The soil quality and health definitions both mention “the continued capacity of a soil to function” (FAO, 2008; Bünemann et al., 2018). Soil functions therefore have a central role in the quality and health debate. EC (2006) defined the following soil functions: (1) biomass production, including agriculture and forestry; (2) storing, filtering and transforming nutrients, substances and water; (3) biodiversity pool, such as habitats, species and genes; (4) physical and cultural environment for humans and human activities; (5) source of raw material; (6) acting as carbon pool; and (7) archive of geological and archaeological heritage. Functions 4, 5 and 7 are not covered in this contribution since, if considered relevant, specific measures have to be taken to set soils apart by legislative measures. The other functions are directly and indirectly related to Function 1, biomass production. Of course, soil processes not only offer contributions to biomass production, but also to filtering, biodiversity preservation and carbon storage. Inter- and transdisciplinary approaches are needed to obtain a complete characterization, requiring interaction with other disciplines, such as agronomy, hydrology, ecology and climatology and, last but not least, with stakeholders and policymakers. Soil functions thus contribute to ecosystem services and, ultimately, to all 17 UN Sustainable Development Goals (e.g., Bouma, 2016, 2014; Keesstra et al., 2016). However, in the context of this paper, attention will be focused on Function 1, biomass production.

Soil physical aspects play a crucial role when considering the role of soil in biomass production, as expressed by Function 1, which is governed by the dynamics of the soil–water–atmosphere–plant system in three ways:

Roots provide the link between the soil and plant. Rooting patterns as a function of time are key factors for crop uptake of water and nutrients. Deep rooting patterns imply less susceptibility to moisture stress. Soil structure, the associated bulk densities and the soil water content determine whether or not roots can penetrate the soil. When water contents are too high, either because of the presence of a water table or of a dense, slowly permeable soil horizon impeding vertical flow, roots will not grow because of lack of oxygen. For example, compact plow pans, resulting from the application of pressure on wet soil by agricultural machinery, can strongly reduce rooting depth. In fact, soil compaction is a major form of soil degradation that may affect up to 30 % of soils in some areas (e.g., FAO and ITPS, 2015).

Soil functions 3 and 6 are a function of the organic matter content of the soil (or % organic carbon – %OC), the quantity of which is routinely measured in chemical soil characterization programs (also in the soil health protocols mentioned earlier that also define methods to measure soil respiration). The organic matter content of soils is highly affected by soil temperature and moisture regimes and soil chemical conditions. Optimal conditions for root growth in terms of water, air and temperature regimes will also be favorable for soil biological organisms, linking soil functions 1, 3 and 6.

When defining soil physical aspects of soil quality and soil health, focused on soil Function 1, parameters that integrate various aspects will have to be defined, such as (1) weather data; (2) the infiltrative capacity of the soil surface, considering rainfall intensities and quantities; (3) rootability as a function of soil structure, defining thresholds beyond which rooting is not possible; and (4) hydraulic and root extraction parameters that allow a dynamic characterization of the soil–water–atmosphere–plant system. This system can only be realized by process modeling, which requires the five parameters listed above, and is therefore an ideal vehicle to realize interdisciplinary cooperation. Simulation models of the soil–water–atmosphere–plant system are ideal to integrate these various aspects.

When analyzing soil quality and soil health, emphasis must be on the dynamics of vital, living ecosystems requiring a dynamic approach that is difficult to characterize with static soil characteristics (such as bulk density, organic matter content and texture) except when these characteristics are used as input data into dynamic simulation models of the soil–water–plant–climate system. Restricting attention to soil physical characteristics, hydraulic conductivity (K) and moisture retention properties (O(h)) of soils are applied in such dynamic models. Measurement procedures are complex and can only be made by specialists, making them unsuitable for general application in the context of soil quality and health. They can, however, be easily derived from pedotransfer functions (PTFs) that relate static soil characteristics such as bulk density, texture and %OC to these two properties, as recently summarized by Van Looy et al. (2017). The latter soil characteristics are available in existing soil databases and are required information for the dynamic models predicting biomass production.

Simulation models of the soil–water–atmosphere–plant system, such as the Soil Water Atmosphere Plant model (SWAP) (Kroes et al., 2008) to be discussed later in more detail, integrate weather conditions, infiltration rates, rooting patterns and soil hydrological conditions in a dynamic systems approach that also allows exploration of future conditions following climate change. The worldwide agronomic yield-gap program (http://www.yieldgap.org, last access: 29 December 2018) can be quite helpful when formulating a soil quality and health program with a global significance. So-called water-limited yields (Yw) can be calculated, assuming optimal soil fertility and lack of pests and diseases (e.g Gobbett et al., 2017; van Ittersum et al., 2013; Van Oort et al., 2017). Yw reflects climate conditions at any given location in the world as it is derived from potential production (Yp) that reflects radiation, temperature and basic plant properties, assuming that water and nutrients are available and pests and diseases do not occur. Yw reflects local availability of water. Yw is usually, but not always, lower than Yp. Yw can therefore act as a proxy value for physical soil quality, focusing on Function 1. Note that Yp and Yw, while providing absolute science-based points of reference, include assumptions on soil fertility and crop health.

Actual yields (Ya) are often lower than Yw (e.g., Van Ittersum et al., 2013). The ratio Ya / Yw is an indicator of the so-called yield gap showing how much potential there is at a given site to improve production (http://www.yieldgap.org) (Bouma, 2002). When multiplied with 100, a number between 1 and 100 is obtained as a quantitative measure of the yield gap for a given type of soil. Yw can be calculated for a non-degraded soil. Ya should ideally be measured but can also be calculated as was done in this exploratory study (in terms of Yw values) on the basis of the assumed effects of different forms of soil degradation, such as subsoil soil compaction, poor water infiltration at the soil surface due to surface compaction, or crusting and erosion. This requires the introduction of a compact layer (plow pan) in the soil, a reduction of rainfall amounts with the volume of estimated overland flow and the removal of topsoil. Each variant of the analyzed soil series represents a phenoform. In this exploratory study Ya values were simulated but, ideally, field observations should be made in a given soil type to define effects of management as explored, for example, by Pulleman et al. (2000) for clay soils and Sonneveld et al. (2002) for sandy soils. They developed phenoforms based on different %OC of surface soil and such phenoforms could also have been included here to provide a link with soil biology, but field data were not available to do so. Field work identifying phenoforms includes important interaction with farmers as also mentioned by Moebius-Clune et al. (2016). Sometimes, soil degradation processes, such as erosion, may be so severe that the soil classification (the soil genoform) changes. Then, the soil quality and soil health discussion shifts to a different soil type.

This approach will now be explored with a particular focus on the Mediterranean environment. Physical soil quality is defined by Yw for each soil, considering a soil without assumed degradation phenomena (the reference) and for three variants (hypothetical Ya, expressed in terms of Yw) with (1) a compacted plow layer; (2) a compacted soil surface resulting in overland flow; and (3) removal of topsoil following erosion, without a resulting change in the soil classification. This way a characteristic range of Yw values is obtained for each of the six soil series, reflecting positive and negative effects of soil management and representing a range of soil physical quality values of the particular soil series considered. Within this range an actual value of Ya will indicate the soil physical health of the particular soil at a given time and its position within the range of values will indicate the severity of the problem and potential for possible improvement.

The ratio (Ya/Yw)×100 is calculated to obtain a numerical value that represents soil health as a point value, representing actual conditions. Health is relatively low when real conditions occur in the lower part of the soil quality range for that particular soil and relatively high when it occurs in the upper range. Again, in this exploratory study measured values (at current climate conditions) for Ya have not been made, so Ya only applies to the three degraded soil forms being distinguished here where hypothetical effects of soil degradation have been simulated as related to the corresponding calculated Yw values. Of course, actual measured Ya values cannot be determined at all when considering future climate scenarios and simulation is the only method allowing exploratory studies. We assume that climate change will not significantly affect soil formation processes until the year 2100. Soil properties will therefore stay the same.

To allow estimates of the possible effects of climate change, the Representative Concentration Pathway (RCP) 8.5 scenario based on the Intergovernmental Panel on Climate Change (IPCC) modeling approach will be applied. Obviously, only computer simulations can be used when exploring future conditions – another important reason to use dynamic simulation modeling in the context of characterizing soil quality and soil health. The approach in this paper extends earlier studies on soil quality for some major soil types in the world that did not consider aspects of soil health nor effects of climate change (Bouma, 2002; Bouma et al., 1998).

The soil–water–atmosphere–plant model (Kroes et al., 2008) was applied to solve the soil water balance. SWAP is an integrated physically based simulation model of water, solute and heat transport in the saturated–unsaturated zone in relation to crop growth. In this study only the water flow module was used; it assumes unidimensional vertical flow processes and calculates the soil water flow through the Richards equation. Soil water retention θ(h) and hydraulic conductivity K(θ) relationships as proposed by van Genuchten (1980) were applied. The unit gradient was set as the condition at the bottom boundary. The upper boundary conditions of SWAP in agricultural crops are generally described by the potential evapotranspiration ETp, irrigation and daily precipitation. Potential evapotranspiration was then partitioned into potential evaporation and potential transpiration according to the LAI (leaf area index) evolution, following the approach of Ritchie (1972). The water uptake and actual transpiration were modeled according to Feddes et al. (1978), where the actual transpiration declines from its potential value through the parameter α, varying between 0 and 1 according to the soil water potential.

The model was calibrated and validated by measured soil water content data at different depths for Italian conditions (Bonfante et al., 2010; Crescimanno and Garofalo, 2005) and in the same study area by Bonfante et al. (2011, 2017). In particular, the model was evaluated in two farms inside of Destra Sele area, on three different soils (Udic Calciustert, Fluventic Haplustept and Typic Calciustoll), under maize crop (two cropping seasons) during the regional project Campania Nitrati (Regione Campania, 2008) (Table 2).

Soil hydraulic properties of soil horizons in the area were estimated by the pedotransfer function HYPRES (Wösten et al., 1999). A reliability test of this PTF was performed on θ(h) and K(θ) measured in the laboratory by the evaporation method (Basile et al., 2006) on 10 undisturbed soil samples collected in the Destra Sele area. The data obtained were compared with estimates by HYPRES and were considered to be acceptable (RMSE = 0.02 m3 m-3) (Bonfante et al., 2015).

Simulations were run considering a soil without assumed degradation phenomena (the reference) and for three variants with a compacted plow layer, surface runoff and erosion, as discussed above. i.

The compacted plow layer was applied at -30 cm (10 cm of thickness) with the following physical characteristics: 0.30 WC at saturation, 1.12 n, 0.004 “a” and Ks of 2 cm day-1. Roots were restricted to the upper 30 cm of the soil.

The Destra Sele study area, the plain of the Sele river (22 000 ha, of which 18 500 ha is farmed), is situated in the south of Campania, southern Italy (Fig. 1). The main agricultural production consists of irrigated crops (maize, vegetables and fruit orchards), greenhouse-grown vegetables and mozzarella cheese from water buffalo herds. The area can be divided into four different landform classes (foothills, alluvial fans, fluvial terraces and dunes) with heterogeneous parent materials in which 20 different soil series were distinguished (within Inceptisol, Alfisol, Mollisol, Entisol and Vertisol soil orders) (Regione Campania, 1996), according to Soil Taxonomy (Soil Survey Staff, 1999). Six soil series were selected in the area to test application of the soil quality and soil health concepts. Representative data for the soils are presented in Table 1.

Decision trees were developed to test whether the selection process of the soil series was based on stable criteria, allowing extrapolation of results from measured to unmeasured locations when considering effects of climate change. While extrapolation in space of soil series data has been a common procedure in soil survey (e.g., Soil Survey Staff, 2014; Bouma et al., 2012), extrapolation in time has not received as much attention. A basic principle of many taxonomic soil classification systems is a focus on stable soil characteristics when selecting diagnostic criteria for soil types. Also, emphasis on morphological features allows, in principle, a soil classification without requiring elaborate laboratory analyses (e.g., Soil Survey Staff, 2014). A given soil classification should, in order to obtain permanent names, not change following traditional management measures, such as plowing. This does, however, not apply to all soils and therefore a different name will have to be assigned.

This way, soil classification results in an assessment of the (semi)permanent physical constitution of a given soil in terms of its horizons and textures. That is why soil quality is defined for each soil type as a characteristic range of Yw values, representing different effects of soil management that have not changed the soil classification.

Future climate scenarios were obtained by using the high-resolution regional climate model (RCM) COSMO-CLM (Rockel et al., 2008), with a configuration employing a spatial resolution of 0.0715∘ (about 8 km), which was optimized over the Italian area. The validations performed showed that these model data agree closely with different regional high-resolution observational datasets, in terms of both average temperature and precipitation in Bucchignani et al. (2015) and in terms of extreme events in Zollo et al. (2015).

In particular, the Representative Concentration Pathway 8.5 scenario was applied, based on the IPCC modeling approach to generate greenhouse gas (GHG) concentrations (Meinshausen et al., 2011). Initial and boundary conditions for running RCM simulations with COSMO-CLM were provided by the general circulation model CMCC-CM (Scoccimarro et al., 2011), whose atmospheric component (ECHAM5) has a horizontal resolution of about 85 km. The simulations covered the period from 1971 to 2100; more specifically, the CMIP5 historical experiment (based on historical greenhouse gas concentrations) was used for the period 1976–2005 (reference climate scenario – RC), while for the period 2006–2100, a simulation was performed using the IPCC scenario mentioned. The analysis of results was made on RC (1971–2005) and RCP 8.5 divided into three different time periods (2010–2040, 2040–2070 and 2070–2100). Daily reference evapotranspiration (ET0) was evaluated according to the Hargreaves and Samani (1985) equation (HS). The reliability of this equation in the study area was performed by Fagnano et al. (2001), comparing the HS equation with the Penman–Monteith (PM) equation (Allen et al., 1998).

Under the RCP 8.5 scenario the temperature in Destra Sele is expected to increase by approximately 2 ∘C respectively every 30 years to 2100 starting from the RC. The differences in temperature between RC and the period 2070–2100 showed an average increase in minimum and maximum temperatures of about 6.2 ∘C (for both min and max). The projected increase in temperatures produces an increase in the expected ET0. In particular, during the maize growing season, an average increase in ET0 of about 18 % is expected until 2100.

Soil physical quality of the six soil series, expressed as calculated Yw values for the reference climate and for future climate scenario RCP 8.5 as well as for three time periods, is shown in Fig. 2. Considering current climate conditions, the Longobarda and Cifariello soils with loamy textures have the highest values, while the sandy soil Lazzaretto is lower. This can be explained by greater water retention of loamy soils (180 and 152 mm of AWC in the first 80 cm for Longobarda and Cifariello respectively) compared to the sandy soil (53 mm of AWC in the first 80 cm for Lazzaretto). The effects of climate change are most pronounced and quite clear for the two periods after 2040. Reductions compared with the period up to 2040 range from 20 %–40 %, the highest values associated with sandier soil textures. This follows from the important reduction of projected rainfall during the cropping season (Fig. 3) ranging from an average value of 235 (±30) mm in the 2010–2040 period to 185 (±26) mm (-21 %) and to 142 (±24) mm (-40 %) in the 2040–2070 and 2070–2100 periods respectively (significant at p<0.01). The figure also includes a value for Yp, potential production (under RC with optimal irrigation), which is 18 t ha-1, well above the Yw values. Only a Yp value is presented for current conditions because estimates for future climates involve too many unknown factors.