Simulating and evaluating best management practices for integrated landscape management scenarios in biofuel feedstock production

Sound crop and land management strategies can maintain land productivity and improve the environmental sustainability of agricultural crop and feedstock production. This study evaluates a strategy of incorporating landscape design and management concepts into bioenergy feedstock production. It examines the effect of land conversion and agricultural best management practices (BMPs) on water quality (nutrients and suspended sediments) and hydrology. The strategy was applied to the watershed of the South Fork Iowa River in Iowa, where the focus was on converting low‐productivity land to provide cellulosic biomass and implementing riparian buffers. The Soil and Water Assessment Tool (SWAT) was employed to simulate the impact at watershed and sub‐basin scales. The study compared the representation of buffers by using trapping efficiency and area ratio methods in SWAT. Landscape design and management scenarios were developed to quantify water quality under (i) current land use, (ii) partial land conversion to switchgrass, and (iii) riparian buffer implementation. Results show that implementation of vegetative barriers and riparian buffer can trap the loss of total nitrogen, total phosphorus, and sediment significantly. The effect increases with the increase of buffer area coverage. Implementing riparian buffer at 30 m width is able to produce 4 million liters of biofuels. When low‐productivity land (15.2% of total watershed land area) is converted to grow switchgrass, suspended sediment, total nitrogen, total phosphorus, and nitrate loadings are reduced by 69.3%, 55.5%, 46.1%, and 13.4%, respectively. Results highlight the significant role of lower‐productivity land and buffers in cellulosic biomass and provide insights into the design of an integrated landscape with a conservation buffer for future bioenergy feedstock production. Published 2015. This article is a U.S. Government work and is in the public domain in the USA. Biofuels, Bioproducts and Biorefining published by Society of Industrial Chemistry and John Wiley & Sons Ltd.

growth, and agricultural inputs and management practices. SWAT simulates buff ers by using two methods. Th e fi rst, the trapping effi ciency method, uses the width of buff ers as a key parameter, where trapping effi ciency is calculated by the reduction of sediment and nutrient loadings transported by in-surface runoff through buff ers. 27 Th e second method, vegetative fi lter strip (VFS), which was developed by Munoz-Carpena, 28 was derived from 22 published studies. Th e VFS calculates buff er effi ciency on the basis of the ratio of total land area to buff er area (area ratio method). 21 Th e trapping effi ciency method allows for convenient application with only one parameter (buff er width), whereas users have more fl exibility with the ratio method, in which runoff concentrations and fractions of water fl ow concentrations can be defi ned as inputs. Previous studies investigated impacts of diff erent buff er widths from 9 to 50 m 29-35 on water quality in agricultural land. A buff er width of greater than 50 m was also considered. 36 Switchgrass and Bermuda grass were oft en selected as buff er crops. 35,37 Most studies showed positive improvement in water quality with the application of a buff er. However, literature on the methodology selection and evaluation of buff ers (vegetative barriers and riparian buff ers) at the watershed scale is limited.
Th e purpose of this study is to (i) evaluate the representation of buff ers in SWAT modeling; (ii) quantify the impacts of vegetative barriers and riparian buff ers on water quality in future biofuel feedstock production scenarios, including land conversion to switchgrass; and (iii) study the application of a riparian buff er. Th e study area is located at the South Fork Iowa River (SFIR) watershed in central Iowa, which covers 80 029 ha, and is predominately agricultural land (Fig. 1). Major crops include corn and soybeans. Th e study further compares the model resolution, assumptions, and application range at the sub-basin scale. Results highlight the capabilities and limitations of characterizing buff er areas on agricultural land and provide insights into integrated landscape design and conservation management for future bioenergy feedstock production.

Introduction
S ince 1985, the US Department of Agriculture (USDA) National Resources Conservation Service (NRCS) has been recommending the development of conservation measures in agricultural landscapes. 1 In the Midwest, conservation is particularly important because nitrate and phosphorus runoff from corn and soybean fi elds potentially aff ects downstream water quality. Historically, corn and soybean crops have been the dominant conventional feedstocks for biofuel production. In recent biofuel feedstock development, researchers began to investigate how we can mitigate potential negative impacts by incorporating cellulosic feedstock production and conservation practices. 2 Conceptually, integrated landscape design and management can serve the purpose of maintaining land productivity while protecting water resources. Th is approach -a combination of land conversion to grow energy crops 3-6 and implementation of best management practices (BMPs) for conservation practices -enables cellulosic biomass production while benefi ting soil and water quality. Switchgrass, a high-energy herbaceous perennial, is regarded as a promising candidate for the production of second-generation biofuels because of its attributes for conservation. Switchgrass has long been selected for use in natural resource conservation programs for its eff ect on the control of soil runoff and nutrient leaching, 7 limiting negative soil property changes, 8 improving soil organic carbon content, 9 and increasing biodiversity. 10,11 One option for switchgrass application is to convert lowproductivity land or idle land to switchgrass farms.
Th e NRCS 12 defi nes conservation buff ers as small areas or strips of land with permanent vegetation, designed to intercept such pollutants as nitrogen, phosphorus, pesticides, pathogens, herbicides, and sediments before they enter water bodies. Conservation buff ers include riparian buff ers, vegetative barriers, and others. A riparian buff er is set adjacent to a stream, whereas vegetative barriers are typically placed in fi eld borders. Buff ers have been applied to a considerable extent in agricultural lands, and water quality improvement has been demonstrated at various scales. [13][14][15][16][17][18][19][20] Th is study focuses on vegetative barriers and riparian buff ers.
Land and hydrologic models have been developed to describe land use and buff ers. SWAT 21 is a physically based, watershed-scale simulation model for assessing hydrologic properties and water quality associated with land cover and land use. 22 Nutrients (nitrogen and phosphorus) and suspended sediments from runoff fl ow and lateral fl ow from the sub-basin and watershed outlet were analyzed. Alamo switchgrass was selected as the buff er species for planting in the riparian buff er. It also grows as a biomass feedstock where the land is converted from idle land and low-productivity land.

Input data
Th e Digital Elevation Model (DEM) with a 30 m resolution was obtained from the National Hydrography dataset (NHD, http://nhd.usgs.gov/). Th e HRUs were defi ned for 5% land use, 10% soil class, and 10% slope over sub-basin areas. A land use map was created from remote sensing data, which were collected on an annual basis from the Crop Data Layer (CDL) database as geographic information system (GIS) raster format fi les (these data are available at http://nassgeodata.  . Th e area ratio method is modeled at the HRU level in SWAT (Supporting information B). Ratios of buff er-implemented area to total fi eld area from 1:10, 1:30, 1:40, 1:60, and 1:100 were evaluated for vegetative barriers, which are equivalent to 10%, 3.33%, 2.5%, 1.67%, and 1% of the agricultural area. For the riparian buff er, additional steps were taken to simulate the buff er area separately from the total fi eld area. Other parameters include 0.5 for the fraction of the HRU that drains to the most concentrated 10% of the fi lter strip area and 0 for the fraction of the fl ow within the most concentrated 10% of the fi lter strip that is fully channeled. An agricultural area with a slope of 2-5% was selected for applying vegetative barriers. Th e riparian buff er areas were calculated by ArcGIS. Th e current SWAT version has limitations on the riparian features. For enhanced deposition associated with riparian buff ers from upstream areas, the following values for the stable stage of the stream channel were assigned in SWAT: 0.2 for channel cover factor (CH_COV2) 37 and 0.1 (natural streams, heavy timber, and brush) for Manning's n value (CH_N2) in the main channel.

SWAT model calibration and validation
Calibration techniques for this study included manual calibration based on sensitivity analysis, auto-calibration by using SWAT calibration, and uncertainty procedures groups (e.g. corn-corn-corn-corn [CCCC], corn-soybeancorn-soybean [CSCS], soybean-corn-soybean-corn [SCSC]). Other signifi cant land uses include a low-density residential area (10.6%) and pasture (8.5%). Management operations for tile drainage and manure application were adapted by Green et al. 39 Th e soil properties were obtained from the Soil Survey Geographic Database (SSURGO) soil database. Climate data (precipitation and maximum and minimum temperatures) from 2000 to 2009 were obtained from the NOAA's National Climate Data Center (NCDC, http://www.ncdc.noaa.gov/ cdo-web/datasets#GHCND). Other weather parameters, such as relative humidity, wind speed, and solar radiation, were generated by the SWAT weather generator. Figure 1 presents the land elevation of SFIR. Figure 2 shows the land-use map (a), weather stations, and USGS gauging stations (b) in the study area.

Partial land conversion to switchgrass
Th e scenario is a proposed landscape design with cornsoybean rotations and newly established switchgrass where low-productivity land and idle land are converted to switchgrass. A proposed landscape design scenario was developed by integrating production of agricultural crops and energy crops (Ian et al. 38 ). Th e scenario covers current and future land use and resulting changes in corn, soybean, pasture, forest, idle land, and switchgrass.

Simulating buffers -trapping effi ciency and area ratio method
In this study, two diff erent empirically based methods (trapping effi ciency and area ratio) in SWAT were applied  Changes in water quality indicators based on the area ratio method are presented in Fig. 3(b). Similarly, sediments had the most pronounced responses to an increase in vegetative barrier area from 1% (1:100) to 10% (1:10), which is represented by its reduction from 17.5% to 32.4%. Th e reduction of organic P (from 10.8% to 18.0%) was similar to reductions in organic N (9.2-17%). Total phosphorus decreased by 9.5-15.7%. Reduction of total nitrogen ranged from 7.0% up to 12.8% and was primarily aff ected by minimal reductions in nitrate (0.9-1.0%). Sahu and Gu 35 reported that vegetative barriers with switchgrass covering 10% to 50% of the sub-basin area resulted in a higher reduction (55-90%) in NO 3 in the Walnut Creek watershed, IA, than is shown in the results from this study. A vegetative barrier was most eff ective in reducing sediment loadings in the watershed (followed by organic N and organic P) and less eff ective in nitrate control in the SFIR watershed.

Effect of riparian buffer
To understand the computation process and be better able to interpret model simulation results, we developed detailed riparian buff er measurements according to the watershed land maps across the entire stream network. As indicated in Fig. 4, implementing riparian buff er reduced the loss of all of the water quality indicators (suspended sediment, organic N, organic P, total N, total P, and nitrate) evaluated. An increase in buff er width (from 30 m to 50 m) resulted in higher reduction rates (Fig. 4). Among them, organic P and total phosphorus showed the largest reduction in a 30 m buff er. At a 50 m riparian buff er, approximately 6% of the nitrogen, phosphorus, and sediments loss can be avoided. Again, nitrate reduction was the lowest -3% by using trapping method (Fig. 4). Note that at the watershed scale, the area ratio of 30 m riparian (CUP), as well as calibrated values from previous studies. 39,[40][41][42][43] Th e SWAT parameters selected for calibration, descriptions of the parameters, and calibrated values are shown in Table S1. Observed and simulated monthly stream fl ows and nitrate loadings are shown in Figs S1(a) and S1(b). Monthly fl ows have seasonal trends, with a peak in May (for most years), in both simulated and observed values. Nitrate loadings show a pattern similar to stream fl ows, with peaks during the growing season. Th e model performance was evaluated by using the coeffi cient of determination (R 2 ), Nash-Sutcliff e effi ciency (NSE), 44 percent bias (PBIAS), and ratio of the root mean square error to the standard deviation of measured data (RSR) ( Table S2). 45 A SWAT model is considered satisfactory if monthly NSE > 0.5, RSR ≤ 0.70, and PBIAS < ±25 for stream fl ow and ±25 ≤ PBIAS < ±40 for nitrogen and phosphorus. For this study, the performance between observed and simulated stream fl ow was generally good or satisfactory for the calibration and validation periods (NSE 0.60-0.67; R 2 0.72-0.85). Th e nitrate-loading performance was good or satisfactory (NSE 0.70-0.71, R 2 0.76-0.80) (Table S2).

Effect of vegetative barrier
Vegetative barriers were evaluated by using the trapping effi ciency and ratio methods. Figure 3(a) shows percent of reduction in suspended sediment yield (SS), organic nitrogen (ORGN), organic phosphorus (ORGP), total nitrogen (TN), total phosphorus (TP), and nitrate (NO 3 ) in the SFIR watershed with diff erent buff er widths determined by using a trapping effi ciency method. Vegetative barriers are most eff ective in control sediments. At a 30 m buff er, sediments decreased signifi cantly by 70.7%, 68.0% for organic N, 63.7% for organic P, 55.5% for total N, and 59.0% for total P. Water quality indicators N, P, and SS are sensitive to initial placement of buff er. Even with a 5-m buff er width, reduction of sediments (42.2%), organic nitrogen (39.9%), organic phosphorus (38.6%), total nitrogen (30%), and total phosphorus (35.2%) could be achieved. Th e extent of reduction is not linear to the buff er width. With the same increment of buff er width (i.e., 5 m), the reductions were more pronounced with a change from zero to 5 m than with a change from 5 m to 10 m. For example, the reduction of sediments was 52% with a 10 m vegetative barrier, which is a small improvement compare with 42% when buff er width changed from zero to 5 m (Fig. 3(a)). Using the same modeling method, Moriasi varied. Th e trapping effi ciency method is noticeably more sensitive to the installment of a riparian buff er at 30 m, whereas the ratio method is quick to respond to a change in buff er width to 50 m. Reduction of nitrate loss increased from 1.3% (30 m) to 3.3% (50 m) when using the trapping effi ciency method, compared with 0.3% (30 m) and 1.8% (50 m) by using the ratio method-a sixfold increase. Th is study showed both methods can bring comparable riparian buff er representation when consistent assumptions are used in modeling. Th e trapping method tends to give slightly optimistic results at a 50 m buff er width. buff er to total watershed land is about 1.9%. With the similar area ratio, vegetative barriers could reduce sediment and nutrient loss by 6-17% (Fig. 3(b), ratio 1:100 [1% area]) and thus would be the preferred choice for this watershed.
When buff ers were applied to riparian areas, similar results were found for the trapping effi ciency and ratio methods with a 30 m buff er (Fig. 4). Th e two methods responded similarly to an increase in buff er width in the riparian area. At the 50 m level, the trapping method yielded slightly higher reductions, although the difference between the two methods is relatively small. Th e projection for nitrate loadings by the two methods

Land area and biomass production in riparian buffer
Switchgrass was grown on the riparian buff ers and could be harvested as feedstock for biofuel production. Figure 6 presents a switchgrass biomass yield map, riparian application areas, and total biomass production at the sub-basin level in the SFIR watershed. Switchgrass yield varies from 15.3 to 23.7 tonnes/ha across the sub-basins ( Fig. 6(a)). Th e location of a riparian buff er depends on the stream network (Fig. 2). At the sub-basin level, buff er areas vary from 0 to 176.6 ha (Fig. 6(b)) when a 30 m riparian buff er was

Scenario comparisons -impact of landscape change on water quality and quantity
To further understand the extent of the benefi ts of land conversion and buff er to water quality, three landscape scenarios were evaluated: current land use (Scenario 1), land conversion to switchgrass (Scenario 2), and riparian buff er application (Scenario 3). Figure 7 shows a land-use map for Scenario 2, where partial land conversion emphasizes a change across agriculture, pasture, forest, urban, and switchgrass in the SFIR watershed. Th is scenario describes a switch of land use from current ( Fig. 8(a)) to an integrated agriculture and biofuel production landscape ( Fig. 8(b)), with a signifi cant increase in switchgrass area (15.2% of the total land area in the watershed). Total agriculture land decreased by 14.2%, of which 6.1% is from continuous corn and 8.1% is from other types of corn/soybean rotations. Changes in urban area, forest, and pasture are minimal (<1%) in Scenario 2 (Fig. 8). With Scenario 3, at 30 m of width, the riparian buff er accounts for 2.4% of total agricultural land and 1.9% of the total land in the watershed. Th e 2.4% buff er area is mostly from agricultural land. Cropland conversion to grow switchgrass and riparian buff er implementation can signifi cantly improve water quality in the SFIR watershed (Fig. 9). Both future scenarios (2 and 3) lead to a reduction of nutrient and sediment loss. When a riparian buff er with a 30 m width was installed (Scenario 3), the reductions at the watershed scale are 1.6% for suspended sediment, 1.3% for nitrate loadings, 1.2% for total nitrogen, and 2.4% for total phosphorus. Th e primary reason for this result is that less land (1.9%) is converted to buff er as compared with the land conversion rate under Scenario 2 (15.2%). Th e reduction of nutrients and sediments is also heterogeneous at the sub-basin installed. On the basis of biomass yield at each sub-basin, a total of 26 074 metric tonnes of biomass can be available from the watershed. Assuming a 70% harvesting rate and a 20% loss during collection, as well as a conversion yield of 80 gal per dry short tonne, this amount of biomass collection translates to 4 million liters (1 059 708 gal) of biofuel production. Similarly, with a 50 m riparian buff er, a total of 48 122 metric tons of switchgrass can be harvested, which contributes to 7.4 million liters (1 955 738 gal) of biofuel production.
For the 30 m buff er, 15.1 km 2 of land are required, and for the 50 m buff er, 25.3 km 2 of land are required, which equate to 1.9% and 3.2% of the total land area in the watershed, respectively. Th ese fi gures translate to 2.4-4.0% of total agricultural croplands. Nonetheless, the stream network is heterogeneous across the sub-basins, and thus the distribution of the riparian buff er land varies substantially. Two sub-basins -35 and 21 -produced the highest amount of biomass ( Fig. 6(c)), although their yields of switchgrass are markedly diff erent ( Fig. 6(a)). Sub-basin 21 is at the high end of the entire watershed, whereas Sub-basin 35 is at the low end. To produce the same amount of biomass, Sub-basin 35 requires more land area as riparian buff er than does Sub-basin 21. Results showed that 177 ha of land were dedicated to riparian buff er in Sub-basin 35 -the largest buff er area among sub-basins in the SFIR, equivalent to 11.7% of the total riparian buff er in the watershed. Sub-basin 21 used 125 ha of buff er area, representing 8.3% of total buff er area in SFIR. Each sub-basin is able to produce about 10% of the biomass as a result of riparian buff er application in the watershed. Th e top three sub-basins with top water quality performance -12, 17, and 27 (sub-basin level analysis) -contribute less to total biomass production because of their relatively small land area and modest switchgrass yield (Fig. 6).  showed the highest reductions watershed-wide, followed by total nitrogen, total phosphorus, and then nitrate. Th e degree of reduction varies substantially across the subbasins. With few exceptions in some sub-basins where nutrients and sediments may increase, the reduction ranged from 9.6% to 81.7% for suspended sediment, from 1.9% to 32.2% for nitrate, 9.1% to 72.5% for total nitrogen, and 1.3% to 66.1% for total phosphorus at the sub-basin level (Fig. 9).
To analyze the eff ect of future scenarios on nutrient distribution, we further analyzed nitrate surface runoff and lateral fl ow under the three scenarios. Nitrate loadings at the watershed outlet accumulated from surface runoff and lateral fl ow are summarized in Table 1. Noticeably, the nitrate loadings in surface runoff stream decreased by 0.223 kg/ha and by 0.055 kg/ha when Scenarios 2 and 3 were applied, respectively. Th e amount of reduction in runoff nitrate is equivalent to 46% according to Scenario 2 relative to current land use (Table 1). On the contrary, changes in nitrate loading in lateral fl ow were minimal in Scenario 2 (0.8%). Th ere were no changes in nitrate in groundwater under both scenarios. Scenario 2 reduced 13.4% of nitrate from the total of surface runoff , lateral fl ow, and ground water. Scenario 3 had similar and lowerlevel reductions (1.3%) of nitrate at a fraction (~10%) of the land conversion in Scenario 2. With an increase of land coverage by riparian buff er, a higher level of nitrate reduction can be expected. Results suggest land conversion to switchgrass represented by Scenario 2 can be most eff ective to curtail nitrate in surface runoff . Furthermore, a riparian buff er can have a positive impact by reducing nitrate loss in both surface runoff and lateral fl ow. level: from 0% to 55.3% for suspended sediment, 0.3% to 14.4% for nitrate loading, 0.4% to 41.2% for total nitrogen, and 0.5% to 45.1% for total phosphorus (Fig. 9). Th ere is a unique distribution pattern of changes in nutrient and sediments across the watershed in Scenario 3. Suspended sediment, nitrate, total nitrogen, and total phosphorus had more reductions in the upstream region than in the downstream region (Fig. 9). Total phosphorus decreased across the watershed, whereas sediments, nitrate, and total nitrogen experienced an increase in a few sub-basins despite a decrease in a majority of the sub-basins.
Th e reduction of watershed loadings was more pronounced with Scenario 2. Converting a selected portion of cropland to switchgrass (Scenario 2) results in signifi cant reductions in suspended sediments (69.3%), nitrate loadings (13.4%), total nitrogen (55.5%), and total phosphorus (46.1%) in the SFIR watershed. Suspended sediments

Conclusion
Th e SWAT model was applied to evaluate buff ers and integrated landscape management for agriculture and biofuel production by quantifying their impacts on stream fl ow, suspended sediment, and nutrients in the SFIR watershed. Both approaches can eff ectively mitigate nitrogen, phosphorus, and suspended sediment loadings for surface streams and runoff s. Nitrate reduction is less extensive. A vegetative barrier is most eff ective in reducing sediments loadings in the watershed, followed by organic N and organic P, and less eff ective in nitrate control in the SFIR watershed. A riparian buff er showed a similar level of nitrate reduction and is most eff ective in removing phosphorus, followed by nitrate and sediments. Th e magnitude of eff ect increases with an increase in the land area covered by switchgrass, as a buff er or as dedicated biomass farmland. SWAT represents a riparian buff er well with both the trapping effi ciency and area ratio methods when consistent assumptions are applied. Under a landscaping design (Scenario 2) in which switchgrass was grown in low-productivity land, suspended sediment, total nitrogen, and total phosphorus were reduced signifi cantly. Installation of a riparian buff er with switchgrass for the entire watershed could have a similar eff ect, whereas water yield (resource) remains unchanged. While land conversion to switchgrass farming is able to capture a signifi cant amount of nitrate in runoff , implementation of a riparian buff er can trap a similar level of nitrate from both lateral fl ow and surface runoff in the watershed studied. Moreover, a 30 m riparian buff er with switchgrass yields 26,074 metric tonnes of biomass feedstock per year, which translates to 4 million liters of biofuel production. Th e study highlights key approaches in integrated biomass and agriculture development. A careful landscape design and management of current agricultural lands while incorporating appropriate BMPs can eff ectively improve water quality and strengthen soil erosion control, with minimal impact on water resources, while producing food, feed, and feedstock for bioenergy and bioproducts. Th e concept can be integrated with other watershed management programs to enhance sustainability of land, water, and the ecosystem.

Acknowledgments
Th is work was funded by the US Department of Energy, Offi ce of Energy Effi ciency and Renewable Energy, Bioenergy Technologies Offi ce (BETO). Th e authors would like to thank Ian Bonner, Kara Caff erty, and Jacob Jacobson from Idaho National Laboratory for developing the land use change scenarios. We also thank Kristen Johnson of