Riparian buffers in tropical agriculture: Scientific support, effectiveness and directions for policy

1Durrell Institute of Conservation and Ecology (DICE), School of Anthropology and Conservation, University of Kent, Canterbury, UK; 2Department of Zoology, University of Cambridge, Cambridge, UK; 3Department of Zoology, University of Oxford, Oxford, UK; 4Lancaster Environment Centre, University of Lancaster, Lancaster, UK; 5Department of Life Sciences, University of Sussex, Brighton, UK; 6Centre for Environmental Sustainability and Water Security (IPASA), Universiti Teknologi Malaysia, Johor Bahru, Malaysia; 7Centre for Ecology and Hydrology (CEH), Edinburgh, UK; 8School of Biological and Chemical Sciences, Queen Mary University of London, London, UK; 9South East Asia Rainforest Research Partnership (SEARRP), Lahad Datu, Malaysia; 10Department of Irrigation and Drainage, Water Resources Management Section, Kota Kinabalu, Malaysia and 11Institute for Tropical Biology and Conservation, Universiti Malaysia Sabah, Kota Kinabalu, Malaysia

Globally, most research on riparian buffers concerns hydrology, water quality and quantity (Allan, 2004;Mayer, Reynolds, McCutchen, & Canfield, 2007;Tabacchi et al., 2000). More recently, there has been a growing interest in provisions for biodiversity, landscape connectivity, and ecosystem services such as pollination, pest control, carbon storage and emissions reduction (e.g. Marczak et al., 2010). However, the scientific evidence for these alleged benefits is often lacking and unavailable to policymakers and practitioners.
With the emergence of sustainability standards, and increased transparency in agribusiness and producer governments, there is a window of opportunity to inform policies in tropical countries.
Strengthened protection of riparian buffers is attracting industry interest, particularly via crop certification schemes, such as the Roundtable on Sustainable Palm Oil; Fair Trade International, and Rainforest Alliance. As producers embrace demands for sustainability, it is timely to evaluate current riparian policies and the scientific evidence base available to inform them.
Riparian policies typically prescribe a minimum width for protection (Supporting Information Table S1). However, much of the research on the ecological impact of buffer width is from North America and Europe (Figure 1). Policies are absent or poorly defined in many tropical countries, particularly the emerging agricultural markets of Central Africa (Supporting Information Table S1). Where policies do exist in tropical countries, they can be vague, highly variable between and within countries, and often loosely based on information from other locations.

| A SS E SS ING THE TROPI C AL E VIDEN CE BA S E
To assess the research and recommendations available for riparian buffers in tropical agriculture, we undertook a rapid evidence appraisal of the scientific literature (see Supporting Information Appendix S1). The search returned 847 publications. After including papers we knew had been missed by the search there were 265 studies that considered the impacts of tropical agriculture on riparian zones and waterways, of which 107 explicitly focussed on the effects of riparian buffers. Most of these 107 studies were from Brazil (31%), Malaysia (14%) and Costa Rica (11%) (Supporting Information Figure S1). Fifty per cent of the 107 studies considered terrestrial ecology, biodiversity and function; 30% hydrology and/or water quality; 18% covered freshwater ecosystems; 15% terrestrial connectivity; 11% agricultural ecosystem services and 4% carbon storage and emissions (some publications covered multiple themes).
Below we summarise the current state of knowledge, drawing on examples from the 107 studies. Very few gave specific recommendations for buffer design or management, but where they did we report them.

| Hydrology and water quality
Riparian areas regulate rainfall and run-off into freshwaters, filter sediments and pollutants, stabilise riverbanks, maintain shading F I G U R E 1 Minimum (light shading) and maximum (dark shading) riparian buffer widths recommended to protect riparian functions in temperate (evidence for North America in Collins et al., 2006) and tropical regions (material in this manuscript and Barclay et al., 2017). The number of studies on which the recommendations are based are in parentheses and low water temperatures, and provide inputs of terrestrial organic matter such as wood, leaves, seeds and insects (Allan, 2004;Tabacchi et al., 2000). Protecting non-cultivated riparian buffers also mitigates flooding, sedimentation, and nutrient run-off in farmland (Mayer et al., 2007;Tabacchi et al., 2000).
In general, buffers with greater vegetation quality provide better hydrological benefits. Across multiple studies and tropical regions, high tree cover is associated with high levels of dissolved oxygen in rivers, and low levels of sediment (Heartsill-Scalley & Aide, 2003), sand , and diseasecausing bacteria (Ragosta et al., 2011). In Malaysia, oil palm plantation streams with high riparian foliage cover are more shaded and cooler, and have more leaf litter (Chellaiah & Yule, 2018b;Luke, Barclay, et al., 2017). In mixed farmland of Nicaragua, buffers with higher leaf area index and decreased grazing intensity also have higher levels of water absorption and slower overall flow (Niemeyer, Fremier, Heinse, Chávez, & DeClerck, 2014). In contrast, the limited available evidence indicates greater forest cover may not directly result in greater nitrogen removal (Chaves et al., 2009;Connor, Nelson, Armour, & Hénault, 2013).
Landscape structure at larger spatial scales may outweigh the impact of localised riparian buffers. Forest quality and anthropogenic activities at the catchment scale were found to be important in both Malaysia and Brazil, particularly where buffer widths are <100 m Mello, Randhir, Valente, & Vettorazzi, 2017). Subtle changes in road layouts or forest cover across a catchment can strongly influence run-off, sedimentation and water temperatures (Leal et al., 2016).
Conclusion: Riparian management policies should account for multiple scales from the riparian to catchment level. Once this is considered it is likely that protecting relatively narrow buffers (c. 5-10 m) will help regulate hydrology in tropical farmland ( Figure 1).

| Freshwater biodiversity
Freshwater biodiversity is heavily affected by upstream and downstream areas as well as surrounding riparian habitat through the influence of nutrient inputs and microclimate (Pusey & Arthington, 2003). Although fish communities in agricultural streams with buffers are typically more similar to those in pristine forest than those without buffers (Giam et al., 2015;Lorion & Kennedy, 2009a), there are mixed effects on species richness, abundance, and biomass reported in the literature. For example, fish that use leaf litter and coarse substrate for hiding and foraging were found to be missing from oil palm rivers without buffers (Giam et al., 2015;Lorion & Kennedy, 2009a). As with water quality, fish diversity responds to both local stream and catchment level conditions, and may also depend on buffer widths (Leal et al., 2018;Tanaka, de Souza, Moschini, & Oliveira, 2016).
Freshwater invertebrates are central to aquatic food webs, contribute to decomposition and therefore support healthy freshwaters (Covich, Palmer, & Crowl, 1999). Macroinvertebrate composition and diversity in buffer-protected rivers is typically intermediate between that of pristine and agricultural sites, although there is notable variation between studies and crop types (Chellaiah & Yule, 2018a;Cunha, de Assis Montag, & Juen, 2015;Cunha & Juen, 2017;Lorion & Kennedy, 2009b;Luke, Dow, et al., 2017;Tanaka et al., 2016). Higher aquatic invertebrate diversity is associated with high levels of coarse particulate organic matter, coarse substrate, dissolved oxygen, low levels of slow-flowing "glide" habitat and ammonium concentrations
As habitat quality and tree species numbers are often greater in wider buffers (Lees & Peres, 2008;Metzger et al., 1997), it is difficult to discern the influence of forest structure on riparian biodiversity. For birds at least, more species are recorded in riparian areas with a more even canopy profile (Lees & Peres, 2008), or greater above-ground biomass (Mitchell et al., 2018). For this reason, exclusion of cattle from riparian buffers has been recommended in Brazil (Mendoza et al., 2014), which leads to vegetation regeneration (Griscom, Griscom, & Ashton, 2009) and improved bird diversity (Lees & Peres, 2008).
Several studies have investigated the role of isolation from forest in structuring buffer communities. Notably, buffers near to large tracts of forest support larger bat populations (Galindo-González & Sosa, 2003), and more diverse dung beetle (Barlow et al., 2010) and bird assemblages (Keir et al., 2015;Lees & Peres, 2008). However, the long-term viability of terrestrial biodiversity in buffers remains open to question as edge effects may cause continual habitat degradation, and so the extent to which buffers act as ecological sinks is unclear (Beier & Noss, 1998).
Conclusion: Positive associations exist between riparian buffer width and terrestrial tropical biodiversity.
A buffer width of 100 m each side of the bank would help support multiple animal and tree taxa regardless of agricultural land use or geographic location. species. This is a key knowledge gap that is in critical need of further research to inform policy.

F I G U R E 2
The proportion of (a) bird and (b) dung beetle species found in riparian buffers of increasing width in oil palm plantations compared to riparian areas in nearby forest (figures redrawn from Mitchell et al., 2018 and Similar trends were apparent in Borneo, although riparian carbon stocks were highly variable (Mitchell et al., 2018). Data from Brazil indicated that effective restoration of degraded riparian habitats could reverse high carbon losses associated with drainage and erosion, and result in a net increase of 70% carbon storage (Wantzen et al., 2012).
The effects of buffers on emissions is limited to a single study, which found similar N 2 O emissions in riparian forest and fertilised maize farms in the dry season, but higher emissions in the buffers in the wet season. However, the buffer still provided positive benefits such as reduced nitrogen inputs to freshwater (Kachenchart et al., 2012).
Conclusion: There are few empirical studies on the carbon dynamics of riparian buffers in tropical agriculture, and only one on the effects of buffers on GHG emissions. Further research is urgently needed.

| Agricultural services
Riparian buffer habitat could improve agricultural yields and production costs via pollination, pest control, decomposition, and water provision services; or agricultural productivity could fall due to increased exposure to pest and predators (Power, 2010;Zhang, Ricketts, Kremen, Carney, & Swinton, 2007). In Costa Rica, pollination rates in coffee farms decreased near riparian forest buffers compared to those by a non-riparian remnant (Ricketts, 2004). In Borneo, oil palm sites near and far from buffers had a similar diversity of ants and dung beetles, as well as similar levels of dung decomposition (Gray, Simmons, Fayle, Mann, & Slade, 2016), ant scavenging (Gray et al., 2015) and defoliating pests . Moreover, the presence of forest remnants, including buffers, had little impact on oil palm yield in Borneo (Edwards, Edwards, Sloan, & Hamer, 2014).
Conclusion: Evidence for 'spillover' of diversity and services from riparian buffers is limited. However, there is likely a balance between services and disservices provided by buffers in tropical farmland.

| D IREC TI ON S FOR SCIEN CE AND P OLI C Y
Although additional research on tropical riparian buffers is clearly needed, several policy-relevant conclusions can be made from the existing literature: 1. Riparian buffers should be maintained and restored. Sufficient evidence exists to confirm buffers improve water quality and hydrological processes, support biodiversity, and contribute to landscape-wide carbon storage in tropical farmland. However, further studies are needed on connectivity, GHG balance and ecosystem service provision. As biodiversity, carbon storage, hydrology and water quality improve when vegetation heterogeneity, canopy cover and biomass in buffers are high, retaining natural vegetation in buffers is essential. Research exploring thresholds or tipping points of habitat quality effects on riparian functions is currently lacking, and would be informative for restoration.

2.
Wider buffers are better than narrow ones. Effective buffer widths will vary by function (Figure 1). Currently, width thresholds are largely based on hydrology and water quality research, with guidelines usually recommending widths of 10-100 m (Supporting Information Table S1). However, biodiversity studies from Latin America and Southeast Asia indicate 40-200 m on each riverbank is needed, depending on the taxon studied, and whether the buffer is isolated within the agricultural matrix. Larger or widerranging species may require large buffer widths, and so decision trees that allow context-specific recommendations are needed.

Catchment-level processes should be considered alongside riparian
processes. The effectiveness of buffers for aquatic functions can be confounded by how land is managed upstream. Similarly, the value of buffers for terrestrial biodiversity is linked to habitat availability over the broader landscape. Efforts should be made to protect habitat in stream headwaters, and the location of roads and agricultural activities should be carefully planned across whole catchments to maximise benefits. The relative roles of riparian-versus catchment-level land cover remains poorly understood, especially in the tropics, and studies that quantify variation on both these scales (Iñiguez-Armijos, Leiva, Frede, Hampel, & Breuer, 2014) will be very valuable to inform policy.
We suggest four critical components needed to implement effective riparian policies in tropical countries: 1. Clear buffer design protocols are needed to decide how much riparian habitat should be retained in tropical agriculture. A wide range of variables are assessed to determine riparian buffer widths in some temperate locations (Figure 1), and could form a basis for similar function-specific policies in the tropics, noting that a one size fits all width threshold is insufficient. For example, the High Carbon Approach (http://highcarbonstock.org) uses a decision tree incorporating patch area as a criterion for forest conversion, but could be expanded by incorporating minimum width thresholds for riparian buffers under varying contexts. Such decision-making tools should facilitate buffer design for the landscape in question, incorporating key factors (e.g. size of river, connectivity and matrix type) and automated computational processes. Examples include the Riparian Zone Estimator Tool (RipZET; https://www.sfei.org/projects/ripzet) 2. Rapid riparian survey protocols to assess and monitor buffer effectiveness should be developed using a suite of standard indicator species and functions. We suggest expanding existing toolkits, such as the forest integrity assessment tool (www.hcvnetwork.org/resources/forest-integrity-assessmenttool) and the Toolkit for Ecosystem Service Site-based Assessment (TESSA) (Peh et al., 2013), to riparian contexts.

Guidelines for rehabilitation and restoration of riparian areas in
tropical agriculture are notably absent from the published literature, but sorely needed. Recent oil palm certification standards offer some suggestions , and experiments in Sumatra are testing various approaches (http:// oilpalmbiodiversity.com/). The Riparian Ecosystem Restoration in Tropical Agriculture (RERTA) project provides a research design template that could be adapted and replicated in other countries and agricultural systems to allow informed guidelines at landscape-scales. We also suggest expanding on existing initiatives such as the Riparian Restoration Plant Database (https://www.ctahr.hawaii.edu/rnre/Riparian_Restoration_ Plant_Database.asp) 4. Local technical support including capacity to map streams and land boundaries, expertise to help with monitoring and restoration, and schemes to increase policy awareness among land managers, are often lacking, meaning that riparian guidelines may fail to deliver benefits on the ground (Nunes et al., 2015). In addition to the open sharing of topographical data to accurately delimit watercourses, historical maps would be particularly useful to overcome shifting baselines, whereby deforested landscapes tend to lose perennial streams that could otherwise retain some functioning if buffered appropriately. Addressing this issue requires closer collaboration and improved data sharing between scientists, policy-makers, environmental managers and local practitioners to build local capacity, and to ensure that riparian science is translated into policy where it is needed most.

DATA ACCE SS I B I LIT Y
Data have not been archived because this article does not use data.