Impacts of Wildre Frequency On Plant Recovery, Soil Properties And Water Storage In Pine Woodlands of North-Central Portugal

Purpose Increasing wildre frequency in Mediterranean-basin together with drought periods expansion could affect plant-soil-water dynamics processes. The goal is to assess the effects of wildre frequency on plant recovery, soil properties, soil moisture content (SMC; %) and effective soil water content (ESWC; %) during the rst hydrological year after a 2012 moderate-severity-wildre. Methods This study was conducted in pine woodlands of North-central Portugal affected by 1-, 4-wildres and unburnt (1975-2012). Soil samples were gathered from plant/bare microsites at top-mid-bottom hillslope positions to determine bulk density, soil texture, soil moisture, soil organic matter content-SOM, pF-curves (available water content-AWC, eld capacity-FC, permanent wilting point-PWP) (n=54) during four dry/wet periods. Soil cover, plant recovery and soil water repellency were measured. On the burnt areas 72 sensors daily/seasonal monitored SMC and ESWC at two depths (2.5/7.5 cm) and two microsites (plant/bare). The 1 re hillslopes showed higher plant recovery than the 4 res hillslopes. SOM was higher in the burnt soils (17-20%) than in the unburnt ones (12-14%). Wildre frequency: i) increased the water stress for plants and led to both maximum and minimum values of SMC/ESWC, respectively, for the wet-/dry-seasons; ii) reduced the capacity of the soils to retain water (decreased FC/AWC, increased PWP), being more accentuated in bare microsites. The increasing wildre frequency and the predicted expansion of drought periods promotes lower water availability for plants in the more frequent bare soil patches. The water-stress window of the dry season happened sooner and extended for longer as increasing wildre frequency.


Introduction
Landscape res burn 3-5 million km 2 of the Earth's surface annually (Jones et al. 2019). In re-prone ecosystems as Portugal continental, wild re frequency is increasing and may challenge the resilience of these ecosystems (Vallejo and Alloza 2015). Warmer and drier scenarios may transform the high-biomass forestry systems into low-biomass shrublands (Acacio et al. 2009; Pausas and Bond 2019). However, although wild re frequency has been highlighted as an important factor in the understanding of the effect of wild res on vegetation and soil properties (Mayor et al. 2016), studies evaluating the cumulative effects of wild re frequency on soil water balance are scarce (Shakesby 2011) and limited by the lack of long-term accurate records of the different wild re events.
During the last decade, the average wild re frequency in the Mediterranean basin has increased due to climatic (drought) and anthropogenic factors (land use changes) ( The mechanism of SMC redistribution during dry and wet periods depends on the different climatic regions. Grayson et al. (2006) de ned the non-local (wet) and local (dry) states of SMC on temperate regions: "Non-local control occurs under wet conditions, with catchment terrain leading to organisation of wet areas along drainage lines. Local control occurs under dry conditions, where soil properties and local terrain (areas of high local convergence) in uenced spatial patterns". The switching mechanisms are highly sensitive, and will probably be affected by the frequency of wild re occurrence. Thus, during the dry periods the soil water balance dynamics depends in a set of soil properties such as texture (Grayson et al. 2006), soil organic matter content (SOM), bulk density, porosity and soil structure (Famiglietti et al. 1998), soil depth (Cardenas and Kanarek 2014; Nolan et al. 2014) and the dynamic state of the soil cover components (Fitzjohn et al. 1998;Cammeraat 2004), while during the wet periods it is dominated by lateral water movement through both surface and subsurface paths, with catchment terrain leading to organisation of wet areas along drainage lines (Grayson et al. 2006).
Topsoil SMC controls the hydrological processes, especially through different states of hydrophilic-hydrophobic topsoil layers (González-Pelayo et al. 2015; Malvar et al. 2016) related to biotic and abiotic soil cover components (Ruiz-Sinoga et al. 2010a). In humid Mediterranean environments, biotic cover factors determined the soil hydrological network (Lavee et al. 1998). The increase in wild re frequency drastically reduce the biotic cover, such as vegetation and litter (Gimeno-García et al. 2007). Additionally, it depletes the topsoil organic layers (Eugenio et al. 2006b) and enhance the remaining abiotic components (bare soil, stones), which act as the key factors controlling SMC patterns (Ruiz-Sinoga et al. 2010a; Gabarrón-Galeote et al. 2013). Furthermore, the increase in wild re frequency will increase soil nutrient losses and change the pattern of bare-vegetation microsites (Mayor et al. 2016;Santana et al. 2016) which will probably affect the SMC patterns (Sarah 2002). However, to the best of our knowledge, no studies have assessed the extent to which an increase in wild re frequency will impact the soil water balance.
The main goal of this research is to assess the effects of wild re frequency on plant recovery, soil properties, soil moisture content (SMC; %) and effective soil water content (ESWC; %) available for plants during the rst post-re year in re-prone areas of North-Central Portugal with three different wild re frequencies: unburnt mature pine forest (no wild res in the period 1975 to 2012), burnt only once (affected by the 2012 wild re) and burnt 4 times (affected by 1978,1985,2005 and 2012 wild res). The speci c objectives are: 1. To assess the effects of three wild re frequencies (unburnt, 1 re vs 4 res) on vegetation recovery (plant height, stem diameter and stem length), soil cover and soil properties (texture, bulk density, pH, soil organic matter, pF-curves, soil water repellence, SMC and ESWC) at the end of the rst year after the 2012 wild re, 2. To assess the effects of two wild re frequencies (1 re vs. 4 res) on the daily readings of SMC and on the effective soil water content (ESWC) for plants during the rst year after the 2012 wild re, 3. To compare the effects of wild re frequency (1 re vs. 4 res), soil depth (2.5 and 7.5 cm) and microenvironment (soil below plant vs. bare soil) on SMC and ESWC.

Study area and sites
This study was conducted near Várzea (40°46'059"N, 7°51'726"W), North-central Portugal, after a wild re on 9th September 2012 that burnt through a Pinus pinaster Ait. forest area. Climate is humid Mediterranean, classi ed as Csb (FAO 2006), with a dry period from June to August and peak rainfalls between October and March. Mean annual rainfall and temperature are 1200 mm and 14°C, respectively (SNIRH 2015). The earliest available wild re records of this re-prone area are from 1975 to present, and no prescribed res have been recorded from this time interval (ICNF 2013). Within this area, we selected three hillslopes for each of the following wild re frequencies ( . The average wild re frequencies corresponded, respectively, to > 39 years, 37 years and 9 years. Soil burn severity for the 2012 wild re was estimated as moderate at the two burnt areas (using soil visual indicators described in Vega et al. 2013) with all the litter being transformed to a uniform black layer of charcoal, although in the hillslopes burnt 4 times the ash colour was grey in some spots ( Table 1). As a consequence of the wild re frequency, pre-re vegetation cover was different between sites. The unburnt and the areas burnt 1 time (1 re) were covered by mature Pinus pinaster woodland, with Pterospartum tridentatum (L.) Willk., Calluna vulgaris (L.) Hull and Agrostis spp., as the dominant understory vegetation. The areas burnt 4 times (4 res) were shrubland, characterized by sparse 7 years old young pines regenerated from the previous wild re in 2005, and dominated by Pterospartum tridentatum shrubs and the co-occurring species Agrostis spp., Calluna vulgaris, Cistus spp., and Halimium spp. (Fig. 1). Table 1 General site characteristics, soil type and properties in the Ah topsoil layer (0-5 cm depth) of each site (unburnt (0x), 1 or 4 times burnt areas). Mean (and standard deviation) for the granulometric composition (%, n = 54), bulk density (g cm − 3 , n = 54), soil organic matter content (SOM in %, n = 54), pH (n = 54), pF-curves (pF1, pF2, pF2.5, pF3.5, and pF4.2, in %, n = 54), and available water content (AWC in %, n = 54) on each area followed by a different letter were statistically different at p < 0. All selected sites had similar elevation (450-550 m.a.s.l.), slope aspect (200-210° azimuths) and steepness (7-8° on top to 14-17° on bottom slope positions), and the soils were developed from pre-Ordovician schist of the Hesperic Massif. According to IUSS (2015), the soils were classi ed at the unburnt areas as an association of Humic cambisol and Epileptic Umbrisol (dystric). In the 1 re areas corresponded to Umbric leptosol and in the 4 res areas to Epileptic Umbrisol. These soils were all acidic sandy loam in the unburnt areas and loam in the 1 and 4 res areas. Soils were somewhat deeper in the unburnt and the 4 res hillslopes (30-40 cm) as compared to 1 re hillslopes (less than 25 cm). Topsoil Ah layers showed similar bulk density and pH, although the unburnt soils showed a somewhat coarser texture and lower SOM than the two burnt soils (Table 1).

Rainfall
The rainfall data were recorded using 3 rainfall automatic gauges (ECRN-100 from Decagon devices) as well as 3 storage gauges in each of the three areas. In order to characterize rain volume and intensity for the whole area, median values were used.
Field measurements During early November 2012, ve resprouting Pterospartum tridentatum shrubs (target specie) and paired bare spots were randomly selected on each slope. In April 2013, three of these ve shrubs (excluding the biggest and the smallest shrubs), together with their paired bare spots (total of 54 spots), were selected as experimental microsites to monitor the direct effects of wild re (plant recovery and soil cover) as well as the indirect effects of wild re (soil physical and hydrological properties). On each shrub, the maximum plant height and the length and diameter of ve stems were measured in May, July and September 2013. On each of the 54 spots, soil samples were taken on May 2013.
Topsoil cover was measured on six 0.5 x 0.5 m plots per hillslope (three on shrub and three on bare microsites (n = 54 plots). Ground cover was assessed twice; three months and one year after the 2012 wild re. Five cover categories were considered and classi ed as biotic (litter and vegetation) or abiotic (ash, bare soil and stones) cover components (Ruiz-Sinoga and Romero-Díaz 2010). Ground cover was quanti ed by laying a square grid of 0.5 m × 0.5 m at a xed position over the plot and recording the cover category at 100 grid intersection points as a percentage (Prats et al. 2019).

Soil sampling
Main soil properties were characterized at each of the nine hillslopes by sampling the topsoil with 5 cm-height steel cores in April 2013. At each hillslope, topsoil samples were taken under the three shrubs and at three bare soil microsites (n = 54), after removing the above ash and/or litter layer. The samples were gathered with 100 cm 3 steel cores for soil moisture content (SMC; % volume) after drying at 105ºC 24 h and soil organic matter (SOM; % volume) determination after sieving at < 2mm and incineration by loss-on-ignition at 550°C 4 h (ISO 11465, 1993). Another set of 54 samples was gathered for pH (ISO 10390, 2005), granulometric determination (Guitian & Carballas 1976) and bulk density analyses with 250 cm 3 steel cores (Malvar et al. 2016). A third set of 54 samples (100 cm 3 steel cores) was used for the assessment of soil water retention curves using a sandbox apparatus (Van der Harst and Stakman 1961) and a pressure plate device (Stolte 1997). The amount of water retained in the soil was calculated at ve suction pressures: -1, -10, -33, -124, and − 1550kPa; corresponding respectively to: pF1 (saturation water), pF2 ( eld capacity, FC), pF2.5, pF3.2, and pF4.2 (permanent wilting point, PWP). Available water content (AWC; %) was calculated as FC -PWP, once water retained at PWP is not accessible for plant uptake (Chen et  Soil moisture probes A total of 72 soil moisture sensors (Decagon Inc.) were installed in November 2012 in order to carry out a detailed assessment of the daily variations in SMC between the 1 time and 4 times burnt hillslopes. These burnt hillslopes were both affected by the 2012 wild re and only differed in the wild re frequency. Therefore, the sensors were installed at two soil depths (at 2.5 and 7.5 cm depth) under two microsites (shrub and bare) in each of the three shrubs selected in each of the six burnt hillslopes, totalizing 72 sensors (2x2x3x6). The EC5 and GS3 dielectric capacitance soil moisture sensors were connected to Em5b dataloggers, and determine volumetric soil moisture content (SMC; v/v) by measuring the dielectric constant of the media using capacitance/frequency domain differences. A capacitance sensor uses the soil as a capacitor element and use the soil charge storing capacity to calibrate to water content. They have a measuring range of 0.0-1.0 (m 3 m − 3 ) and an accuracy of ± 0.03 m 3 m − 3 typical in mineral soil solutions that have an EC < 8 dS/m. Besides volumetric water content, GS3 sensors can also measure bulk electrical conductivity in a range of 0-25 dS/m. To avoid in uence of the hillslope aspect, all probes were oriented to 200-210° azimuths. SMC was measured at 5-min time intervals, which were averaged to obtain the daily SMC (n = 350, from 21 Nov. 2012 to 6 Nov. 2013). All the EC5 and GS3 sensors were calibrated before installation and the offset differences between sensors were minimized by making use of individual 4-point calibration in uids with known apparent dielectric permittivity (following user´s guide available at www.metergroup.com). In this way, the mutual differences between sensors, which is especially important in the dry range, could be minimized and a high relative measuring accuracy could be obtained (van den Elsen et al. 2014).

Data analyses
The daily average SMC (%) and results obtained from pF-curves (FC, PWP) were used to calculate the effective soil water content (ESWC; %). The ESWC approach is intended to assess the water volume that is effectively available for plant uptake. It was assumed that the availability of soil water to plants is at best at FC, and it declines with decreasing SMC (Chen et al. 2007). At PWP, it is generally accepted that the soil water is no longer available for plants. The effective soil water content (% of the total available water) was calculated using the following formula (Porporato et al. 2002): Where SMC actual is the average daily SMC measured with the soil moisture sensors (% volume). FC ( eld capacity) and PWP (permanent wilting point) were calculated from the pF-curves assessed from the soil samples adjacent to each sensor location (Van Genuchten et al.

1991).
Linear mixed-effects statistical models (Littell et al. 2006) were used to assess the differences in mean plant height, stem diameter, stem length, SMC and SOM with wild re frequency (unburnt, 1 re, 4 res) as the xed factor, and the plant/ topsoil sample as the random factor. Similarly, linear mixed-models were constructed to assess the differences in the daily SMC and ESWC among four xed factors: wild re frequency (1, 4 res), soil depth (2.5, 7.5 cm), microsite (plant and bare), and season (autumn 2012, winter 2012-2013, spring 2013, summer 2013, autumn 2013). The individual soil moisture sensors were included as the random factor. The covariance structure of the repeated measures was modelled using a compound symmetry function or autoregressive heterogeneous variance, as it gave the lowest − 2 Res Log likelihood model-tting values (Littell et al. 2006). In order to assure a normal distribution of the model residuals, plant variables, SOM and SMC were fourth-root transformed, while daily SMC and ESWC were log-transformed. Explanatory continuous variables were tested as covariates in a forward selection procedure, including daily rainfall, maximum 30-min rainfall intensity (I 30 ), and accumulated rainfall amount from 1 (Ant_rain_1) and until 14 days (Ant_rain_14) before the daily SMC measurement. As the multiple rainfall characteristics are related, they were tested independently and only the variable with highest F-value was included in the model, following the principle of forward selection.
Linear regressions and coe cient of determinations (R 2 ) between the explanatory variables and the daily SMC and ESWC were calculated.
Differences in ground cover, bulk density, pH, pF-values and on SWR, were tested using mixed-effects models with or without repeated measures similar to the models described earlier, except no covariates were used. If the assumptions for equal variance and normality were not met, as it was the case of SWR, relative frequencies of each class were calculated, and the non-parametric Wilcoxon test was used to assess the differences between wild re frequency and soil depths.
Comparisons among the xed effects, as well as differences between the levels of the factors were tested by least-squares means and adjusted by the Tukey-Kramer method (Kramer 1956). All statistical data analyses were carried out using the SAS 9.4 software package (SAS Institute, Inc. 2008), and all statistical tests used α = 0.05.

Rainfall
From the analysed 350 days, 138 were rainy days, with a total rainfall of 1473 mm. Maximum rainfall intensities in 30 minutes were measured in March and September 2013 (14 and 29 mm h − 1 ). Peak daily rainfall happened also in autumn 2013 (Fig. 2). Out of the summer drought (July, August and September 2013), the average and the median number of days between rain events were 4 and 3, respectively, while maximum days without rain amounted to 26.
Plant recovery, soil cover and topsoil properties The plant response at the two burnt areas was very similar, although at the end of the last dry period of September 2013, before the rainfall season, the 4 res hillslopes showed a slightly lower plant height, stem diameter and stem length, as compared to the 1 re hillslopes, although without signi cant differences. Plant height (Pterospartum tridentatum) was very similar in the unburnt hillslopes through time, and signi cantly higher than the 1 and 4 res hillslopes ( Table 2).  Ground cover differed between unburnt and burnt areas (Fig. 3). Immediately after re, abiotic soil surface components such as ashes, bare soil and stones amounted to 81% at the 1 re and 99% at the 4 res hillslopes. One year later, the unburnt hillslopes remained completely covered by (biotic) litter and vegetation, while in the 1 re and 4 res hillslopes amounted to only 49% and 16% of the biotic soil cover, respectively (Fig. 3).
Soil water repellence (SWR) did not display statistically signi cant differences between 1 and 4 res area in the overall data pool (Fig. 4). However, in the unburnt area, moderately/strongly SWR levels were more frequent as compared to the burnt sites. Overall, annual SWR frequency at 2-3 cm depth was non-repellent (MED class 0) in 44 and 43% of the measurements on the 1 and 4 re areas respectively, and only 17% on the unburnt areas. The 4 res areas showed very strong and extremely repellent categories in 21% of the measurements, while the 1 re and the unburnt areas were close to 40%.
The pF-curves revealed no differences in the gravitational soil water content (SWC) (< pF2 and eld capacity (FC, pF2), due to wild re frequency (Table 1). However, the soil water retained at higher pressures (pF3.2 and pF4.2 or PWP) increased with wild re frequency. Consequently, AWC (AWC = pF2-pF4.2) was lower on the 4 res hillslopes (Fig. 5) as compared to 1 re site and unburnt site, although the differences were not signi cant.
The unburnt, 1 and 4 res areas re ected differences in SOM, SMC and ESWC (Table 2). SOM did not vary between the ve soil sampling campaigns, but the unburnt hillslopes showed signi cantly lower SOM contents than the burnt hillslopes. SMC decreased consistently during the dry periods, and markedly so in the 4 res hillslopes in July 2013, which were statistically dryer than the 1 re and the unburnt hillslope ( Table 2). On the contrary, during the wet sampling of November 2013, the 4 res hillslopes showed a signi cantly higher SMC (17% vs. 13%). ESWC varied in the wake of SMC. ESWC for the 4 res hillslopes was signi cantly lower during July 2013 (16%), and signi cantly higher during November 2013 (43%), as compared to the 1 re hillslopes.

Factors affecting soil moisture sensors readings
The overall mean SMC values for the soil moisture sensors at the 1 and 4 res hillslopes during the rst post-re year were not statistically different (Table 3), and reached, respectively, 12.7% and 13.0% (S1). There was not a signi cant microsite effect (the overall mean SMC value for plant and bare microsites was the same, see S1) but there were strong depth and season effects, and both were statistically signi cant ( Table 3). The presence of signi cant interactions indicated that the effect of wild re frequency was not straightforward for all the levels of the factors. For example, wild re frequency and microsite showed complex relations: the plant microsite was wetter than the bare soil microsite on the 1 re hillslopes, but the opposite was found on the 4 res hillslopes (Table 4). These patterns were consistent throughout all the seasons, but any of the within-season differences were signi cant. On the other hand, the ESWC values showed a tendency for keeping more effective water into the 1 re area, under plant microsites and deeper soil layer (7.5 cm depth, S1), although differences were not signi cant (Table 4). There was also a signi cant effect of wild re frequency on ESWC, with the 1 re hillslopes (9.0%) exhibiting signi cantly higher values than the 4 re areas (7.3%; S1). Plant microsites registered higher overall ESWC values than bare soil microsites (9.9% vs. 6.4%), but only for the 1 re hillslopes, and the differences were not signi cant. The effect of soil depth was signi cant and the surface soil layers exhibited much lower ESWC than the deeper soil layers (-4.1% vs. 20.4%; S1).
The mean seasonal SMC were signi cantly lower for summer 2013, due to the drought period, and highest for winter 2013 (Table 4). ESWC was signi cantly lower for the same dry season, when the effective soil water was below 0%; or in other words, below PWP and thus, plants were under water stress. The daily variations of SMC and ESWC showed that the hillslopes burned 1 time were more stable than the hillslopes burned 4 times (Fig. 6). Threshold ESWC of 0% (no water available for plants) was reached 17 days sooner in the 4 res hillslopes than on the 1 re hillslopes, and also 10 days sooner on the bare than plant microsites (respectively 15 and 3 days, for the 1 re and 4 res hillslopes). Although the differences were small, both SMC and ESWC were signi cantly higher during some of the driest days of summer 2013, and signi cantly lower during some of the wettest days of winter 2013 on the 1 re hillslopes as compared to the 4 res hillslopes. The plant microsites were slightly wetter than the bare soil microsites, although the daily differences were marginal (Fig. 6). The seasonal variations were highly correlated with rainfall amount. However, the most important covariate for both SMC and ESWC was the cumulated 7-days rainfall amount (Fig. 7). The worst coe cient of correlations happened during the dry seasons of autumn 2012 and summer 2013 (r2 = 0. 53 and 0.32), when the soils were much dryer and hydrophobic.

Discussion
Effects of wild re frequency on plant recovery and soil properties The increase in wild re frequency had important effects on vegetation composition and structure. Vegetation physiognomy changed from pine woodland on the unburnt and 1 re hillslopes to open shrubland on the 4 res hillslopes. The occurrence of repeated wild res killed the pine trees before they reach the reproductive age, which left a footprint that constrains vegetation recovery and compromise forest productivity (Kowaljow et al. 2018). Plant recovery (Pterospartum tridentatum) was slightly faster on the 1 re than the 4 res hillslopes, which can be attributed to the higher litter and vegetation cover on the 1 re than the 4 res hillslopes. These cover categories protect the topsoil water from high soil temperatures and evaporation, while ashes and bare soil increase the albedo and water evaporation (Sankey et al. 2012). However, exceptions were found between the higher water demand of plants during water stress periods as compared to bare agriculture lands (Bellot et al. 1999) or the lower SMC measured on the vegetated areas as compared with burnt gorse shrubland (Soto and Díaz-Fierros 1997).
The higher plant recovery on the 1 re hillslopes can be also a consequence of the higher soil water availability for plants, which in turn is a function of soil texture, bulk density and SOM (Vereecken et al. 1989; Ankenbauer and Loheide 2016). Wild res can have an effect on these parameters and decrease the clay fraction, via soil erosion, increase the bulk density, via loss of soil structure and loss of SOM (Boix-Fayos 1997; Certini 2005). However, none of these changes were observed in our study (Table 1). We observed that wild re frequency increased the silt fraction and the SOM. The ning of the particle size distribution after burning can possibly be explained by . Furthermore, the increase in SOM in our burnt hillslopes did not lead to higher water retention ( Table 2). SWR appeared to decrease at increasing wild re frequency (Fig. 4), and was higher in the unburnt than the 1 or the 4 res hillslopes. These patterns can be partially attributed to a destruction of repellent substances by re ( We reduced at maximum the sources of variation on our hillslopes (same aspect, steepness, hillslope length, parent material) and maximized the amount of measurements (plant and bare microsites in each of 3 positions along each of the 3 hillslopes per wild re frequency treatment sampled 4 times) and the differences were small. The 1 re hillslopes showed the highest soil water retention (more gravitational, eld capacity and wilting point water), followed by the unburnt and by the 4 res hillslopes (Fig. 5; n = 54 soil samples). We hypothesize that the increased silt fraction can explain the higher water contents at high pF tensions observed on both the 1 and 4 res soils as compared to the unburnt soils. On the one hand, the higher gravitational water on the 1 re soils can be attributed to the incorporation of ash to the soil, which enhanced the water retention at low pF tensions more effectively than the soil itself (Ebel 2012). On the other hand, the 4 re soils had comparatively less fuel loads and lower re temperatures, and fuels were likely transformed into charcoal, due to incomplete combustion. Consequently, the char incorporation to the soil can explain the lower gravitational water on the 4 res soils, as charcoal retained less water as compared to the soil itself ( Microsites exerted an effect on SMC and ESWC, but this effect differed between areas. Plant retained more soil water than bare microsites in the 1 re hillslopes, but the opposite was true for the 4 res hillslopes (Table 4). In his review, Sankey et al. (2012) found that plant microsites accumulated more SOM, nutrients and soil water than bare microsites, the effect being stronger on the wild re-affected steppes than in the unburnt areas, as it happened in our 1 re hillslopes. The existence of soil moisture radial gradients between plant to bare microsites have been well described for Mediterranean arid (Sarah 2002 The SMC and ESWC results of the seasonal soil sampling campaigns tted in well with the soil moisture sensor data series. The wet season soil sampling campaigns (Table 2) showed that increasing wild re frequency led to a loss of the soil buffer capacity to store extreme SWC and to accommodate greater quantities of water during autumn. On the other hand, the dry soil sampling of July showed that the more frequentlyburnt soils had di culties to retain soil water during the drying summer. Sankey et al. (2012) described that the increase in re frequency implied that plants afforded less time to recover between res, and soils were subjected to decreases in their water accommodation capacity. The same conclusions arose from the soil moisture sensor data series, both at seasonal (Table 4) and daily (Fig. 6) periods: an increase in wild re frequency led to maxima SWC during winter periods, and minima SWC during dry periods, with longer periods of soil water stress for plant growing. During the rst post-re year our study shows that increasing wild re frequency increases the water stress period of summer in 17 days. The microsite effect was also important, having the plant microsites 10 days less of water stress than the bare microsites. Sankey et al. (2012) hypothesized that this contrasting effect will depend on plant recovery to pre-re levels, which depends on both wild re frequency, and plant physiology to recover after disturbances. Plant water uptake play a major role in the complex post-re seasonal behaviour of SMC during wet/dry seasons, but the few studies assessing this reached different results.  (Bochet 2015). Additionally, the more frequent bare space between shrubs, have a lower capacity to capture and hold resources (water, sediments, nutrients, seeds, etc.), thereby being unable to function as sink areas for a longer time after re (Mayor et al. 2016). Finally, the temporal patterns of the hillslopes burnt at high wild re frequency are these of longer water stress periods during the dry seasons. Vegetation recovery can take decades or hundreds of years to reach pre-re levels (Cammeraat and Imeson 1999). Post-re managers should follow strategies to preserve and restore pre-re vegetation structures, due to their important role in water conservation and redistribution and prevent catastrophic tipping points towards deserti cation (Lavee et al. 1998;Ludwig et al. 2005) even in humid Mediterranean forestlands.

Conclusions
The conclusions about the effect of wild re frequency on plant recovery, soil properties and soil water dynamics throughout the rst year after the 2012 wild re in unburnt hillslopes, hillslopes affected by 1 re and 4 res, are as follow: i) The increase in wild re frequency switched the vegetation community, from pine woodland to shrubland, and delayed the plant recovery response of the shrub community, ii) The increase in wild re frequency signi cantly decreased the biotic ground cover, signi cantly increased the SOM content and the silt soil fraction, and slightly decreased SWR, iii) Increasing wild re frequency did not affect the overall annual/seasonal soil moisture content (SMC), although soil water retention was signi cantly lower at eld capacity (FC) and signi cantly higher at the permanent wilting point (PWP) on the 4 res as compared to the 1 re soils, resulting in a lower available water content (AWC) as increasing wild re frequency. iv) However, the threshold for some effective soil water available for plants (ESWC > 0%) extended for longer (17 days more) in the 4 res hillslopes than on the 1 re hillslopes, and also in the bare than plant microsites (10 days). As a consequence, the effective soil water content (ESWC) was lower on the 4 res than in the 1 re area.
v) Both SMC and ESWC were signi cantly higher at deeper soil layers than on the super cial ones and were also signi cantly affected by seasonal in uences. The microsite was not a signi cant factor, although overall mean values revealed higher ESWC under plant, as compared to bare microsites. Antecedent rainfall in 7 days was a signi cant covariate for both SMC and ESWC.
Increasing wild re frequency led to both maximum and minimum values of SMC, respectively for the wet and dry seasons. Sites burned more frequently had higher SMC during the wet seasons, when water was not a constraining factor for plant growing, and on the other hand, experienced a more pronounced drop of SMC and ESWC during the dry season. The opening of the water-stress window of the dry season happened sooner and extended for longer in these hillslopes. These ndings showed that present and future global warming scenarios, are triggering catastrophic tipping points towards deserti cation even in humid Mediterranean forestlands.

Con icts of interest
Not applicable.
Availability of data and material Not applicable.
Code availability Not applicable.

Figure 1
Experimental set-up showing the unburnt (green circles), 1 re (yellow circles) and 4 res hillslopes (red circles) as well as a scheme of the microsites (plant and bare) and soil depths (2.5 and 7.5 cm) monitored in this study. Bottom pictures (from left to right) corresponds to a 4 res (4x), a 1 re (1x) and an unburnt (0x) hillslope. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.  Soil water repellence (SWR) determined by the molarity ethanol drop test (MED) for the unburnt, 1 re and 4 res areas. SWR classes were grouped as follows: Class 0, very wettable; 1, 2 and 3, wettable; 4 and 5, moderately and strongly; 6 and 7 very strongly; 8 and 9, extremely repellent.

Figure 5
Soil water retention curves (% volume) for the unburnt, 1 re and 4 res hillslopes at a range of selected tensions (pF-values from 0 to 4.2; n=54). The curves were tted to all data points following Van Genutchen et al. (1991).