A secondary assessment of sediment trapping effectiveness by vegetated buffers

Abstract

Vegetated buffers and filter strips are a widely used Best Management Practice (BMP) for enhancing streamside ecosystem quality and water quality improvement through nonpoint source pollutant removal. Most existing studies are either site-specific, rely on limited data points, or evaluate buffer width and slope as the only design variables for predicting sediment reduction, not considering other parameters such as soil texture, vegetation types, and runoff loads that can significantly influence the buffer efficiency. In this paper, we carry out a meta-analysis of published studies and fit regression models to explore the sediment removal capacity of riparian buffers. We compiled 905 data points from over 90 studies (including data from an online BMP database) documenting sediment trapping by vegetated buffers and recorded data regarding buffer characteristics such as buffer width, slope, area, vegetation type, sediment loading, water flow rates, and sediment removal efficiency. We found that an exponential regression model describing the relationship between sediment removal efficiency by the buffer and water inflow/outflow volume ratio explained 44% of the variance. Adding the square root of roughness increased the R² to 0.50. The model performance was compared with other sediment reduction regression models reported in the literature. The results point towards the importance of considering flow parameters in vegetative buffer design. The improved empirical relationships derived here can be used at local scales to understand sediment trapping potential by vegetated buffers for water quality mitigation purposes and can be built into extant hydrologic models for improved watershed-scale assessments.

Introduction

Naturally occurring riparian forests and streamside vegetation play a critical role in intercepting and purifying pollutant laden runoff, but their degradation by humans coupled with nonpoint source pollutant export has contributed to the deterioration of over 50% of stream and river lengths in the US (Sweeney and Newbold, 2014). Sediment pollution can clog waterways, cause flooding, and reduce water quality for domestic uses (drinking, cooking), recreational uses and/or municipal and industrial uses (Ribaudo et al., 1999). Sediment–laden water also negatively affects aquatic biodiversity and can destroy aquatic habitat by decreasing light-penetration in water, increasing water temperatures, clogging fish gills, and covering spawning areas and smothering aquatic biota (Cooper, 1993; Ribaudo et al., 1999). The US EPA ranks siltation as the leading cause of pollution of streams and rivers in the United States (US Environmental Protection Agency, 1998; Ribaudo et al., 1999).

Nonpoint source pollutants can either be managed at the source or intercepted to filter out nutrients and sediment before they reach surface waters (Dillaha et al., 1989; Ribaudo et al., 2001). The establishment and maintenance of vegetative filter strips (VFS) and riparian buffers have gained immense popularity as a cost-effective interception strategy for mitigating water quality and improving riparian ecosystem quality by nonpoint source pollutant removal (Lowrance et al., 1997; Webber et al., 2010; Sweeney and Newbold, 2014; Cole et al., 2020). Vegetative filter strips (VFS) are bands or areas of closely grown vegetation that receive and purify runoff from upslope areas such as croplands or pastures or other pollutant source areas (Dillaha et al., 1988). Vegetative filter strips and buffers perform a wide array of functions - they filter out sediments and nutrients from runoff by promoting processes such as infiltration, adsorption, plant uptake, sedimentation, and pollutant degradation through numerous biogeochemical processes (Webber et al., 2010; Pinho et al., 2008; Rahman et al., 2014). They prevent streambank erosion and improve habitat and biodiversity (Sweeney and Newbold, 2014; Lind et al., 2019; Cole et al., 2020). Several studies have documented the effectiveness of vegetated filter strips for sediment trapping. Le Bissonnais et al. (2004) reported as much as 98% decrease in sediment loads from a field using a 6 m grass strip. Duchemin and Hogue (2009) reported total suspended solid load reductions of 87% using grass strips and 85% using mixed grass and tree buffer strips. Lee et al. (1999) documented 77% and 66% sediment load reduction from adjacent crop fields using 6 m and 3 m grass buffers, respectively. Since vegetated buffers form an integral part of watersheds, either as on-site mitigation features in the form of grass hedges/vegetated filter strips at field (or other pollutant producing sites) edges or as end-of-pipe features such as riparian buffers, field and watershed scale models often need algorithms to better simulate hydrology and/or water quality through buffer systems. Hence there is a need to assess buffer effectiveness and/or load reductions from these systems through quantitative methods.

Regression (statistical) models are useful tools for water quality prediction and for making management decisions regarding buffer maintenance and pollutant attenuation (Mayer et al., 2007). Published literature has identified various factors affecting the load mitigation performance of vegetated buffers including buffer width, slope, area ratio (pollutant source area: buffer area), and hydrological flow conditions (Arora et al., 1996; Abu-Zreig et al., 2004; Boyd et al., 2003; Barfield et al., 1998l; Daniels and Gilliam, 1996; Dillaha et al., 1989; Dosskey et al., 2008, Dosskey et al., 2011; Duchemin and Hogue, 2009; Gharabaghi et al., 2006; Lee et al., 1999, Lee et al., 2000; Deletic and Fletcher, 2006; Alemu et al., 2017; Saleh et al., 2017). A few studies have conducted meta-analysis assessments of sediment trapping-effectiveness by vegetated buffers. Liu et al. (2008) evaluated data from 31 studies and concluded that regardless of the area ratio of buffer to agricultural field, optimum sediment trapping was obtained when buffer width was 10 m and had a slope of 9%. In a meta-analysis study by Zhang et al. (2010), buffer width alone captured 37% of the total variance in sediment removal efficiency. In that study, a 30 m buffer with a slope ≈ 10% removed >85% of the sediment. These studies primarily evaluated buffer width and slope as design variables for predicting sediment reduction. While these are important design variables to consider for sediment reduction as well as for estimating costs related to buffer installation and maintenance (Dosskey et al., 2008), there is a need for evaluating buffer width and slope impacts in light of various site conditions such as soil textures, vegetation types, and runoff loads that can significantly influence buffer efficiency.

Conducting secondary analysis studies on buffer efficiency can be quite daunting because of the variability in buffer parameterization across different studies. Quantifying loads and runoff with consistent dimensions and interpretation across experimental studies poses a significant challenge because of the large variations in the objectives and site conditions in which the vegetative filter strip trials are tested. Sediment trapping efficiency (or sediment reduction) is represented by the following equation:

$$R_m=\left\{\frac{M_{\mathit{in}}-M_{\mathit{out}}}{M_{\mathit{in}}}\right\}\times 100\%$$

where R_m is the percent removal efficiency, M_in is the sediment mass entering the buffer, and M_out is the sediment mass leaving the buffer. However, there is a large variability in quantifying M_in and M_out across different studies. Some studies such as Dillaha et al., 1988, Dillaha et al., 1989 and Lee et al., 1989, Lee et al., 2000, Lee et al., 2003 compare a control erosion plot with a buffered erosion plot, where the dimensions of the erosion plots are the same in both cases. Other studies such as Uusi-Kämppä and Jauhiainen (2010), Tingle et al. (1998), and Thayer et al. (2012) compare different area ratios of erosion plot and buffer with each other to determine % removal. Studies such as Abu-Zreig et al. (2004) strictly evaluate loads entering and leaving the buffer area. Consequently, the units in which loads have been calculated in the various studies differ. Most studies quantify loads as load per unit area. i.e., kg/ha, kg/m², tons/ha/yr, etc. However, a significant portion of them does not specify the area over which the load was calculated. For some studies such as Lee et al. (2000) and Uusi-Kämppä and Jauhiainen (2010), the reported loads were divided by the entire area of the plot which included the erosion plot and the buffer. For other studies such as Abu-Zreig et al. (2004), the loads reported were divided over the buffer area only. This can cause inconsistencies in the reporting of the loads and calculation of sediment trapping efficiency. Moreover, many experimental approaches are used to generate runoff required for testing the effects of the buffer. One approach is to let natural rainfall generate runoff (Lee et al., 2003; Arora et al., 1996; Duchemin and Hogue, 2009; Daniels and Gilliam, 1996); which in many cases did not produce enough water for this purpose (Hay et al., 2006). Other approaches include simulated rain events (Barfield et al., 1998; Dillaha et al., 1989; Coyne et al., 1995; Chaubey et al., 1994), simulating inflows (Van Dijk et al., 1996; Deng et al., 2011; White et al., 2007), or both (Schmitt et al., 1999). These create challenges in the quantification of inflows, runoff, and rainfall, if measured at all. Meta-analysis studies have mostly overlooked these inconsistencies, which can have implications on the structure of the developed model.

Stand-alone models such as the process-based model vegetative strip model (VFSMOD; Muñoz-Carpena and Parsons, 2004) and Riparian Ecosystems Management Model (REMM; Lowrance et al., 2000) have been used to evaluate sediment reduction for different site designs and vice versa. For instance, Dosskey et al., 2008, Dosskey et al., 2011 used VFSMOD to develop graphical design aids for width and area ratio to achieve specific sediment reduction targets under a broad range of agricultural site conditions. However, VFSMOD algorithms are complex and require detailed inputs and significant computing resources to run the models and interpret results, and as such, are not used in site planning (Dosskey et al., 2008). Simpler mathematical models for buffer impacts based on theoretical equations, simplified mathematical abstractions, or regressions have been used in commonly utilized watershed models such as the Soil and Watershed Assessment Tool (SWAT). Earlier versions of SWAT considered a vegetated buffer model where trapping efficiency was solely a function of filter strip width (Neitsch et al., 2002). SWAT ver. 2012 considers an improved sediment reduction model where sediment trapping is a function of sediment loading to the buffer and runoff reduction actuated by the buffer. This model was developed by White and Arnold (2009) at the field scale using data from published literature, supplemented with data from VFSMOD simulations to deal with lack of inflow and runoff data. VFSMOD simulations were used to develop an empirical runoff reduction model in which runoff reduction was calculated as a function of runoff loading to the buffer and saturated hydraulic conductivity of the soils. A sediment reduction model was formulated based on measured data from 61 entries, and sediment reduction was quantified as a function of sediment loading to the buffer and runoff reduction. They observed that sediment loading to the buffer alone accounted for 41% of the variability in buffer sediment trapping, which increased to 64% when runoff reduction was added to the model. While predictions by VSMOD, like other physically-based models, are an inevitable manifestation of our limited understanding of reality and thus at best approximations, observed flow and sediment data are almost undoubtedly influenced by other equally important factors or processes not accounted for in the model.

Most studies on buffer sediment removal efficiency are limited to small-scale evaluation of filter strips and/or site-specific assessment of riparian areas. A comprehensive dataset can provide insights into generalizations about factors that are crucial for improving sediment removal and can inform best management practices in areas where data are scarce. This study aimed to compile a comprehensive dataset and develop a regression model while addressing the challenges in data quantification explained above. The objective was to understand if this exercise points towards similar sensitive parameters for buffer sediment trapping efficiency, as identified in the literature, while using a larger dataset.

We conducted a detailed analysis of published studies on sediment trapping, evaluated the inflow and runoff conditions using different assumptions to glean realistic field relationships between buffer characteristics and trapping efficiency. Our specific objectives were to 1) construct a database from published literature and online databases with detailed site-specific buffer characteristics such as width, slope, flow volumes/rates, sediment loading, and sediment reduction, 2) develop a sediment reduction model using multi-regression analysis as a function of various design characteristics, and, 3) compare and assess model performance to other published sediment reduction regression models for vegetated buffers. The overall objective was to obtain improved relationships that can be used at local scales to understand sediment trapping potential by vegetated buffers for water quality mitigation purposes and potentially be built into extant hydrologic models for improved watershed-scale assessments.