Impact of climate, soil properties and grassland cover on soil water repellency

Impact of climate, soil properties and grassland cover on soil repellency. Numerous soil water repellency (SWR) studies have investigated the possible causes of this temporal phenom- enon, yet there remains a lack of knowledge on the order of importance of the main driving forces of SWR in the context of changing environmental conditions under grassland ecosystems. To study the separate and combined effects of soil texture, climate, and grassland cover type on inducing or altering SWR, four sites from different climatic and soil regions were selected: Ciavolo (CI, IT), Cs ´ olyosp ´ alos (CSP, HU), Pwllpeiran (PW, UK), Sekule (SE, SK). The investigated parameters were the extent (determined by repellency indices RI, RI c and RI m ) and persistence (determined by water drop penetration time (WDPT) and water repellency cessation time, WRCT) of SWR, as well as field water ( S w ) and ethanol ( S e ) sorptivity, water sorptivity of hydrophobic soil state ( S wh ) water sorptivity of nearly wettable soil state ( S ww ) and field hydraulic conductivity ( K ). Our findings showed an area of land has a greater likelihood of being water repellent if it has a sandy soil texture and/or a high frequency of prolonged drought events. Water infiltration was positively correlated with all the sorptivities ( r = 0.32 – 0.88), but was mostly negatively correlated with RI ( r = – 0.54 at CI), WDPT ( r = – 0.47 at CI) and WRCT ( r = – 0.58 at CI). The importance of natural and synanthropized vegetation covers with regards to SWR was not coherent; moving to regions having coarser texture or moving to drier climatic zones led to higher risk of SWR conditions. Climate change has been predicted to lead to more frequent extreme weather events and prolonged dry periods across Europe, which will most likely increase the extent of SWR-affected areas and increase the role of SWR in water management of grassland ecosystems. Therefore, there is a need to determine SWR risk zones to prevent decreases in soil moisture content, soil fertility, carbon and nitrogen sink potentials, as well as biomass production of the related agro-ecosystems.


Introduction
More than one fifth (20.7%) of EU-28 ′ s area is covered by temperate grasslands (EUROSTAT, 2015). These grasslands are characterized by considerable plant diversity depending on environmental conditions, management practices (grazing, mowing) and land use intensity (Fischer et al., 2015). Vegetation can have beneficial effects on surface hydrologic properties, including increasing infiltration capacity and erosion control (Moreno- de Las Heras et al., 2009). Mechanisms associated with these effects include soil organic matter accumulation and soil aggregation, the prevention of soil crusting, development of sub-surface channels in association with plant rooting (Alaoui et al., 2011;Beven and Germann, 2013;Clark and Zipper, 2016). Besides the listed beneficial effects of vegetation cover on hydrologic properties, the hydrophobic compounds released from roots (Robinson et al., 2010) may significantly change soil-water interactions at the soil surface with potentially negative effects on soil hydrological processes (e.g., Doerr et al., 2000).
SWR has positive effects on the stability of soil aggregates (Fér et al., 2016;Goebel et al., 2005) and may reduce losses of soil water by evaporation (Bachmann et al., 2001). Rye and Smettem (2017) also concluded that water repellent surface layers can bring about significant reductions in net evaporative moisture loss, in which case the phenomenon of SWR has positive impacts on water budgets, the grassland ecosystem and its services. On the other hand, SWR can lead to irregular moisture patterns, e.g. preferential flow (Täumer et al., 2006), or to a diminished infiltration enhancing overland flow, and consequently increases the risk of surface runoff, soil erosion and surface waters pollution (Diehl, 2013;Doerr et al., 2000;Orfánus et al., 2014Orfánus et al., , 2016. Due to reduced soil-water redistribution resulting in poor germination and plant growth, reduced agricultural productivity can be found in water repellent soils (Roper, 2005;Ward et al., 2015). Furthermore, rewetting of water-repellent soils could have an enhanced effect on pathogen (such as Rhodococcus erythropolis) release and transport (Sepehrnia et al., 2019). Additional studies showed that SWR can be significantly decreased by adding lime, kaolinite clay, or wax degrading bacteria to the soil (Lichner et al., 2002;McKissock et al., 2000;Roper, 2005;2006).
Changing global environmental conditions (e.g. temperature and precipitation) also have a significant effect on SWR and plant community structure (Kneitel, 2010), and thus indirectly affect worldwide food production and security. Temperature change is among the key drivers of our future climate, and global mean land surface air temperatures over terrestrial areas have already increased by 1.53 • C from 1850 to 1900 to 2006(IPCC, 2019. More frequent extreme arid years are predicted for the future, with longer dry spells, drought periods (Meehl et al., 2007) and heatwaves (Fischer and Schӓr, 2010). As the changes in temperature and precipitation will not be uniform, the outcomes will vary between different climate zones. In dry summers, available soil water content (SWC) for vegetation is also reduced by the phenomenon of SWR, which shows significant seasonal variability (Müller et al. 2014) and spatial heterogeneity (Sándor et al., 2015). SWR is generally found to be the most extreme when soils are dry, declining and eventually disappearing as soils become wet (de Jonge et al., 1999;Dekker and Ritsema, 1994;Fér et al., 2016;Lichner et al., 2013). This is the main reason why heavy rainfalls after longer dry periods may result in considerable surface runoff (Imeson et al., 1992) and soil erosion (Doerr and Moody 2004). Müller et al. (2018) identified the SWR as a factor leading to runoff and phosphorus losses, where about 20% of the applied 45 kg P ha − 1 fertilizer was lost to runoff. Thus, the predicted increase in annual mean global temperature and heatwave frequency is expected to influence SWR and the related soil water balance processes in the future (Goebel et al., 2011). The phenomena of SWR is not only present in arid and semiarid regions; e.g. Lozano-Baez et al. (2020) detected it in the Atlantic Forest region of Brazil. For this reason, it is essential to clarify those criteria which could turn a balanced or carbon sequestrating grassland ecosystem into a carbon releasing environment owing to enhanced effects of SWR related to climate change.
The objectives of this study were to demonstrate the effect of different factors on SWR for sites with different soil, climatic and vegetation characteristics by: i) analysing the effect of the two most important factors (soil texture and climate), ii) describing the differences in SWR as a consequence of different grassland cover across the investigated soil types and climatic regions, iii) identifying the circumstances with potentially high and low chance of water repellency in order to support the reduction of the negative effects of SWR on the ecosystem.

Material and methods
To determine the effect of different climate and soil properties on the severity of SWR, four locations were selected across Europe: in Italy, Hungary, Slovakia and the UK (Fig. 1). This approach to site selection provided opportunities to test different combinations of climate (short vs long dry periods) and soil texture (coarse vs fine texture). The two Eastern-European sites (cf. Fig. 1) were characterised in the greatest detail, particularly with regards to vegetation cover. SWR experiments were carried out in September of 2016 at Pwllpeiran and Ciavolo, and in June and September of 2017 at Sekule and Csólyospálos sites, respectively. All cases the measurements were taken after at least five consecutive dry days. Daily meteorological datasets were available from 1980 to 2017 for all the sites. Each experimental area was investigated on a pedon scale, which is generally considered to be 'homogenous'. Approximately 0.5-1 cm thick soil layer was removed before the measurements to reduce the influence of, e.g., aggregates compacted by rain or vegetation residues on water repellency (Rodríguez-Alleres and Benito, 2011). An area of around 1 m 2 area was divided into 20 cm × 20 cm cells, where SWC, water drop penetration time (WDPT), field unsaturated hydraulic conductivity (K), water sorptivity (S w , S ww , S wh ), repellency indexes (RI, RI c , RI m ) and water repellency cessation time (WRCT) were measured in 5-25 replications, while ethanol sorptivity (S e ) was measured in 5-11 replications. Sand, silt, clay, CaCO 3 , and organic carbon contents were determined in the laboratory using sieved (2-mm-mesh sieve), mixed, and air-dried (at 30 • C) samples collected from the topsoil (0-5 cm) of the experimental sites in 3 to 10 repetitions. The initial SWC was measured from each cells of the field measurement quadrat using a HH2 Soil Moisture Meter combined with WET-2 sensor (Delta-T Devices, Cambridge, UK). Potential WDPT (WDPT P ) of the PW soil was measured in laboratory on dried samples (SWC =~0.00 m 3 m − 3 ) in five repetitions, because the SWC value was too high there during the field measurements.

Study sites
The first experimental site is located in Ciavolo on the island of Sicily, Italy (37 • 45 ′ 40.6 ′′ N, 12 • 34 ′ 09.0 ′′ E), in a pine forest glade (site CI) vegetated with spontaneous annual grasses. According to the Kӧppen-Geiger climate classification, the region has a Mediterranean temperate climate with a dry and hot summer (Csa) (Kottek et al., 2006) and winter domination of precipitation (rainfall seasonality index <    Table 1). The multi-year monthly average temperature and precipitation values also show summer drought events (Fig. 2). Elevation at the experimental site is 105 m a.s.l. and the surface slope is low (4.4%). The soil is a Calcaric Chromic Endoleptic Cambisol (WRB, 2014) with a depth of 0.40-0.60 m and the parent material is calcareous sandstone. According to Soil Survey Division Staff (1993) the soil texture is clay loam ( Table 2) and covered by Mediterranean species listed in Table 3. The second experimental site (with two grassland sub-sites) is located in Csólyospálos, Hungary (46 • 25 ′ 05 ′ ' N, 19 • 50 ′ 28 ′ ' E). According to the Kӧppen-Geiger climate classification, the region has a temperate climate without a dry season, hot summer (Cfa) (Kottek et al., 2006), however some years show the characteristic of humid continental climate (Dfb). Table 1 shows the most important climate details and Fig. 2 demonstrates the monthly distribution of precipitation and temperature over years. The average annual rainfall is ~ 520 ± 195 mm with ~ 60% dominance of summer precipitation, while the rest 40% of the investigated period characterised by uniform distribution. Elevation at the experimental site is 102.6 m a.s.l. Sub-site CSP17, abandoned for 17 years, was covered by strongly synanthropized vegetationpoor grassland dominated by perennial weedy grass and annual grasses (listed in Table 3). Sub-site CSP44, abandoned for 44 years, was covered by successionally more advanced grassland dominated by perennial grasses with an admixture of mostly weedy herbs (listed in Table 3). The soil is an Arenosol (WRB, 2014) and has sandy texture (Soil Survey Division Staff, 1993) (Table 2) the parent material is aeolian calcareous nonstructural sand.
The third experimental site is located at Pwllpeiran, mid Wales, UK (52 • 21 ′ 53 ′ ' N, 03 • 49 ′ 39 ′ ' W), on an upland pasture with a long agricultural history, vegetated mostly with Lolium perenne and other species listed in Table. 3 (site PW). According to the Kӧppen-Geiger climate classification, the region has a moderate marine climate (type Cfb; Kottek et al., 2006) with quite cold summers, ~14.6 ± 1.0 • C, and precipitation all year round (Table 1, Fig. 2). Elevation at experimental site is 310 m a.s.l. The soil is Typical Brown Podzolic soil (WRB, 2014) and has a silt loam texture (Soil Survey Division Staff, 1993) (Table 2), with a bedrock depth of between 0.6 and 1 m.
The fourth experimental site (with two grassland sub-sites) is located at Sekule (48 • 37 ′ 10 ′ ' N, 17 • 00 ′ 10 ′ ' E) in the Borská nížina lowland (southwest Slovakia). According to the Kӧppen-Geiger climate classification, the region has a temperate climate without dry season, warm summer (Cfb) (Kottek et al., 2006). Mean annual precipitation is ~542 mm, which is summer-dominant (Table 1) as is shown in Fig. 2. Elevation is 158 m a.s.l. and surface slope is negligible. Sub-site SeN with natural vegetation undisturbed by human activities was dominated by the grasses (Table 3). Of cryptogams, the mosses Bryum capillare, Hypnum cupressiforme, Brachythecium albicans, and the lichen C. furcata were recorded. Sub-site SeS with synanthropized vegetation formed by secondary succession triggered by some anthropogenous disruptions was covered by species listed in Table 3. The soil is an Arenosol (WRB, 2014) and has sandy texture (Soil Survey Division Staff, 1993) and the parent material is aeolian siliceous sand.

Field methods
All the field measurements were carried out on the surface of the studied soils after at least five consecutive rainless days. Field water and ethanol infiltration measurements were performed with a minidisk infiltrometer (MDI, Decagon Devices, 2012) under a pressure head value of h 0 = -2 cm. A pressure head of -2 cm was selected, because tension infiltration experiments are preferred to ponded ones to exclude the contribution of macropores that may overwhelm soil hydrophobicity (Nyman et al., 2010). The sorptivity, S (m s − 1/2 ), was estimated from the cumulative infiltration, I (m) (Clothier et al., 2000): Eq. (1) was used to calculate both the water (S w (m s − 1/2 )) and ethanol sorptivity (S e (m s − 1/2 )) from the cumulative infiltration vs. square root of the time relationships taken from the minidisk infiltrometer measurements . Zhang (1997) proposed using the first two terms of the Philip infiltration equation to estimate the soil hydraulic conductivity K(h 0 ) (m s − 1 ):

Site
Year  Fig. 3. The hockey-stick-like relationship of the cumulative infiltration of water (I) against the square root of time (SQRT t) for sandy soil from the Csólyospálos site. The water sorptivity (S wh ) for water-repellent state of soil and the water sorptivity (S ww ) for nearly wettable state of soil were estimated, respectively, from the less steep and steeper parts of hockey-stick-like relationship. The WRCT was estimated from the point of intersection of two straight lines, representing the I = f(SQRT t) relationships for water-repellent and nearly wettable states of the soil.
where C 2 is the coefficient of the quadratic term of the parabola when the cumulative infiltration data are fitted as a function of the square roort of time and A is a dimensionless coefficient that's value is relating the van Genuchten retention parameters for a given soil to suction at the infiltration surface and the infiltrometer disk radius.. Eq.
(2) was used to estimate the soil hydraulic conductivity K(-2 cm) in this study. The values of A were obtained from the Minidisk Infiltrometer User's Manual (Decagon Devices, 2012).
In the standard method of estimating the repellency index, RI (-), S e and S w were measured in pairwise arrangements and the standard repellency index RI was calculated as follows (Hallett et al., 2001):RI = 1.95 S e / S w (3) In the second method of estimating the repellency index, the combination of all the ethanol (S e (m s − 1/2 )) and water sorptivities (S w (m s − 1/2 )) was used to calculate a combined repellency index, RI c (-), i. e., m × n values of RI (-) were calculated from m values of S w (m s − 1/2 ) and n values of S e (m s − 1/2 ) (Pekárová et al., 2015).
In the third method of estimating the repellency index, the water sorptivity for water-repellent state of soil, S wh (m s − 1/2 ), and the water sorptivity for nearly wettable state of soil, S ww (m s − 1/2 ), were calculated, respectively, from the shallower and steeper parts of hockey-sticklike relationship (Fig. 3). This relationship was used to calculate the modified repellency index RI m (-) (Alagna et al., 2017;Sepehrnia et al., 2016).

Statistical analyses
Daily, monthly and yearly aggregated precipitation, minimum, maximum temperature database were analysed between 1980 and 2017 to estimate climate indexes such as rainfall seasonality, aridity index and to determine the multi-year average annual temperature, summer temperature, annual precipitation and summer precipitation values which are reported in Table 1. For determining the aridity categories of the sites, the De Martonne-Gottmann index (b) was calculated at each site (De Martonne, 1942) using the following categories: 5 ≤ b ≤ 14: arid, 15 ≤ b ≤ 19: semi-arid, 20 ≤ b ≤ 29: sub-humid and 30 ≤ b ≤ 59: humid. To describe the seasonality of the precipitation, the rainfall seasonality index (SI P ) were calculated, where -0.13 ≤ SI P ≤ 0.13 category describes uniform distribution, more negative values show wetter winters than summers, while more positive values mean wetter summers (Walsh and Lawler, 1981). To complete the study with further meteorological analyses, the length of dry spells (number of consecutive dry days) were summarized and ordered. Rainy days got zero values and were excluded from the dry spell length calculations, and the number of the dry days were added until the next precipitation event. This approach helped to demonstrate the severity of water deficit periods. These indexes and calculations were used to described and distinguish the sites' climatic circumstances and differences.
Measured WDPT and WRCT values of the measuring quadrat were analysed by using percent exceedance probability distributions, approximated as [m/(n + 1)] × 100, where m is a rank order of each estimate, with m = 1 as the largest and m = n for the lowest, with n being the number of available data (number of available data is shown in Table 4). These analyses provide the opportunity to establish the severity of SWR at each site.
For establishing further relationships between the measured soil chemical (CaCO 3 , organic carbon contents and pH) and hydrological properties (such as soil water content, WDPT, hydraulic conductivity, sorptivity and repellency index) at the sites we applied Pearson's correlation and scatterplot correlation matrix for characterizing the differences and similarities. Different combinations of variables for different locations were selected, because not all sites had the same number of measured hydrological properties (see in Table 4, last column), because, for instance, if a site was not water repellent there was no available RI or S e data. PW site provided soil texture measurements from each soil hydrology measuring cell, but these measurements were not available for every point of the quadrat of other sites.
R Studio (R Studio Team, 2016) was used for statistical computing and graphical visualization.

Results and discussion
The SWC (m 3 m − 3 ), of the upper 5 cm of soil profile had the highest values at the PW site, while other sites' water contents were below 5% (Table 4). At the PW site WDPT measurements were carried out on dried samples resulting in WDPT P = 0.7 ± 0.2 s, which confirmed the PW site to be non-repellent. There was high variability between and within sites and land covers in WDPT, field hydraulic conductivity(K)(-2 cm), both water, S w (-2 cm) and ethanol sorptivity, S e (-2 cm), and repellency index prior to the infiltration experiment. For instance, the highest variability of WDPT was recorded at the CSP site, with the greatest average values at the CSP17 site. However, the maximum WDPT value was observed for the CSP44 site. The cumulative infiltration of water showed a hockeystick-like relationship over time (Fig. 3) at the CSP and CI sites, where the water repellent (S wh ) and wettable (S ww ) state of soil were demonstrated by different infiltration rates. This relationship was not observed at the SE and PW sites, although the soil physical properties were similar: SE vs CSP and PW vs CI (Table 2), but there are some differences in their pH values, soil organic matter, bulk density and CaCO 3 contents and vegetation abundance. Therefore, the differences were more likely caused by the interactions between climate, soil chemical properties and vegetation cover.

Effect of climate and seasonal field conditions on SWR
Analyses of the frequency of dry days (Fig. 4) during the summer half year (between 1st of April and 31st of October) showed that the PW site had the shortest dry spell length, 2.9 d across years, with the longest drought period 31 d during the period 1980 to 2017. The SE site's average dry spell length was 3.6 d, and the longest dry spell was 37 d. The CSP site experienced more arid conditions, with a 4.7 d average length of day period across investigated years. The longest dry period here lengthened to 50 d, whilst it was 99 d at CI. The Mediterranean site's average annual dry day length was 4.9 d, while it was 8.2 during the summer term. According to these analyses, the chance of precipitation is 2.5 times greater at PW than at other sites, which means around 62% of the year is rainy, compared to 32-36% at other sites. This estimate corresponds to the annual climatic conditions (see in Table 1, and Fig. 2). The CSP and CI sites showed the highest chance of dry periods; their De Martonne-Gottman aridity (b, De Martonne, 1942) categories are semi-arid and arid (b = 14.4 ± 6.5 and 10.6 ± 2.9) and also the precipitation is relatively low (~310 ± 134 and ~160 ± 64 mm, respectively) during the growing season, which is associated with high average maximal summer temperatures (~36 ± 2 and ~37 ± 3 • C, respectively). Owing to the high temperature and rainless periods at the CSP, SE and CI sites, the soil moisture content was low, below 0.05 m 3 m − 3 on average (see SWC in Table 4). This low, near zero, SWC value could considerably decrease and in some cases halt the rate of evaporation, as a dry surface layer functions as an impeding layer. Thus, the dry surface layer is able to significantly diminish the net evaporative moisture loss; according to Rye and Smettem (2017) the near surface layer's evaporative losses decreased by 70-80 % compare to a wettable control soil. Lichner et al. (2020) found that a 2 cm thick layer of duff or water repellent sand saved up to 58-59% of water from evaporation in a clay loam soil and sand soil, respectively. The length of the rainless period has been reported as being positively correlated with the severity of SWR (de Jonge et al., 1999;Doerr et al., 2000), which corresponds to our findings; the maximum length of the dry period was highest at the CSP and CI sites. Considering the wetting-drying history three months before the measurements, CI site had the lowest amount of precipitation (~17 mm), after SE and CSP sites (~161 and ~162 mm, respectively) and PW (~387 mm). To explore the chances of finding strong or extreme water repellent points at the sites in more detail, cumulative distribution functions (Fig. 5) were calculated, which demonstrate the exceedance probability of WRCT and WDPT values. In addition, the importance of vegetation cover was investigated for the CSP and SE sites.

Effect of vegetation on SWR
In terms of WRCT, which is estimated from the intersection of the linearized hydrophobic and hydrophilic sections of the water infiltration curve, the CSP44 site showed the greatest WRCT (Table 4, Fig. 5), followed by CSP17 site. This suggests that strongly synanthropized vegetation or poor grassland dominated by perennial weedy grass and annual grasses has enhanced the SWR at the CSP site. The SE and PW sites did not exhibit hydrophobic stage of the infiltration, therefore no S wh values were estimated. Thus, the chance of finding an extremely strong water repellent measuring point (WDPT > 3600 s), was ~20% at the CSP44 site, ~26% at SeS, and above 40% at the CSP17 and SeN sites, whereas it was ~0% at PW and CI. The natural vegetation at the SE site showed much higher SWR values than under synanthropized vegetation, which contradicts what was found at the CSP site. This paradox can be resolved by attributing more influence to climate, and the length of dry spells in particular, in developing SWR. However, the greatest WDPT values were measured at CSP44 (13630 s, Table 4), which has more natural grassland species similar to SeN. This also suggests that the effect of the climate factor on SWR could be even greater than the effect of the vegetation cover, which corresponds to the fact that the water repellency is not a static soil property and has a strong seasonal variance (Dekker et al., 2009). For instance, the entire topsoil can dry out and become water repellent during prolonged hot and dry periods, which induces incomplete wetting during subsequent rainfall events. According to Täumer et al. (2005) the total effective cross section area is being reduced to 20% to 40% therefore the potential rate of infiltration is much smaller than would result from the soil texture. A previous study (Sándor et al., 2015) at CSP17 site, when the site was 12 years abandoned, showed that SWR was not a function of the organic matter type, but rather caused by the large quantity of non-degraded organic compounds of the local vegetation. In this case, the SWR could be attributed to the drought related withering and degradation of vegetation, rather than the actual constituents of "fresh" vegetation.

Correlations between soil hydrological properties
The correlation scatterplot analyses highlight the connections   6. Scatterplot Pearson-correlation matrix of hydrophysical properties at the CSP17 (Csólyospálos, Hungary) site, where K: field unsaturated hydraulic conductivity; S w : field water sorptivity; S wh : water sorptivity of hydrophobic soil state; S ww : water sorptivity of nearly wettable soil state; WRCT: water repellency cessation time; WDPT: water drop penetration time, SWC: soil water content, RI and RI m : repellency indices.
2 Fig. 7. Scatterplot Pearson-correlation matrix of hydrophysical properties at CSP44 (Csólyospálos, Hungary) site, where K: field unsaturated hydraulic conductivity; S w : field water sorptivity; S wh : water sorptivity of hydrophobic soil state; S ww : water sorptivity of nearly wettable soil state; WRCT: water repellency cessation time; WDPT: water drop penetration time, SWC: soil water content, RI and RI m : repellency indices.
between the different measured values and variables (Figs. 6-11), where soil hydraulic conductivity, K, is positively correlated with all the sorptivity values (S w , S wh , S ww ) for all the sites. Values for K have a negative correlation with WRCT at all sites, and with RI and WDPT at most of the sites. Also, a slight negative correlation was found between K and moisture content. Although K was relatively more variable at SE and PW sites with coefficients of variation (CV) in the range 1.05 to 1.39, the spatial heterogeneity of K was within the values reported in literature, where, for example, K reached three orders of magnitude difference within the pedon scale at CSP site (Sándor et al., 2015). The repellency index mostly demonstrates a negative correlation with sorptivity (S w , S wh and S ww ) values. In general, WRCT and WDPT positively and significantly correlates with S w and S wh (r is between 0.28 and 0.81), but a negative correlation was found with the water sorptivity of nearly wettable soil state, S ww , at the CI site (r = -0.50 and -0.37, respectively) ( Fig. 11). At the PW site, the soil texture was also compared with the infiltration values (Fig. 8), where only the sand content showed a high positive correlation (r = 0.56) with hydraulic conductivity, whilst the silt and clay content negatively correlated (r = -0.49 and r = -0.002, respectively) with it. Clay content was negatively correlated with SWC values (Fig. 8). Regardless, our moisture content measurements reveal generally negative correlations with SWR related values (e.g. Fig. 9), although SWC alone is not a reliable predictor for the appearance and disappearance of SWR (e.g. Müller et al., 2014) owing to its limitation and high seasonal variability and spatial heterogeneity.
Overall, there is a general existence of a correlation between hydrophysical properties (K and sorptivity in the early stage of infiltration, S w and S wh ) and repellency indices (RI, RI c and RI m ). However correlations are not always significant, but the sign is in most cases negative (always for S). This is meaningful as one should expect that infiltration is prevented in SWR soils. As our regional scale investigations demonstrate, to establish the SWR risk of an area, meteorological factors, soil properties and land cover type also play key roles alongside the SWC.

Conclusions
Separate and combined effects of two main factors (climate and soil texture) on SWR were investigated at four sites from different climatic and soil regions across Europe. Based on our results, sites with finer texture soil under humid climate are unlikely to be water repellent. The PW site shows no signs of SWR thus no factor known to cause repellency is present or strong enough to induce repellency. Regions with coarser texture soils (CSP and SE sites) with high or, conversely, low organic matter content and regions with drier climatic zones (CI site) characterized by longer dry periods have a greater likelihood of containing water repellent soils. Hydraulic conductivity showed a positive correlation with all the sorptivities (S w , S wh , S ww ), but was generally negatively correlated with the repellency index, WDPT and WRCT. The fact that climatic or soil textural differences could indicate differences in SWR despite the variations in either the pH or the soil organic matter content demonstrates that climate and soil texture are two of the main factors influencing SWR. This conclusion was supported by the finding that climatic factors can override or supress the impact of vegetation on SWR as it was observed at PW site where SWR was not induced under grass cover even during or after long dry spells.
Our findings support identifying localized non-repellent areas (such as PW) and distinguishing them from potentially water repellent places, providing opportunities to develop site-specific solutions. If a soil has coarse texture (SE site) or a high frequency of water stress periods during the vegetation period (CI site) it could induce slight SWR, whilst the combination of these factors could cause strong and some places extreme water repellent circumstances (CSP site). Within the current study the SWR characteristics of the natural, perennial grass dominated surface cover provided more favourable conditions under strong and extreme water repellent conditions at CSP. Therefore, there is an increasing need to take SWR into account to a greater extent in grassland ecosystems. There are recent initiatives to simulate the soil-water dynamics of water repellent soils (e.g. Brown et. al., 2018), where the phenomena of preferential flow is also included. However, the SWR is more complex, as it has effect on surface runoff, nutrient loss, soil    11. Scatterplot Pearson-correlation matrix of hydrophysical properties at CI (Ciavolo, Italy) site, where K: field unsaturated hydraulic conductivity; S w : field water sorptivity; S wh : water sorptivity of hydrophobic soil state; S ww : water sorptivity of nearly wettable soil state; WRCT: water repellency cessation time; WDPT: water drop penetration time, SWC: soil water content, RI and RI m : repellency indices. nitrogen and carbon pools, and productivity, as well as biodiversity. Future study will focus on modelling SWR phenomenon what could help us to determine potentially water repellent areas and define the most sustainable grassland community of each risk zone in order to optimize their water management and increase their productivity and carbon sink potential.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.