Recognizing economic value in multifunctional buffers in the lower Mississippi river basin

Integrating conservation practices with bioenergy has been recommended as a promising strategy to improve bioeconomy and water quality but the literature on the economics of this strategy is limited. This study evaluated the value proposition of reducing nutrient loss from cropland by implementing switchgrass riparian buffers in the Lower Mississippi River Basin (LMRB). Nutrient loss was simulated by using the Soil and Water Assessment Tool. The value proposition of nutrient abatement was quantified by estimating (1) value of trapped nutrients as fertilizer and (2) potential net returns of harvesting switchgrass as bioenergy feedstock at different farm‐gate prices. Results suggest that switchgrass buffers may reduce mean annual total nitrogen and total phosphorus loads from cropland in the LMRB by 23% and 31%, respectively. The value of trapped nutrients is considerable (mean = $69 ha−1 year−1) but far less than the cost of implementing a switchgrass buffer (mean = $163 ha−1 year−1). At biomass prices of $20, $40, $60, and $80 per dry‐ton, mean net returns of switchgrass buffers (without considering land‐use change from cropland to buffers) were estimated to be around −$66, $199, $463, and $727 ha−1 year−1, respectively. Total net returns for the LMRB may be reduced by 20% if switchgrass is grown without the addition of commercial fertilizer. The results highlight the potential of switchgrass buffers for improving water sustainability of both agricultural and bioenergy production. The value proposition of switchgrass buffers is nevertheless sensitive to future feedstock price. The impact of fertilizer prices change and forgone income on benefit analysis is also presented. © 2018 The Authors. Biofuels, Bioproducts, and Biorefining published by Society of Chemical Industry and John Wiley & Sons, Ltd.


Introduction
E xcessive nutrient loads from cropland in the Mississippi River Basin (MRB) are considered to be the major factor contributing to the seasonally hypoxic zone in the northern Gulf of Mexico, 1,2 which is the second largest in the world. The Gulf of Mexico Hypoxia Task Force set up an action plan in 2008 and updated its goal framework in 2015, aiming to reduce nitrogen and phosphorus loading by 20% by 2025 and to reduce the area extent of the hypoxic zone to less than 5000 km 2 by 2035. 3,4 If the 5000 km 2 goal is to be met, a 60% reduction in nutrient loads is needed. 5 Among possible nutrient abatement strategies, conservation practices 2,6,7 and improving spatial land use management have been recommended. 1,[8][9][10] For instance, the US Department of Agriculture (USDA) National Resources Conservation Service (NRCS) has promoted the adoption of conservation practices on cultivated cropland since 1985. In particular, perennial grasses, such as switchgrass, are seen as promising cellulosic biofuel feedstock. 11 Switchgrass has also been demonstrated to be a good choice for riparian buffers because of its relatively highly effective nutrient removal. [12][13][14][15] Recent developments have focused on integrating agricultural conservation practices and cellulosic biomass production through a landscape-management approach. 10,13,16 While conservation practices like riparian buffers, constructed wetlands, and bioreactors are effective at reducing nutrient lost through surface runoff, the costs associated with conservation practices are typically viewed implicitly as a financial burden to the public sector. 2,17,18 Although studies have tried to justify the costs of conservation practices by monetizing the benefits obtained through enhanced water quality and ecological services, [19][20][21] the economic benefits for in situ recovery of leached nutrients are rarely considered. When integrated with biomass production, the intercepted nutrients may be reused by energy crops, like switchgrass, to generate economic returns. 22 In recent years, integrated landscape management strategies, such as using riparian buffers with perennial grasses or woody biomass crops have been gaining increased attention 13,23-25 because they can take advantage of both the effectiveness of conservation practices in reducing nutrient loss and the economic potential of cellulosic biomass as bioenergy feedstock.
Although riparian buffers for biomass feedstock are of great interest, the literature on the economics of largescale implementation is limited. 25 Ssegane et al. 24 analyzed the costs and benefits of growing shrub willow as a bioen-ergy buffer in an agricultural watershed in central Illinois and found that a willow buffer has negative net returns if biomass is marketed at the price of $71.5/dry ton, largely because of the high land costs in the Corn Belt region. In the Chesapeake Bay region, Woodbury et al. 8 found that subsidizing switchgrass can be more cost-effective than planting cover crops for nitrogen loading reduction, but the biomass value of switchgrass was not assessed. In the MRB region, the USDA NRCS has assessed the effects of various conservation practices on cultivated cropland but not the economic costs and benefits of doing so. In recent years, a number of studies have estimated the cost effectiveness of selected common conservation practices [26][27][28] and cropland retirement 1 on nutrient abatement in the MRB. Nonetheless, few studies to date have evaluated the value proposition of reducing nutrient loss from cropland by integrating switchgrass riparian buffers with bioenergy feedstock production, especially in the LMRB. Although some studies have conducted economic analysis of switchgrass as a crop for nutrient abatement purposes, 15,29,30 establishing switchgrass as a multifunctional riparian buffer can be quite different. 23,31 For instance, understanding the economic benefits of growing biomass in riparian buffers requires a detailed analysis of factors that affect costs (e.g. establishment, biomass cultivation, and harvest) and benefits (e.g. fertilizer prices, biomass market price). Given the importance of the LMRB to the Gulf of Mexico and its prominence in future large-scale biofuel feedstock production, 11 there is a clear need to understand the economics and interactions among conservation practices, bioenergy feedstock production, and water-quality improvement.
The objective of this study is to develop a spatially explicit estimation of the water quality and economic benefits of implementing switchgrass-based riparian buffers in the LMRB. The effects of riparian buffers on nutrient load reductions were simulated by using the Soil and Water Assessment Tool (SWAT) model. 32,33 The economic benefits of switchgrass buffers were quantified by (1) measuring the value of nutrients trapped in the buffer zone as fertilizers and (2) estimating the potential economic returns of growing switchgrass as bioenergy feedstock in riparian buffers, both with and without fertilizer additions. The total economic benefits of switchgrass based riparian buffers include direct economic benefits from biomass production as well as the other benefits attributed to reduced nutrients and top soil loss, decreased sedimentation, improved water quality, habitat for native species, pollination services, reduced greenhouse gas emission, and recreational benefits. The benefits are channeled to farmers as well as other stakeholders at local and global level. Here we focus on the economic benefit to the farmers only, so the value proposition represents a lower bound estimate of total benefits.

Study area
The LMRB drains an area of nearly 253 279 km 2 and flows through seven states in the Mississippi River Basin: Arkansas, Illinois, Kentucky, Mississippi, Missouri, Tennessee, and Louisiana. It receives streamflows from four major river basins, including the Arkansas River Basin, Missouri River Basin, Upper Mississippi River Basin, and the Ohio-Tennessee River Basin. Major land-use types in the LMRB include forest (48.9%) and cultivated croplands (24.8%), followed by pasture/grassland (17.6%), urban land (4.9%), shrubland (2.5%), and water bodies (1.3%) ( Fig. 1(a)). Based on land-use and terrain characteristics, the LMRB can be divided into three major clusters: the central valley has a flat landscape (slope < 1%) and land-use type is dominated by cropland (Fig. 1). Forests and pasture/grasslands dominate the western sub-basins, and the slope is higher than 5% for most of the area in the cluster. The eastern sub-basins has a mix of cropland and forests, and slope ranges from 1% to 5% in most parts of the cluster (Fig. 1).

SWAT model for the LMRB and riparian buffer simulation
The SWAT model 32 is a continuous, physically based watershed model developed by the USDA Agriculture Research Service (ARS). The model has been widely used to evaluate the impacts of land management on water, sediment, and nutrient cycling for large river basins. 32,33 Major input data include terrain, stream network, soil properties, climate, and land-use and management practices. In this study, the LMRB boundary was delineated using eight-digit hydrologic unit codes (HUCs) and divided into 76 sub-basins and 6724 hydrologic response units (HRUs).
Model parameterization and calibration methods have been detailed in Ha et al. 34 The simulation period is 21 years from 1990 to 2010. The LMRB SWAT model was calibrated and validated for streamflow, sediment, and nutrients (total nitrogen (TN) and total phosphorous (TP)). Data source and inputs for model development are summarized in Table 1. More details on model calibration and validation can be found in supporting information (Tables S1-S4).
Switchgrass-based riparian buffers were implemented along the edges of agricultural fields adjacent to streams in the LMRB. According to USDA NRCS, the width of the riparian herbaceous buffer should range from about 9 to 37 m. 38 Thirty-meter buffers were selected because previous studies found that a 30 m buffer is more effective than a narrower one, 39 but reduction rates remain flat when the buffer width increased beyond 30 m. 13 The SWAT uses the vegetative filter strip (VFS) model to simulate the effect of riparian buffers on hydrology and nutrient loads. 40 By overlaying a land-use map with flow lines obtained from the National Hydrography Dataset, 36 buffer-implemented areas along streams were calculated by using the ArcGIS software. 41 The ratio of buffer area to total field area is about 0.1. We also assumed that the fraction of HRU area that drains to the most concentrated 10% of the buffer zone is 0.5 and that none of the concentrated flow is fully channelized, which is consistent with previous studies. 13,42 Fertilizer value of nutrient stored in riparian buffers The economic value of the nutrients trapped by riparian buffers was estimated by using market fertilizer prices at the sub-basin level (Eqn (1) where v i ($/year) represents the value of nutrients (N and P) intercepted by riparian buffers in sub-basin i; F N,i ($/kg N) and F P,i ($/kg P) are nitrogen and phosphorous fertilizer prices in sub-basin i, respectively; and N reduc,i (kg N ha −1 year −1 ) and P reduc,i (kg P ha −1 year −1 ) represent perhectare reductions in mean annual N and P loads in sub-basin i, respectively. Finally, F rb,i (ha) refers to the size (ha) of the buffer-implemented area in sub-basin i.

Economic value of growing switchgrass as bioenergy feedstock in riparian buffers
Annualized economic returns (π i ($/year)) of growing switchgrass as biomass feedstock in riparian buffers were calculated as the difference between potential revenues from biomass sales and total cost of switchgrass riparian buffer implementation at the sub-basin level (Eqn (2)): where P SWG,i represents the potential biomass price ($/dry ton) in sub-basin i; Y SWG,i represents estimated mean switchgrass yield (dry ton/ha) in sub-basin i; C RB,i ($/ha) represents establishment costs of a riparian buffer in subbasin i; M SWG,i represents maintenance costs ($ ha −1 year −1 ) for switchgrass in sub-basin i, including fertilizer and herbicides, and H SWG,i is the cost of harvesting switchgrass biomass ($ ha −1 year −1 ). Harvest cost was assumed to be $114/ha. 11 Switchgrass was assumed to have a lifespan of 10 years from the initial establishment year to the final harvest year. 11 A conversion factor (0.925) was used to adjust annualized yields because we assumed a 50% harvest in year 1, a 75% harvest in year 2, and a 100% harvest in years 3 through 10. 11 In this study, market prices (farm-gate prices) for biomass feedstock were assumed to be $20 and $40 per dry ton for low-price scenarios, and $60 and $80 per dry ton for high-biomass-price scenarios. The $40, $60, and $80 per dry ton price scenarios were used in the 2016 U.S. Billion Ton Report (BT16) 11 to cover a range of possible price scenarios for analyzing the economic availability of feedstock. In this study, we added $20 per dry ton as a low-price scenario. Fertilizer application rates for switchgrass were calculated based on the assumption that 4.54 kg N and 1.81 kg P inputs would be needed per dry ton of biomass removed. 11 Thus, N and P rates vary by sub-basins, depending on projected yields for each sub-basin (Fig. 2). The VFS module used in the current SWAT model does not support the simulation of plant growth within riparian buffers. For this reason, simulated yields for lowland switchgrass were obtained from the WATER model (http://water.es.anl.gov/). The estimation was originally based on the 2011 US Billion Ton Update report (BT2) 43 and Jager et al. 44 County-level mean lowland switchgrass yields were then aggregated to the sub-basin level by using an areal-weighting method, weighting factors were (1) calculated based on the area of the sub-basin that fell inside each county separately and then (2) aggregated to obtain area-weighted mean values at the sub-basin level. Further assumptions were that nutrient demands would be met by nutrients trapped in buffers first, and that fertilizers would be applied only if there were any remaining nutrient deficit. The fertilizer cost reduces to zero if reusing trapped nutrients is sufficient to meet switchgrass needs, or switchgrass is produced without fertilizer additions.
We calculated establishment costs both with and without the consideration of forgone income (i.e. revenue loss due to land conversion to riparian buffers) because the land  close to the riverbank may be prime land rather than marginal land in certain areas. Establishment costs of switchgrass (Table 2) were represented by grass-based riparian buffer implementation costs and annualized establishment cost was calculated by dividing total cost by 10 years, which is the lifespan of a riparian buffer. State-level riparian buffer implementation costs were obtained from the USDA NRCS Field Office Technical Guide (FOTG) ( Table 2). Establishment costs include equipment installation (chemical, seeding, tillage, and tractor), materials, labor, and mobilization. A previous study estimated that about 3% of cropland in LMRB has adopted vegetative buffers as conservation practices. 52 For economic benefit analysis, we considered potential switchgrass biomass from both existing and additional riparian buffers in cropland; buffers in other land uses (e.g. pasture) were not considered. Implementing 30 m riparian buffers in LMRB would occupy about 10% of cropland. For the case that forgone income was considered, we assumed that additional riparian buffers (7% of existing cropland) would carry the burden of forgone income as establishment costs.

Switchgrass production without fertilizer addition
Actual switchgrass yields may be lower than potential yields if fertilizers are not applied. The yield response curve might be quadratic, or a hybrid function in which the yield increases proportionally with fertilizer inputs until a plateau is reached (Eqns (3) and (4)). [53][54][55] The hybrid function is adopted in this work to estimate switchgrass yield when fertilizers are not applied. Recent field experiments in Tennessee, 53, 56 Mississippi, 57 and Oklahoma 58,59 found that the mean annual yield of switchgrass may be reduced by 22-36% if N fertilizer is not applied. As a conservative estimate, we assumed in the present study that the yield of switchgrass without any N input is 60% of the projected potential yield. As riparian buffers will provide some, if not all, N input for switchgrass through cropland runoff, estimated switchgrass yield would always be higher than 60% of projected yield regardless of whether commercial N fertilizer is added or not. We did not consider yield response to P fertilizer rates because previous studies found switchgrass is not very sensitive to P fertilization, 60,61 at least in the short term.
where Y SWG,i,NF represents estimated mean switchgrass yield (dry ton/ha) in sub-basin i when commercial fertilizer is not added; N rb,i is the average N trapped in riparian buffers (kg N ha −1 year −1 ) in in sub-basin i, simulated by  11 and it was included in switchgrass production cost rather than riparian buffer implementation cost. b Mississippi data were based on grass filer strips, as grass riparian buffer data were not available.  (Table S5) and $0.67/lb P 2 O 5 -P to $2.00/lb P 2 O 5 -P (Table S6). In this study, 2016 fertilizer prices (Table 3) were used to reflect more recent market conditions. The phosphate price was converted to the phosphorus price by a factor of 2.29. For economic analysis, sub-basins located within the state boundary share the same data. Note that for sub-basins that crossed the boundaries of more than one state, mean values at the sub-basin level were calculated using the area-weighting method.

Results and discussion
Effects of riparian buffers on reducing nutrient loads from cropland The SWAT simulated results suggest that baseline mean annual (1990-2010) TN and TP loads from cultivated cropland in the LMRB were around 98.6 billion ton (BT)/ year and 15.2 BT year −1 , respectively. At the sub-basin level, mean annual TN and TP loads ranged from 0.32 metric ton (MT) N/year to 10.42 BT N year −1 (Fig. S1(a)) and 0.09 MT P year −1 to 1.59 BT P year −1 (Fig. S1(b)), respectively. Sub-basins with higher nutrient loads are clustered in the central valley (Fig. S2), which is reasonable because most cropland is located within that area ( Fig.  1(a)). Simulation results suggest that installing switchgrass riparian buffers on the edge of cropland may reduce mean annual TN and TP loads in the LMRB by 22.85 BT N year −1 , or 23%, and 4.75 BT N year −1 , or 31%, respectively. Nutrient intercepting efficiency, measured as the percentage of reductions in nutrient loads, is lower in the central valley than neighboring sub-basins (Fig. 3). Lower trapping efficiency in the central valley can be attributed to the flat landscape (slope < 1%). According to USDA NRCS, the drainage area above the vegetative buffer or filter zone should have a slope of 1% or greater 69 for the buffer zone to achieve its designed performance. In terms of the amount of reductions in nutrient loss, sub-basins in the central and eastern LMRB presented the highest numbers (Fig. 3). At the sub-basin, reductions in mean annual TN and TP loads (Fig. 3) ranged from 0.2 MT N year −1 to 1.89 BT N year −1 and 0.07 MT P year −1 to 422.2 MT P year −1 , respectively. Although sub-basins in the western basin presented the highest percentage reductions, the amount of TN and TP load reductions were negligible (< 100 Mt/year/sub-basin) when compared to those in the central and eastern basin, largely because there is not much agricultural runoff to be filtered out in the western basin. On a monthly basis, reductions in nutrient loads were most evident in April and May for TN and for TP (Fig. S2), respectively. Monthly reductions were strongly correlated to monthly nutrient loads (Fig. S1(c) and (d)) because nutrient intercepting efficiencies did not vary much across months: between 1990 and 2010, mean monthly TN and TP intercepting efficiency in the LMRB ranged from 25.6% to 28.9% and 30.8% to 36.2%, respectively.

Value of intercepted nutrients
When measured by chemical fertilizer prices, the value of nutrients (N and P) stored in riparian buffers ranged from $11 to $260 ha −1 year −1 (mean = 69, SD = 47.1) ( Table 4). At the sub-basin level, the value of trapped nutrients ranged from $301/year to $2.18 Million (M)/year (Fig. 4), with sub-basins in the central valley presenting the highest values (Fig. 4). Between N and P, contributions from N dominated all sub-basins (Fig. 4). This is because annual T loads were five times more than annual P loads in the LMRB, while the average P fertilizer price in 2016 ($0.97/lb. P2O5-P) is only about two times higher than N fertilizer price ($0.35/lb. N) ( Table 3) . In the LMRB, nutrient loading from croplands peaks between April and May (Fig. S2), which generally matches with nutrient demand of switchgrass, since field experiments suggest that increases in switchgrass biomass are highest in May and June. 57,70 Although the value of trapped nutrients as fertilizer is considerable ($28 M year −1 for the LMRB), the costs of riparian buffer implementation are much higher. On a per-hectare basis, annualized riparian buffer implementation costs (i.e. dividing overall cost by 10 years, which is the designed lifetime of riparian buffer) ranged from $144 to $180 ha −1 year −1 (Fig. 5(a)). At the sub-basin level, total costs of riparian buffer implementation ranged from $197/year to $10.6 M year −1 (Fig. 5(b)). This large variation can largely be explained by differences in buffer area across sub-basins. For instance, riparian buffer area in the sub-basin with a cost of $197 year −1 and $10.6 M year −1 were estimated to be around 0.01 and 622.8 km 2 , respectively, in this study. The total cost of installing riparian buffers on all cropland in the LMRB would be around $101.6 M year −1 , which is more than three times higher than the value of trapped nutrients ( Table 4). The high   costs of riparian buffers suggest that preserving nutrients alone is unlikely to convince farmers to install conservation buffers on their farmland. Either additional revenues from buffers (e.g., biomass production) or incentive payments from the public sector would be needed to encourage broader participation. Given that financial resources for conservation efforts are limited, integrating bioenergy production with conservation buffers may be preferred because revenues through biomass feedstock production may offset, at least partially, the costs of riparian buffers.

Economics of switchgrass riparian buffers for bioenergy
Our analysis suggests that whether harvesting switchgrass as biomass feedstock can offset the costs of riparian buffers or not largely depends on the projected feedstock prices (Fig. 6).    (Table 4). This means substantial financial support or incentives would be needed to make up the difference in order to prompt wide adoption of this conservation practice. Still, if nutrient reduction is the major objective, switchgrass riparian buffers may still be helpful because revenues from biomass production could reduce the cost of riparian buffer implementation from $163 to $66 ha −1 year −1 , on average. Notice that the economic analysis in this section excludes forgone income. The impact of forgone income on net returns was discussed above. One of the advantages of growing switchgrass as riparian buffers on the edge of cropland is that it avoids fertilizer costs. Net returns of switchgrass production mentioned above reflected the value of trapped nutrients by deducting avoided commercial fertilizer costs from switchgrass production costs, so that value of intercepted nutrients was not double counted. According to nutrient interception simulated by SWAT and market prices for chemical fertilizers for the LMRB, our results suggest that farmers may save around $58.17 ha −1 year −1 (SD = 27.1) on average on fertilizer costs if trapped nutrients can be used by switchgrass grown in riparian buffers. For the LMRB, total avoided fertilizer costs would be around $26.2 M year −1 , which means more than 90% of the fertilizer value of intercepted nutrients ($28 M year −1 ) may be recovered. These results demonstrate that integrating perennial grasses with conservation practices could be a synergic strategy for economic development of bioenergy and water quality improvement.
Geospatially, there are significant spatial variations in both riparian buffer costs and economic returns (Fig. 5) estimated in this study. The economic returns of switchgrass buffers at the sub-basin level are highest in the central valley of the LMRB (Fig. 5(a) and (b)). Spatial variations in per-hectare riparian buffer costs (Fig. 5(a)) seem to have little impact on sub-basin-level economic returns. This is because buffer in agricultural land and therefore the biomass potential is closely related to the area of cropland and the location of stream lines in that sub-basin. For the same reason, sub-basin level net returns is less sensitive to spatial variation in switchgrass yields ( Fig.  2(a)) than differences in riparian buffer area. For instance, at a biomass price of $60 or $80 per dry ton, sub-basinlevel total economic returns would be higher than $20 or $35 M year −1 , respectively, in the central valley (Fig. 5(c) and (d)), where cropland is the dominate land-use type. In contrast, sub-basin-level economic returns outside the central valley are mostly less than $10 M year −1 .
In terms of both nutrient abatement and economic returns, simulation results suggest that implementing switchgrass riparian buffers in the central sub-basins tended to be more effective than in other sub-basins (Fig.  5). Among the 76 sub-basins in the LMRB, riparian buffers can only reduce TN loads in seven sub-basins by more than 850 Mt year −1 (Fig. 3). Specifically, five of the seven sub-basins are located in the central valley and the other two are located in the eastern basin (Fig. 3). Although sub-basins in the northern basin are also dominated by agricultural production, simulated results suggest that both nutrient loads and the amount of reductions that can be achieved by riparian buffers would be significantly lower than those for sub-basins in the central and eastern basin. For the seven sub-basins mentioned above, the total cost of riparian buffers estimated in this study was around $39.8 M year −1 , which is about 40% of total riparian buffer Figure 6. Comparison of riparian buffer (RB) implementation costs versus potential economic returns of growing switchgrass (SWG) as bioenergy feedstock in the RB zone (basinwide mean values). Per-hectare RB costs ($ ha −1 year −1 ) were calculated by dividing total establishment cost by 10 years. Potential economic returns of SWG (lowland) production were calculated assuming a marketing price of $20, $40, $60, and $80 per dry ton, respectively. Annual maintenance and harvest costs were included in net return calculations for SWG production scenarios.
Modeling and Analysis: The value proposition of multifunctional riparian buffers H Xu, M Wu, M Ha cost in the LMRB. However, when biomass production is considered, net returns from the seven sub-basins alone also constitute about 40% of total net revenues in the LMRB.

Growing switchgrass without fertilizer additions
Growing switchgrass without commercial fertilizer addition in the LMRB may reduce biomass potential moderately from 14.29 dry ton/ha (SD = 1.8) (Fig. 2(a)) to 12.79 dry ton ha −1 (SD = 2.2) (Fig. 2(b)), on average. The reduction is moderate (10%) because nutrients trapped in riparian buffers can largely meet the needs of switchgrass. At the sub-basin level, average N fertilizer needed for switchgrass production is about 59.9 kg ha −1 year −1 (range = 37.9-71.8), whereas TN intercepted in buffer zone would be around 54.3 kg ha −1 year −1 (range = 8.8-182.4). Nonetheless, our results suggest that the impacts of trapped nutrients on yields varied significantly across sub-basins. The changes are more evident in the northern and central sub-basins ( Fig. 2(b)) than in other areas, because the amount of N trapped by buffers is estimated by the SWAT model to be lower in these sub-basins on a per-hectare basis ( Fig. 7(a)). The spatial pattern of N deficit, which is calculated as the difference between required N fertilizer rates and the amount of N trapped in riparian buffers, clearly demonstrates that relying on N trapped in buffers alone is not sufficient to achieve the desired switchgrass yields in the northern and central sub-basins. In the central valley, the estimated N deficit ranged from 17.7 to 38.6 kg ha −1 year −1 for most sub-basins ( Fig. 7(b)). Estimated N deficits in the northern sub-basins were found to be substantially higher (mean = 45 kg ha −1 , SD = 7.5) than in other regions, because of the combined effects of higher switchgrass yields (Fig. 2(a)), which means higher N fertilizer demands and lower N availability from riparian buffers (Fig. 7(a)).
In terms of economic returns, growing switchgrass without fertilizer addition may still generate substantial revenues, after accounting for reductions in yields, if feedstock price is higher than $40/dry ton. At a biomass price of $20/dry ton, $40/dry ton, $60/dry ton, or $80/dry ton, we found annualized mean economic returns would be around −$55, $182, $418, or $654 ha −1 year −1 , respectively (Fig. 6). With the same biomass price scenarios, simulation results suggest that total net economic returns for the LMRB ranged from −$46 to $357.8 M year −1 ( Table  4). Compared to the switchgrass production scenarios with fertilizer applications, growing switchgrass without fertilizer additions may actually increase total net returns by 13% when feedstock price is $20/dry ton. With higher feedstock prices (i.e., $40/dry ton −$80/dry ton), total net returns for the LMRB may be reduced by up to 20% (Table 4). At the sub-basin level, changes in economic returns vary significantly. For instance, at a biomass price  (Fig. 8(a)), or from −44% to 7% (mean = −10%) ( Fig. 8(b)). Across the LMRB, reductions in net returns are most evident in the northern and central sub-basins (Fig. 8), which is consistent with the spatial pattern of yield changes mentioned above. The results suggest that, when the biomass feedstock price for switchgrass is higher than $40/dry ton, establishing low-input switchgrass production without fertilizer additions in conservation buffers could be an economically viable solution to reconcile the tradeoffs between water quality and agricultural production. However, when feedstock price is low (e.g., $20/dry ton), increased yields are unlikely to compensate for the extra cost of chemical fertilizer.

Impact of forgone income and fertilizer price variation on economic analysis
Depending on local soil and field conditions, cropland near riverbanks may be prime land rather than marginal land. If forgone income (i.e. revenue loss due to conversion of productive land to riparian buffers) is considered, then the average establishment cost would increase sig-nificantly from $163 to $683 ha −1 year −1 (SD = 126) ( Table  4). This means net returns for switchgrass production would decrease by around $520 ha −1 year −1 (SD = 118.8) on average (Table 4) under all four feedstock price scenarios. Mean net returns (with commercial fertilizer addition) under $40/dry ton and $60/dry ton scenarios would be −$321 ha −1 year −1 (SD = 110.7) and −$57 ha −1 year −1 (SD = 110.8), respectively. In this case, substantial financial incentives might be needed to encourage wide adoption of switchgrass riparian buffers. Nonetheless, if feedstock price can reach $80/dry ton, net returns would still be around $207 ha −1 year −1 (SD = 120.4). When forgone income is taken into account, we estimated LMRB total net returns (with fertilizer additions) for the four price scenarios (i.e. $20, $40, $60, $80) to be $−384.9, $−222.9, $−60.9 , and $ 101.1 M year −1 , respectively (Table 4).
Uncertainty due to fertilizer prices change is another important factor to consider. Using fertilizer price data between 2004 and 2016, we calculated 95% confidence intervals of N and P fertilizer prices for each state (Tables  S5 and S6). At the LMRB level, upper (95%) and lower bounds (5%) of historical N fertilizer prices are $0.42/lb N and $0.53/lb N (Table S5); 95% confidence intervals for P fertilizer prices are $0.89/lb P2O5-P (lower bound) and $1.28/lb P2O5-P (upper bound) (Table S6). We found that  (Fig. 9). This is because fertilizer prices in 2016 were more or less the lowest since 2004 (Tables S5 and S6). Among the seven states contributing to the LMRB, increase in the value of trapped nutrients (with upper bound prices) ranges from $13.3 ha −1 year −1 in Illinois to $54.9 ha −1 year −1 in Tennessee (Fig. 9). The increase is lower for sub-basins within Illinois because the amount of nutrients intercepted by riparian buffers is much lower than other sub-basins (Fig. 7), owing to their flat terrain (the slope is close to zero). Across the subbasins, the average value of intercepted nutrients (with upper bound prices) will increase from $69 ha −1 year −1 (SD=47.1) to $104 ha −1 year −1 (SD=71.6). The LMRB-level total nutrient value will increase from $28 to $42 M year −1 .
When the lower bounds of fertilizer prices were used, perhectare and LMRB total nutrient value were estimated to be around $75 ha −1 year −1 (SD=49.6) and $31 M year −1 ( Table 4) in this study, respectively.

Limitations and future work
Using switchgrass-based riparian buffers as an example, the analysis presented in this study quantified the water quality and economic benefits obtainable through integrating cellulosic biomass production with conservation practices. The economic value of buffer-trapped nutrients can be estimated on the basis of fertilizer prices, although fertilizer prices vary significantly across years. The estimations of net returns of growing switchgrass for bioenergy purposes are also highly uncertain. At this time, observed data on large-scale switchgrass production are not available and estimated biomass production costs vary substantially in the literature. 11,[71][72][73] Depending on the assumptions made on biomass price and production cost, estimated economic returns could be quite different. For instance, mobilization ($367/ha) is a major part of the riparian buffer implementation cost. Mobilization cost covers transportation of equipment, and this cost may be lower if it can be spread among large areas, assuming the same equipment can be shared in adjacent fields. Furthermore, the value proposition was based on possible farm-gate feedstock prices 11 but off-farm factors (e.g., transportation costs and capacities of biorefineries) are also important factors to be considered. 30 The actual costs of biomass from riparian buffers could also be higher than those from dedicated large biomass farms because the logistics of managing small amounts of biomass generated in a distributed way across the landscape are also more complex than production in contiguous large farms. 24 For future studies, the analysis can be improved in a number of ways. For instance, in addition to the economic benefits obtained through in situ nutrient recycling, improvements in water quality can also reduce water treatment costs downstream. Including savings in municipal water treatment costs would be helpful for a more complete value proposition analysis. Furthermore, switchgrass yield was estimated by using an off-line model. Improvements in the VFS model, which is used by SWAT to simulate riparian buffers, are desirable to enable the dynamic simulation of nutrient cycles and plant growth in the riparian buffer zone.

Conclusion
Combining conservation practices with bioenergy feedstock productions in the MRB to address the reoccurring hypoxia problem in the Gulf of Mexico is of growing interest. However, studies on the economics of the largescale implementation of this integrated land-management strategy are rather limited. In this study, we evaluated the value proposition of switchgrass-based riparian buffers in the LMRB. The value proposition of implementing switchgrass riparian buffers in the LMRB was quantified by estimating (1) the value of nutrients trapped in riparian buffers as fertilizers, and (2) the potential economic returns of producing switchgrass as biofuel feedstocks in riparian buffers, both with and without fertilizer addi-  (Tables S5 and  S6)  . If forgone income is considered, net returns will decrease by $520 ha −1 year −1 (SD = 118.8). Spatially, subbasin-level analysis demonstrated that future conservation planning should target fields located in the central and eastern sub-basins because fields in these areas offered the highest potential for both nutrient reductions and economic returns. Growing switchgrass without fertilizer additions result in moderate reduction of yield (< 20%) in the agriculture region in LMRB because most of nutrient demand of switchgrass can be met by trapped nutrients. In fact, utilizing trapped nutrients to grow switchgrass buffer may save $26.2 M year −1 in fertilizer cost. Our results suggest that when there is a mature market for cellulosic biomass and the feedstock price is higher than $40 ha −1 year −1 , integrating perennial bioenergy crops into agricultural conservation practices might be an economically viable strategy to reconcile the tradeoffs among energy, agriculture, and the environment. However, if biomass price is kept at a low level (e.g., $20 ha −1 year −1 ), it is unlikely that switchgrass riparian buffers will be profitable, and financial support or incentives might be required to encourage more farmers to adopt switchgrass riparian buffers on their cropland. When forgone income is considered, biomass price needs to reach a high level ($80/ dry ton) for switchgrass riparian buffers to be profitable. Because the estimated economic value from switchgrass based riparian buffer does not include all economic benefits (e.g. water quality improvement), the value proposition presented in this study is the lower bound estimate. Evaluation of the other benefits may change the scenario to a higher benefit-to-cost ratio. The spatially explicit value proposition presented in this study can help decision makers to take the value proposition of nutrient reductions achieved by integrated land management strategies, such as switchgrass riparian buffers, into consideration when planning for water sustainable agricultural and bioenergy industry development.
Modeling and Analysis: The value proposition of multifunctional riparian buffers H Xu, M Wu, M Ha

Hui Xu
Dr Hui Xu is a postdoctoral appointee in the Energy Systems Division at Argonne National Laboratory, USA. His research interests include modeling the environmental sustainability of bioenergy feedstock production, integrated spatial land management, and lifecycle analysis of biofuels. His research integrates computational modeling (e.g., watershed modeling) with geospatial and economic analysis to identify more efficient and sustainable natural resource-management strategies. He holds a PhD in natural resources and environment from the University of Michigan, Ann Arbor.

May Wu
Dr May Wu is a principal environmental system analyst in the Energy Systems Division at Argonne National Laboratory, USA. As a principal investigator of a multiyear bioenergy sustainability project supported by the US Department of Energy, Dr Wu's research addresses water resource use and availability and water quality by developing biofuel water footprints and SWAT models for Mississippi River Basin tributaries. She is a member of the Global Bioenergy Partnership (GBEP) Bioenergy and Water working group and has extensive experience in water and wastewater treatment and analysis. Dr Wu holds several US patents, has authored 50+ publications, and holds a dual PhD in environmental engineering and environmental toxicology from Michigan State University.

Miae Ha
Dr Miae Ha is an assistant hydrologist in the Energy Systems Division at Argonne National Laboratory, USA. She has extensive experience in watershed modeling, which examines water quality and hydrology in the agricultural landscape. Her research interests include development of hydrologic models at multiple scales, integrated land use and management, and simulation of best management practices (BMPs) for sustainable bioenergy production under various climate scenarios. She holds a PhD in water management and hydrological science from Texas A&M University.