Impacts of different vegetation in riparian buffer strips on runoff and sediment loss

Abstract Buffer strips continue to feature in the management of agricultural runoff and water pollution in many countries. Existing research has explored their efficacy for reducing environmental problems in different geoclimatic settings but, the evidence on the efficacy of different vegetation treatments is less abundant than that for other buffer strip characteristics, including width, and is more contradictory in nature. With policy targets for various environmental outcomes including water or air quality and net zero pointing to the need for conversion of agricultural land, the need for robust experimental evidence on the relative benefits of different vegetation types in buffer strips is now renewed. Our experiment used a replicated plot scale facility to compare the efficacy of 12 m wide buffer strips for controlling runoff and suspended sediment loss during 15 sampled storms spanning 2017–2020. The buffer strips comprised three vegetation treatments: a deep rooting grass (Festulolium cv. Prior), a short rotation coppice willow and native broadleaved woodland trees. Over the duration of the monitoring period, reductions in total runoff, compared with the experimental control, were in the order: willow buffer strips (49%); deciduous woodland buffer strips (46%); grass buffer strips (33%). The corresponding reductions in suspended sediment loss, relative to the experimental control, were ordered: willow buffer strips (44%) > deciduous woodland buffer strips (30%) > grass buffer strips (29%). Given the 3‐year duration of our new dataset, our results should be seen as providing evidence on the impacts during the establishment phase of the treatments.

Buffer strips have been utilized as a means of reducing the movement of pollutants from agricultural land into watercourses for many years (Barling & Moore, 1994;Hickey & Doran, 2004). The form of the vegetation may take that of a grass verge at the edge of the field where no targeted planting of chosen species is undertaken and natural colonization is allowed to determine the dominant form of vegetation. Alternatively, targeted planting of specific grasses and woody plants has been utilized to vegetate buffer strips, with consequent effects on landscape aesthetics, biodiversity and interactions with the local watercourse and its ecology (Cole et al., 2020). Choices of the type of plants that can be deliberately planted within a buffer strip range from herbaceous grasses and forbs to small woody shrubs with multiple stems to taller woody tree species. The physiognomy of the plants may affect the runoff, the movement of pollutants including fine-grained sediment, or both (Roberts et al., 2012). The interactionpotential the buffer strip has for removing pollutants from the runoff leaving the field from upslope may thus change depending on the form of planting used to vegetate the buffer strip. Here, the form of planting chosen may affect the ability of the buffer strip to remove a priority pollutant within a local area, and as a result, some degree of potential exists to optimize buffer vegetation to ameliorate specific local concerns over particular pollutants or, alternatively, to address multiple issues .
Water pollution and flooding events associated with the movement of agricultural run-off have been reduced due to the interaction of water and vegetation within buffer strips. However, the ability of a buffer strip to provide such services continuously may be reduced or lost over time if the buffer strip becomes saturated with fine-grained sediment or nutrients (Valkama et al., 2018). To alleviate the potential for saturation of nutrients, planned removal of buffer strip vegetation can be implemented. For grass buffer strips, mowing and/or grazing can reduce the standing crop within the strip and cause compaction by trafficking or trampling. Access to strips may negate the possibility of using machinery in some circumstances (e.g., steeply sloped land), and refusal by grazing animals to consume standing vegetation may affect the amount of vegetation removed. The age of a grass dominated buffer strip may need to be considered if grazing animals are the only option available to reduce the standing crop. Woody plants can be harvested for their timber within a planned management system, and act as a means of both removing nutrients from the strip as well as reinvigorating plant growth rates, and thus facilitating the further removal of nutrients entering the buffer strip.
In England, implementation of water pollution interventions on farms, including buffer strips, is driven by a combination of regulation, incentivization in the form of agri-environment schemes and the delivery of on-farm advice for win-wins. Here, improved uptake rates by farms can be encouraged by robust scientific evidence on the efficacy of buffer strips for controlling runoff and pollution losses. Existing work examining the efficacy of buffer strips for environmental good has focussed on both external and internal factors (Eck, 2000). The former encompass the phase (i.e., particulate, dissolved) and delivery pathway (i.e., surface, subsurface) of the incoming pollution, whereas the latter include buffer width and vegetation cover. Advice delivery has tended to focus on buffer width in the case of internal controls since this is the easiest component of management to influence via farm management and existing evidence on varying efficacy for runoff and water pollution reduction, including width, can be readily extracted from a number of comprehensive reviews (e.g., Barling & Moore, 1994;Collins et al., 2009;Dorioz et al., 2006;Hickey & Doran, 2004;Kay et al., 2009). Beyond buffer strip width, the existing evidence on the effects of different vegetation cover remains less easy to generalize. Some work suggests that for the same buffer strip width, different vegetation cover impacts efficacy for pollution control by at most 20% (Dorioz et al., 2006). Other studies report very limited or no effect of vegetation cover (e.g., Schmitt et al., 1999;Uusi-Kämppä et al., 2000). In other cases, the results of investigations comparing herbaceous and woody vegetation in buffer strips report both a lack of (Daniels & Gilliam, 1996;Syversen, 1995) and detectable (Cooper et al., 1986;Parsons et al., 1994) differences in pollution reduction, with the latter suggesting better performance by herbaceous cover.
Given the above context, the new study detailed herein was undertaken to assess the impact of three different vegetated buffer strips on runoff and sediment loss to contribute to the evidence base.
The research project was planned to provide replicated evidence on buffer strip efficacy and to engage multiple stakeholders with this evidence given the ongoing inclusion of buffer strips in agricultural policy in the United Kingdom and beyond. This paper reports the preliminary results for the efficacy for reductions in runoff and sediment loss using our new dataset.

| Study site description
The assessment of buffer strip efficacy was undertaken on experimental plots located at the Rothamsted Research North Wyke site in Devon, UK (50 46 0 31.3 00 N 3 55 0 41.6 00 W). This site has a mean annual rainfall of 1032 mm y À1 , a mean maximum temperature of 13.5 C and mean minimum temperature of 6.7 C (1982-2019; see Table 1 for mean monthly values).
The site is situated upon a bedrock of clay bearing shales of the Carboniferous Crackington Formation, with overlying soils represented by a poorly drained Hallsworth series pelo-stagnogley soil (FAO classified as Stagni-Vertic Cambisol; Harrod & Hogan, 2008).
The stony clay loam topsoil comprises 16%, 48% and 36% of sand, silt and clay, respectively. Below the topsoil layer ($0 to 30 cm), the subsoil ($30 to 160 cm) is impermeable to water and is seasonally waterlogged; most excess water moves by surface and sub-surface lateral flow across the clay layer (Orr et al., 2016), with the experimental area having a slope of 8 .  Table 2.

| Experimental instrumentation
Runoff from each of the 12 experimental plots passed into a gravel- 3 | RESULTS

| Rainfall event information
Information for the individual rainfall events that were sampled is provided in Table 3. The largest rainfall event (

| Reductions in runoff for the buffer strips with different vegetation
Discharge data (m 3 ) for all treatments spanning the entire monitoring period (2017-2020) are presented graphically as a timeseries in Figure 2 and summarized on an annual basis in Table 4.
The percentage changes in discharge for the each buffer strip treatment for each rainfall event, compared with the control, are given in Table 5 and were calculated using the formula: Where x c is the control treatment and x t is one of the other buffer strip treatments. Positive values represent a reduction in volume of discharge compared with the control whereas negative values represent an increased volume of discharge; that is, no reduction. During the first year of the establishment of the buffers, except for the grass buffer in event number 1, there was no reduction in discharge for the first 7 events that were sampled. Event number 8 (12 November 2018) showed a reduction in discharge from all buffer treatments, compared with the control, in the order of willow > grass > woodland.
T A B L E 2 Activity dates and rates of fertilizer applications on the agricultural plots upslope of the buffer strip treatments. It is not clear as to why there was a marked change in performance of all the buffer treatments other than the total rainfall for this event was 26.7 mm following a 5-day antecedent period (P5) of rainfall totalling 50.2 mm (Table 3) In the case of the grass buffer strip treatment, the reduction in discharge (Table 6)

| Discharge:rainfall ratios
The discharge:rainfall ratios for each rainfall event sampled are given in Table 7 and show that in the first year following their establishment, discharge in the buffer strip treatments occasionally exceeded rainfall and generated more flow than the control (event numbers 3, 5), though the reason for these results is unclear. These early data are most probably artefacts associated with the settling down of the experimental set-up including the instrumentation, and possibly also due to soil disturbance in the process of setting up the buffer treatments. From event number 5 onwards, ratios varied between 0.11 and 0.73 with an overall average value of 0.33, though statistical analysis showed that any differences between the treatments were not statistically significant (data not shown). The average value is slightly lower than the reported annual standard percentage runoff (SPR) associated with the on-site soil series, which is 40% (Boorman et al., 1995).

| Time to peak discharge compared with time to peak rainfall
Analysis of the data for time to peak discharge compared with that of peak rainfall for each of the sampled events showed that differences between the treatments were not statistically significant (data not shown). This is attributed to the short length of the flow paths being monitored. While time to peak discharge can be important for flood mitigation, the current plot set-up was not designed to examine this specific effect.
T A B L E 3 Rainfall characteristics of the sampled events.

| Reductions in suspended sediment loss for the buffer strips with different vegetation
The maximum concentration of suspended sediment (mg L À1 ) for each event is shown in Figure 3 The efficacy for sediment load reduction for each sampled event (Table 9) followed a similar pattern as was observed with the discharge data in that there was high event-based variation but an indication of an improvement in efficacy with buffer maturity, especially in the case of the woody treatments.

| DISCUSSION
Overall, our results suggest that reductions in discharge in the individual sampled events (Table 5) indicate that the buffer treatments are reaching a stage of maturity that has begun to have a positive impact on hydrology. Continuation of the study over more years may have been able to confirm this. The storm period results suggest that the performance of the types of vegetated buffers tested in this study may be influenced by the amount of rainfall combined with antecedent soil moisture conditions. There was no clear treatment effect on time to peak discharge which is surprising given that hydraulic conductivity rates were greater under the woody buffers.
The results for the establishment phase of the buffer treatments therefore suggest that woody treatments improve sediment trapping  could be expected to increase as the buffers continue to mature. Previous work has underscored the potential for reductions in sediment loss to be strongly influenced by deposition of incoming sediment along the upslope leading edge of buffer strips due to the initial reduction in runoff velocity and sediment transport capacity (Ligdi & Morgan, 1995;Pearce et al., 1997). Such edge effects were not observed during our experiment.
Excess sediment loss from agricultural land remains a global issue despite the uptake of best management practices. For England and T A B L E 8 Annual and total losses (kg ha À1 equivalent) of suspended sediment (standard deviations in parentheses). Wales, for example, such elevated sediment losses due to current structural land cover have been estimated to result in £523 M of environmental damage costs annually, with the uptake of best management practices on farms only reducing those societal costs to £462 M . Buffer strips continue to feature in the mix of best management practices implemented on farms to protect water quality and their uptake by farmers can be facilitated by robust evidence on the efficacy for reducing water pollution. Agricultural runoff encountering a buffer strip meets a more porous and rougher surface, resulting in a reduction in runoff velocity and sediment transport capacity. Here, the vegetation cover generates increased resistance to runoff and sediment transport and the root systems increase the permeability of the soil surface, thereby encouraging infiltration and deposition (Magette et al., 1989;Rose et al., 2003).
Buffer strips can also assist in the management of the sediment problem by stabilizing and reducing the erosion of riverbanks (Bowie, 1995;Kemper et al., 1992) and by displacing sediment generating land management away from watercourses (Wenger, 1999). The beneficial effects of displacement are often, however, less pronounced on heavy meandering watercourses where channel migration drives bank erosion (Williamson et al., 1992). In England Wales, eroding channel banks have been estimated to contribute 22% of the total fine-grained sediment load delivered rivers and streams (Zhang et al., 2014). The potential beneficial impacts of buffer strips on reducing bank erosion across England and Wales, as well as sediment loss from utilized agricultural land, should therefore be borne in mind given the important role of bank erosion in the excess sediment problem nationally.
When interpreting evidence for buffer strip impacts on sediment loss, it is important to acknowledge various issues which can confound efficacy. Buffer strips can be prone to silting up, especially when soils are saturated Hayes et al., 1979).
Under such conditions, deposited sediment is likely to remain unconsolidated and prone to remobilisation, especially when a sequence of extreme storm events occurs or buffer strips are breached by concentrated runoff in preferential flow paths. Sediment trapping by buffer strips is commonly particle size selective with coarser particles preferentially retained (Hayes et al., 1996;Robinson et al., 1996;Hickey & Doran, 2004). Here, particle size selectivity is often buffer width dependent, with narrow 1 m buffer strips only trapping the coarsest particles . Vegetation management can influence buffer strip efficacy for reducing incoming sediment loads since, for example, long grass is more prone to lodging, which can permit preferential flow routes and reduced efficacy. Incoming flow mechanisms can influence efficacy for reducing sediment loads with, for example, concentrated flows reducing efficacy (Blanco-Canqui et al., 2004;Dillaha et al., 1986;Dosskey et al., 2002). At our experimental site, however, pervasive raindrop-impacted saturation-excess overland flow has been identified as a primary mechanism for sediment mobilization and delivery, rather than concentrated runoff (Pulley & Collins, 2019). Finally, in real-world settings, buffer strips serving agricultural land can be bypassed by field drains (Haycock & Muscutt, 1995;McKergow et al., 2003), meaning that the reductions in sediment loads relate to the surface runoff pathway. In England and Wales, a considerable proportion of farmed land has assisted underdrainage in support of productive agriculture (Robinson & Armstrong, 1988), and field drains represent an important sediment delivery pathway (Deasy et al., 2009;Zhang et al., 2016). If assisted underdrainage exists, this means that more engineered buffer strip solutions will be required to deliver multi-pathway control of sediment pollution in many parts of England and Wales. Such solutions might, for instance, include the cutting back of field drains to permit the construction of artificial wetlands (Lenhart et al., 2016) thereby delivering a "treatment-train" strategy combining buffer strips and wetlands. Where woody vegetation on buffer strips is harvested, the timing of such management activities will be critical to minimize compaction issues since these could reduce sediment trapping efficacy.

| CONCLUSIONS
Our results herein clearly indicate that the initiation of different buffer strip vegetation treatments can disturb soils and negate sediment trapping efficacy initially which should be borne in mind, especially when communicating early impacts to land managers. Thereafter, the grass treatment matured faster than the willow and deciduous woodland treatments for reducing sediment loss. Regardless of this timeline, all three vegetation treatments delivered some capacity for reducing sediment loss and our results provide new evidence for farmers, catchment managers and policy teams. Clearly, our results in this paper only report reductions in sediment loss delivered by the different buffer strip treatments, but positive impacts on additional priority pollutants for the agricultural sector, including nutrients and pesticides are likely.