Native woodland establishment improves soil hydrological functioning in UK upland pastoral catchments

Extreme rainfall and flood events are predicted to increase in frequency and severity as a consequence of anthropogenic climate change. In UK upland areas, historical over‐grazing and associated soil compaction have further exacerbated peak flood levels and flash‐flood risk along many river catchments. As a result, the reinstatement of upland woodland is increasingly seen as a key component of an integrated suite of options forming part of natural flood management (NFM) associated with a 'public money for public goods' approach to European agriculture. Nevertheless, understanding the impact of native woodland establishment on upland soil hydrology remains relatively poor. We compare physical and hydrological properties from the surface soils of establishing woodland and grazed pasture across four flood vulnerable upland headwater catchments in Dartmoor National Park, SW England. We show upland native woodland establishment is a viable soil recovery option, with a doubling of soil saturated hydraulic conductivity, increased 'wetness threshold' and reduced surface soil compaction and bulk density within 15 years of establishment. Our study supports the establishment of native woodland as an effective tool to improve the hydrological functioning of soils in upland pastoral catchments and the provision of flash‐flood mitigation 'ecosystem services'. We caution, however, that land managers and policymakers must consider past and present management, soil type and catchment location when planning new NFM schemes if environmental benefits are to be maximised and 'public money for public goods, are to be commensurate with outcomes.

Like many regions, the United Kingdom has experienced notable summer and winter flood events in recent years, resulting in significant economic and environmental damage along river catchments (Chatterton et al., 2016;Marsh et al., 2016;Marsh & Hannaford, 2007;Schaller et al., 2016). The UK is set to see an increase in flash-flood events due to projected increases in winter, spring and autumn precipitation and more intense rainfall events (Bevacqua et al., 2019;Lavers et al., 2013;Murphy et al., 2018). The 'uplands', typically >250-300 m amsl in the UK (Bunce, Wood, & Smart, 2018), are particularly vulnerable and important for managing this risk. Not only have these areas experienced greater increases in precipitation compared with lowland sites (Burt & Holden, 2010;Murphy, Hanley, Ellis, & Lunt, 2019), as the source of 68% of the UK's freshwater, they represent the principal areas of river flow generation (Robinson, Rodda, & Sutcliffe, 2013;Van der Wal et al., 2011).
Historic degradation of upland areas (Bardgett, Marsden, & Howard, 1995;Bunce et al., 2018) is, therefore, of particular concern. A legacy of soil compaction from long-term over-grazing, combined with high but falling livestock numbers, has left many upland soils in poor condition Sansom, 1999;Silcock, Brunyee, & Pring, 2012). Structural degradation and soil compaction result in the loss of macro-porous structures within the soil profile, which is of key importance for flood risk management (Alaoui, Rogger, Peth, & Blöschl, 2018;Palmer & Smith, 2013). Indeed while macro-pores typically consist of 10-15% of the soil volume, they account for 74-100% of the water movement (Alaoui & Helbing, 2006). The loss of connectivity between near-surface and sub-surface macro-pores, and the alteration of pore distribution, changes the water saturation states of soils, subsequent runoff and hydrographic characteristics after rainfall events (Dixon, Sear, Odoni, Sykes, & Lane, 2016;Meyles, Williams, Ternan, & Dowd, 2003). Heavily grazed, compacted areas can become 'active source areas' for runoff generation by lowering the threshold between dry and wet soil states (wetness threshold) Meyles, Williams, Ternan, Anderson, & Dowd, 2006). Increased runoff leads to unnaturally high flows in wet periods and decreased river base-flow in dry periods (Sansom, 1999;Shuttleworth et al., 2019), representing a significant challenge for mitigating the impacts of future seasonal precipitation regimes expected with climate change.
As part of a move towards natural flood management (NFM), woodland creation is increasingly seen as a way to deliver flood mitigation and attenuate peak river flows (Dadson et al., 2017;Lane, 2017;Nisbet, Silgram, Shah, Morrow, & Broadmeadow, 2011;Stratford et al., 2017). NFM attempts to deliver multiple ecosystem services and public benefits, including, carbon sequestration, habitat creation, water purification and public health, while minimising the social, environmental and economic costs (Burgess-Gamble et al., 2017;Iacob, Brown, Rowan, & Ellis, 2014;Lane, 2017). Trees offer NFM potential via three mechanisms: (a) higher water use (transpiration) increasing the water absorption capacity within soils and reducing surface water runoff; (b) greater hydraulic 'roughness'" and canopy interception increasing water losses via evaporation, and reducing the velocity of surface runoff through temporary floodwater retention (including via woody debris); and (c) amelioration of soil structure enhancing water infiltration, increasing water storage and reducing runoff (Birkinshaw, Bathurst, & Robinson, 2014;Nisbet et al., 2011;Nisbet & Thomas, 2006;Robinson et al., 2003).
It is potential impacts on soil properties that are of most relevance for the mitigation of extreme flood events. Evidence from the Pontbren catchment (Wales, UK) suggests woodland creation in former pasture systems had significant and rapid (<10 years) impacts on soil infiltration properties and flood risk (Carroll, Bird, Emmett, Reynolds, & Sinclair, 2004;Marshall et al., 2014). Nonetheless, our knowledge of how applicable results are to other upland catchments is limited (Burgess-Gamble et al., 2017) and our understanding of how trees affect soil hydraulic properties is more generally, surprisingly poor (Archer et al., 2013;Chandler, Stevens, Binley, & Keith, 2018;Rogger et al., 2017;Stratford et al., 2017). This knowledge seems particularly pertinent, given the recent commitment by the UK government to plant 11 million trees by 2050 (DEFRA, 2018), a move which is part of a growing interest in native woodland restoration more widely, linked to a 'public money for public goods' approach to European and UK agricultural policy (Baldock, Hart, & Scheele, 2019;Bateman & Balmford, 2018). Consequently, there is a pressing need for improved understanding on the impact native woodland creation has on soil infiltration and physical properties, especially in the upland pastures where they look set to be established.
In this study, we test the hypotheses that woodland establishment is associated with (a) lower surface soil compaction, (b) higher soil water infiltration and (c) increased soil macro-porosity. We exam-   (Perry, 2014), being, on average, 127 m higher than the surrounding lowland sedimentary basin (County of Devon; www.en-gb.topographicmap.com). Woodland in this area was cleared in prehistory since when the area has primarily been used for grazing livestock (Fyfe & Woodbridge, 2012). Consequently, vegetation in DNP is dominated by acid grassland and Atlantic heath with relatively sparse tree cover over most of the area (Mercer, 2009). In addition to this long history of (over) grazing and associated soil compaction (Sansom, 1999), the area naturally receives high levels of precipitation, with extreme rainfall events set to increase into future decades associated with climate change (Fowler & Wilby, 2010;Murphy et al., 2019). The many small streams and rivers that rise on the open moorland form 'flashy' (or 'torrential') catchments, naturally vulnerable to spate flooding (Perry, 2014). Indeed, the recent flood events in this area (Devon County Council, 2013

| Hydraulic conductivity and soil compaction
Differences in water infiltration capacities between adjacent establishing woodland and grazed pasture areas were examined over 5-weeks between September and early October 2018. A total of 8-14 sample locations were selected for each establishing woodland and pasture area within catchment sites (sample number matched for each pair)( Figure 1). Micro-topographic slope variation between sample pairs, an important variable of hillslope runoff (Marshall et al., 2014;Thompson, Katul, & Porporato, 2010), was minimised (within 5 range) and recorded using an electronic spirit level. Saturated hydraulic conductivity (Ksat) of soils was quantified using a portable, single ring infiltrometer (100 mm × 130 mm) inserted 6 cm into the soil surface (Carroll et al., 2004). Soil water infiltration was measured until a 'steady state' (Ksat) was reached (i.e., <10% difference between three consecutive readings) (Eijkelkamp, 2018), with minimal possible time gaps between comparable readings. The 'wetness threshold' (Meyles et al., 2003), defined as the volume of water (cm) required for soils to transition from a 'dry state' to a 'wet state', was calculated. In the present study, we define this as the volume of water required for the transition from initial infiltration rates at soil field capacity (soil moisture readings, Appendix A, Table A2) to Ksat.

| Soil physical analysis
Six soil cores (60 mm × 55 mm) were collected from establishing woodland and grazed pasture areas at all four catchments using a Pittman corer (0200 Soil Core Sampler; www.soilmoisture.com) in December 2018. Cores were divided into upper (2-5 cm) and lower (6-9 cm) sections, using a sharp knife, to avoid smearing. Separate cores were then saturated for 3 days, before being weighed and placed on sand suction tables at 0.05 bar (−50 cm pressure head) until a constant mass was reached; i.e., defined as no more than 100 mg between readings (Hall, Reeve, Thomasson, & Wright, 1977). Samples were re-weighed before drying at 100 C for 24 hours, and re-weighed afterwards. The particle density of fine earth soils from each core was determined using the density bottle method (British Standards, 1990).

| Soil classification
The upper soil layers (0-25 cm) were classified by digging a hexagonal-shaped pit (40-50 cm deep), and slices of undisturbed soil were used to assess the structural condition (Palmer & Smith, 2013).
Soil structure is characterised by the shape, size and degree of development of primary soil particles into naturally or artificially formed structural units, as well as the presence of voids (pores) between and within aggregates [see Hodgson (1997)]. Structural degradation assessments were paired to surface compaction readings during field visits conducted at each catchment location for improved confidence in sheer vane methodology (Appendix A, Table A3). Samples were assigned to a 'soil series' (Clayden & Hollis, 1987) by inspection of surface and sub-surface layers within pits and a 5-cm wide Edelman auger to assess deeper layers. Soil classification was conducted within adjacent establishing woodland and grazed pasture plots at each site to confirm sample locations were true pairs. threshold) parametric testing was applied to avoid the potential for type II error. Results for these variables were treated and interpreted with caution (significance determined at p < .05 rather than p ≤ .05) (Dytham, 2011). Differences between the establishing woodland and pasture areas ('land use'), between catchment and for interacting catchment vs 'land use' impacts, were assessed via two-way ANOVA.
Data are available via Murphy (2020).

| RESULTS
Ksat (1.8-fold), initial infiltration (2.7-fold) and 'wetness threshold' (1.6-fold) were significantly higher in establishing woodland than grazed pasture areas ('Land Use' effect p < .001 for all responses) ( Table 2). For Ksat and initial infiltration, the impact of woodland establishment was dependent on the catchment site (i.e., we found a  Table A4) and historic stock density higher (Appendix A, Table A1). The impact of woodland establishment in lowering surface soil compaction was greatest where SOM was higher (i.e., Colly Brook and Dean Burn) but the site with the youngest establishing woodland (Erme) evidenced the most marked changes in BD, SOM and M porosity (i.e., lower BD and increased SOM and M porosity in establishing woodland areas).
For SOM and M porosity, we find catchment differences resulted in no overall 'land use' impact for these soil properties. We also observed differences in compaction, BD, percentage of small stones (%), macroporosity (M porosity) (%) and SOM of surface soils between catchments (Table 3).
T A B L E 2 The influence of newly established woodland on surface soil hydrological properties along the four study river catchments located in Dartmoor, SW England

| DISCUSSION
The degraded nature of soils in many UK upland pastoral catchments Sansom, 1999), alongside elevated precipitation trends in these areas (Murphy et al., 2019) highlights the importance of hydrological integrity and soil recovery in flood risk management. Our results show that native woodland establishment in upland pasture areas offers a viable, and potentially rapid (7-15 years) means to reduce surface soil compaction and bulk density with concomitant benefits to Ksat and 'wetness threshold"'(i.e., soil water holding capacity).
Difference in soil wetness thresholds has considerable impact on the steepness of river flow hydrograph peaks, with wet state 'active source areas' quickly converting rainfall into either saturated overland flow or subsurface flow runoff (Meyles et al., 2003). During this 'wet state,' the water at hillslope scale can become highly connected, with topography and slope angle dominating (Appendix A, Table A5). This connectivity results in the rapid conversion of water to stream T A B L E 3 The influence of newly established woodland on surface soil physical properties along the four study river catchments located in Dartmoor, SW England  discharge via a network of ephemeral channels and rapid flow pathways, often associated with animal tracks and areas of high compaction (Meijles, Dowd, Williams, & Heppell, 2015;Meyles et al., 2006). It is likely, therefore, that establishing woodland offers effective NFM by reducing the number of wet source areas, the connectivity of hillslope moisture and the conversion of rainfall to stream discharge.
Our study is the first to measure comparable differences in the water infiltration rates of soils between establishing native woodland and pasture sites across multiple (more than two) UK upland catch-  (Chandler & Chappell, 2008).
The ability of soils to accept rainwater is highly dependent on soil type, with less permeable ('stagnant') soils reaching 'wet state' quickly, and freely draining soils, such as brown earth podzols, rarely reaching saturation if in good condition (Brady & Weil, 2008). Indeed, the catchment specificity of our results, specifically the negligible recovery of Ksat for the site with highest past grazing intensity (Holy Brook), suggests soils can reach a 'point of no return' if the elastic limit is exceeded. It is also worth bearing in mind that recovery from longterm soil compaction can range from just 6 months to more than 50 years. Specifically, the resistance (vulnerability) and resilience (recovery) of soils to compaction depend on management impacts and natural soil properties (Brady & Weil, 2008;Gregory et al., 2009), involving a complex interaction between clay content (%), SOM, water content, soil texture, biological activity, and past and previous management (Bonetti, Anghinoni, de Moraes, & Fink, 2017;Gregory et al., 2009;Pulleman, Six, Uyl, Marinissen, & Jongmans, 2005).
The most noticeable beneficial effect of woodland establishment on Ksat occurred where SOM was highest (i.e., Colly Brook and Dean Burn), supporting the view that higher SOM is linked to increased soil resilience (Bonetti et al., 2017 This finding highlights the important role of historic stock density, soil type and management in determining potential woodland NFM outcomes and soil remediation requirements. Moreover, it should not be assumed that woodland creation will always improve soil health, hydrology and peak flows (Soulsby, Dick, Scheliga, & Tetzlaff, 2017).
Indeed, while long-term evidence from upland catchments typically links higher tree cover to reductions in river discharge (Birkinshaw et al., 2014;Evaristo & McDonnell, 2019;Robinson et al., 2003), records for effective attenuation of peak flows for the most severe river flooding by woodland at catchment scale is limited (Burgess-Gamble et al., 2017;Dadson et al., 2017;Soulsby et al., 2017).
Consequently, it is important that realistic NFM expectations are communicated to the public and policymakers.
Our study demonstrates the potential of native woodland restoration in upland pasture systems to improve the hydrological functioning of soils needed to mitigate the increasing flash-flood risk expected with climate change. Establishment of native woodland where naturally freely draining soils have suffered long-term soil compaction through (over) grazing offers the highest potential increases in infiltration rates. These compacted soils are typical of mid-slope valley pastures in UK catchment headwaters. Here, not only may native woodland establishment reduce surface runoff on site, but crucially, 'soak up' runoff generated further up the hillslope (Chandler et al., 2018). Indeed, the strategic placement of native woodland will be critical to reduce surface runoff generated by both saturation excess (runoff when soil is saturated) and infiltration excess (rainfall intensity greater than infiltration rate). Flash-flooding can result from both these processes (individually and in combination), and the flood mitigation potential of native woodland will be defined by its placement (soil type, soil condition, slope angle), character (extent, tree density, tree species, management), and the seasonal climate (rainfall patterns, evaporation potential) at respective catchments.
In addition, changes in land management will likely demand tradeoffs between ecosystem service benefits (Cord et al., 2017;Iacob et al., 2014). While our study shows grazing cessation and woodland

| CONCLUSIONS
Our study provides support for the establishment of native woodland as an effective tool to improve the hydrological functioning of soils in upland pastoral catchments and the provision of flash-flood mitigation 'ecosystem services'. We caution, however, that land managers and policymakers must consider past and present management, soil type and catchment location when planning new NFM schemes if environmental benefits are to be maximised and public money for public goods' are to be commensurate with outcomes.
Despite the likelihood that upland land-use policy will increasingly promote woodland establishment within pasture systems to mitigate lower-catchment flooding, it is vital that land managers and policymakers consider the context in which NFM outcomes are expected. We recommend long-term monitoring of river flows in upland catchments to clarify and refine realistic NFM outcomes associated with native woodland establishment. Furthermore, consultation and cooperation with farmers and land managers with local soil knowledge will be essential if the environmental benefits associated with woodland are to be maximised, and to more widely ensure the sensitive implementation of nature-based solutions to climate change (Seddon et al., 2020).

CONFLICT OF INTEREST
The authors declare that they have no conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available in