Physical soil quality indicators for monitoring British soils

Soil condition or quality determines its ability to deliver a range of functions that support ecosystem services, human health and wellbeing. The increasing policy imperative to implement successful soil monitoring programmes has resulted in the demand for reliable soil quality indicators (SQIs) for physical, biological and chemical soil properties. The selection of these indicators needs to ensure that they are sensitive and responsive to pressure and change, e.g. they change across space and time in relation to natural perturbations and land management practices. Using a logical sieve approach based on key policy-related soil functions, this research assessed whether physical soil properties can be used to indicate the quality of British soils in terms of their capacity to deliver ecosystem goods and services. The resultant prioritised list of physical SQIs was tested for robustness, spatial and temporal variability, and expected rate of change using statistical analysis and modelling. Seven SQIs were prioritised: soil packing density, soil water retention characteristics, aggregate stability, rate of soil erosion, depth of soil, soil structure (assessed by visual soil evaluation) and soil sealing. These all have direct relevance to current and likely future soil and environmental policy and are appropriate for implementation in soil monitoring programmes.


Introduction
In recent years soil quality and its measurement have increasingly been based on soil functions (Loveland and Thompson, 2002;Ritz et al., 2009;Rosa, 2005).These functions determine the ability of a soil to deliver and support ecosystem goods and services, which have been linked to human health and wellbeing, but are often difficult to quantify.Soils are typically recognised for their role in provisioning goods such as building materials, fresh water, fuel, fibre and food (Robinson et al., 2013).Soils also interact with other environmental components (air and water) and provide a platform for infrastructure.The ecosystem services that rely on these functions include regulation of climate and hydrology, contaminant transformation, biocontrol of plant pathogens and parasites (Sylvain and Wall, 2011), and water filtration/runoff reduction/purification (Breure et al., 2012).Supporting services provided by soils include soil formation, soil fertility, biogeochemical cycling (C storage and nutrient cycling), decomposition of organic materials and plant available water.A number of cultural services are also supported such as recreational surfaces (Robinson et al., 2013) and heritage services including preserving historic artefacts, burial grounds and non-monetary values, such as containing information of events that have occurred in historical and geological timescales (Costantini and L'Abate, 2009).There are a number of databases and protection strategies in place for soils with cultural heritage (otherwise known as pedosites, geosites and geoparks) (Serrano and Ruiz-Flaño, 2007;UN-ESCO, 2016).In summary, soil-derived goods are tangible and include the food, fibre and fodder produced from soil.Soil-derived services are intangible and examples include water and climate change regulation through carbon sequestration.
For many soil ecosystem services, there are significant gaps in the valuation literature, especially related to soil regulation and formation, nutrient regulation and water regulation (Costanza et al., 2006;Jonsson and Davidsdottir, 2016).Soils deliver both economic and ecological benefits, the latter being less amenable to quantification and assessment.Better understanding of soil properties and associated functions will provide information for improved valuation of the services and goods delivered by soil.
However, the ability of soil to function and deliver these essential goods and services can be threatened by soil degradation processes.These include soil erosion, compaction, loss of organic matter content, loss of biodiversity, soil surface sealing, acidification and salinisation (European Commission, 2006).These processes can be natural and/or caused by human actions, such as using land beyond its inherent capability.Soil degradation processes result in changes in soil physical, biological and chemical properties, so affecting soil functions and soil quality.
In order to measure soil quality and functions, soil quality indicators are commonly used.Indicators of soil quality are required for environmental monitoring/reporting and provide the basis for many soil protection policies and monitoring programmes (Pulleman et al., 2012).Indicators help assess human and natural impacts on soils, including degradation processes and identify the effectiveness (or otherwise) of sustainable land management practices (Doran and Parkin, 1994;Karlen and Stott, 1994;Schipper and Sparling, 2000).They have also been used to evaluate the effectiveness of restoration strategies (Costantini et al., 2016).In order to assess soil quality, a combined approach is required in which the soil's biological, chemical and physical attributes and their interactions are assessed (Bone et al., 2010;Seybold et al., 1998).In this respect, monitoring is defined as a method to determine the quality and condition of the soil environment over space and time.This is measured by determining actual values of the attributes of interest.
There have been a few studies that have discussed and attempted to prioritise the most appropriate soil quality indicators (SQIs) for biological (Masto et al., 2015;Pulleman et al., 2012;Ritz et al., 2009) and physico-chemical indicators (Arshad and Coen, 1992;Asensio et al., 2013;Karlen and Stott, 1994;Masto et al., 2015;Rickson et al., 2012).This study focusses on a systematic process of selection for physical SQIs and then explores their potential for use in national monitoring schemes (e.g.England and Wales) (Loveland and Thompson, 2002;Merrington et al., 2006).In par-ticular, practical aspects such as sampling design and size, the use of proxies and pedotransfer functions, and the application of sensor technology are explored.
The role or function of a soil system can be defined differently, depending on the stakeholder/user groups and their specific objectives (e.g.whether they are interested in the production, regulation or cultural services of soil) (Rickson et al., 2012).As such, indicators are usually selected on the basis of the function(s) of interest, and observed and measured to infer the capability of a soil to perform that particular function (Bone et al., 2010;Ditzler and Tugel, 2002;Doran and Parkin, 1994).Once selected, effective SQIs need to meet the following criteria: be meaningful, interpretable and sensitive (and measurable) to natural and human-induced pressures and change (Burger and Kelting, 1999;Loveland and Thompson, 2002); reflect the desired condition or end point for a particular soil and/or land use and/or function (Loveland and Thompson, 2002); be relatively cheap, practical and simple to monitor (Loveland and Thompson, 2002); be responsive to corrective/management measures (Burger and Kelting, 1999); be applicable over large areas and different soils/land use types (Burger and Kelting, 1999); be capable of providing continuous assessment over long timescales (Burger and Kelting, 1999).
Selected physical SQIs need to be sensitive to pressure and able to reflect change in soil quality status (the capacity of the soil to function) at any given location and time (Burger and Kelting, 1999;Loveland and Thompson, 2002;Rickson et al., 2012).As such, an effective physical SQI would need to detect "meaningful change" in a given soil property and associated function(s) and be responsive to this change in the light of expected changes in soil quality.In other words, does the physical SQI change sufficiently that it can be detected, and is this change indicative of a significant loss/gain in soil quality?In order to evaluate the effectiveness of the indicator, the criteria for what constitutes a "meaningful change" need to be set.
In some instances, where the indicator itself may drive the change (for example the effect of bulk density on crop growth), "meaningful change" may be as simple as ascertaining the SQI value at a particular location and comparing this to a critical value or target value or range.This approach is taken by Merrington et al. (2006) and whilst simple in its approach, it does not capture the dynamic relationships between SQI and soil functions.These relationships may differ between soil functions, land uses and soil types (Jones, 1983).As such, there needs to be a focus on the dynamic relationships between soil functions and SQIs: however, information in the literature is sparse.Physical SQIs also need to be meaningful in terms of the soil processes that they represent.A change in the SQI needs to relate to a change in the processes that are taking place in the soil and therefore how the soil functions.For example, a change (increase) in bulk density would result in a change in processes operating in the soil (e.g.restriction to root elongation) and therefore a change in soil function (reduced crop yield).
Soil properties are spatially and temporally variable as a result of land use and management, parent material and climate.For example, there are over 1800 soils in Britain alone.Geological diversity means that these soils range from porous sands to impermeable clays.They include upland and lowland peat areas, surface and ground water gleys, shallow calcareous lithomorphic soils and free-draining brown soils.They range from acid to alkaline.Some soils are directly related to the parent material below them and others are unrelated, having been previously transported by wind, water and glacial action.Soil depths range from shallow (< 10 cm), typically over chalk, to deeper soils (> 150 cm), typically sandy silt loam, silt loam to silty clay loam soils.The soils in lowland Britain have largely been modified by anthropogenic activity, as most are intensively managed for agricultural production and have therefore been affected by tillage and some by artificial drainage.
Inevitably, this variability introduces "noise" into the signal response (signal : noise, i.e. meaningful change) in two ways.Firstly, there is the consideration of the spatial unit over which the soil quality is assessed.The spatial variability within this unit (e.g.plot, field, farm, catchment, national scale) will introduce variability to the SQI, irrespective of whether there are any changes in soil function(s).Secondly, there is the consideration of the impact of a particular land management practice on the effectiveness of a SQI to indicate soil quality.
Based on the above criteria and considerations, it has been argued that it may not be possible to achieve a single, affordable, workable soil quality index (Sojka and Upchurch, 1999) or a consensus on a standardised methodology which would be appropriate across different soil and land use types (Karlen and Stott, 1994).Furthermore, soils can frequently perform several functions simultaneously, although these can be diverse and often conflicting, but they must still be taken into account (Bone et al., 2010;Schoenholtz et al., 2000).
The ultimate aim of this study is to inform the design and implementation of a national-scale soil monitoring programme.This will centre on the identification of a number of meaningful indicators of physical soil quality that detect and reflect changes in soil properties over space and time, the associated changes in soil functions, and the goods and services delivered by soil.The study uses a multi-stage approach in the selection and prioritisation of physical SQIs that meet the required criteria and conditions outlined above.It consists of a systematic review and selection procedure, followed by assessment of the selected SQIs and how they could be best applied in a national-scale monitoring programme.The final priority list should indicate the soil's capacity to deliver ecosystem goods/services and is therefore indicative of soil quality.

Methodology
The process of physical SQI selection takes a multi-stage approach as outlined in Fig. 1.In the first stage, potential physical SQIs were identified from the available literature, including those defined by Loveland and Thompson (2002) and Merrington et al. (2006).Other physical SQIs (and the methods used to measure them) that had not been considered previously were also included to produce an up-to-date list.In the second stage, the candidate physical SQIs were prioritised using a logical sieve (Ritz et al., 2009) and a scenariobased approach.Here, the logical sieve was interrogated by running three scenarios based on typical priorities of different stakeholders, by applying weightings to the scoring system used to prioritise SQIs.As such, the approach was used to prioritise a specific soil function or degradation process of interest (Rickson et al., 2012).
Finally, the priority physical SQIs were tested for robustness (statistical reliability and accuracy as well as practicability), spatial and temporal variability, and expected rate of change using statistical analysis and modelling.This involves determining appropriate sample numbers for defining meaningful change, as well as proxy methods that can be used to make the physical SQI measurements operational and feasible.For example, where measurement and monitoring of a standard physical SQI may be time or resource intensive in a large-scale monitoring programme, a proxy that is easier and cheaper to measure may exist that could make that physical SQI feasible for inclusion into such a programme.

Identification of potentially meaningful physical SQIs
The identified physical SQIs were derived from the literature with consideration of recent scientific advances and developments in soil policy.Loveland and Thompson (2002) identified 22 direct and 4 indirect physical SQIs.Direct indicators refer to those that are associated directly with a soil function, whereas indirect indicators refer to those that are indirectly related to a soil function.Merrington et al. (2006) give the example of soil water storage following rainfall as a direct indicator, whereas a catchment hydrograph is an example of an indirect indicator of rainfall interception and water storage by soils.Merrington et al. (2006)   Where an indicator could be measured using alternative techniques/approaches, sub-categories reflecting this were created, ensuring that the indicator and its different measurement methods were scrutinised by the logical sieve.

Prioritisation of candidate physical SQIs
The 42 physical SQIs were evaluated in terms of the following criteria: Criterion 1 Soil function: does the candidate SQI reflect all soil function(s)?In this case, the four main functions, as described in the Millennium Ecosystem Assessment (Millennium Ecosystem Assessment, 2005), were used (i.e.provisioning, regulating, cultural and supporting).
Criterion 2 Land use: does the candidate SQI apply to all land uses found nationally?The range of land uses considered was based on the Centre for Ecology and Hydrology's land cover map (Morton et al., 2011) that also reflected differences in land use resulting from differences in land management practices (e.g.cultivations on arable land as opposed to pasture).
Criterion 3 Soil degradation process: can the candidate SQI express the impacts of soil degradation processes?This refers to the range and representation that each physical SQI gives to the main soil degradation processes as identified in the Thematic Strategy for Soil Protection (European Commission, 2006, Table S4) was considered.This approach captures whether the SQIs reflect the effect of potential degradation threats on soil functions.
Criterion 4 Challenge criteria: does the candidate SQI meet the challenge criteria used by Merrington et al. (2006)?These challenge criteria were developed for a nationalscale soil monitoring scheme and were integrated with criteria used to identify the inverse of soil quality indicators from the ENVASSO project (Huber et al., 2008).For example, are the measurements of the indicator practicable?Can the indicator be measured cost effectively?Is the indicator policy relevant?
Each of these criteria categories (and constituent factors) was considered separately and each of the physical SQIs was scored numerically, with weighting factors using the approach outlined by Ritz et al. (2009).The criteria are presented in Tables S2-S5 in the Supplement; the methodology for weighting, scoring and ranking in Methods S6; and an example of the logical sieve assessment in Table S7.Three scenarios were run to test the logical sieve.
Scenario 1 involved no weightings applied (all factors are equally important).For example, when considering Category 1 (soil functions category), all soil functions (provisioning, regulating, cultural and supporting) are equally important.In scenario 2, a higher priority was applied to the provisioning and regulation soil functions (factors).These two soil functions were selected as they are considered high priorities in current soil policy as highlighted in the Natural Environment White Paper "The natural choice: securing the value of nature" (DEFRA, 2011a), the Soils Evidence Plan (DEFRA, 2011b).Scenario 3 used a weighting factor to normalise values across all categories.As such, differences in the number of factors in each category would not affect the outcome (e.g.Category 1 (soil function) includes four factors to consider, Category 2 (land use) has seven factors to consider, and so on).
These scenarios represent the types of questions that may be asked by different stakeholder groups.The results from the three scenarios (top 25 % cumulative scores, as well as any of the physical SQIs that survived the sieving process by scoring > 0 in all factors of all categories) resulted in 18 candidate physical SQIs (from the original 42; Table S1; Methods S8).These "short-listed" indicators were further filtered by taking a more "numerical" approach, i.e. are data available to test the robustness of the candidate SQIs through statistical/modelling analysis?Also, the remaining SQIs were rationalised in terms of duplications, overlaps/double counting and linkages.Selection was also made on the basis of whether there was scientific evidence regarding the following: As a result of this process, seven physical SQIs were selected for further analysis: soil packing density/bulk density, soil water retention characteristics, soil sealing, depth of soil, visual soil evaluation, rate of soil erosion, soil aggregate stability.

Assessment of priority physical SQIs
These final physical SQIs were tested for uncertainty in their measurement, the spatial and temporal variability in the indicator (as given by observed distributions), and the expected rate of change (for a given soil function in light of expected changes in soil quality) in the indicator.For each SQI, the following points were addressed: whether the SQI could be directly related to soil functions, what constitutes meaningful change in the SQI by determining the relationships between the SQI (and how it changes) and soil processes, the spatial variability of the SQI and the implications for sampling using spatial statistics and power analyses.

Statistical analyses and modelling approach
The type of analyses conducted on the SQIs depended on the type of soils data available.Where full data were available, quantitative methods such as power analysis or pedo-transfer functions were used.Statistical power is the probability that a specific difference will be detected at a specified level of confidence.It allows the determination of a sample size required to detect an effect of a given size with a degree of confidence (Rickson et al., 2012).Otherwise analysis was carried out (semi)qualitatively (e.g. using remote sensing) or qualitatively (where no data exist).Three of the selected priority physical SQIs will be discussed in this paper, as sufficient monitoring data were readily available for each (Table 1).This allowed geostatistical analysis and application of pedotransfer functions, including evaluations based on quantitative or semi-quantitative methods.
Where there was substantive quantity of data (e.g.packing density), we explored the sampling intensity required to detect a change in the SQI.In other words, for the SQI to be effective as an indicator, it needs to be sensitive to changes in soil quality, and sufficiently responsive to be detectable above the natural variability of the soil (meaningful change) without requiring an impractical number of samples to determine this change.We estimated the natural variability of a particular property in two ways: (i) through a natural stratification by land use and soil types and (ii) through geostatistics (see Methods S9), where we used block kriging to estimate the within block variance of blocks sized 5, 10, 25 and 50 km 2 , roughly approximating management units of increasing size (e.g.field, farm, landscape).
Where the particular property is obtained from complex analytical methods, such as soil water retention characteristics, we explored the use of pedotransfer functions, in particular a multiple regression model and multiple additive regression splines, which are described in Methods S10. 3 Results and discussion

Packing density
Packing density is a measure of soil porosity and an indirect measure of soil functions such as water regulation, biomass production and habitat support.It also provides a good estimate of soil compaction due to reduced total porosity.Compaction is generally associated with land degradation (the inverse of soil quality, Huber et al., 2008) and can result in decreases in water holding capacity, water infiltration, microbial functions and biogeochemical cycling (Edmondson et al., 2011;Gregory et al., 2015a).It is derived by measuring dry bulk density (BD) modified by clay content (C) and is a very useful parameter for spatial interpretations that require a measure of the compactive state of soils (Jones et al., 2003).Bulk density (from which packing density is derived) is most commonly measured using a Kopecky ring.This method is easy, convenient and cheap, but results can be unrepresentative over large spatial areas due to the small diameter of the ring or cylinder, and depth of measurement (usually 5 cm).A number of proxies exist that overcome some of the issues regarding sampling effort using the traditional Kopecky ring method.These allow a higher resolution of measurements (1500-2500 samples) per hectare over larger areas and include online (mobile) and non-mobile systems (Rickson et al., 2012).The methods used require multiple sensors and advanced techniques for data analysis (Mouazen and Ramon, 2006) such as a combination of visual and near-infrared (vis-NIR) measurements, combined with Theta probe determinations for soil moisture or with soil resistance (penetrometer measurements) and vis-NIR measurements to determine BD (and thus PD, when combined with clay content).
Measurements of packing density (PD) can detect relatively large changes in soil physical properties.PD has been used to detect differences in soil compaction between different management practices, such as contrasting tillage systems (da Silva et al., 2001;Dam et al., 2005).For example, in no-till systems, BD can be 10 % higher compared to conventional tillage systems, particularly in the 0-10 cm layer (Dam et al., 2005).
The power analysis based on land use by soil strata from national data (Fig. 2) clearly demonstrates the trade-off between the sample size required to detect a meaningful change in packing density (i.e. a change that impacts on soil functioning).Approximate sample sizes for a national monitoring programme can be determined based on expense and desired power.In terms of sampling effort, this suggests that a different sampling regime would be required for different geographical areas to ensure statistical robustness, taking into account the different land use/soil and climate combinations.
The influence of spatial scales on sample size was calculated using a model-based approach where the variation of different regions (size of spatial unit) was obtained from a variogram in Methods S9.The sample size needed if a change is to be determined over different spatial scales areas (i.e.field: 5 and 10 km 2 ; farm: 25 km 2 ; and landscape level:  50 km 2 ) was determined (Fig. 3).The graphs suggest that as spatial area increases, the number of samples also needs to increase in order to determine change within a given size of spatial area.If other factors that contribute to spatial variability of PD (such as land use) are included, fewer samples are required.

Soil water retention characteristics
Soil water retention characteristics (SWRCs) encapsulate a number of important capacity-based physical SQIs, includ-ing plant available water capacity (PAWC), air capacity (AC), relative field capacity (RFC), macroporosity (M), soil porosity (Reynolds et al., 2002(Reynolds et al., , 2009) ) and the soil physical quality index, the Dexter "S" value (Dexter, 2004a, b, c).Of these, PAWC, M and the Dexter S value are related to root growth and therefore directly to provisioning soil functions such as crop production.PAWC refers to the soil's capacity to store and provide water that is available for uptake by plant roots.M represents the volume of macropores with an equivalent pore diameter ≥ 300 µm, indicating the capacity of the soil to drain excess water quickly and facilitate root growth (Reynolds et al., 2009).The Dexter S value is a measure of the micro-porosity of the soil (Dexter, 2004c) and has been linked to a number of soil physical processes and soil quality indicators, including bulk density.It is also related to root growth in soil (Dexter, 2004a).Generally, the higher the value of S, the higher the soil physical quality.It is recommended that the S value be used in combination with other capacity-based indicators.This is because, in some soils, values may be overestimated (e.g.sands with unimodal and narrow pore size distributions) (Reynolds et al., 2009).
The Dexter S value and other capacity-based physical SQIs are related to pore volume and pore size distribution (Reynolds et al., 2009).They are derived from soil hydraulic behaviour and therefore are likely to be more sensitive to temporal and spatial changes in soil condition and soil quality compared to other less dynamic indicators which look solely at pore volume, such as bulk density (Dexter, 2004a;Merrington et al., 2006;Naderi-Boldaji and Keller, 2016).The optimum values for each of the relevant physical SQIs for the provisioning function are displayed in Table 2 and are assumed to represent a meaningful change in the physical SQI as changes of this magnitude are expected to affect root (and therefore crop) growth.
In order for the soil water retention characteristics SQI to be meaningful, it needs to be indicative of soil functions that operate at different spatial scales (i.e.laboratory to field to catchment).However, O'Connell et al. (2004O'Connell et al. ( , 2007) ) and Beven et al. (2008) discuss uncertainties and inconsistencies in the measurement of rainfall and flow data between years, which tend to dominate over the impacts of land use and management change on flow characteristics at the catchment scale over time.These include uncertainties in estimates of precipitation inputs to a catchment, uncertainty in measurements of stream discharges (particularly during flooding events), and the uncertainty in characterising land use/management patterns in space and time.Also, significant impacts at the local scale may not have significant impact at catchment scales, due to poor landscape connectivity (Rickson et al., 2012).As such, there are gaps in connecting soil hydrological processes and the physical properties that influence them at the larger scale, and this influences any sampling efforts.
In terms of sampling effort, the standard Soil Survey of England and Wales method for determining soil water reten-  S g < 0.020 "Very poor" soil physical quality 0.020 ≥ S g ≤ 0.035 "Poor" soil physical quality 0.035 ≥ S g ≤ 0.050 "Good" soil physical quality S g ≥ 0.050 "Very good" soil physical quality tion characteristics is to collect three undisturbed soil samples per horizon in winter or spring, when the soil is near field capacity (Avery and Bascomb, 1982).This involves using a coring device that reduces compaction during sampling.The laboratory measurement of soil water retention characteristics can be lengthy and requires considerable effort.The process involves saturation of the soil samples, allowing soils to reach equilibration, determining bulk density and finally calculating the volumetric water content at different soil water suctions.For the current analysis, soil water retention curves were calculated from soil water retention data from the Lan-dIS database (http://www.landis.org.uk/;see Table 1).The method used is shown in Methods S11 in the Supplement.
As an alternative, pedotransfer functions (PTFs) can be a proxy technique that can be used to derive these properties from simple to measure soil characteristics such as BD and soil carbon (C) (Matula et al., 2007;Mayr and Jarvis, 1999).Two types of PTFs were considered: the first represents a standard type PTF that is derived using multiple linear regressions (MLRs); the second is an extension of the MLR approach in which categorical data such as "soil series" and "land use" can be considered.Multiple additive regression splines (MARS) is a nonparametric regression technique that combines both regression splines and model selection meth-ods (Friedman, 1991).The general method used for the PTFs is described in Methods S10.
The results of the PTF were compared for fit (Table 3) and show a high level of agreement.MARS regression approaches tended to perform better than the standard regression approaches.The predicted values of the SWRC indicators were compared against the observed values calculated from the Land IS database.Again, there was good agreement amongst the PTFs (Fig. 4), and as such these approaches are feasible as a proxy for SWRC.It has been recommended that for a plot of 20 by 20 m, 25 bulked samples would be required for the measurement of BD and organic C that are required for the input data for the PTFs (Rickson et al., 2012).

Soil sealing
Soil sealing refers to the impermeabilisation of soils resulting from natural factors (Pulido Moncada et al., 2014) and human activities (for example road construction) (Xiao et al., 2013).In the context of this work, soil sealing refers to the covering of soil surfaces by expanding urban infrastructure.
Soil sealing has been identified as one of the greatest threats to soil functions in the UK (Rawlins et al., 2013) and worldwide (García et al., 2014;Jie et al., 2002).The growth of these impervious areas is regarded as an indicator of land  degradation (Munafò et al., 2013) as it results in interruptions to gaseous, water and energy exchanges in soils (for example water regulation); decreased biomass production; and increased concentrations of soil pollutants (Scalenghe and Marsan, 2009).Soil sealing also has a climatic impact by altering surface albedo and air temperature, and can impact on soil biogeochemical cycles (Gregory et al., 2015b;Zhao et al., 2012).In order to observe and assess changes in these soil functions, the change in the proportion of sealed surfaces must also be monitored.
There are a number of methods to evaluate soil sealing that have been used in the past, ranging from statistical analysis of national cadastral maps to aerial photo interpretation (Rickson et al., 2012).Currently, remote sensing techniques are favoured as they have a large spatial and temporal coverage, have improved certainty of measurements and also provide baseline data on the proportion of sealed soils within urban areas.The extent of the built environment can be estimated using a number of remote sensing techniques including high-resolution satellite imagery (< 1 m ground resolution) and aerial photography.Both these methods allow for the inclusion of narrow corridors such as roads and rail tracks, as well as providing accurate estimates of unsealed soil areas surrounding urban areas (i.e.green spaces).By integrating remote sensing with other, existing databases such as soil maps, even finer spatial resolutions can be achieved (Rickson et al., 2012).
In terms of measurement, the key indicators for soil sealing are (1) the absolute area of sealed soil (ha) and ( 2) the change/growth rate of area of sealed soil (ha yr −1 , ha d −1 , % change to baseline).With the first indicator, the levels of soil sealing can depend on a number of factors including policy decisions, individual's choice and the degree of coverage (Meinel and Hernig, 2005) ranging from 100 % sealed (roofs, concrete, asphalt) to 70 % sealed (paving slabs with seep-able joints) to 50 % sealed or less (green roofs, gravel, crushed stone, porous pavements).The measure of change/growth rate associated with the second indicator must also incorporate any de-sealing (or negative sealing) that would reduce the extent of sealed soil (for example installation of green roofs, porous block paving, porous tarmac and geotextiles used in car parks).This would require very high- resolution (< 1 m) monitoring data as the areas can be small and fragmented.
There are a number of earth observation data that can be used (Table 4) for identifying, classifying and monitoring soil sealing.They all have advantages, disadvantages and considerations for the user in terms of sampling/data analysis effort required.One of the most important considerations relates to spatial and temporal scale.The use of remotely sensed information allows population estimates to be made in the imaged area at the pixel resolution.As such there is usually a trade-off between the resolution and area that is covered.In terms of the spatial scale for urban areas, very high-resolution data (< 1 m) are recommended to monitor smaller sealed and fragmented areas such as domestic driveways.This scale is also recommended for determining de-sealed surfaces which tend to be small scale.
Regarding appropriate temporal scales of measurement and monitoring, soil sealing in urban areas can occur on the timescale of months to years depending on what is being built.Furthermore, the capture of remote sensing data, in particular very high-resolution imagery usually occurs only every 3-5 years (Rickson et al., 2012) and therefore a monitoring schedule would have to fit around this.If medium-to high-spatial-resolution imagery is to be used, sampling could take place annually (data are collected more frequently and has a larger spatial coverage) (Rickson et al., 2012).

General approach
The multi-stage approach used in this study proved to be flexible, and whilst there was paucity in the data, it can be altered according to the needs of the end user/monitoring body/policy maker and what they want to get out of a soil monitoring programme.These diverse needs to be reflected in the priorities set in the logical sieve process, cost considerations, sample numbers and/or what constitutes meaningful change for that end use.In order to test for meaningful change in the selected indicators, spatial and temporal data are required to reflect the variability of each property (signal : noise ratio).However, in the examples given, recommendations for a sampling effort were based solely on the scientific literature.In this case, the evidence base was poor in terms of data that are meaningful (i.e.degree of change in the SQI that will affect soil processes and functions) and detectable (sample size required to detect the meaningful signal from the variability in the signal) (Rickson et al., 2012).In order to overcome this, further work is required to build up the evidence base in terms of spatial and temporal data on the key SQIs.Where other sampling issues were identified, suitable proxies or modelling functions were tested and proved to be effective in terms of how well they correlated with the standard measurements for the indicator and any time/labour issues associated with its measurement.

Conclusions
This study has demonstrated a multi-stage process that prioritises and analyses the suitability of physical SQIs for monitoring soil quality and function.In the first stage a logical sieve and alternative scenario approach were used to prioritise candidate physical SQIs from the literature.These were then assessed for uncertainty in their measurement, spatial variability, expected rate of change and impacts on soil processes and functions.Of the seven prioritised physical SQIs, three were selected here as case studies representing the varying degrees of analysis and modelling that could be applied depending on the evidence base.
By emphasising the current key soil functions related to current soil and environmental policy in the UK (i.e.provisioning and regulating functions), the prioritised SQIs can be related to soil processes, soil functions and consequent delivery of ecosystem goods and services.These are likely to shape any future soil and environmental policy in the UK, as well as efforts to develop soil monitoring programmes that aim to evaluate soil physical quality.
Data availability.Data used in this analysis pertains to the LandIS datasystem and can be accessed through www.LandIS.org.uk and has is recorded on data.gov.ukas NSRI-LANDIS-NATMAP1000.
Competing interests.The authors declare that they have no conflict of interest.

Figure 1 .
Figure 1.Multi-stage approach taken in the selection of meaningful physical soil quality indicators (SQIs).

Figure 2 .
Figure 2. Power analysis on national soil packing density data based on land use by soil strata.The spatial distribution of the data points superimposed on the land use/soil classification is taken from Graves et al. (2011).

Figure 3 .
Figure3.Power analysis using a model-based approach in which the variability was estimated given a particular block size using the variogram described in Methods S9.Soil water retention characteristics.

Figure 4 .
Figure 4. Biplots representing the predicted SQI versus observed SQI based on the MARS pedotransfer functions.(a) Relative field capacity.(b) Drainable porosity.(c) Plant available water.
is the candidate SQI indicative of (i.e.what function is being degraded)?
2. What it is responsive to?How responsive is it (i.e.sensitivity, responsiveness)?3.What factors may mitigate or accentuate the response (e.g.soil type, land use)?4.Is this indicator a first-order indicator (i.e. the SQI is a direct measure of the change in soil quality) or a second-, third-, etc. order indicator (i.e. the change in the SQI is measured indirectly, e.g. by remote sensing)?5.Are there existing or suspected data holdings for the indicator?6.How is it measured?7.What sampling support does it need?8.What is the sampling intensity required?

Table 1 .
Datasets and analyses for selected physical SQIs.

Table 2 .
(Dexter, 2004a) (2009)haracteristics indicators, optimum values and impacts on the provisioning soil function.Values for PAWC, M and RFC taken fromReynolds et al. (2009).Values for Dexter S value taken from(Dexter, 2004a).θ FC : volumetric moisture content at field capacity, occurring at 0.5 or 1 m pressure head; θ sat : saturated moisture content at 0 m pressure head; θ PWP : moisture content at permanent wilting point, occurring at 150 m pressure head; θ m : porosity of the soil matrix occurring at 0.1 m pressure head.

Table 3 .
Soil water retention characteristics: fit results from PTFs based on LandIS data (BD, texture (clay, silt and sand) and organic C content).S v is related to S g through the soil bulk density ρ b S v= ρ b S g .RSQ conc R RSQ conc R RSQ conc R RSQ conc R RSQ: R 2 statistic; conc R: correlation coefficient.