where no conservation practices were implemented. The authors also compared N and P loads delivered to the bay for the level of conservation implementation during the baseline years 2003 to 2006 with conditions in 2011. Relative to the baseline conditions, N and P loads decreased by 6% and 5%, respectively (table 4). The SPARROW models in this study cannot be used to produce simulations of the no conservation practice scenario, for comparison with the SWAT model simulations, due to lack of robustness and reliability (described in the Materials and Methods section) of SPARROW predictions when simulations of the BMP variables approach zero. Qualitative comparison of the results from this study to the conditions in 2011 presented in the CEAP study is possible, however, and shows there are some points of general agreement. The studies agree that increasing the implementation of cover crops is an effective conservation practice for controlling N delivery to the Chesapeake Bay; however, the studies disagree on the effects of tillage on N delivery. The studies agree that increasing tillage conservation practices reduces P delivery to the Chesapeake Bay. Differences in the methodologies between the two studies explain discrepancies in the magnitude and nature of the potential changes in the delivered loads that could come from more aggressive adoption of BMPs.

Results from the Northeast SPARROW models present a complicated set of circumstances to be navigated by managers because of the contrasting effects of cover crops and reduced tillage BMPs on N loads, and the contrasting effects of reduced tillage BMPs on N versus P loads. These contrasting effects would be consistent, however, with the adoption of more cover crops to mitigate N, and additional reduced tillage to mitigate P loads. The results of both studies demonstrate that the potential changes in N and P delivery to the Chesapeake Bay fall short of the reduction targets outlined by the USEPA (2010) based on the amount of cropland area where these conservation practices could be implemented. The challenges associated with controlling nutrient pollution in the Chesapeake Bay is further exacerbated by fluxes of nonagricultural nutrient sources. The results of this study suggest that limiting only agricultural nutrient sources have little influence on the overall nutrient loads delivered to the Chesapeake Bay. In the Chesapeake Bay, fluxes from wastewater treatment, urban runoff, and septic systems contribute the largest fraction of N and P to the bay (Ator 2019). Therefore, achieving nutrient targets will require aggressive actions, which target both agricultural and nonagricultural sources of nutrients.

Quantifying Regional Effects of Best Management Practices on Nutrient Loading to the Gulf of Mexico. The Gulf Hypoxia Taskforce was created to address water quality issues in the Gulf of Mexico and throughout its watershed. The taskforce identified agricultural nutrient sources as a significant driver in the size of the hypoxic zone in the Gulf of Mexico. The use of agricultural BMPs was identified as an integral part of the strategy to reduce nutrient inputs to the Gulf by 30% of historic levels (Mississippi River/Gulf of Mexico Watershed Nutrient Task Force 2001). In their 2017 report (USEPA 2018), the taskforce set a goal of a 20% reduction of nutrients in the Gulf by 2025 and full realization of the nutrient reduction target by 2035. In this study N and P loads delivered to the Gulf of Mexico were simulated by the Midwest and Southeast models. The Midwest model accounted for N and P inputs from the Mississippi-Atchafalaya River Basin and the Southeast model accounted for inputs along the Gulf Coast. As stated in the previous section, increasing BMPs by 50% in the Midwest N model increased loads, and in the P model loads decreased. Using the Midwest models for the Mississippi-Atchafalaya River watershed, increasing BMPs 50% could result in the delivery of an additional 42.0 × 10⁶ kg of N to the Gulf, which is a 2.73% increase (table 4), and a decrease of 6.8 × 10⁶ kg of P, which is a 3.58% decrease (table 4). Using the Southeast models for the direct drainages along the Gulf Coast, increasing BMPs 50% could result in a 4.73 × 10⁴ kg decrease of N delivered to the Gulf, which represents a 0.27% decrease (table 4). The amount of P delivered to the Gulf from the direct drainages could decrease by 7.62 × 10⁴ kg, which represents a 3.22% decrease (table 4). If the model results for the two regions (Midwest and Southeast) are combined, the amount of N delivered to the Gulf could increase by 2.46% and the amount of P delivered to the Gulf could decrease by 6.8%.

In an earlier study of BMPs in the Upper Mississippi River basin, N reductions attributed to BMPs ranged from 5% to 34%, and P reductions ranged from 1% to 10% (Garcia et al. 2016). According to a CEAP study that used SWAT models to estimate load changes from increased BMP intensity between 2003 to 2006 and 2011, the N load delivered to the Gulf of Mexico should have decreased by 11% and the P load should have decreased by 16% (table 4) (USDA NRCS 2014). As noted in discussions of the other CEAP studies earlier in this section, differences in estimates of the potential load reductions between the CEAP studies and this study are expected due to different frameworks and time periods, and more detailed information describing BMP adoption rates and a wider variety of BMPs included in the CEAP studies. Nevertheless, these comparisons are meaningful because they provide additional insights with respect to the trajectory of conservation activities toward reaching nutrient reduction targets. The results of this study and the other studies described in this section demonstrate the complexity in assessing the active roles of BMPs in the transport of N and P at large spatial scales, interactions between multiple BMPs, and the net effects of upstream practices on N and P delivery to major drainages. The consensus of all the studies is that additional actions are needed, beyond the conservation practices considered here, to reach nutrient targets in all regions. These actions could include nutrient management plans, structural BMPs, more aggressive (than 50%) increases in cover crops and tillage practices, or reducing fertilizer use. In addition to expanding and improving current conservation practices and technologies, it will also be important to understand the economic implications of this growth. Because of the changing global economy, increasing urbanization, and growing industrialization in emerging economies, aligning the goals of nutrient management within the context of these societal and economic factors will only become more important. The impacts of climate change on nutrient management will also be an important consideration in the management of water resources and the development of nutrient targets. Studies have demonstrated the potential for increasing nutrient availability (Feuchtmayr et al. 2009; Frey et al. 2007; Jones 2005) resulting from increasing global temperatures. Extreme storm events driven by climate variability are projected to increase runoff (Wang et al. 2018; Rotz et al. 2016) containing larger nutrient concentrations. The variable nature of climate change will not influence water quality uniformly across a range of spatial and temporal scales. Nevertheless, the findings of these studies emphasize the importance of more aggressive implementation of effective strategies to mitigate negative environmental and economic outcomes in the future.

Limitations in Best Management Practice Analysis. In practice, implementation of conservation practices typically involves adopting multiple BMPs simultaneously to increase the potential to reduce nutrient losses (Wade et al. 2015). This study focuses on a subset of vegetative practices, which is a simplification of what was actually implemented in the catchment areas. Improving our understanding of BMPs and our assessments of their effectiveness should consider all conservation practices used in developing regional-scale nutrient reduction targets to reduce nutrient pollution. Accounting for input reduction practices that are part of nutrient management programs is difficult because these types of conservation actions do not typically change land cover as is the case with vegetative BMPs, rather they attempt to reduce nutrient losses by reducing the mass of nutrients applied to the surface of soils. Meeting the goals of nutrient management plans is most often the responsibility of farmers who must determine the most feasible practices to implement on their farms. Accounting for each individual farm-scale management practice is a challenge for large-scale regional models because the availability of these data is based on individual reporting requirements, which may vary. Since they do not trigger large-scale changes, the effects of local-scale management actions are not captured unless it is assumed that their influence is inherent to the observations used in model calibration. Therefore, many large-scale models, including SPARROW, assume that source data inputs include the effects of large-scale nutrient management practices, which include fertilizer application reduction plans.

The present analysis is also limited in the representation of more dynamic practices and the contributions of legacy nutrient masses or contributions from sediment bound N and P that is trapped in soils, stored in stream channels, or dissolved in groundwater. The nutrient source data used in the SPARROW models represent a specific base year. Because of the static nature of the SPARROW models used in the study, simulating transient effects of BMPs was outside of the scope of this study. As a result, it should not be assumed that the models represent equilibrium conditions in which the full effects of source control BMPs or structural BMPs are realized. Assessing the lag between the implementation of BMPs and temporal changes in nutrient loads is best done using dynamic constructs of the SPARROW model or by using multiple static SPARROW models representing different points in time (Ator et al. 2019).

This analysis of the regional effects of BMPs is a novel application of the SPARROW model. Assessments of this type may be valuable to state and federal agencies in the development of effective nutrient management strategies. In subsequent BMP assessments using the SPARROW model, increasing the quantity and quality of BMP data would help to improve future models and their usefulness as decision support tools. Having an accurate accounting of catchment areas where specific vegetative or structural BMPs are adopted is an important example for several reasons, such as assessing the return on investment in programs that incentivize BMP adoption. Facilitated by advancements in remote sensing, the ability to accurately quantify BMPs has greatly improved, but opportunities remain to further advance the quality and quantity of BMP data. Improving the resolution of BMP data beyond the county level data used in this study represents a significant advancement. County level BMP data are very coarse and can introduce additional model uncertainty when these data are extrapolated to the catchment level. Apportionment of the amount of BMP area within a catchment relative to county totals is complicated because catchments generally do not adhere to political boundaries and distributing BMP area to catchments by some weighting methodology based on county area may oversimplify and homogenize the distribution of BMPs on the landscape, which potentially dilutes the effects of BMPs with smaller footprints. This represents an important limitation of the models used in this study, but currently this approach represents the best viable option to conduct these types of assessments with SPARROW.