Riparian buffer effectiveness as a function of buffer design and input loads

Abstract

Although many agricultural watersheds rely heavily on riparian buffer adoption to meet water quality goals, design and management constraints in current policies create adoption barriers. Based on focus group feedback, we developed a flexible buffer design paradigm that varies buffer width, vegetation, and harvesting. Sixteen years of daily-scale nutrient and sediment loads simulated with the Soil and Water Assessment Tool (SWAT) were coupled to the three-zone Riparian Ecosystem Management Model (REMM) to compare the effectiveness of traditional, policy-based buffer designs with designs that are more flexible and integrate features important to local farmers. Buffer designs included (i) 10 m grass, (ii) 15 m grass, (iii) 15 m deciduous trees, (iv) 30 m grass and trees, (v) 30 m grass and trees with trees harvested every 3 yr, and (vi) 30 m grass and trees with grass harvested every year. Allowing harvesting in one zone of the buffer vegetation (either trees or grasses) minimally affected water quality, with annual average percent reductions differing by <5% (p > .05; 76–78% for total nitrogen [TN], 51–55% for total phosphorus [TP], and 68% for sediment). Under the highest input loading conditions, buffers with lower removal efficiencies removed more total mass than did buffers with high removal efficiencies. Thus, by focusing on mass reduction in addition to percent reduction, watershed-wide buffer implementation may be better targeted to TN, TP, and sediment reduced. These findings have important implications for informing flexible buffer design policies and enhanced placement of buffers in watersheds impaired by nutrient and sediment.

Abbreviations

BMP

best management practices

CREP

Conservation Reserve Enhancement Program

REMM

Riparian Ecosystem Management Model

SWAT

Soil and Water Assessment Tool

TMDL

total maximum daily load

TN

total nitrogen

TP

total phosphorus.

1 INTRODUCTION

Agricultural nonpoint source pollution is a global problem that causes tremendous environmental and economic damage (Mateo-Sagasta, Zadeh, Turral, & Burke, 2017). Excessive nitrogen (N) and phosphorus (P) losses from agricultural fields are major contributors to eutrophication in coastal systems around the world (Howarth, 2008; King et al., 2015; Smith, 2003; Withers, Neal, Jarvie, & Doody, 2014). In 2008, 415 eutrophic coastal areas were identified worldwide, with only 13 in recovery due to less intensive agricultural activities and better industrial and wastewater controls (Selman, Greenhalgh, Diaz, & Sugg, 2008). Sediment load from agricultural lands into waterbodies contributes to water quality degradation by increasing turbidity and reducing the amount of sunlight reaching aquatic vegetation. Overall, about 30% of the total global land area is degraded, and the annual cost of this degradation to agricultural lands is about US$300 billion (Nkonya et al., 2016). In the European Union, 12.5% of agricultural lands are estimated to experience moderate to high erosion, classified at more than 5 t ha⁻¹ yr⁻¹ (European Commission, 2018).

The Chesapeake Bay watershed in the Mid-Atlantic region of the United States is home to approximately 18 million people and provides enormous economic value to many industries, such as fishing, recreation, and tourism. However, the Bay suffers from eutrophication and water quality impairment as a result of excess nutrients (N and P species) and sediment (Boesch, Brinsfield, & Magnien, 2010). In 2010, the USEPA issued mandatory water quality improvement goals through the Chesapeake Bay total maximum daily load (TMDL) (USEPA, 2010). The mandate requires reduction of N, P, and sediment loads to the Bay from various contributing sectors, including agriculture. Of the six states within the Bay watershed, Pennsylvania, as the largest contributor of both nutrient and sediment loads, has fallen behind other states in meeting required milestones and remains especially off track toward meeting the load reduction goals for the agriculture and urban/suburban sectors (USEPA, 2018). The USEPA urged Pennsylvania to focus on the implementation of agricultural conservation plans, such as nutrient management plans and structural and nonstructural agricultural best management practices (BMPs), and to report findings and results from the actions taken.

To be compliant with the TMDL goals, state and local governments are required to implement BMPs that mitigate the adverse environmental impacts of agricultural activities. A riparian buffer, also known as a streamside buffer, is a commonly adopted BMP used to reduce nutrients and sediment loads reaching surface water bodies from adjacent agricultural fields. Riparian buffers can reduce sediment load transported by surface runoff by increasing surface roughness and thus slowing the water movement and increasing the amount of sediment and sediment-bound P trapped (Lee, Isenhart, & Schultz, 2003; Lee, Isenhart, Schultz, & Mickelson, 2000; Vought, Pinay, Fuglsang, & Ruffinoni, 1995). Riparian buffers act as an N sink by promoting plant uptake and increasing denitrification (Hill, 1996; Mayer, Reynolds, McCutchen, & Canfield, 2007).

Due to the significant environmental and ecological benefits that buffers can provide, Pennsylvania has proposed to meet ∼25% of its remaining N reduction goals and more than 80% of its remaining P reduction goals using forest and grass buffers (Pennsylvania Department of Environmental Protection, 2019). However, the current programs from which farmers and landowners receive monetary compensation for implementing riparian buffers on their properties, such as the Conservation Reserve Enhancement Program (CREP), provide limited options in buffer design and management. A survey conducted in central Pennsylvania revealed that riparian landowners would be more likely to adopt a buffer if they had a say in the buffer design (Brooks et al., 2011). By developing economic models to estimate the most cost-effective buffer contract designs, Li, Zipp, Shortle, Gall, and Jiang (2019) concluded that flexibility in buffer design (vegetation type and width), management practices (e.g., harvesting), contract lengths, and penalties for early termination can increase buffer adoption and enhance environmental benefits. Landowners in agriculturally dominated landscapes in the midwestern United States also suggested that flexible buffer designs should include harvest options contributing to various market and nonmarket goods (Schultz, Isenhart, Simpkins, & Collett, 2004). Therefore, the adoption rate of riparian buffers is likely to increase if incentive programs allowed for more flexible buffer management. These constraints have been recognized by the Pennsylvania Department of Conservation and Natural Resources, which has created a multifunctional buffer program to help increase adoption to the state's goal of 385 km² of riparian buffers by 2025 (Pennsylvania Department of Conservation and Natural Resources, 2019).

Core Ideas

• Grass buffers outperform tree buffers in trapping nutrients and sediment.
• Buffer zone harvesting, which may promote adoption, does not affect load reduction.
• Prioritizing buffer reduction rate does not infer maximized pollutant mass reduction.

There is some concern that policy changes that increase the flexibility of buffer design management options may increase the uncertainty of water quality benefits from the adopted buffers, although these changes may increase adoption rates. Performance of conventional buffer designs is highly variable over time and space and is a function of hydrologic and site-specific characteristics (Gall et al., 2018; Hickey & Doran, 2004; Liu et al., 2017; Liu, Yang, & Wang, 2007; Liu, Zhang, & Zhang, 2008; Wallace et al., 2018). In some cases, riparian buffers have even been found to act as P sources (Hoffmann, Kjaergaard, Uusi-Kämppä, Hansen, & Kronvang, 2009; Kleinman et al., 2015). Given the high variability in conventional designs, it is possible that more flexible buffer design constraints will not significantly affect overall water quality benefits at a watershed scale, especially if they increase adoption rates (Pennsylvania Department of Conservation and Natural Resources, 2019).

To better predict riparian buffer effectiveness for sediment and nutrient removal, many modeling efforts have been conducted. The Riparian Ecosystem Management Model (REMM), developed by the USDA-ARS, is a commonly used tool to simulate physical, chemical, and biological processes in the riparian buffer zone (Lowrance et al., 2000). Early development and validation of REMM was conducted on a field site in Georgia (Inamdar, Lowrance, Altier, Williams, & Hubbard, 1999). The uncalibrated model was able to represent the measured N and P outflux and the denitrification rate. In other studies where upland input data were hard to obtain, REMM was integrated with upland models to obtain the required input. So far, REMM has been successfully integrated with the Groundwater Loading Effects of Agricultural Management Systems (GLEAMS), the Agricultural Policy/Environmental eXtender (APEX), the Soil and Water Assessment Tool (SWAT), and the Annualized Agricultural Non-Point Source Pollution Model (AnnAGNPS) to study the impact of riparian buffers on water quality at basin or watershed scales (Liu, 2015; Liu et al., 2007; Stone, Gerwig, Williams, Watts, & Novak, 2001; Williams et al., 2013).

Previous applications of REMM and riparian buffer modeling work have mainly focused on validating the model's ability to represent the observed nutrient loads leaving the buffer. Few REMM studies have tested buffer performance with different buffer designs and under different land use and land management scenarios. Here, we propose several flexible buffer designs and explore their performance in removing sediment and nutrients. Specifically, the objectives of this study are (a) to determine the effectiveness of traditional and flexible buffer designs that represent the existing policy recommendations and integrate features important to farmers in the study watershed and (b) to evaluate the effectiveness of the buffer designs in reducing nutrient and sediment loads leaving agricultural fields as a function of buffer attributes and input loads. It is expected that these results will help to inform flexible riparian buffer policies and to evaluate the impacts of such policy recommendations on stream water quality.

2 MATERIALS AND METHODS

2.1 Study site overview

The Spring Creek Watershed (Hydrologic Unit Code 02050204), located in Centre County, PA, is characteristic of the karst topography of the Appalachian Ridge and Valley physiographical province, with limestone residuum in valleys and shale/sandstone colluvium on ridges. Elevations range from 280 to 675 m, with a surface water contributing area of 370 km² and a groundwater boundary of 450 km². The climate is temperate humid, with annual average precipitation of 1,050 mm. The runoff/precipitation ratio varies across the watershed by topographic index but ranges from approximately 0.1 to 0.5. Forty-three percent of the watershed is forested (primarily mountain ridges and some stream corridors), 23% is developed, and 34% is agricultural, with commercially focused beef and dairy farms as well as smaller family farms (Figure 1). The two major agricultural soils in the watershed, Hagerstown (fine, mixed, semiactive, mesic Typic Hapludalfs) and Opequon (clayey, mixed, active, mesic Lithic Hapludalfs), are well-drained and productive silt loams (Table 1) that are distributed throughout the valley region of the watershed. The Opequon soils (0.5 m depth) are generally underlain by fragipans, making the rooting zone much shallower than in the Hagerstown soils (0.5 vs. >2 m depth). The major riparian soil in the watershed is Lindside (fine-silty, mixed, active, mesic Fluvaquentic Eutrudepts), which is a well-drained silt loam with an approximate depth of 1.5 m (Table 1). Typical crop rotations in the watershed include (a) 3 yr of corn (Zea mays L.) grain followed by 1 yr of soybean [Glycine max (L.) Merr.], (b) 3 yr of corn and nonharvested winter wheat (Triticum aestivum L.) cover crop followed by 1 yr of soybean, and (c) continuous hay (Dactylis glomerata L.) (Amin, Veith, Collick, Karsten, & Buda, 2017; Dr. Charles White, personal communication, 2019). Row crops are typically farmed on the contour with minimum tillage. Classified as a cold-water fishery, the Spring Creek is degraded due to high nutrient and sediment loads from agricultural activities (Pennsylvania Department of Environmental Protection, 2016) and is currently a priority watershed in the Chesapeake Bay for increasing adoption of agricultural BMPs to meet the Bay's TMDL by 2025 (Pennsylvania Department of Environmental Protection, 2019).

FIGURE 1

Land use map in Spring Creek watershed (National Land Cover Database, 2011)

TABLE 1. Properties of major agricultural soils in the Spring Creek watershed

Soil series	Soil horizons	Thickness	Ksata	Bulk density	Organic C	Clay	Silt	Sand
		cm	mm h^−1	g cm^−3	%
Hagerstown	A	20	83.8	1.3	1.74	20	56	24
	BE, Bt, and BCt	94	33.0	1.4	0.29	42	29	29
	C	77	33.0	1.4	0.15	35	36	29
Opequon	A	15	27.9	1.35	1.45	36	45	19
	Bt1 and Bt2	26	27.9	1.55	0.15	55	42	3
	R	10	50.8	1.55	0.10	55	42	3
	A	20	33.0	1.36	3.00	21	68	11
Lindside	BA, Bw, and BC	92	37.9	1.36	0.25	21	68	11
	C	30	78.8	1.37	0.05	20	43	37

• Note. Source, Web Soil Survey (2019).
• ^a Saturated hydraulic conductivity.

2.2 Model description and set-up

2.2.1 Nutrient and sediment loads from upland crop fields using SWAT

SWAT (Arnold, Srinivasan, Muttiah, & Williams, 1998) is widely used to predict discharge, water quality, and crop yield in a watershed as a function of climate, topography, soils, and management practices (Arnold et al., 1998; Neitsch, Arnold, Kiniry, & Williams, 2011). Amin et al. (2017) modeled and vetted the typical agricultural conditions (2007–2015) of the Spring Creek watershed while exploring the environmental impacts of various alternative dairy cropping strategies (Amin, Karsten, Veith, Beegle, & Kleinman, 2018) and evaluating the cost-effective placement of BMPs (Amin, Veith, Shortle, Karsten, & Kleinman, 2019). Using the Spring Creek SWAT project (setup, calibration, and validation detailed in Amin et al. [2017]) to provide a system-level representation of the entire watershed, the current study focuses on six typical agricultural fields comprising the dominant three crop rotations and two riparian upland soils in the Spring Creek watershed. Simulation of these six fields, represented in SWAT by six unique hydrologic response units, over 16 yr (1999–2015) provided surface runoff, subsurface flow, and nutrient (N and P) and sediment loads from upslope, adjacent agricultural fields into the riparian zones. The simulation period of 16 yr was set up to capture four cycles of 4-yr crop rotations. These six fields were set up to have 100 m length and 100 m width, with a total field area of 1 ha.

2.2.2 Simulating riparian buffer processes using REMM

REMM (Lowrance et al., 2000) is a field-scale, process-based, two-dimensional model that simulates the water quality benefit of riparian buffers. REMM sets up a three-zone buffer system, where Zone 1 is nearest to the stream and Zone 3 is farthest. Each buffer zone is represented with one surface litter layer and three soil layers. Users can specify the width, vegetation, and management practice of each buffer zone. The REMM database includes photosynthesis and growth parameters to represent the average vegetative conditions in the northeastern United States (Allison, Fatula, & Wolanski, 2006). Management practice options in REMM include burning, harvesting, mowing, and fertilizing. REMM uses daily input of upland sediment, surface and subsurface water, and nutrient loads to simulate the biological, physical, and chemical processes in each buffer zone (Inamdar et al., 1999). In this way, it tracks the transport of water, nutrients, and sediment through each buffer zone and from the buffer system to the stream.

A Lindside (fine-silty, mixed, active, mesic Fluvaquentic Eutrudepts) soil series was selected as the representative riparian buffer soil in this watershed (Table 1). General soil properties, including soil texture, bulk density, and initial soil organic carbon (C) content, were obtained from the SSURGO database (Soil Survey Staff, 2019). The initial soil C/N ratio and partitioning of the soil organic C across the C pools were estimated based on Kelly et al. (1997). The same daily weather data and simulation time period were used in both REMM and SWAT. The slope of the simulated riparian buffers (5%) was determined based on the average slopes within 30 m of streams using 1-m resolution digital elevation models from the USDA-NRCS geospatial data gateway (https://gdg.sc.egov.usda.gov/GDGOrder.aspx).

Daily SWAT outputs from the six fields into the buffer provided REMM inputs of surface runoff; subsurface runoff; nitrate in surface runoff and subsurface flow; and soluble phosphorus, sediment-bounded phosphorus, organic N, and organic P in surface runoff. Other daily upslope inputs required by REMM but not obtainable from SWAT include C/N and C/P ratio in surface and subsurface runoff, which were estimated based on the REMM database as suggested by a previous study that also integrated SWAT and REMM (Liu et al., 2007).

2.2.3 Riparian buffer design and management scenarios

Six buffer designs (Table 2) were formulated based on current design recommendations in Pennsylvania and feedback received from landowners during a focus group meeting conducted in the study watershed. The buffers were each set up to have a length (parallel to stream) of 100 m, which was the same length the fields. The deciduous trees and perennial grasses in the buffer zone represent the native species in the northeastern United States and do not refer to a specific species.

TABLE 2. Six buffer designs (width, vegetation) and harvesting operations simulated using the Riparian Ecosystem Management Model

Buffer ID	Total width	Buffer/field area ratio	Zone 3^a	Zone 2	Zone 1
	m
Buffer i	10	0.1	grass	NA^b	NA^b
Buffer ii	15	0.15	grass	NA^b	NA^b
Buffer iii	15	0.15	deciduous trees	NA^b	NA^b
Buffer iv	30	0.3	grass (6 m)	deciduous trees (19 m)	deciduous trees (5 m)
Buffer v	30	0.3	grass (6 m)	deciduous trees (19 m; harvested every 3 yr)	deciduous trees (5 m)
Buffer vi	30	0.3	grass (6 m; harvested annually)	deciduous trees (19 m)	deciduous trees (5 m)

• ^a Buffer Zone 1 is located next to the stream; Zone 3 is next to the contributing area/agricultural field and farthest from the stream; Zone 2 is in the middle. ^bNot applicable.

The Pennsylvania Department of Environmental Protection Riparian Forest Buffer Guidance suggests a minimum forested buffer width of 10 m for water quality maintenance purposes (Pennsylvania Department of Environmental Protection, 2010). The Pennsylvania Riparian Forest Buffer Handbook for CREP requires a minimum width of 15 m for landowners to be eligible for payment from the state (Noto & Linsenbigler, 2017). The Pennsylvania Stormwater BMP Manual (Pennsylvania Department of Environmental Protection, 2006) suggests a three-zone buffer system, with Zone 3 composed of natural grass with a minimum width of 6 m, Zone 2 covered by managed forest with a minimum width of 19 m, and Zone 1 containing undisturbed or native forest with a minimum width of 5 m. The forested zones in REMM are represented using the average photosynthesis and growth parameters of native vegetative conditions in the northeastern United States (Allison et al., 2006).

A focus group with landowners and environmental agencies in the study watershed in March 2018 revealed that landowners preferred grasses over trees in riparian buffer zones due to the lower maintenance demand and higher success rate. Additionally, they expressed interest in harvesting part of the buffer vegetation for economic benefits. Therefore, we modeled two buffer operation regimes: tree harvesting at the end of September every 3 yr and grass harvesting at the end of September every year. The harvesting removes most aboveground vegetation but leaves some residues (about 10%) to maintain ground cover. Although harvesting buffer vegetation will have an impact on subsequent months as harvested vegetation regrows, REMM is able to represent the regrowth of harvested perennial grasses but not that of wood. Rather, it assumes that the wood is permanently removed from the buffer once it is harvested and does not have an option to regrow trees the next year. Therefore, we had to start new simulations for the years subsequent to the harvesting year, which resulted in simulations that assumed trees had grown back fully by the next year. We made this assumption because bioenergy feedstock vegetation, such as willow, grows rapidly, with harvestable heights reached within 3 yr (Jacobson, 2013; Volk, Verwijst, Tharakan, Abrahamson, & White, 2004). Harvesting is recommended every 3 yr afterward for up to seven harvests without needing to replant (Jacobson, 2013).

2.3 Data analysis

Buffer performance for reducing sediment and nutrient loads was evaluated in two ways: (i) the mass of sediment and nutrients removed by the buffer, calculated as the difference between the mass into and out of the buffer, and (ii) the reduction rate or removal efficiency of the buffer, calculated as the percentage of mass removed by the buffer. These calculations were conducted at an annual and monthly scale. Mass of reduction was used for evaluating the effectiveness of the same buffer when receiving different upland loads. In this evaluation, a single buffer type was analyzed under input from six upland fields simulated in SWAT. The removal efficiency was used for evaluating the effectiveness of different buffer designs when receiving the same upland loads. In this second evaluation, a single upland field provided input for four buffer designs (Buffers i–iv). The harvested buffers (Buffers v and vi) were compared in this second way with the nonharvested buffer (Buffer iv) to determine whether harvesting reduced the ability of the buffer to mitigate nutrients and sediment. The ANOVA and Tukey honestly significant difference (HSD) tests were conducted in R Studio 1.2 to detect significant differences among buffer designs and management practices at a confidence level of 95%.

3 RESULTS AND DISCUSSION

3.1 SWAT simulation results: Upland sediment and nutrient loads

SWAT simulation results showed that crops grown on Opequon soil generated three to four times the sediment load as did Hagerstown soil fields with the same crop rotation (Table 3). Within each soil type, average annual losses predicted from continuous hay were lower than those from the corn/winter wheat–soybean rotations, which were lower than those from the corn–soybean rotations without a cover crop. In particular, adding a winter cover to the corn–soybean rotation decreased sediment-bound nutrients by about 50% on Hagerstown soils and 33% on Opequon soils. Seasonal differences showed similar trends by crop rotation and soil type (Supplemental Figure S1).

TABLE 3. Annual average nutrients and sediment loss over 16 yr (1999 – 2015) from agricultural fields of three dominant crop rotations and two dominant agricultural soils in Spring Creek watershed, as simulated by the Soil Water Assessment Tool (SWAT)

	Hagerstown			Opequon
Average annual constituent load	3C-1S	3C/ww-1S	Hay	3C-1S	3C/ww-1S	Hay
Particulate organic N, kg ha^−1 yr^−1	7.88	4.67	0.35	13.92	9.68	1.28
Nitrate-N, kg ha^−1 yr^−1	2.41	1.85	1.34	2.37	2.22	0.97
Particulate organic P, kg ha^−1 yr^−1	2.16	1.22	0.11	3.68	2.43	0.4
Particulate mineral P, kg ha^−1 yr^−1	2.30	1.25	0.11	3.71	2.36	0.39
Dissolved mineral P, kg ha^−1 yr^−1	0.56	0.44	1.54	0.28	0.26	0.98
Sediment, kg ha^−1 yr^−1	1,621.25	1,082.50	19.00	4,760.00	2,757.75	87.63

• Note. 3C-1S, 3 yr of corn followed by 1 yr of soybeans; 3C/ww-S1, 3 yr of corn and winter wheat cover crop followed by 1 yr of soybeans; Hay, 4 yr of continuous hay.

Nutrient and sediment losses simulated using SWAT (Table 3) are comparable to literature values. Harmel et al. (2006) reported a median annual nutrient loss from corn and soybean fields in the United States of 2.7–3.02 kg ha⁻¹ for dissolved N, 7.27–21.9 kg ha⁻¹ for organic N, and 0.85–9.6 kg ha⁻¹ for particulate P. USDA-NRCS (2006) estimated N loss from corn and soybean fields in the northeastern United States as 3.6 and 3.3 kg ha⁻¹ for dissolved N and 26.1 and 14.9 kg ha⁻¹ for organic N. USDA-NRCS (2010) estimated the average sediment loss from cropland in the northeastern United States in 2007 as 6.6 ± 0.5 t ha⁻¹. Sediment loss simulated in this study was relatively lower than the values reported by USDA-NRCS (2010), likely due to increased adoption of minimum tillage in the study watershed. The current study also shows that hay fields, which receive multiple manure applications in the study watershed to meet the farm-level nutrient management requirements, generate higher dissolved P loss than corn–soybean fields, which receive fertilizer application based on agronomic demand.

3.2 Buffer performance as a function of buffer design

REMM simulations revealed that grass buffers were more efficient than tree buffers of the same width in reducing surface runoff, sediment, total N (TN), and total P (Figure 2). For buffers with the same vegetation, the removal efficiency of sediment and nutrients increased with buffer width (Figure 2). All buffer removal efficiencies are provided in Supplemental Tables S1–S9; only key findings are highlighted here. The 15-m grass buffer had significantly (p < .05) higher removal efficiencies of sediment and nutrients than the 10-m grass buffer (Figure 2). Additionally, the 15-m grass buffer was significantly more effective in reducing runoff, sediment, and nutrients than the 15-m tree buffer (p < .05). The 30-m buffer, which was composed of one grass zone (6 m) and two zones of trees (Buffer iv) following Pennsylvania Department of Environmental Protection (2006) recommendations, was slightly more efficient than the 10-m grass buffer in nutrient removal, but the difference was not statistically significant (p > .05) (Figure 2). These results are consistent with previous research, and a more detailed discussion regarding these results is provided in the Supplemental Material.

FIGURE 2

Box and whisker plot showing annual reduction rates (i.e., percent reduction) over a 16-yr period of (a) surface water, (b) sediment, (c) total N (TN), and (d) total P (TP) by four buffer designs. The box represents the first to third interquartile of annual reduction rates, the whiskers represent the minimum and maximum annual reduction rates, the line within the box represents the median value, and the x represents the mean. Values outside of the box and whisker plots represent outliers. All buffers received nutrient and sediment loads from a corn–soybean rotation on Hagerstown soil. Different letters above boxes indicate a significant difference between buffer designs (p < .05, ANOVA)

3.3 Comparison between policy-driven and farmer-preferred buffer designs

Buffers i through iv represent various buffer designs supported by state and national policies for grass and riparian buffers. The 30-m grass plus tree buffer (Buffer iv) consisting of 6 m of grass and 24 m of tree is the recommended buffer design by the Pennsylvania Stormwater BMP Manual (Pennsylvania Department of Environmental Protection, 2006), whereas the 10-m grass buffer is the narrowest option with farmer-preferred vegetation. For nutrient and sediment removal efficiency, the 30-m grass plus tree buffer (Buffer iv) was slightly more efficient than the 10-m grass buffer (Buffer i), but the differences were not significant (p > .05), except for sediment removal (Figure 2). Closer inspection of simulation results revealed that most of the sediment and nutrients removed by the 30-m grass plus tree buffer was removed in Zone 3 (6 m grass) (Figure 3), which is the zone closest to the source and furthest from the stream. For example, for the loads generated by the corn–soybean rotation on Hagerstown soil, 59% of the TN reduction by the buffer (8.0 kg mass reduction) occurred in Zone 3 (6 m grass), 29% (3.9 kg) occurred in Zone 2 (19 m tree), and 12% (1.6 kg) in Zone 1 (5 m tree) (Figure 3a). This result agrees with the findings from Inamdar et al. (1999), who also found that most sediment and nutrients deposited and were removed in Zone 3 (7.6 m grass), instead of Zone 2 (53 m tree) or Zone 1 (15.2 m tree) of the buffer.

FIGURE 3

Annual average flux of (a) total N (TN), (b) particulate P, and (c) sediment going through each zone of the 30-m grass + tree buffer when the buffer received loads from corn–soybean rotation on Hagerstown soil. Error bars depict SD

The 15-m grass buffer (Buffer ii) represents a combination of the minimum width for landowners to be eligible for the CREP payment (15 m; USDA Farm Service and Forest Service, 2007) and farmer-preferred vegetation (grass). It was found to be equally effective as the 30-m grass plus tree buffer (Buffer iv) for sediment and nutrient removal, with no significant difference (Figure 2).

Including harvesting in the 30-m buffer design did not significantly (p > .05) change the buffer's ability to reduce incoming nutrient and sediment loads from adjacent fields (Figure 4). This is likely because harvesting was selected to occur at the end of September when the biomass was at its peak. This approach can provide the highest biomass yield and leaves the vegetation in the buffer for as long as possible prior to frost. Also, harvesting does not remove all aboveground biomass and usually leaves parts of vegetation on the ground. Furthermore, because harvesting vegetation removes nutrients from the buffer, the simulation results suggested that the harvested buffer sometimes achieved a slightly (2–3%) higher nutrient reduction rate than the nonharvesting buffer. This finding is consistent with research conducted by Hubbard and Lowrance (1997), who found that coastal riparian forests were effective in removing nitrate under clearcutting or thinning management. Lowrance et al. (1984) suggested periodical forest harvesting to ensure that buffers act as a long-term sink for nutrients. Harvesting of plant biomass has also been found to be effective in removing P retained in buffer systems (Dodd & Sharpley, 2016; Hoffmann et al., 2009; Satchithanantham, English, & Wilson, 2019).

FIGURE 4

The influxes and outfluxes of (a) total N (TN), (b) total P (TP), and (c) sediment of a 30-m grass plus tree buffer under three management practices: no harvesting operation (Buffer iv), harvesting of Zone 2 (Buffer v), and harvesting of Zone 3 (Buffer vi). The harvesting of trees in Zone 2 occurred once every 3 yr, and therefore a bar for Buffer v is graphically displayed every third year. Harvesting of grass in Zone 3 occurred every year for Buffer vi. All buffers shown here received nutrient and sediment loads from corn–soybean rotation on Hagerstown soil

3.4 Buffer removal efficiency as a function of input loads

For any given buffer design, comparison across the six SWAT simulated fields revealed that greater upland loads into the buffer resulted in lower buffer reduction rates of nutrients and sediment; however, the actual masses of nutrients and sediment that were removed by the buffer were greater when the input loads were higher (Figure 5). For example, when the 30-m buffer received runoff from the corn–soybean rotation grown on Opequon soils, the average annual total N mass reduction was 12.8 kg yr⁻¹, and the average annual removal efficiency was 62%. When the same buffer design received a lower load from the continuous hay grown on Hagerstown soils, the average annual total N mass reduction was 7.4 kg yr⁻¹, and the average annual reduction was 82%. The buffer removal efficiency was higher in the latter scenario, but the mass reduction was higher in the former scenario. All mass reductions and percent reductions are provided in Supplemental Tables S1–S9. Supplemental Tables S5–S9 contain simulation results for individual N and P species. Disagreement between mass reduction and percent reduction was also found in seasonal and event-based results (Supplemental Figures S2 and S3).

FIGURE 5

Riparian buffer (30-m grass plus tree buffer, Design iv in Table 2) performance when receiving loads from the six fields represented in Table 3. The height of the entire bar is the annual average nutrient and sediment input to the buffer (kg yr⁻¹). The shadowed bars (top) are the outflux mass from the buffer. The dotted bars (bottom) are the masses retained by the buffer. 3C-1S, 3 yr of corn followed by 1 yr of soybeans; 3C/ww-S. 3 yr of corn and winter wheat cover crop followed by 1 yr of soybeans; Hay, 4 yr of continuous hay, TN, total N; TP, total P

3.5 Management implications

From the perspective of reduction mass and reduction percentage, placing buffers based on highest load reduction may lead to greater water quality benefits at the watershed scale than placing buffers based on highest percent reduction. This strategy can help to inform cost-effective watershed-scale BMP implementation plans (Gitau, Veith, Gburek, & Jarrett, 2006) such that buffers are placed where they will function to reduce the highest loads rather than where they will achieve the highest percent reduction. The former approach may yield an average field-scale percent reduction that is lower than the latter approach, despite achieving greater load reduction at the watershed scale. This potential mismatch in field and watershed-scale approaches to decision-making highlights the importance of communicating field-scale benefits (targeting based on field-scale percent reduction) versus watershed-scale benefits (targeting based on total load reduction).

3.6 Limitations

Although it is known that concentrated flowpaths are a major factor that increases sediment and nutrients loss from crop fields and decreases buffer effectiveness in nutrients and sediment reduction (Blanco-Canqui, Gantzer, Anderson, & Alberts, 2004; Pankau, Schoonover, Williard, & Edwards, 2012), they were not considered in this study. Wallace et al. (2018) concluded that the presence of concentrated flowpaths reduces the effectiveness of forested buffers by 54% in the study watershed. Here, we assumed that the contributing area of each buffer scenario was 1 ha, which results in a buffer area ratio of 0.1–0.3 for the range of buffer widths (10–30 m) we simulated. These numbers are similar to or greater than the buffer area ratios for buffers that already exist in the study watershed. Wallace et al. (2018) estimated the buffer area to effective contributing area ratio to be about 0.1 in the study watershed. However, our simulation results likely overestimate buffer effectiveness, particularly for the buffer designs that include trees (Buffers iii–vi), by excluding concentrated flowpaths in the simulation, which are known to reduce buffer performance (Wallace et al., 2018) in the study watershed.

We were unable to calibrate and validate the various hydrologic and biogeochemical processes simulated by REMM due to a lack of field measurement data. Insufficient field measurement is also a limitation of many BMP evaluation projects (Liu et al., 2017). As such, we focused on comparisons between buffer designs and scenarios instead of accurately simulating reduction rates of certain buffers. Moreover, unlike SWAT, REMM does not heavily rely on user-input parameters. In previous REMM applications, calibrations primarily focused on the denitrification parameter (Tilak, Youssef, Burchell, Lowrance, & Williams, 2017).

4 CONCLUSIONS

In this study, we tested the effectiveness of four different buffer designs (variations in vegetation and width) and two alternative buffer harvesting scenarios (grass and trees) that represented combinations of policy-driven and farmer-preferred options. Nutrient and sediment loads from three crop rotations and two soils were simulated in SWAT and used as inputs to REMM to better understand how the effectiveness of a buffer changes as a function of input load and buffer design and management. Simulation results suggest that for buffers of the same width, the farmer-preferred vegetation (grass) outperformed policy-preferred vegetation (trees) for sediment, N, and P removal. Additionally, simulation results revealed that harvesting a portion of the buffer zone did not significantly (p > .05) decrease buffer effectiveness in nutrient and sediment removal and, in some cases, can even increase nutrient reduction rate. Finally, simulation results concluded that riparian buffer systems removed a greater mass of sediment and nutrients when receiving higher upland loads despite a lower percentage removal efficiency. Therefore, evaluating a buffer's performance by percentage removal efficiency may be misleading when the input nutrient loads and therefore total masses removed/retained by the buffer are different. Our results can provide policymakers with information on flexible buffer systems and help to evaluate the water benefit tradeoffs of such a flexible buffer paradigm. However, prior to adoption of such a flexible buffer paradigm, tradeoffs between water quality provision and ecosystem services not explored in this research, such as streambank stabilization, habitat, and stream shading, should be pursued.

ACKNOWLEDGMENTS

The authors thank Matt Royer, Director of the Penn State Agriculture and Environment Center, and Tina Hoy from the Penn State Survey Research Center for facilitating focus group meetings and all of participating stakeholders for their valuable input. This research was supported by USDA Grant #2017-77019-26374. H.E. Preisendanz (Gall) is supported in part by the USDA National Institute of Food and Agriculture Federal Appropriations under Project PEN04574 and Accession number 1004448. R. Cibin is supported in part by the USDA National Institute of Food and Agriculture Federal Appropriations under Project PEN04629 and Accession number 1014132. P.J. Drohan is supported in part by the USDA National Institute of Food and Agriculture under Project PEN04573 and Accession number 1004449. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by The Pennsylvania State University or the USDA. All entities involved are equal opportunity providers and employers.

CONFLICT OF INTEREST

The authors declare no conflict of interest.