Modelling the Hydrological Effects of Woodland Planting on Infiltration and Peak Discharge Using HEC-HMS

: Woodland planting is gaining momentum as a potential method of natural food management (NFM), due to its ability to break up soil and increase infltration and water storage. In this study, a 2.2 km 2 area in Warwickshire, England, planted with woodland every year from 2006 to 2012, was sampled using a Mini Disk infltrometer (MDI). Infltration measurements were taken from 10 and 200 cm away from the trees, from November 2019 to August 2021. Two individual hydrological models were built using the US Hydraulic Engineering Center Hydrological Modelling System (HEC-HMS), to model the effects of infltration change on peak fows from the site throughout the summer and winter. The models were calibrated and validated using empirical data; the Nash and Sutcliffe Effciency (NSE) was used as an indicator of accuracy. Results from this study show that woodland planting reduced peak fow intensity compared to impermeable land cover by an average of 6%, 2%, and 1% for 6-h, 24-h, and 96-h winter storms, respectively, and 48%, 18%, and 3% for 6-h, 24-h, and 96-h summer storms, respectively. However, grassland simulations show the greatest reduction in peak fows, being 32%, 21%, and 10%, lower than woodland for 6-, 24-, and 96-h winter storms, respectively, and 6%, 3%, and 0.5% lower than woodland for 6-, 24-, and 96-h summer storms, respectively. Peak fows the are 6.52% higher than the for 50%, 10%, 2%, and 1% AEP events, respectively; however, only showing a difference of 1.28%, 0.98% for 50% and 10% AEP events, and 0 change for the 2%, and 1% AEPs in the summer. Peak fows show less variation between summer and winter for all land cover types, compared to the trends seen in the 6- and 24-h duration simulations.


Introduction
Urbanisation and the replacement of permeable and vegetated surfaces to impermeable surfaces, such as asphalt and concrete, reduces lag times and increases peak fows in receiving watercourses, infuencing the likelihood and severity of high-fow or fooding events across the UK [1,2]. Coupled with this, the global climate is predicted to change in ways unseen in recorded history [3]. In the UK, sea levels are expected to rise, the frequency of extreme weather events will increase, summers will become hotter and drier, and winters will become warmer and wetter [3,4]. Consequently, authorities responsible for managing food risk in the UK have increased investment in alternative, more sustainable methods of mitigating fooding, such as natural food management (NFM) techniques [2,5,6].
The design and operation of any NFM feature is based primarily on emulating the natural hydrology of a catchment as it was prior to human interaction, with the intention of reducing fuvial food risk [1,7]. Common NFM methods can be categorised into those that (a) reduce hydrological or hydraulic connectivity; (b) create storage; or (c) increase infltration [8]. Examples of these methods include vegetation planting to increase infltration and interception, changing animal grazing and farming routines to reduce compaction and increase lag time, and reconnecting or introducing offine marshlands and mudfat areas, to slow the fow of fooding water during a storm event [7,[9][10][11].
Whilst continuous academic investigations into the real-world applicability of NFM methods are ongoing, NFM implementation is slow, primarily due to the lack of long-term evidence-based studies [1,12]. One method of NFM that is assumed valuable, although under-investigated, is woodland planting [4,13,14]. Tree roots break up the surrounding soil, increasing infltration rate and water storage capacity, whilst simultaneously offering a higher opportunity for interception and evapotranspiration [15,16]. The value of woodland planting has been identifed by the UK government, who have allocated GBP 4 million to various organizations to increase woodland planting for food risk reduction, and the Environment Agency (England), who have been awarded GBP 1.4 million for the same purpose [17]. However, whilst some studies have attempted to investigate the link between woodland planting and soil infltration characteristics [4,15,18], evidence-based studies focusing solely on the impacts of changing infltration as a result of woodland planting are sparse, posing a challenge to researchers in this discipline.
In light of the changing climate, and the predicted increases to food frequency and severity, the policy shift towards NFM methods, and the lack of evidence-based studies investigating the feasibility of woodland planting as a method of NFM, this study aims to determine the extent to which woodland planting has infuenced infltration at a site in central England [1][2][3][4]. Two hydrological models were built using HEC-HMS, calibrated and validated using the NSE method, and simulations were undertaken to predict the ability of woodland to increase infltration and reduce peak runoff to the receiving watercourse.

Infltration Data Collection
Infltration data were collected once every two weeks from specifc areas of a 2.2 km 2 site in Warwickshire, UK (52. Tremula). Additionally, infltration data were collected from a plot planted in cc.1900 (Quercus Petraea), and a control site consisting of a grassland area that pre-exists the HofE forest. The grassland area was sampled for comparison with the woodland areas, and the cc.1900 area was sampled to provide information on the infltration characteristics of mature woodland, and for comparison to more recently planted areas. Figure 1 shows the locations of the infltration sample plots and sampling locations. Infiltration rates are influenced by soil texture [20][21][22]. Therefore, soil samples were extracted from the surface (~5 cm depth) of the soil surrounding the area of MDI measurement using a trowel. The soil texture of these samples was then determined using a LaMotte soil texture testing kit [23]. The percentiles of sand, silt, and clay for each soil were compared against the UK soil texture triangle to determine the classification name of the sample soils. The percentiles and soil texture classifications of the sample area are shown in Table 1.   Infltration rates are infuenced by soil texture [20][21][22]. Therefore, soil samples were extracted from the surface (~5 cm depth) of the soil surrounding the area of MDI measurement using a trowel. The soil texture of these samples was then determined using a LaMotte soil texture testing kit [23]. The percentiles of sand, silt, and clay for each soil were compared against the UK soil texture triangle to determine the classifcation name of the sample soils. The percentiles and soil texture classifcations of the sample area are shown in Table 1. Infltration measurements were collected (10 and 200 cm away from the base of the sample trees) using a Mini Disk infltrometer (MDI) [24]. The 10 cm proximity was chosen to represent the infuence of the tree on infltration directly adjacent to the trunk, and the 200 cm proximity was chosen to account for potential root spread due to tree growth [25][26][27]. As the MDI required a watertight seal with the sample soil, vegetation was removed from the surface of the soil before infltration measurement proceeded. See  Collecting infiltration data from two proximities allowed for the comparison and representation of both proximities in hydrological modelling. Infiltration measurements were carried out until three consecutive volumes were recorded (~10 min) and replicated three times at both proximities around the sample tree [15,28].
It is acknowledged that, in addition to infiltration, woodland can influence hydrology through interception and evapotranspiration, which is also accounted from in this study (and discussed in Section 2.5.3) [7,9,10]. These additional factors are important for justifying the use of woodland as a method of NFM, and are accounted for throughout the modelling and results of this study (this is discussed further in Section 2.5.3). A total of 1617 individual infiltration measurements were collected from October 2019 to August 2021; 888 from the 10 cm proximity (including a grassland control), and 729 from the 200 cm proximity. Infiltration data were not collected from March to July 2020 due to the UK Collecting infltration data from two proximities allowed for the comparison and representation of both proximities in hydrological modelling. Infltration measurements were carried out until three consecutive volumes were recorded (~10 min) and replicated three times at both proximities around the sample tree [15,28].
It is acknowledged that, in addition to infltration, woodland can infuence hydrology through interception and evapotranspiration, which is also accounted from in this study (and discussed in Section 2.5.3) [7,9,10]. These additional factors are important for justifying Water 2021, 13, 3039 4 of 22 the use of woodland as a method of NFM, and are accounted for throughout the modelling and results of this study (this is discussed further in Section 2.5.3). A total of 1617 individual infltration measurements were collected from October 2019 to August 2021; 888 from the 10 cm proximity (including a grassland control), and 729 from the 200 cm proximity. Infltration data were not collected from March to July 2020 due to the UK national COVID-19 lockdown.

Hydrometric Data Collection and HEC-HMS Modelling
Rainfall data (in mm) were collected via a tipping bucket rain gauge (accuracy ±4% between 0.2 and 50 mm) located at the NextGen (2020) waste water treatment plant, 1 km north of the HofE site (see Figure 1). Whist not located directly on the HofE site, the rain gauge is the only one located within eyeshot of the study site and is representative of local rainfall [29][30][31][32]. The stage (in mm) of the study site tributary was recorded every 15 min via a pressure transducer (accuracy ± 0.05% FS) located at the downstream end of the site watercourse ( Figure 1). Upon instillation, the transducer was calibrated to a fow meter, enabling the generation of a ratings curve to determine watercourse fow (in L/s) relative to measured water pressure [33,34].
The US Hydrologic Engineering Center Hydrological Modelling System (HEC-HMS) allows the application of various numerical methods to each stage of the rainfall-runoff process, meaning a model can be tailored to serve a very specifc purpose dependent on the required output [35][36][37][38]. The software is also capable of modelling simple hydraulic elements, allowing watercourses to be inputted as either user-defned open channels, or specifed-shape culvers/pipes.
For this study, ArcMap 10.6.1 was used in conjunction with a 1 m resolution digital terrain model (downloaded from the Department for Environment, Food and Rural Affairs) [39], and the shapefle of the study site tributary to delineate the watershed and generate a fow accumulation fle. The area defned by the watershed delineation was overlain with HofE feld boundaries. Each feld boundary was treated as a separate subcatchment model input, and these were individually digitised to represent the 10 cm woodland proximity, the 200 cm proximity, and the grassland areas individually. This was implemented so that the model would account for the collected infltration data of both the 10 and 200 cm proximities and the grassland (taken from the control site) data separately.
The watercourse of the HofE site was added to the model in a series of reaches and junctions, the dimensions of which were validated from cross-section measurements taken on site. After the watercourse, sub-catchment nodes were added. Each node represented a different land cover type of each sub-catchment, meaning there were up-to three nodes for each catchment (10 cm proximity, 200 cm proximity, grassland, and impermeable). ArcGIS, the fow accumulation, and personal knowledge of the site (feld visits, observations of fow paths during storm events, topography) were used to determine where the nodes representing the different plots should connect to the tributary. The HEC-HMS model is shown in Figure 3.
taken on site. After the watercourse, sub-catchment nodes were added. Each node represented a different land cover type of each sub-catchment, meaning there were up-to three nodes for each catchment (10 cm proximity, 200 cm proximity, grassland, and impermeable). ArcGIS, the flow accumulation, and personal knowledge of the site (field visits, observations of flow paths during storm events, topography) were used to determine where the nodes representing the different plots should connect to the tributary. The HEC-HMS model is shown in Figure 3. The Muskingum-Cunge routing method (Equations (1)-(6)) was used for modelling tributary flow [38,40,41]: The Muskingum-Cunge routing method (Equations (1)-(6)) was used for modelling tributary fow [38,40,41]: where Q is discharge, j is a spatial index, n is time index. C 0 , C 1 , and C 2 are calculated as follows [42]: K and X are calculated as follows [42]: where Δx is reach length, c is food wave celerity, q is unit width discharge, and S o is channel bed slope [42] Water 2021, 13, 3039 6 of 22 The initial and constant loss method (Equation (7)) was used to simulate the collected infltration data (Section 2.1), the constant rate element is calculated as follows: where pe t is excess overland fow, p 1 is precipitation depth, f c is the maximum potential rate of precipitation, and I a is initial loss. The initial and constant method (equation) was chosen for use in this study due to the nature of collected infltration data; however, it is acknowledged that other authors have modifed the Richards equation to account for root water uptake in soils [43][44][45]. Whilst these physics-based infltration models can account for variations in soil texture, the wetting front and unsaturated hydraulic conductivity [45][46][47]; they can be prone to error and require in-depth data of the soil column for accurate execution compared with the initial and constant method used here. The Snyder unit hydrograph transform (Equations (8) and (9)) was used to simulate the observed runoff and lag times of the catchment areas, Snyder's equation for lag time is [48]: where T lag is the catchment lag time (hours), C t is the catchment gradient coeffcient, L is fow path length (km), and L c is length of fow path from outlet to closest point of the catchment centroid (km). For peak discharge, is [48]: where Q p is peak discharge related to 1 cm of effective rainfall (m 3 s −1 ), A is catchment area (km 2 ), and C p is an empirical coeffcient of peak intensity.
The constant monthly basefow method was used to simulate antecedent basefow of the site, which applied a user-defned constant fow to all models as required (see Section 2.5.4) [40,49,50].
Two identical models, 'winter' and 'summer' were constructed and independently calibrated and validated (see Sections 2.3 and 2.4) to generate the results for this study. This approach was decided as a result of observed hydrological variations across the site from dry-to-wet seasons. The winter model is representative of hydrological data (infltration, telemetry, rainfall) from October to March (2019/2020 and 2020/2021), and the summer model from April to September (2019/2020 and 2020/2021). These timeframes are based on UK average annual rainfall and temperature data, as defned by the Met Offce (2021).

Model Calibration
Model calibration involved setting the initial basefow to match that of the observed tributary value for the selected event, then gradually adjusting unobserved model parameters until the modelled output best simulated those of the observed values [35,37,51,52]. Regarding the observed model parameters, infltration was the key parameter for the hydrological model, it had been collected from October 2019 to August 2021 (with a break from March to July 2020 due to , and this parameter could not be changed during the calibration process. The same applied to rainfall and basefow, as these had been observed through use of the rain gauge and in-channel telemetry. This meant the only adjustable parameters were the lag times and peaking coeffcients of the Snyder unit hydrograph transform (Equations (8) and (9)), so these parameters were adjusted through trial-and-error until one set of Snyder values (based on site observations and observed and simulated model output) could be used across all events and produce a similar outcome to the observed fow. This process was undertaken for both the summer and winter models, using available data from the time periods specifed in Section 2.2. The Nash and Sutcliffe (1970) Effciency (NSE) method was used to determine the closeness-of-ft between the simulated and observed values in calibration. The NSE equation is displayed as follows: where Y obs is the observed discharge, Y sim is the simulated discharge, and Y mean is the i i mean of observed discharge. Table 2 shows the timeframes of the model calibration for winter and summer models, in addition to the individual and mean NSE values. The NSE method produces a value between 0 and 1, as an indication of how well the simulated dataset (Y sim ) fts the observed dataset (Y obs ) [53,54]. A value of 1 indicates a perfect ft i i between the two datasets, whereas a value of 0 (or a negative value) indicates a poor ft. The mean NSE of both the winter and summer calibrations are 0.65. Seen from Table 2, the shorter duration events (24-and 72-h) showed a lower calibration NSE output compared with longer duration events (96-and 120-h). Across the summer calibration events, the lowest NSE value of 0.29 was produced by the 96-h duration, infuencing the average NSE.

Model Validation
The model was validated, again, using the NSE method for determination of relationship; however, different storm events were used to those used in calibration (using the same durations) (Table 3) [37,55]. Table 3 shows the timeframes of model validation for winter and summer models, in addition to the individual and mean NSE values. The mean NSE of the winter validations is 0.87, and summer 0.44. The NSE average for validation events in the winter is high (0.87) indicating that the winter model is very effective at modelling the observed response from the site; whereas the summer validation NSE is 0.44, indicating that the output from the summer models is less accurate than the The mean NSE of the winter validations is 0.87, and summer 0.44. The NSE average for validation events in the winter is high (0.87) indicating that the winter model is very effective at modelling the observed response from the site; whereas the summer validation NSE is 0.44, indicating that the output from the summer models is less accurate than the winter model. Figures 6 and 7 show the observed and simulated discharge flow graphs for winter and summer model validation events.   The chosen calibration and validation datasets were selected primarily due to the scarcity of collected telemetry data from the study site and the timescales from which the data needed to be collected. Therefore, replication of calibration and validation events for the purposes of accuracy was not possible in this instance. Data scarcity for calibration and hydrological modelling is a common occurrence in the field of hydrological model- The chosen calibration and validation datasets were selected primarily due to the scarcity of collected telemetry data from the study site and the timescales from which the data needed to be collected. Therefore, replication of calibration and validation events for the purposes of accuracy was not possible in this instance. Data scarcity for calibration and hydrological modelling is a common occurrence in the feld of hydrological modelling. The methods presented throughout this section could be adopted by other researchers in the discipline aiming to simulate similar scenarios to those presented here.

Precipitation and AEP Events
The Flood Estimation Handbook (FEH) was used to generate the design storms used in simulations [56]. Annual exceedance probabilities (AEPs) representing 50% (1 in 2), 10% (1 in 10), 2% (1 in 50), and 1% (1 in 100) rainfall intensities were simulated over 6-, 24-, and 96-h durations [56][57][58]. The 24-and 96-h durations were chosen to test the short-to-medium scale impacts of woodland planting on infltration. The 6 h duration was chosen due to the requirement of all UK sustainable drainage systems (SuDS) to be tested to this level [59,60]. The rainfall intensities were chosen for similar reasons: the modelled results would enable further understanding regarding the true ability of woodland planting to mitigate runoff from low intensity (50% AEP) to very high intensity (1% AEP) storms, offering insight in to their use as a method of NFM.

Infltration Data
The collected infltration data (Section 2.1) were interpreted and included as a primary focus of the modelling process. As this study focuses on the impacts of woodland planting on runoff, the mean infltration rate (in mL) from every sample site at 10 (Table 1).

Interception
As the sampled woodland is deciduous, interception needed to be considered as it would vary seasonally across the study site [61][62][63]. Interception loss was not empirically monitored for this study; however, it was accounted for, considering the interception loss for grassland to be negligible (<10%) [11,64], and the interception from broadleaves to be between 10-34% (mean 24.25%) [65,66]. FEH values were adjusted in the summer model to account for the rainfall loss due to interception, as simulating the site in both winter and summer with uniform rainfall would not account for any interception loss encountered. This method allows for the inclusion of interception loss in the model without the use of specialised equipment or continuous monitoring, and was an important process, as the infuence of interception loss would vary seasonally across the site, both at present and in the future.

Basefow
Antecedent basefow had to be calculated, as neglecting to consider this parameter could result in the total discharge from each simulated storm being inaccurate, and not represent true site conditions [67,68]. To calculate the basefow, telemetry data from both winter and summer periods (Section 2.2) were averaged, the average basefow for winter s −1 models is 0.284 m 3 s −1 and for summer models is 0.239 m 3 .

Hydrological Simulations
Three scenarios were simulated using both the winter and summer models with the intention of simulating the peak outfow of the site dependent on varying observed infltration. Infltration values for the frst simulation scenario were derived from the collected infltration data from the HofE site (Section 2.1); these data were simulated to refect the "current HofE (woodland) land cover" as it is at present, and would allow a representation of current site peak discharge. Infltration values for the second simulation scenarios were altered to represent sites discharge if it was impermeable land cover (developed). The rationale for this originates in Section 1, where it is acknowledged that urbanisation is a key infuence of rising food risk in the UK [1,2]. Infltration for scenario three were adjusted to represent infltration collected from the grassland control site (Figure 1). This was to enable a comparison (Section 3) between peak fows from current woodland cover, impermeable land cover, and grassland land cover. It was decided to use the above scenarios as they could be based on collected infltration data from the site, and provide an accurate representation of the hydrological variations of the sampled study site compared to using published values. Figure 8 shows the peak discharge of the 6-h summer and winter simulations, Table 4 shows the tabulated data with the discrepancy between land cover types shown in comparison to the current HofE site as a percentage.       Figure 8 and Table 4 show that peak runoffs from the impermeable land cover simulations are higher than current and grassland simulations across all modelled storm events in both summer and winter; however, the discrepancy in the winter was slight. Peak fows from HofE land cover were 9.84%, 5.95%, 4.39%, and 3.82% lower than impermeable for a 50%, 10%, 2%, and 1% AEP events, respectively. However, in the summer this difference was higher, with peak fows being 94.55%, 47.62%, 27.91%, and 22.49% greater across 50%, 10%, 2%, and 1% AEP events for impermeable land cover compared to HofE. Grassland peak fows for winter and summer differ signifcantly, with summer peak fows being 84.71%, 41.67%, 23.52%, and 18.67% higher for 50%, 10%, 2%, and 1% AEP events, respectively. In the winter, grassland shows a 47.54%, 34.52%, 25.44%, and 22.14% reduction in peak fows compared to HofE land cover; this reduction is less in the summer, being 10.91%, 5.71%, 4.07%, and 3.35% for 50%, 10%, 2%, and 1% AEP storms. Figure 8 and Table 5 show that impermeable cover produces the highest peak fows, compared to the HofE and grassland simulations, over both summer and winter. Current HofE site values are similar to impermeable values throughout the winter; however, this trend is not seen in the summer, where HofE site values are much lower. Grassland produces the lowest peak fows overall; however, grassland values are more similar to current HofE site values in the summer.  Figure 9 shows the peak discharge of the 24-h summer and winter simulations, Table 5 shows the tabulated data with the discrepancy between land cover types shown in comparison to the HofE site as a percentage.

Results
HofE site values are similar to impermeable values throughout the winter; however, this trend is not seen in the summer, where HofE site values are much lower. Grassland produces the lowest peak flows overall; however, grassland values are more similar to current HofE site values in the summer. Figure 9 shows the peak discharge of the 24-h summer and winter simulations, Table  5 shows the tabulated data with the discrepancy between land cover types shown in comparison to the HofE site as a percentage.   Figure 9 and Table 5 show the highest peak runoff of all simulated land cover types over both winter and summer to be generated by the impermeable land cover; being on average 2.07% higher than HofE land cover in the winter, and 18.31% higher in the summer over all AEPs. The impermeable cover, however, produces the same peak runoff as the HofE site for the 50% AEP in the winter, and peak flows for the impermeable land  Figure 9 and Table 5 show the highest peak runoff of all simulated land cover types over both winter and summer to be generated by the impermeable land cover; being on average 2.07% higher than HofE land cover in the winter, and 18.31% higher in the summer over all AEPs. The impermeable cover, however, produces the same peak runoff as the HofE site for the 50% AEP in the winter, and peak fows for the impermeable land cover for 10%, 2%, and 1% AEPs in the winter are on average higher than that of the HofE site land cover by only 1.39%. A similar trend is seen in the summer, where peak fows from HofE land cover are only 0.24% lower than impermeable values for all AEPs. All 24-h events show less of a discrepancy between the HofE and impermeable land covers in the winter compared to the 6-h simulations, showing that, as storm duration increases, the fows from impermeable and HofE land cover become more similar in the winter months. Similar to the 6-h duration, grassland peak fows are lower than impermeable and HofE site land cover throughout the winter, however HofE site and grassland peak fows vary less in the summer. In the winter, grassland peak fow is 32.10%, 23.15%, 16.78%, and 14.29% lower than HofE land cover for 50%, 10%, 2%, and 1% AEPs; however, in the summer, grassland is only 4.65%, 2.99%, 2.08%, and 1.79% lower than the current site for all respective AEPs.
Results from the 24-h simulations are similar in trend to those of the 6-h simulations. Impermeable cover produces the highest peak fows over both summer and winter. These values are similar to HofE site discharge in the winter, but not the summer. Grassland produces the lowest peak fows; however, grassland values are more similar to HofE site values in the summer. Summer peak fows for all land cover types are higher than winter values. Figure 10 shows the peak discharge of the 96-h summer and winter simulations; Table 6 shows the tabulated data with the discrepancy between land cover types shown in comparison to the HofE site as a percentage. Impermeable cover produces the highest peak flows over both summer and winter. These values are similar to HofE site discharge in the winter, but not the summer. Grassland produces the lowest peak flows; however, grassland values are more similar to HofE site values in the summer. Summer peak flows for all land cover types are higher than winter values. Figure 10 shows the peak discharge of the 96-h summer and winter simulations; Table 6 shows the tabulated data with the discrepancy between land cover types shown in comparison to the HofE site as a percentage.   As apparent from Figure 10 and Table 6, the results from the 96-h simulations show similar trends to 6-and 24-h simulations; however, the discrepancy in peak fow between land cover simulations are less notable, as are the seasonal variations in peak fows. Impermeable cover exhibits the highest peak fows compared to HofE and grassland cover; however, this difference is marginal. Impermeable peak fows are 1.25%, 0.99%, 0.80%, and 0.72% higher in the winter, and 5.13%, 2.94%, 1.54%, and 1.39% higher in the summer than HofE land cover for 50%, 10%, 2%, and 1% AEP events, respectively. There is more variation between HofE land cover and grassland in the winter than the summer. Peak fows from the grassland cover are 16.25%, 10.89%, 7.20%, and 6.52% higher than the HofE site in the winter for 50%, 10%, 2%, and 1% AEP events, respectively; however, only showing a difference of 1.28%, 0.98% for 50% and 10% AEP events, and 0 change for the 2%, and 1% AEPs in the summer. Peak fows show less variation between summer and winter for all land cover types, compared to the trends seen in the 6-and 24-h duration simulations.
The results displayed in Figure 10 and Table 6 show that the 96-h duration simulations show less variation in peak fows across all land cover types. There is very little variation in peak fows between land cover types; however, impermeable is slightly higher compared to both current HofE land cover and grassland cover over both summer and winter.
Overall, Figures 8-10 and Tables 4-6 show that, at present, impermeable cover produces the highest peak fows over all durations and storm intensities compared to other simulated land cover types. However, this is somewhat expected as it is known that the increase in impermeable surface cover is driving the push towards NFM (see Section 1) [1,2,10]. The current land cover of the site shows less of a peak fow compared to impermeable; however, the discrepancy is small and reduces with increased storm duration and intensity. The grassland simulations result in the lowest peak fows, regardless of season or storm scenario. The summer simulations show signifcantly higher peak fows compared to winter values across all land cover types in the lower duration storms (6-and 24-h); however, this is less signifcant in the higher duration simulations (96 h).

Discussion
The results of the HEC-HMS models presented throughout Section 3 show that woodland planting across the HofE site has reduced peak runoff compared to if the entire site was impermeable, by an average of 6% for 6-h, 2% for 24-h, and 1% for 96-h duration events in the winter; and 48%, 18%, and 2.7% for 6-, 24-, and 96-h durations in the summer. This fnding reinforces the beneft that woodland planting can have on increasing the surrounding soils infltration potential, storage potential, resultantly reducing peak fow from the study area. Whereas the current HofE site does reduced peak fow compared to the impermeable scenario, an entirely grassland catchment shows the greatest reduction in peak fow, being 32%, 21%, and 10% lower than the current site in the winter, and 6%, 2%, and 0.5% lower than the HofE site in the summer for 6-, 24-, and 96-h duration storms, respectively.
It is worth considering that the reduction in peak fows exhibited by grassland compared to both the impermeable and current HofE site reduces as storm duration and intensity increases. This can be explained through considering both the age, and relative root spread of the woodland species sampled [69,70]. As discussed in Section 2.1, aside from the cc.1900 (Quercus Petraea) woodland, the oldest trees sampled were planted in 2006 (Betula Pendula) and the youngest in 2012 (Populus Tremula). Therefore, the 2006 trees have only been developing for 15 years, and the 2012 trees for 9. Betula Pendula reaches its ultimate height at around 60 years from planting, and can live for up to 100 years in total, meaning that the infltration data collected and simulated in this study is only representative of the beginning of the likely effects that this tree will have, and infltration will only improve as the tree (and its root system) develop, as the tree approaches maturity [71][72][73][74]. A similar rationale can be applied to the Populus Tremula (2012) tree; the growth of Aspen in the UK slows at around 30 years, and trees can live for 100-120 years [75][76][77]. Thus, if this project was to be replicated in 20 years, tree roots would have developed, breaking up the surrounding soil, and infltration would likely be greater due to the increased porosity [15,16]. Consequently, the modelled data show that, whilst some reduction in runoff is possible during the growth phase of trees, it will not be until they mature that the total potential reduction is demonstrated in comparison to grassland peak fow reduction.

Woodland Planting Mentality
The fndings of this project are signifcant when considering the way in which woodland areas are currently managed, regarding the growth and felling of trees and the removal of mature woodlands to make way for either newer areas of woodland or, more signifcantly, impermeable developments [4,15,78]. It is not uncommon for mature(ing) woodland to be removed to make way for impermeable developments, which signifcantly alters the local hydrology of an area, sealing-off once permeable areas and excluding them from participating in infltration processes [4,15,79,80]. Felled trees as a result of development are commonly 'balanced out' by planting saplings in alternate locations, however newly planted saplings will not have a comparable moderating impact on food risk compared with the felled mature trees [18,74,80,81]. This project has demonstrated the above through presenting modelled results, showing that young trees, whilst they do reduce peak fows compared to impermeable land cover, have not yet developed the root systems, and infuenced surrounding infltration rates, to the extent they will with maturity.

The Infuence of Precipitation, Interception, and Model Calibration
Apparent throughout Section 3 is that summer peak fows over all simulated land cover types are higher than the corresponding winter values, this discrepancy reduces in the 96-h duration storm; however, it is more apparent in the 6-and 24-h duration storms. As discussed in Section 2.5.3, the hydrological effects of interception were accounted for in the summer through reducing the precipitation received. Therefore, it would be assumed that summer peak fows would be reduced; however, this is not the case. A reason for this output may be effect of surface pooling (as discussed above); however, another explanation may be due to the parameters used in calibration. The summer and winter models were calibrated individually (Section 2, Table 2), and a set of parameters used for each. The purpose of calibration is to align the observed and simulated outputs as closely as possible over varying events, leaving a fnal set of values that will produce a reliable output [82,83]. In this case, the only variables that could be adjusted were the components of the Snyder transform method, and the summer model had a quicker lag and a higher peaking coeffcient than the winter model. It is possible that this caused the variation seen between summer and winter; however, the calibration was comprehensive, resulting in a fnal calibration and validation NSE of 0.65 and 0.44 for summer.

Antecedent Conditions and Results
Grassland continually showing the greatest reduction in peak fow may have been infuenced by several external factors. The grassland control site comprises of a sandier soil texture than the pedology of the other sample sites (with the exception of 2010, see Table 1), making it more permeable [84]. This means that when the grassland infltration data are applied to the whole site to the represent grassland coverage, it is not accounting for variations in soil texture across the site [85][86][87]. Additionally, Table 1 shows that the cc. 1900, 2008, and 2012 sites are comprised of a clay-heavy soil texture, meaning that they are naturally less permeable due to the smaller particle sizes of clay compared with sand [84]. This may be an indication as to why the current HofE site peak fow is higher than that of grassland, particularly in the winter. The winter of 2020 was the ffth wettest on record (329.4 mm/143% higher than the 1981-2010 baseline), and February of 2020 was the wettest ever recorded, with 155 mm of precipitation (258% higher than the 1981-2010 baseline) [88,89]. These dates are within the time periods that winter infltration data were collected, and the excess rainfall received would have contributed to the study site saturation, infuencing infltration data collection. See Figure 11.
As seen in Figure 11, the variation between antecedent rainfall and clay-saturation throughout winter, and cracking and drying-out throughout summer may have led to infltration rates being signifcantly higher than the grassland for the current site simulations in the winter. These effects may also explain the higher peak fows observed throughout summer compared with winter; the increase in surface pooling of the sample sites in the winter (Figure 11a,b) held water in place across the study site. Infltration would have been slowed due to the clay-geology and antecedent conditions of the site [90,91], meaning runoff was slowed, creating more of a lag between precipitation and peak fow in the winter. Whereas in the summer, the cracking of the ground (Figure 11c,d) reduced infltration, with the dry clay acting similarly to an impermeable surface. Due to this, rainfall was able to runoff into the watercourse, causing a quicker lag time and a higher peak. [84]. This may be an indication as to why the current HofE site peak flow is highe that of grassland, particularly in the winter. The winter of 2020 was the fifth wet record (329.4 mm/143% higher than the 1981-2010 baseline), and February of 2020 w wettest ever recorded, with 155 mm of precipitation (258% higher than the 198 baseline) [88,89]. These dates are within the time periods that winter infiltration dat collected, and the excess rainfall received would have contributed to the study sit ration, influencing infiltration data collection. See Figure 11. As seen in Figure 11, the variation between antecedent rainfall and clay-satu throughout winter, and cracking and drying-out throughout summer may have infiltration rates being significantly higher than the grassland for the current site s tions in the winter. These effects may also explain the higher peak flows ob throughout summer compared with winter; the increase in surface pooling of the s a b c d

Study Applications
Whilst antecedent weather conditions and soil texture have infuenced the trends seen between woodland and grassland land cover (see Section 3), the clay-heavy soil textures of the sample site are representations of the conditions of many sites around the UK. With the woodland sites being clay, and the ongoing comprehensive infltration data collection, this project provides an assessment of the impact of NFM (woodland planting) over one of the most impermeable soil types throughout the UK.
Area calculations regarding the coverage of superfcial alluvium, clay, peat, and fuvial deposits throughout the UK show that 15% (39,269.24 km 2 ) of UK superfcial geology is similar in infltration to the geology of the HofE site [92]. This demonstrates that the results of this study are signifcant, and are representative of 15% of UK geology, demonstrating that the results found throughout this study with regard to woodland planting and their runoff reduction capabilities can be extrapolated, furthering the usage of NFM across the UK. This therefore shows that woodland planting can be considered as a method of NFM throughout other areas of the UK, and once could expect to fnd similar positive results to those found in this study.

Conclusions and Future Work
The results from the simulations undertaken in this study have shown that woodland can reduce peak fows when compared to impermeable cover; however, at present, grassland reduces peak fow most in both winter and summer. The values of the data collected throughout is that; if woodland can reduce peak fows at present over a predominantly clay (impermeable) geology, then the value of woodland planting on a site with slightly more permeable geology would show greater results than this study. This presents an opportunity for this research to be extrapolated and applied to other geologies and soil textures, to potentially aid in justifying the use of NFM (woodland planting), as a viable method of increasing infltration and reducing ruff peaks to watercourses.
Woodland planting is gaining momentum as a method of NFM due to its ability to break up soil and increase infltration and water storage. This study took infltration measurements at 10 and 200 cm away from woodland planted at a 2.2 km 2 area in Warwickshire, England, from November 2019 until August 2021. Infltration data were incorporated into two calibrated and validated HEC-HMS models, and the reductions in peak fow for woodland, grassland, and impermeable land cover were quantifed. Results of this study show that, across a predominantly clay-site:

•
Woodland planting across the HofE site shows less of a peak fow compared to impermeable land cover simulations; however, the discrepancy is small and reduces with increase storm duration and intensity.

•
The grassland simulations result in the lowest peak fows, regardless of season or storm scenario. • Impermeable land cover produces the highest peak fows throughout all durations and storm intensities compared to woodland and grassland; however, this is somewhat expected as it is known that the increase in impermeable surface cover is aiding the push towards NFM) [1,2,10].

•
The summer simulations show signifcantly higher peak fows compared to winter values across all land cover types in the lower duration storms (6-and 24-h); however this is less signifcant in the higher duration simulations (96-h).
The quantifed results of this study show woodland to have a positive impact on peak fow reduction after only 15 years (since 2006), and indicate that the impacts will become more signifcant with root spread as the site matures [16,93]. This study is also representative of a clay-textured site, the same soil texture as 15% of the UK, indicating that if woodland can show a reducing in peak fow across this study site, similar results will be seen in other similar sites. Further to this, the results will likely be more signifcant in areas inherent of a more permeable soil texture [84]. Additionally, this study has provided insight into how to collect and extrapolate infltration data and model such information in HEC-HMS. Additionally, it has provided a methodology regarding the calibration and validation of HEC-HMS models where empirical data are sparse. This will enable other authors in the feld of hydrology to use this project as a framework when contributing to the knowledge base regarding infltration, NFM, woodland planting, and hydrology as a whole.
Future work will involve developing a method of projecting the collected infltration data, with the intention of using the HEC-HMS model to project the ability of woodland planting to mitigate fow and overland runoff into the future, regarding precipitation and basefow increases in light of climate change.

Data Availability Statement:
To access the data collected and analysed in this study, please contact the corresponding author: Nathaniel Revell (revelln@uni.coventry.ac.uk).