Improving the identification of hydrologically sensitive areas using LiDAR DEMs for the delineation and mitigation of critical source areas of diffuse pollution

Published Date
Science of The Total Environment
15 June 2016, Vol.556:276–290, doi:10.1016/j.scitotenv.2016.02.183
Open Access, Creative Commons license

• Author 

• I.A. Thomas ^a,b,,
• P. Jordan ^a,b,
• P.-E. Mellander ^a,
• O. Fenton ^c,
• O. Shine ^a,
• D. Ó hUallacháin ^c,
• R. Creamer ^c,
• N.T. McDonald ^a,
• P. Dunlop ^b,
• P.N.C. Murphy ^d,

• aAgricultural Catchments Programme, Teagasc, Johnstown Castle, Wexford, Co., Wexford, Ireland
• bSchool of Geography and Environmental Sciences, Ulster University, Coleraine, Northern Ireland, United Kingdom
• cTeagasc, Environmental Research Centre, Johnstown Castle, Wexford, Co., Wexford, Ireland
• dEnvironment and Sustainable Resource Management Section, School of Agriculture and Food Science, University College Dublin, Dublin 4, Ireland

Received 6 January 2016. Revised 25 February 2016. Accepted 25 February 2016. Available online 12 March 2016. Editor: D. Barcelo

Highlights

• •
  A new HSA Index allows for mitigation of pollutant transfers.
• •
  Flow sinks caused 17–33% of catchment areas to become hydrologically disconnected.
• •
  HSA sizes were empirically estimated using rainfall-quickflow measurements.
• •
  HSAs represented 2.9–8.5% of catchment areas during upper quartile storm events.
• •
  Targeting riparian buffer strips at HSA delivery points would reduce costs.

Abstract

Identifying critical source areas (CSAs) of diffuse pollution in agricultural catchments requires the accurate identification of hydrologically sensitive areas (HSAs) at highest propensity for generating surface runoff and transporting pollutants. A new GIS-based HSA Index is presented that improves the identification of HSAs at the sub-field scale by accounting for microtopographic controls. The Index is based on high resolution LiDAR data and a soil topographic index (STI) and also considers the hydrological disconnection of overland flow via topographic impediment from flow sinks. The HSA Index was applied to four intensive agricultural catchments (~ 7.5–12 km²) with contrasting topography and soil types, and validated using rainfall-quickflow measurements during saturated winter storm events in 2009–2014. Total flow sink volume capacities ranged from 8298 to 59,584 m³ and caused 8.5–24.2% of overland-flow-generating-areas and 16.8–33.4% of catchment areas to become hydrologically disconnected from the open drainage channel network. HSA maps identified ‘breakthrough points’ and ‘delivery points’ along surface runoff pathways as vulnerable points where diffuse pollutants could be transported between fields or delivered to the open drainage network, respectively. Using these as proposed locations for targeting mitigation measures such as riparian buffer strips reduced potential costs compared to blanket implementation within an example agri-environment scheme by 66% and 91% over 1 and 5 years respectively, which included LiDAR DEM acquisition costs. The HSA Index can be used as a hydrologically realistic transport component within a fully evolved sub-field scale CSA model, and can also be used to guide the implementation of ‘treatment-train’ mitigation strategies concurrent with sustainable agricultural intensification.

Graphical abstract

Abbreviations

• CSA, critical source area
• DEM, Digital Elevation Model
• HAA, hydrologically active area
• HSA, hydrologically sensitive area
• LiDAR, Light Detection and Ranging
• RBS, riparian buffer strip
• SMD, soil moisture deficit
• STI, soil topographic index
• TWI, Topographic Wetness Index

Keywords

• Hydrologically sensitive area
• Critical source area
• Diffuse pollution
• LiDAR DEM
• Agriculture
• Mitigation

---

1 Introduction

Diffuse pollution from agricultural land to waterbodies has been identified as a major cause of eutrophication and water quality degradation worldwide (Jarvie et al., 2013and Daniel et al., 1998), with mitigation measures part of wide ranging and international environmental policies (Schoumans et al., 2014, Mcdowell and Nash, 2012 and Murphy et al., 2015). Catchment areas at highest transfer risk of pollutants are termed critical source areas (CSAs) and are often identified as land use conflict areas (Valle Junior et al., 2014). More specifically, CSAs are where pollutant sources coincide with areas of high mobilisation potential and hydrologically sensitive areas (HSAs) that have the highest propensity for surface runoff generation, pollutant transport and delivery via hydrologically connected pathways (Pionke et al., 2000, Walter et al., 2000 and Agnew et al., 2006) (Fig. 1). HSAs are a water quality concept relating saturation-and-infiltration-excess mechanisms of overland flow generation and hydrological connectivity concepts to associated pollutant transport, delivery and CSAs (Fig. 1; Walter et al., 2000 and Agnew et al., 2006). They must be accurately identified if mitigation measures and best management practices aimed at reducing or offsetting diffuse pollution are to be cost-effectively designed and targeted (Sharpley et al., 2011 and Doody et al., 2012).

Fig. 1. Conceptual diagram of pollutant transfer continuum (adapted from Haygarth et al., 2005), HSAs and CSAs (adapted from Agnew et al., 2006).

Recent research has demonstrated the importance of accurately identifying HSAs when identifying and mitigating CSAs. Catchment hydrology has been found to be an important part of CSAs of phosphorus transfers in agricultural catchments (Campbell et al., 2015, Ulén et al., 2007, Heckrath et al., 2008 and Shore et al., 2014). In some of these studies, HSAs were a dominant CSA factor which outweighed source and land management pressures (Jordan et al., 2012, Buda et al., 2009, Mellander et al., 2015, Needelman et al., 2004 and Kleinman et al., 2011; Fig. 1). Catchment areas were hydrologically sensitive to rainfall because of the prominence of poorly drained soils (with low permeability and/or infiltration capacity), or impermeable subsurface soil layers such as fragipans that caused perched water tables. Such findings are consistent with the concept of the pollutant transfer continuum (Haygarth et al., 2005and Lemunyon and Gilbert, 1993), whereby pollutant sources are only delivered to receiving waters if transport pathways exist (Fig. 1).

In certain CSA definitions, watercourse proximity is typically used as a proxy of runoff propensity (Lemunyon and Gilbert, 1993, Gburek et al., 2000, Sharpley et al., 2003and Srinivasan and Mcdowell, 2007). Land adjacent to watercourses is assumed to be a HSA, and as such is considered a CSA if source pressures exist (Campbell et al., 2015 and Gburek et al., 2000). This approach is convenient for implementing measures within agricultural policy, and is the basis of riparian buffer strip (RBS) measures which, from a water quality point of view, are designed to impede surface runoff and reduce pollutant delivery. However, the approach is also an extreme simplification of reality, as overland flow tends to channelise and converge due to topographic and microtopographic influences (Thomas et al., 2014, Marjerison et al., 2011 and Qiu, 2003) and does not always flow uniformly downslope as a sheet (Ó hUallacháin, 2014 and White and Arnold, 2009). Thus some CSA definitions overestimate the size of the HSA at the stream and underestimate HSAs upslope, and hence poorly define pollutant transport potential in surface runoff (Sharpley et al., 2013 and Srinivasan and Mcdowell, 2009).

More scientifically robust approaches to delineating HSAs include topographic indices which use Digital Elevation Models (DEMs), such as the Topographic Wetness Index (TWI) by Beven and Kirkby (1979). It is defined as TWI , where α is the cumulative upslope drainage area per unit contour length and tanβ is the surface slope gradient. Larger upslope drainage areas and shallower slopes will produce larger TWI values which indicate higher runoff propensity (Quinn et al., 1991). A modification called the soil topographic index (STI) also accounts for the soil water storage capacity, and is defined as STI , where D is the local soil depth in metres to the restrictive layer (e.g. bedrock or fragipan) and K_s is the mean saturated hydraulic conductivity of the soil profile in metres per day above the restrictive layer (Walter et al., 2002). Shallower soils and those with lower saturated hydraulic conductivities will have lower soil water storage capacities and higher runoff propensities. Thus the approach accounts for hydrological disconnection of overland flow from the open drainage channel network due to reinfiltration at unsaturated soils which have larger soil water storage capacities. Topographic indices have been found to improve predictions of soil moisture, HSAs and pollutant loads from diffuse sources compared to approaches that do not consider topography, such as watercourse proximity (Buchanan et al., 2014, Agnew et al., 2006 and Hahn et al., 2014).

However, topographic indices such as the TWI or STI do not consider the hydrological disconnection of overland flow from topographic impediment within flow sinks such as depressions, hummocks, hedgerow banks or other microtopographic features. Thus they do not differentiate between hydrologically active areas (HAAs; overland-flow-generating-areas) and HSAs (runoff-generating-areas). This is because topographic indices are typically derived from hydrologically corrected DEMs which remove (fill) all flow sinks so that flow pathways are continuous and hydrologically connected to the catchment outlet (Jenson and Domingue, 1988 and Maune et al., 2007). This improves the modelling of subsurface and groundwater flow pathways and associated propensity for overland flow generation. However, removing flow sinks incorrectly assumes that all of these features do not exist, and are a result of DEM vertical error (Lindsay and Creed, 2006 and Wechsler, 2007). In fact many flow sinks are real topographic features which have important influences on the pathways and hydrological connectivity of overland flow once it is generated (Li et al., 2011 and Lane et al., 2009). This is important because once overland flow is impeded, it will reinfiltrate and deposit and immobilise dissolved or entrained pollutants, and hence the upslope drainage area will not be a HSA or CSA (Fig. 1). These microtopographic features often dominate agricultural catchments, and must be considered when identifying HSAs/CSAs and targeting mitigation measures as they could represent existing mitigating features in the landscape (Thomas et al., 2014 and Sherriff et al., 2015).

High resolution DEMs with high vertical accuracies derived from Light Detection and Ranging (LiDAR) technology can now accurately capture these microtopographic flow sinks (Maune et al., 2007, Lindsay and Creed, 2006 and Li et al., 2011). Furthermore, they allow modelling of HSAs and CSAs at optimal resolutions, accounting for microtopographic controls on surface runoff pathways, hydrological connectivity and soil erosion (Vaze et al., 2010, Djodjic and Villa, 2015, Galzki et al., 2011 and Petroselli, 2012). As such, breakthrough points and delivery points along surface runoff pathways where pollutants are transported between fields or delivered to the open drainage network, respectively, can now also be accurately identified (Thomas et al., 2014). LiDAR DEMs could therefore significantly improve the identification of HSAs and targeting of mitigation measures such as RBS within agricultural policies to reduce diffuse pollution from CSAs.

To improve the identification of HSAs, in order to accurately delineate CSAs of diffuse pollution and target mitigation measures, this study had three objectives: (1) to develop a GIS-based HSA Index which uses LiDAR DEMs and the STI, and accounts for hydrological disconnection of overland flow via topographic impediment from flow sinks; (2) to validate the HSA Index using rainfall-quickflow measurements; (3) to identify cost-effective locations at identified HSAs where sub-field scale diffuse pollution mitigation measures such as RBS could be targeted.

2 Materials and methods

2.1 Study sites and deriving STI maps

Four agricultural catchments in Ireland were selected for this study (Fig. 2). Catchment details are fully described elsewhere (Wall et al., 2011, Jordan et al., 2012and Shore et al., 2013), and the main hydro-physical details are summarised in Table 1.

Fig. 2. Locations of the agricultural catchments in Ireland used in the study.

Table 1. Catchment characteristics.

	Arable A	Arable B	Grassland A	Grassland B
Area (ha)	1116	948	758	1207
Land use	Arable (54%) Grassland (39%)	Arable (33%) Grassland (49%)	Arable (6%) Grassland (84%)	Arable (20%) Grassland (77%)
Soil drainage class	Well drained soils	Mixture of well, moderately, imperfectly and poorly drained soils	Well drained soils	Poorly drained soils Well drained soils in uplands
Dominant soil types	Typical Brown Earths (88%), Gleyic Brown Earths (5%), Typical Groundwater Gleys (5%)	Stagnic Brown Earths (35%), Typical Surface-water Gleys (25%), Typical Brown Earths (22%)	Typical Brown Earths and Typical Brown Podzols (84%), Typical Surface-water Gleys (5%), Humic/Typical Alluvial Gleys (4%)	Typical Surface-water Gleys or Groundwater Gleys (71%) Typical Brown Earths (29%)
Dominant hydrological pathway for storm flow	Subsurface	Surface and subsurface	Subsurface	Surface and subsurface
Average annual rainfall (mm) from 2010 to 2014 hydrological years	1021	913	1117	1078
Geology	Slate and siltstone	Calcareous greywacke and mudstone	Sandstone, mudstone and siltstone	Rhyolitic volcanic and slate

To derive TWI and STI maps, 2 m resolution LiDAR DEMs were utilised (Supplementary Fig. 1). These represented optimal grid resolutions for modelling flow pathways in catchments dominated by microtopography, and were resampled from 0.25 m resolution LiDAR DEMs with horizontal and vertical accuracies of 0.25 m and 0.15 m, respectively (Thomas et al., 2014). A multi-step method (work-flow described in Supplementary Fig. 2) was used to hydrologically correct DEMs and derive TWI maps. DEMs were hydrologically corrected in SAGA GIS v.2.1 by ‘burning’ a field-mapped open drainage channel network into the DEM. To model fully connected flow pathways to the catchment outlets, flow sinks were identified and filled using the method for LiDAR datasets by Wang and Liu (2006). The Deterministic Infinity method by Tarboton (1997) was used to model multiple flow directions and upslope drainage areas, and slope was modelled using the method by Zevenbergen and Thorne (1987). To derive STI maps, soil subgroup maps from the Irish Soil Information System (Creamer et al., 2014) were imported into ArcGIS v10.0 and improved using additional soil sampling, expert knowledge and the DEMs. The dominant soil series for each soil subgroup was identified based on these data and assigned to each soil subgroup. The Irish Soil Information System was then used to extract soil series properties used for D (soil depth) and K_s (soil texture, bulk density and soil horizon thickness). A pedotransfer function was used to determine bulk densities of soil horizons where data were unavailable (Reidy et al., 2016). K_s values were calculated using the Retention Curve model (van Genuchten et al., 1991), and a ln(K_sD) raster was then created. As texture, and hence K_sD, is not determinable for peat soils, a K_sD value of 0.083 was assigned which was the equivalent value of the Kilrush soil series (Typical Surface-water Gley) (R. Creamer, pers comm). Similarly, roads identified using orthophotos were assigned an arbitrary K_sD value of 0.002, representing their impermeability and high runoff risk. Other linear features such as farm tracks and wheelings were not considered due to variability in permeability, location and lack of information. Also, although subsurface artificial drainage may increase K_s, this was not considered in this study, as field surveys of drainage ditches suggest few exist in these catchments, and high variability may exist in their design, age and effectiveness. An STI map was then created by subtracting the ln(K_sD) raster from the TWI raster using the raster calculator tool.

To derive TWI and STI maps, 2 m resolution LiDAR DEMs were utilised (Supplementary Fig. 1). These represented optimal grid resolutions for modelling flow pathways in catchments dominated by microtopography, and were resampled from 0.25 m resolution LiDAR DEMs with horizontal and vertical accuracies of 0.25 m and 0.15 m, respectively (Thomas et al., 2014). A multi-step method (work-flow described in Supplementary Fig. 2) was used to hydrologically correct DEMs and derive TWI maps. DEMs were hydrologically corrected in SAGA GIS v.2.1 by ‘burning’ a field-mapped open drainage channel network into the DEM. To model fully connected flow pathways to the catchment outlets, flow sinks were identified and filled using the method for LiDAR datasets by Wang and Liu (2006). The Deterministic Infinity method by Tarboton (1997) was used to model multiple flow directions and upslope drainage areas, and slope was modelled using the method by Zevenbergen and Thorne (1987). To derive STI maps, soil subgroup maps from the Irish Soil Information System (Creamer et al., 2014) were imported into ArcGIS v10.0 and improved using additional soil sampling, expert knowledge and the DEMs. The dominant soil series for each soil subgroup was identified based on these data and assigned to each soil subgroup. The Irish Soil Information System was then used to extract soil series properties used for D (soil depth) and K_s (soil texture, bulk density and soil horizon thickness). A pedotransfer function was used to determine bulk densities of soil horizons where data were unavailable (Reidy et al., 2016). K_s values were calculated using the Retention Curve model (van Genuchten et al., 1991), and a ln(K_sD) raster was then created. As texture, and hence K_sD, is not determinable for peat soils, a K_sD value of 0.083 was assigned which was the equivalent value of the Kilrush soil series (Typical Surface-water Gley) (R. Creamer, pers comm). Similarly, roads identified using orthophotos were assigned an arbitrary K_sD value of 0.002, representing their impermeability and high runoff risk. Other linear features such as farm tracks and wheelings were not considered due to variability in permeability, location and lack of information. Also, although subsurface artificial drainage may increase K_s, this was not considered in this study, as field surveys of drainage ditches suggest few exist in these catchments, and high variability may exist in their design, age and effectiveness. An STI map was then created by subtracting the ln(K_sD) raster from the TWI raster using the raster calculator tool.

2.2 Empirically estimating HSA sizes using rainfall-quickflow measurements

To validate the HSA Index, high resolution rainfall-quickflow measurements from 2009 to 2014 were used to empirically estimate the size of runoff-generating-areas (HSAs) during storm events in each catchment. Daily rainfall depths were calculated from gauging stations within the catchments. Quickflow volumes were estimated by separating hydrographs of hourly discharge (measured at 10 minute intervals from automatic gauging stations at catchment outlets), using a graphically interpreted hydrograph separation method described by Mellander et al., 2012 and Mellander et al., 2015. Although quickflow volumes were likely a sum of surface runoff, preferential flow and tile and ditch drainage, all quickflow was assumed to be surface runoff in this study. This assumption was supported from field surveys of the drainage ditches within the study catchments which showed little evidence of subsurface drainage. The Soil Moisture Deficit (SMD) model by Schulte et al. (2005) was used to calculate daily SMDs for each soil drainage class for the same period using meteorological data. Quickflow events were selected during winter and spring months on days with saturated conditions (i.e. ≤ 0 mm SMD depending on the soil drainage class), when saturation-excess (and potentially infiltration-excess) HSAs would become most active. To minimise noise from old water contributions (i.e. time lag; Fenton et al., 2011), only events that were not preceded by heavy rainfall/quickflow were selected. For each selected event, the daily quickflow volume (m³) was divided by the daily rainfall depth (m) to estimate the HSA size (m²) generating the observed quick flow (runoff) volume. The HSA size as a proportion of the catchment was then calculated for median, lower quartile (LQ) and upper quartile (UQ) rainfall-quickflow events.

2.3 Developing the HSA Index by accounting for flow sinks

An HSA Index (dimensionless) was developed by modifying the STI to account for the effects of flow sinks on hydrological connectivity of overland flow. This modification involved reducing STI values in upslope drainage areas of flow sinks large enough to topographically impede and trap overland flow generated within an UQ storm event (method shown in Supplementary Fig. 2). To do this, flow sinks were extracted from the DEM, and their depth and area used to calculate their overland flow volume capacity. These sink maps were derived from 0.25 m LiDAR DEMs to improve accuracy and then resampled to 2 m grid resolutions to optimise modelling of flow sink upslope drainage areas and reduce computational demands. Flow sinks within lakes and the open drainage channel network were removed as they were assumed hydrologically connected to the catchment outlet. Very small flow sinks (< 0.05 m depth and < 1 m³ volume capacity) were also removed as they would likely have a negligible effect on impeding overland flow and could also represent DEM vertical error (flow sinks < 1 m³ numbered between 2772 and 7782).

An HSA Index (dimensionless) was developed by modifying the STI to account for the effects of flow sinks on hydrological connectivity of overland flow. This modification involved reducing STI values in upslope drainage areas of flow sinks large enough to topographically impede and trap overland flow generated within an UQ storm event (method shown in Supplementary Fig. 2). To do this, flow sinks were extracted from the DEM, and their depth and area used to calculate their overland flow volume capacity. These sink maps were derived from 0.25 m LiDAR DEMs to improve accuracy and then resampled to 2 m grid resolutions to optimise modelling of flow sink upslope drainage areas and reduce computational demands. Flow sinks within lakes and the open drainage channel network were removed as they were assumed hydrologically connected to the catchment outlet. Very small flow sinks (< 0.05 m depth and < 1 m³ volume capacity) were also removed as they would likely have a negligible effect on impeding overland flow and could also represent DEM vertical error (flow sinks < 1 m³ numbered between 2772 and 7782).

To calculate the overland flow volume generated within the flow sink upslope drainage area, the size of the HAA had to be known. The proportion of the catchment which was an HAA during an UQ storm event was estimated by arbitrarily increasing the UQ HSA size (estimated from rainfall-quickflow data) by 20%. The HAA was spatially distributed within the catchment by selecting the catchment areas with the highest STI values until the catchment area selected was the HAA size. The lowest STI value within this selection was identified as the STI threshold value for delineating HAAs.

To calculate the overland flow volume generated from the HAA within the flow sink upslope drainage area, the HAA size (m²) was multiplied by the UQ rainfall depth (m) to calculate the overland flow volume (m³). If the overland flow volume was less than the flow sink volume capacity, the flow sink would topographically impede the overland flow and not ‘fill and spill’, and the whole upslope drainage area would by hydrologically disconnected from the open drainage network.

The HSA Index was created by reducing STI values within these hydrologically disconnected flow sink upslope drainage areas by 75%. Thus high STI values (HAAs) within these areas were now low HSA Index values and not considered as HSAs. HSAs were spatially distributed within the catchment by selecting the catchment areas with the highest HSA Index values up to the same STI threshold value used to delineate HAAs. If the catchment area selected was larger or smaller than the UQ HSA size estimated from rainfall-quickflow data, the HAA size (originally estimated as 20% larger than the HSA size) was refined and the processes repeated. HSA maps for UQ as well as LQ and median storm events were then created.

2.4 Validating the HSA Index

Value distributions of slope, upslope drainage area, TWI, STI, flow sink depth (≥ 0.05 m), flow sink volume capacity (≥ 1 m³) and the HSA Index were analysed for each catchment. The HSA Index was validated by comparing the HSA size predicted by HSA Index threshold values with the HSA size estimated by the rainfall-quickflow data using correlation analysis. A range of HSA Index threshold values were tested, from the lowest threshold value which matched the observed HSA size of a catchment, to the highest. Thresholds ranged from 13.3–14.7 for LQ storm events, 11.9–12.8 for median storm events, and 10.9–11.7 for UQ storm events. The same correlation analysis was repeated for the STI using the same threshold values for comparisons in performance of the two indices.

2.5 Using HSA maps to identify cost-effective locations for targeting mitigation measures

HSA maps were used to identify breakthrough points and delivery points at HSAs as these were deemed to be cost-effective locations for targeting diffuse pollution mitigation measures if source pressures existed at present or in the future. These locations were prioritised based on the size of the upslope HSA draining into that point (a similar concept to upslope drainage area) and whether it would become active during LQ, median or UQ storm events. Furthermore, the costs of targeted and blanket implementation approaches over 1 and 5 years were estimated for each catchment using the RBS measure from the current agri-environment scheme in Ireland (Green Low-carbon Agri-environment Scheme, GLAS; DAFM, 2015) as a case study. This is a realistic comparison as current RBS policies and the HSA approach here are based on hydrological risk only and do not consider pollution sources as in fully evolved CSAs. The total RBS length was calculated along the whole field-mapped open drainage channel network for the blanket approach, and at HSA delivery points for the targeted approach. RBS establishment costs were derived from current GLAS payment rates (€/m/yr) for each of the four margin widths available (3, 6, 10 and 30 m). These generic RBS establishment costs are included for case-study comparisons only and do not take into account likely variations in costs associated with establishing RBS in arable and grassland systems. The costs of the targeted approach also included an additional one-off cost of acquiring a LiDAR DEM for a 10 km² size catchment which ranged from €10,000–40,000, based on Ó hUallacháin (2014) and this study. Costs of targeted and blanket implementation of RBS were then compared.

3 Results

3.1 STI, flow sinks and HSA sizes

STI and flow sink maps are shown in Fig. 3a and b. Distributions of values of slope, upslope drainage area, TWI and STI are shown in Fig. 4. Due to the high resolution (2 m) LiDAR DEMs used, individual surface flow pathways are clearly identifiable. A larger proportion of Grassland B had higher STI values (i.e. ≥ 10) compared to the other catchments due to the dominance of poorly drained soils with low K_sD values (see Supplementary Fig. 3a and b). This was despite it having the lowest TWI value distributions due to small upslope drainage areas from a dense drainage network. Conversely, Arable A had the highest TWI distributions due to large upslope drainage areas, but the lowest STI values (with Grassland A) due to the dominance of well drained soils with high K_sD values. Arable B, which has a mixture of soil drainage classes, showed a larger proportion of high STI values compared to well drained Arable A and Grassland A. High STI values in these latter catchments tended to be found in concentrated areas where more imperfectly or poorly drained soils existed, particularly in lower hillslope positions.

Fig. 3. Maps of the STI (a) and flow sinks (b) for each catchment.

Fig. 4. Distributions of values of slope, upslope drainage area, TWI, STI, HSA Index, flow sink depths (≥ 0.05 m) and flow sink volumes (≥ 1 m³) for each catchment. Correlation analysis between the proportion of the catchment predicted to be an HSA using the STI or HSA Index and the proportion estimated using rainfall-QF data is also shown.

STI and flow sink maps are shown in Fig. 3a and b. Distributions of values of slope, upslope drainage area, TWI and STI are shown in Fig. 4. Due to the high resolution (2 m) LiDAR DEMs used, individual surface flow pathways are clearly identifiable. A larger proportion of Grassland B had higher STI values (i.e. ≥ 10) compared to the other catchments due to the dominance of poorly drained soils with low K_sD values (see Supplementary Fig. 3a and b). This was despite it having the lowest TWI value distributions due to small upslope drainage areas from a dense drainage network. Conversely, Arable A had the highest TWI distributions due to large upslope drainage areas, but the lowest STI values (with Grassland A) due to the dominance of well drained soils with high K_sD values. Arable B, which has a mixture of soil drainage classes, showed a larger proportion of high STI values compared to well drained Arable A and Grassland A. High STI values in these latter catchments tended to be found in concentrated areas where more imperfectly or poorly drained soils existed, particularly in lower hillslope positions.

Daily rainfall, quickflow and SMD data (for well drained soils) during 2009–2014 for each catchment are shown in Fig. 5. Grassland B had much larger, flashier and more frequent runoff events, followed by Arable B. Arable A and Grassland A showed similar rainfall-quickflow responses, which were the lowest magnitudes of the four catchments. Arable A also showed more prolonged periods of elevated quickflow following rainfall, indicating old water contributions (i.e. not quickflow) and time lag dynamics. All catchments showed broadly similar magnitudes and temporal dynamics of rainfall and SMD, reflecting the dominant influence of North Atlantic weather systems.

Fig. 5. Daily rainfall and quickflow events measured during 2009–2014 for each catchment, and SMD values (for well drained soils only).

HSA sizes as a proportion of each catchment, estimated using the rainfall-quickflow data for selected events during saturated conditions, are shown in Table 2. HSA sizes during LQ, median and UQ events were similar for Arable A and Grassland A, and represented relatively small proportions of the catchment (≤ 3.2%). However, during the largest rainfall-quickflow events, HSA sizes represented significant catchment proportions (6.2–15.1%). Although LQ-median events in Grassland B also indicated relatively modest HSA sizes (2.1–3.4%), UQ and maximum sizes (8.5% and 19.1%, respectively) demonstrate the hydrological sensitivity of this catchment. HSA sizes in Arable B were the second highest of the four catchments.

For further details log on website :
http://www.sciencedirect.com/science/article/pii/S0048969716303941