1. Study sites

The observational and manipulative experiments were conducted in the six CASCADE study sites, which are all located in southern Europe. Two are stressed by fire: Várzea in North-Central Portugal and Ayora in Eastern Spain, and four by grazing: Albatera-Santomera in Eastern Spain, Castelsaraceno in Southern Italy, Messara in Southern Crete and Randi in Southern Cyprus.

For general descriptions of the study sites including information on geography, soils, climate, land use, degradation drivers, socioeconomic status and evolution of vegetation see

»Várzea Portugal: Description of site and main causes of degradation
»Albatera, Spain: Description of site and main causes of degradation
»Ayora, Spain: Description of site and main causes of degradation
»Castelsaraceno, Italy: Description of site and main causes of degradation
»Messara, Greece: Description of site and main causes of degradation
»Randi Forest, Cyprus: Description of site and main causes of degradation

A brief summary of the study site characteristics is provided in Table 1.

Table 1. Summarized description of the study sites

2. Target species

For all six CASCADE sites, we identified the most relevant plant species to be considered as target species. The species selected were either shrubs or perennial grasses (Table 2). The selection was based on the importance of their role (abundance; functional role) in the ecosystems of study. The target species in Várzea, Pterospartum tridentatum, is a dominant shrub species in the study site, which quickly resprouts after fire and plays a critical role in the post-fire recovery of the vegetation. Rosmarinus officinalis, the target species in Ayora, represents a different post-fire regeneration strategy, a post-fire seeding shrub, which is very common in fire-prone shrublands developed in old agricultural fields in the Mediterranean. Anthyllis cytisoides and Calicotome villosa, the target species in Albatera and Randi, respectively, represent abundant palatable shrubs and therefore represent the species that are the most likely to be affected by grazing intensity. Similarly, Brachypodium rupestre and Stipa austroitalica in Castelsaraceno and in Hyparrhenia hirta Messara represent the common palatable grasses in these grazed areas.

Table 2. Selected target species

Species	Species description
	Name: Pterospartum tridentatum (L.) Family: Fabaceae Functional group: Shrub Functional strategy: Resprouter species Site: Várzea (Portugal)
	Name: Anthyllis cytisoides (L.) Family: Fabaceae Functional group: Subshrub Functional strategy: Drought deciduous species Site: Albatera-Santomera (Spain)
	Name: Rosmarinus officinialis (L.) Family: Lamiaceae Functional group: Shrub Functional strategy: Seeder species Site: Ayora-Mariola (Spain)
	Name: Brachypodium rupestre (Host) Roem. & Schult. Family: Poaceae Functional group: Herb Functional strategy: Perennial grass Site: Castelsaraceno (Italy)
	Name: Stipa austroitalica (Martinovský) Family: Poaceae Functional group: Tussock grass Functional strategy: Perennial grass Site: Castelsaraceno (Italy)
	Name: Hyparrhenia hirta (L.) Family: Poaceae Functional group: Subshrub Functional strategy: Perennial grass Site: Messara (Crete)
	Name: Calicotome villosa (Poir.) Link Family: Fabaceae Functional group: Shrub Functional strategy : Drought deciduous species Site: Randi (Cyprus)

 Note: For full references to papers quoted in this article see

» References

Increasing fire recurrence in Europe and impact on soil quality

Since the mid of the last century fire recurrence (number of fire events that occur at a site in a given period of time) has increased in the Iberian peninsula and the overall Mediterranean basin (Pausas and Fernández-Muñoz, 2012). This occurs due to fuel accumulation from land abandonment and extensive reforestation (Koutsias et al. 2012) and to extreme weather events (Camia and Amatulli 2009, Carvalho et al. 2011, Hoinka et al. 2009, Koutsias et al. 2012). The future warmer and drier climate projected for this region will further increase the risk of wildfire occurrence and of increasing fire recurrence (Giorgi and Lionello 2008). Future wildfire risk is projected to increase in Southern Europe (Lindner et al. 2010, Carvalho et al. 2011, Dury et al. 2011). The annual burned area is projected to increase by a factor of 3 to 5 in Southern Europe compared to the present under the A2 scenario by 2100 (Dury et al. 2011).

There is ample literature on the effects of fire recurrence on vegetation. Recurrent fires can lead to long-term cumulative effects at plant community level such as changes in plant composition and structure (Lloret et al 2003, Eugenio et al 2006, Santana et al 2010), losses in plant productivity (Díaz-Delgado et al 2002, Eugenio and Lloret 2004 Delitti et al 2005), and delays in post-fire plant regeneration. Such changes in vegetation are likely to be associated to changes in soil quality. For instance, it has been suggested that loss of plant productivity with subsequent fires is associated with a cumulative reduced availability of nutrients in mineral soils (Ferran et al 2005). However, although the impact of wildfires on soil nutrient content in Southern Europe has been extensively studied, only a few studies have assessed this impact on the basis of fire recurrence (Caon et al 2014).

One of the most common changes in plant communities driven by high fire recurrences in Southern Europe is the replacement of pine woodlands by shrublands. Despite the high post-fire resilience of the most common pines in the Mediterranean (Pinus halepensis, P. brutia and P. pinaster), these trees fail to regenerate when time interval between fires is shorter than the time needed to accumulate a sufficient seed bank, i.e. around 15 years (Eugenio et al. 2006, Santana et al 2010). Surprisingly, it has been poorly studied whether a shift from pine woodlands to shrublands is associated with a shift in soil fertility. Most of the available research assessing the impact of fire recurrence on soil fertility is performed in ecosystems dominated by species with resprouting ability after fire (i.e., Quercus suber woodlands or Q. coccifera shrublands), and thus, with no major shifts in plant community (e.g., Trabaud 1991, Carreira et al 1994, Guenon et al 2001, Ferran et al 2005). One of the few works studying effects of repeated burning on soils in Mediterranean pine woodlands found that, nine years after the last fire, sites burned twice in an interval of 18 years had less developed organic horizons but similar mineral soils than sites only burned once in that period (Eugenio and Lloret, 2005). The authors attributed this response to the lower vegetation development in twice- than in one-burnt areas. The lack of cumulative effects of recurrent fires in mineral soils could however not be concluded as pooled soils up to 20 cm depth were sampled, whereas maximum depth commonly affected by fires is around 5 cm (Giovannini, 1994).

Early warning indicators of soil functioning

A significant decline in soil quality has occurred throughout the entire world as a result of adverse changes in its physical, chemical, and biological properties, caused by human activity and climate change (Van Camp et al. 2004). Soil degradation processes in drylands are particularly acute due to the fragility imposed in these areas by water scarcity in combination with large human and climatic pressures. Thus, soil degradation in drylands is one of the main environmental problems worldwide, including Europe (32% of the land mass are drylands, home to 25% of the population). Moreover, this problem is expected to get worse in the face of current global change (Millennium Ecosystem Assessment, 2005; Reynolds et al., 2007), where Europe is forecasted to be one of the world’s regions most impacted. Further, both theoretical developments and empirical data provide evidence that healthy drylands can shift to a degraded state in response to small increases in human and climatic pressure once a threshold has been surpassed (Scheffer and Carpenter 2003, Rietkerk et al 2004, Schroder et al 2005, Gao et al 2011). This implies the possibility for sudden major and difficult-to-recover ecological and economic losses, what explains the research emphasis on identifying early warning indicators (Dakos et al. 2012, Kéfi et al. 2014). In this context, the identification of early warnings of changes in ecosystem functioning, including changes in soils, is prioritary for identifying areas with higher risk of degradation in response to specific pressures, being recurrent wildfires one of the most common pressures in European forest and shrublands.

Note: For full references to papers quoted in this article see

» References

The main objectives of the measurements were

1. to determine whether fire recurrence levels promoting shifts from pine woodlands to shrubland communities in Southern Europe are associated with shifts in soil fertility,
2. to assess if different types of soil surface cover (i.e., that below shrub patches and that in the openings between shrubs) have different sensitivity to fire recurrence effects, and
3. to identify the most sensitive indicators of changes in soil functioning in response to repeated fires.

Study sites

Two of the CASCADE study sites, Várzea (Portugal) and Ayora (Spain), are affected by increased wildfire recurrence and high fire frequencies have promoted a transition from pine woodlands to shrublands. The two sites are representative of fire-prone regions in Southern Europe, Ayora under fully Mediterranean climate and Várzea under Atlantic climate with some Mediterranean influence.

For a general description of these sites see

»Várzea Portugal: Description of site and main causes of degradation
»Ayora Spain: Description of site and main causes of degradation

The sites include plots characterized by different fire histories that are summarized in Table 1. We combine a diachronic approach (Várzea) for assessing short-term fire effects with a synchronic approach (Ayora) for long-term effects of fire recurrence. In Várzea, two fire recurrence areas (1 and 4 fires), both last burned by a wildfire in the summer of 2012, and a reference fire-free area for the last 35 years were selected. Three plots (30 x 30 m) were set up in each of the three areas (total of 9 plots). The study site in Ayora has plots affected by natural and experimental fires from previous studies and represents a chronosequence in fire recurrence and time since the last burn. It includes three fire recurrence levels (1, 2, and 3 fires) and a reference fire-free for the last 30 years in three different areas (Ayora, Alcoy, and Onil). Three plots (30 x 30 m) were set up in each fire level and area (total of 12 plots). In each site, the plots were selected to have physiographic and edaphic properties as comparable as possible (Table 1).

Table 1

Sampling and chemical analysis

In both study sites, microsites beneath and between shrubs (hereafter, shrub and intershrub) of the most representative species were identified and three 1-m² subplots per plot were randomly located in each microsite. Shrub microsites were Pterospartum tridentatum (resprouter) in Várzea, and Quercus coccifera (resprouter) and Rosmarinus officinalis (obligate seeder) in Ayora. One sample of mineral soil at 0-5 cm was taken in each subplot in spring 2013. For both study sites, soil samples were analysed for total organic carbon (SOC), total nitrogen (N), NH₄, NO₃, Potentially Mineralizable Nitrogen (PMN), and available phosphorus (P_ava). Additionally, hot-water extractable carbon (HWC) and dissolved organic carbon (DOC) were analysed in Várzea and Ayora, respectively. HWC was determined as the amount of dissolved organic carbon that is released during incubation of a soil sample in hot water during 16 hours at 80°C (Ghani et al, 2003). This is a measure of easily decomposable (labile) organic carbon. The HWC fraction of organic matter is rich in amorphous polysaccharides (mucigel) which originate mainly from microbial exudates and to a lesser extent from plant exudates. This fraction is highly available to microorganisms and is also regarded as one of the key labile components of organic matter responsible for soil micro-aggregation, which is an important soil physical parameter to consider in terms of soil quality (Ghani et al. 2003, Haynes 2005). Total organic C was determined using the potassium dichromate oxidation (Walkley-Black) method (Nelson and Sommers, 1982). Total N was determined by the Kjeldahl method (Bremmer and Mulvaney, 1982), and available P by the NaHCO₃-extractable Pi (Olsen-Pi) as described by Watanabe and Olsen (1965). Potentially mineralizable N was determined by anaerobic incubation of a soil sample under water for 1 week at 40°C (Keeny and Nelson, 1982; Canali and Benedetti, 2006) These warm and anoxic conditions are optimal for a quick mineralization of organic matter by anaerobic bacteria. The lack of oxygen prevents conversion of released NH₄⁺ to NO₃^- (nitrification) and uncontrolled N losses by denitrification cannot occur. The amount of mineral nitrogen (NH₄-N) released is a measure of the quality (N-content and decomposability) of the organic matter, and thus for biological soil fertility.
In addition to the total values, we calculated the next ratios expressing the values of different nutrients relative to their source: HWC:SOC, PMN:N, NH₄:N, NO₃:N, and P_ava:SOC. These ratios are used as indicators of the quality of the soil organic matter and allow comparisons between sites with contrasting amounts of soil organic matter, as it is the case for the two study sites.

Statistical analysis

An analysis of variance was performed for each of the soil variables measured, using recurrence and microsite as fixed factors. Additionally, plot (nested in recurrence), and area and plot (this latter nested in the interaction recurrence X area) were used as random factors in the Várzea and Ayora datasets, respectively.

Principal component analysis (PCA) was used to reduce the original set of variables of the soil organic matter quantity matrix (SOC, N, NH₄, NO₃, PMN, P_ava, and HWC or DOC) and quality matrix (HWC:SOC, PMN:N, NH₄:N, NO₃:N, and P_ava:SOC) into a smaller set of uncorrelated components that represent most of the information found in the original variables. Variables were transformed if needed to fit normal distributions. Statistical analyses were performed with SPSS vs. 20.

For the detailed results from the two sites see

» Várzea Portugal: Critical changes preceding a catastrophic shift
» Ayora, Spain: Critical changes preceding a catastrophic shift

Note: For full references to papers quoted in this article see

» References

Short-term effects of fire recurrence on soil nutrients

In general, the fertility of the top mineral soil (0-5 cm) one year after fire was more influenced by the occurrence of the last fire than by the number of previous fires (1 or 4 fires in the last 37 years). Still, there was a decreasing trend with increasing fire recurrence in soil organic C and total N, as well as with their labile fractions, although this latter trend was not significant (Figure 2)

Figure 1
Figure 2

Thus, the occurrence of the last fire had a positive effect on soil organic C, total N, NH4, and available P, which were higher in burned than in unburned soils eight months after the occurrence of one or even four fires. The short-term post-fire pulse in NH4 and available P is commonly found in the literature and is associated with NH4 addition with ash deposition and soil heating, and with the pyro-mineralization of organic P during the fire, respectively (Serrasolsas and Khanna 1995, Certini 2005). Short-term (< 5 years) increases in soil organic matter following fire have also been reported in other woodland ecosystems in southern Europe (Rashid 1987, Dumontet et al 1996, Perez et al 2004, Knicker et al 2005) and are commonly attributed to the decomposition of partially burned woody fragments and charcoal formation (Almendros et al. 1990, Knicker et al 2006, Certini et al. 2011). However, the higher values of N in burned than in unburned soils does not support the idea of relevant quantities of fire-deposited charcoal as this component has a very low content in N. Despite the positive impact of fire occurrence on the total amount of soil nitrogen, its potentially mineralizable fraction did not increase resulting in a lower potentially mineralizable N to total N ratio in burned than in unburned soils (Figure 1 and 2). Thus, fire increased the quantity of soil organic matter but decreased its quality, and to a similar degree for both soils burned once or four times (Figure 2).

The fire-induced increase in soil organic C and total N, mainly represented by the first component  performed with the quantity matrix, was mitigated by fire frequency as this component was lower in soils burned four times than in soils only burned once. These results may be attributed to a lower woody biomass before the 2012 fire in plots burned with higher fire frequency, relative to plots burned only once, and thus a lower increase in soil organic matter due to root death.

Figure 3

The fertility of the soils in the microsites assessed, shrub patches of P. tridentatum and interspaces in between these shrubs, generally showed the same short-term response to fire. The intershrub microsite was however more sensitive to the occurrence of fire, as it showed significant higher soil organic C and total N in burned than in unburned soils, but also significant lower values for C and N labile fractions in burned than in unburned soils (Figure 3).

Long-term effects of fire recurrence on soil nutrients

Soil organic C and most nutrients of the top mineral soil (0-5 cm) in Valencia showed higher values in long unburned soils (at least for 30 years) than in soils burned two or three times (20 and 8 years after the last fire, respectively). This trend was however not significant due to the large variation in the response of the blocks analysed (Figure 4). Although two of the blocks, Alcoy and Ayora showed decreasing trends in soil fertility with increasing fire recurrence (1, 2, or 3 fires), the third block, Onil, only showed this trend for dissolved organic C.

Figure 4

In agreement with our results, most related studies in Mediterranean Europe measuring the long-term response following fire (> 5 years) found that shrublands or woodlands burned with high fire recurrence, that is three or four times in a period of up to 57 years, had significantly lower soil organic matter than sites only affected by one or no wildfire during that period (Carreira et al. 1994, Guénon et al. 2011, Tessler et al. 2013). A negative effect of fire recurrence in the quantity of mineral soil organic matter is associated with the cumulative effects of recurrent fires in reducing soil organic horizons (Eugenio et al. 2006). In our study site, long-term changes in soil organic C and  nutrients related to fire impact on vegetation are also possibly related to changes in the composition of vegetation along post-fire succession since: unburned and  1-fire were pine forests, whereas 2- and 3-fires plots were shrublands. Litterfall quantity and quality, and decomposition rate, may be different for both groups of plots.  

Long-term post-fire reductions in soil C may limit microbial activity and constrain gross production of NH₄ over time (Koyama et al 2010) as we observed in our study area (Figure 5). Furthermore, nitrogen availability may decline over secondary succession as plant and microbial competition for N increases and regenerating vegetation alters the soil microclimate. On the other hand, significant amounts of N can be volatilised or lost due to erosion. Thus, despite temporary post-fire enhanced mineral organic C and N availability, repeated burning in Mediterranean steep areas may lead to substantial depletion of the total N reserves, causing marked changes in the long-term pattern of N cycling (Carreira et al 1995, 1996).

Figure 5

As happened in the Várzea study site, the surface of the intershrubs in Valencia seemed to be more sensitive to fire impact than that of the shrub patches analysed in this site (Q. coccifera and R. officinalis), as intershrub microsites did show a significant decrease of soil organic matter with increasing fire recurrence (Figure 3).

Indicators of changes in soil functioning in response to fire recurrence

Our results suggested a higher sensitivity of the soil surface between shrubs, relative to the surface below shrub patches, to the fire occurrence and recurrence. We attributed this result to differences between the two microsites in the recovery rate after fire. The canopies of shrub patches recover fast after the fire (e.g., Malanson and Trabaud, 1987; Pausas et al., 1999). This entails a higher accumulation of organic matter content in shrub patches, particularly for resprouter shrubs as supported by their higher soil organic C in both of ours study sites. The accumulation of organic matter in shrub patches is not only due to plant litter from the same patch but also through its sink role by trapping soil, seeds and litter from upslope areas during the first post-fire rainstorms (Cammeraat and Imeson, 1999; unpublished observations). Contrastingly, the openings between shrubs, either partially bare or covered by short herbaceous species, have a lower capacity to trap resources and function more like source areas.

Soil organic C and total N gave a similar response to fire recurrence at both the short- and long-term following fire, that is, a decreasing trend with increasing fire recurrence (1, 2, 3 or 4 fires in 37 years). However, the comparison with reference areas at the short term may give biased information due to the transient fire-induced pulse of soil fertility. Thus, relative to long unburned areas, a single fire may increase the quantity of soil organic matter in areas burnt with low or even high fire recurrence the first years following the fire. It, however, may still not increase the bioavailability of soil C and nutrients. In this regard, labile fractions of soil organic matter may be more robust indicators of changes in soil functioning in response to fire recurrence.

Our results also suggest a shift in mineral soil fertility associated with a fire-induced shift of pine woodlands to shrublands. Drops in soil fertility quantity and quality were generally higher between areas burned once and twice, corresponding with the replacement of woodlands by shrublands, than between areas burned twice and thrice, both of them shrublands (Figure 6).

Figure 6

The detailed results from the two fire-affected sites can be found here

» Várzea Portugal: Critical changes preceding a catastrophic shift
» Ayora, Spain: Critical changes preceding a catastrophic shift

Note: For full references to papers quoted in this article see

» References

The stress-gradient experiment was carried out in the four sites in which grazing was defined as the environmental stressor. The sites included plots with different grazing intensities. Relative grazing intensity at each site was quantified by different methods, including

• direct measurements of livestock density,
• tracking herd routes to estimate distances to shelters, and
• dropping counts along transects or within sampling plots.

Counts of livestock dung provide a good index of the amount of time that livestock spent grazing in a particular area (Riginos & Hoffman, 2003), and the frequency of goat visits to a particular area decreases with the distance from the shelter, so both variables can be used as comparable indicators.

Based on different indicators, we selected plots that followed a linear gradient of grazing pressure in each site (Figure 1). For this grazing gradient,

• ungrazed (or barely grazed) plots were considered as control plots (hereafter, Low Stress, LS),
• plots under medium grazing intensity were considered as Medium Stress plots (hereafter, MS) and
• plots under high grazing intensity were considered as High Stress plots (hereafter, HS).

Although we established a linear grazing gradient in all sites, there were differences between sites regarding both the type of livestock and the overall grazing pressure. Thus, goats were the main livestock in Randi and Santomera, while livestock in Messara combined goats and sheep, and livestock in Castelsaraceno included goats, sheep and cattle. Overall grazing pressure during the study period was very high in Randi and Messara, moderate in Castelsaraceno and relatively low in Santomera.

For the soil variables, in addition to the grazing-stress factor, we also assessed the effect of the soil microsite (i.e. soil underneath plant patches and soil in the bare-soil interpatches). This factor is relevant in dryland sites, where vegetation is arranged in patches interspersed in a matrix of more or less bare soil. Therefore, we assessed this as an experimental factor in the CASCADE dryland sites with patchy vegetation, Santomera, Messara and Randi, but not in the wettest site (Castelsaraceno), where patches and interpatches could not be clearly distinguished.

For a general description of the sites see

»Albatera, Spain: Description of site and main causes of degradation
»Castelsaraceno, Italy: Description of site and main causes of degradation
»Messara, Greece: Description of site and main causes of degradation
»Randi Forest, Cyprus: Description of site and main causes of degradation

Experimental design

The experimental design for each site therefore consisted of two factors:

• stress level (HS, MS and LS) and
• microsite type (vegetation patch and bare-soil interpatch).

In each study site, three plots of 30 x 30 m size, approximately, were set up at each of the three stress levels and, for each of these plots, 5 vegetation-patch microplots (P microsite) and 5 interpatch microplots (IP microsite) were selected for repeated soil and plant measurements. In each study site, the plots were either interspersed or following a block design, depending on the local conditions. Two main sampling campaigns (spring and autumn) were carried out during approximately one year of study period (mostly in 2013), plus some preliminary exploratory campaigns to define the general soil and plant characteristics of each experimental plot.

Plant and soil measurements

In order to harmonize the sampling and measurement procedures for the assessment of the selected soil-plant system variables (Table 1), an experimental protocol was agreed among the partners. To assess plant performance, we measured

• Plant Height (PH),
• Plant Biomass (PB),
• Plant Cover (PC),
• Twig Basal Diameter (TBD),
• Twig length (TL),
• Branch Basal Diameter (BBD),
• Branch length (BL),
• Specific Leaf Weight (SLW) 
• Chlorophyll content (SPAD).

To assess soil functioning, we measured variables related to soil texture, pH, soil carbon and soil nutrient status. The specific variables measured were

• Cation-Exchange Capacity (CEC),
• Soil Organic Carbon (SOC),
• Potentially Mineralisable Nitrogen (PMN),
• Available Nutrients (nitrate, ammonium, phosphate),
• Dissolved Organic Carbon (DOC) and
• Hot Water Extractable Carbon (HWC).
• Soil water content (SWC).

Table 1. Vegetation and soil variables monitored at each site for the grazing stress gradient experiment. The sets of parameters measured per site could slightly differ, because of plant/soil/ecosystem properties and available expertise/equipment.

   Acronyms list: ● Vegetation growth: TBD: Twig Basal Diameter; TL: Twig length; BBD: Branch Basal Diameter; BL: Branch length. ● Soft ecophysiological traits: SLW: Specific Leaf Weight; RWC: Relative Water Content; SPAD: Chlorophyll content. ● Soil characterization: CEC: Cation-Exchange Capacity SOC: Soil Organic Carbon. ● Nutrient availability: PMN: Potentially Mineralisable Nitrogen. Available P (Olsen) ● Labile pool of soil C: DOC: Dissolved Organic Carbon; HWC: Hot Water Extractable Carbon. ● SWC: Soil Water Content.

Statistical analyses

Because of the observational character of the stress-gradient experiment, we performed a principal component analysis (PCA) to identify which soil and plant variables were mostly affected by the two factors stress level (S) and microsite (M). In addition, the plant and soil variables were analysed using a General Linear Model (GLM) with two fixed factors (Stress level (S), and Microsite (M)) for the soil data and one fixed factor stress level (S) for the plant variable. All data met normal distribution of residuals and homoscedasticity assumptions. All analyses were carried out by using Statistical package SPSS version 23.0 (SPSS Inc., Chicago, IL, USA).

 Note: For full references to papers quoted in this article see

» References

1. Effect of grazing-stress on soil quality

Results

Figure 1a summarizes the effect of grazing pressure on the overall soil condition in the four CASCADE study sites for which grazing pressure is the main degradation driver; only results for autumn season are shown, but spring data showed similar patterns (see Annex I for detailed description of effects on individual variables and seasons).

Figure 1a
Figure 1b

Data represent means ± SE of axis scores for each stress level. Arrows represent soil variables significantly correlated (|ρ| ≥ 0.6) with the first two axes. LS: low stress (diamonds); MS: moderate stress (triangles); HS: high stress (bullets).

The multivariate (PCA) analysis performed on soil data yielded two main axes in all sites. However, the particular variables that correlated with each axis varied between sites. In Castelsaraceno and Santomera, the wettest and driest site respectively, soil condition did not significantly vary between grazing-stress levels, showing higher variation within each stress level than between levels. However, in Santomera, all soil variables combined in the two PCA axes showed a gradual gradient towards higher nutrient availability and larger amounts of soil organic N and C from high-stress to low-stress plots. Conversely to the previous two sites, soil data from Messara and Randi showed clear variations between stress levels. PC1 in Messara (which positively correlated with total N, SOC, and nitrates) showed a gradual increase from low-stress to high-stress, while PC2 in Randi (which also positively correlated with total N, SOC, and nitrates, and negatively correlated with available P) showed a sharp degradation from low- and moderate-stress to high-stress. In Messara, PC2 correlated with available P and ammonia and showed higher values for moderate-stress than for the other two levels. Similarly, in Randi, PC1, which correlated with ammonia, PMN and labile forms of organic matter, showed higher values for moderate-stress than for the other two levels.

Figure 1b summarizes the effect of grazing pressure on the overall soil condition of patch versus interpatch microsites (only for the study sites with patchy vegetation: Santomera, Randi, and Messara). Results for Messara were scattered, but in Santomera and Randi patch and interpatch microsites were clearly separated, with patch microsites showing higher values for soil quality variables such as PMN, DOC and Available P (Santomera) and Organic C (SOC), Available P, NO3 and Total N (Randi). For these two sites, overall differences in soil condition in response to the stress gradient were higher for patch microsites than for interpatch microsites.

For each CASCADE study site with patchy vegetation, Figure 2 shows the variation in the patch/interpatch ratios for soil variables as a function of the grazing stress level. All soil variables showed higher values in soils underneath plant patches than in the interpatches, with the exception of Total N in Messara site and in medium- and low-grazing areas in Santomera site, which showed ratios close to 1. The ratios patch/interpatch were especially high in Randi site. In Randi, patch/interpatch ratios significantly increased from low stress to medium or high stress levels for SOC, ammonium and PMN (Table 1). The same trend (p<0.1) was observed for SOC in Messara and for total N in Santomera. Conversely, DOC ratio in Santomera and available P ratio in Messara were significantly (p<0.1) higher for low grazing-stress than for medium or high stress. The rest of variables did not show any significant effect of the grazing-stress level (Table 1).

Figure 2

Table 1. Statistics results (F and P-values) of the analysis of variance (ANOVA) on the patch/interpatch ratios of the soil variables assessed for Santomera, Messara and Randi sites as a function of the grazing-stress level.

SOC: Soil organic carbon; NH4: ammonium; NO3: nitrates; Avail-P: available phosphorus; PMN: potentially mineralisable nitrogen; and DOC: dissolved organic carbon. Significant results (p<0.05) are highlighted in bold and marginally significant results (p<0.1) in italics.

Discussion

It is assumed that the interplay between small scale facilitation (local facilitation) and large scale competition (global competition) between plants underlie catastrophic shifts in drylands (Rietkert et al. 2004). Local facilitation is especially relevant under relatively harsh conditions (He et al. 2013). However, under certain (high) levels of the pressure facilitation may no longer counterbalance the overall competition for scarce resources, leading to a sudden shift towards a degraded state. Facilitation is exerted through multiple mechanisms that, in general, imply improved soil conditions underneath and near plant patches, particularly higher water infiltration capacity (Mayor et al. 2009) and nutrient cycling (Mayor et al. 2016). Here, we assessed changes in the degree of soil improvement as a function of a grazing-stress gradient. In doing so, we aimed to identify functional thresholds (Bestelmeyer, 2006) that might occur after a critical decrease in the soil-plant system took place.

In two of our study sites, Castelsaraceno and Santomera, soil quality parameters did not significantly vary between grazing-stress levels. These two sites represented the wettest and the driest CASCADE sites, respectively, and an overall moderate-low grazing pressure. The relatively low grazing intensity of the ranges considered in both cases could be the most plausible cause for the lack of effect of the stress gradient. Mediterranean rangelands are quite resilient against moderate grazing pressures, which do not significantly hamper ecosystem services provision in these areas (Papanastasis et al. 2015).

An alternative or additional possible explanation for the lack of effect of the grazing level in these two sites might be that the effect of grazing interacts with climatic effects, with the importance of grazing pressure somehow reduced under the most extreme climatic conditions. Castelsaraseno is a relatively wet site with high vegetation cover. These conditions might make the soil ecosystem robust enough to deal with the current range of grazing levels. Santomera is a very dry site, and for this site one might argue that soil conditions are mainly affected by water as the limiting factor for vegetation cover and growth, which in turn determine soil quality parameters.

Conversely, Messara and Randi showed clear effects of grazing-stress levels with some soil quality indicators (PMN, ammonium, available P and labile C forms) peaking under moderate grazing stress and further decreasing under high grazing stress in both sites, and some other indicators (SOC and total N) either decreasing (in Randi) or increasing (in Messara) with increasing grazing pressure. In Randi, we also found the largest differences between the values in the patches compared to the interpatches. For several soil variables, these differences increased with the level of grazing pressure. Only in the case of soil organic carbon, we observed a unimodal response, with the highest contrast between patch and interpatch values at medium level of stress and a further decrease in this contrast at the highest stress level. In the proximity of a critical shift, we expect increased spatial (and temporal) variance (Guttal and Jayaprakash 2009), which will be then reduced after the shift to a degraded state. According to this framework, our results suggest that the range of grazing pressure assessed in the Randi site includes conditions that may drive the system to a sudden shift into a degraded state.

2. Effect of grazing-stress on plant performance

Results

Figure 3 gives the results found for canopy area (CA) and twig basal diameter (TBD) of the shrub target species in Santomera and Randi sites. In Santomera, the canopy of the species Anthyllis cytisoides clearly increased in spring time (3 months of monitoring), regardless of the grazing-stress level (Figure 3). After this peak, the canopy area sharply decreased during the summer period. While the variation in canopy area with time was highly significant, this variable did not show any significant effect of the stress level (Table 2). In the case of the shrub species Calicotome villosa, the target species in Randi, canopy growth showed significant differences among stress levels (Table 2), with the highest growth in canopy observed for the low-stress level, whereas the lowest values were observed for the high-stress level (Figure 3). In Santomera, Anthyllis twigs grew during the first 6 months of monitoring, until the beginning of the summer season. After that, only plants under high stress level continued growing while plants at low and medium stress levels became stagnant. Conversely, for Calicotome plants in Randi, the change in the branch basal diameter followed the same pattern as the overall canopy size, growing significantly more for low and moderate grazing stress, and barely growing for high-stress plots (Figure 3; Table 2).

Figure 3
Figure 4

Figure 4 shows the changes in plant cover and biomass for the perennial grass species Brachypodium rupestre and Hyparrhenia hirta in Castelsaraceno and Messara, respectively. The change in plant cover for B. rupestre, relative to the initial values, did not show any clear differences among stress levels during the first six months of monitoring (Figure 4). However, the recovery of plant cover after the seasonal decay was significantly higher (Table 2) for the low-stress level than for the other two levels. In Messara, the changes in plant biomass did not show any significant effect of the stress gradient (Figure 4; Table 2)

Table 2. Statistics results (F and significance, P) of the Repeated Measures Analysis of variance (ANOVA) for the variables “change in canopy relative to the initial”, “change in twig basal diameter (TBD) relative to the initial”, “change in plant cover relative to the initial” and “change in plant biomass relative to the initial” for Santomera, Castelsaraceno, Messara and Randi respectively. We used Time (T) as a within-subject factor and Stress level (SL) as an intersubject factor. Significant results (p<0.05) and highlighted in bold and marginally significant results (p<0.1) in italics.

Discussion

In Santomera, canopy cover and basal twig diameter of Anthyllis cytisoides did not vary noticeable with stress level although there was a trend towards higher twig growth for the high-stress plots, probably due to compensatory growth (Oesterheld and McNaughton 1991) under grazing intensities that were not particularly stressful in absolute terms. In Randi, Calicotome villosa canopy cover and branch basal diameter showed higher values for low and intermediate stress level and low values for high stress. In Castelsaraceno and Messara, plant cover and biomass of the two target species Brachypodium rupestre and Hyparrhenia hirta respectively, did not show clear differences among stress levels, except for a slightly higher recovery in plant cover after the seasonal decay for low-stress plots in Castelsaraceno.

An explanation for these results might be found along a similar kind of reasoning as for the grazing effects on soil quality. Santomera is a water controlled ecosystem with, in our study, relatively small effects of grazing. Castelsaraseno is a relatively vegetation rich ecosystem with a high robustness against stress imposed by grazing. Messara showed response in soil quality to stress through grazing, but apparently no response by the vegetation. Only the Randi site show a comprehensive pattern of results with sensitivity for grazing on both the level of soil quality an on vegetation performance.

 Note: For full references to papers quoted in this article see

» References

The drought-stress manipulative experiment aimed to evaluate the effect of imposed drought on the plant-soil system. It was established for moderate-stress level of either grazing intensity or fire frequency, depending on the experimental site. The experiment was carried out in the six CASCADE study sites, which ranged in natural annual precipitation from 268 mm in Santomera (Spain) to 1290 mm in Castelsaraceno (Italy). Moreover, weather variability between the study sites was particularly high during the experimental period (September 2013- November 2015), with the wettest sites having a very wet period (sites in Portugal and Italy) while the driest sites experienced particularly dry years (sites in Spain and Cyprus).

Experimental design

The drought-stress treatment was applied on plots of around 2 m² in size. For each of these plots, we considered two microsites: soils under the target plant, i.e., vegetation-patch microsite (P microsite), and the respective upslope (bare-soil) interpatch microsite (IP).

The drought experiment was set-up using special roofs constructed ad hoc that captured rainfall to induce drought below the roof (Figure 1). In order to minimize greenhouse effects on the experimental plots due to the rainfall-exclusion plots, we built a fully translucent cover, supported by metallic poles. To minimize the lateral influx of water from overland flow from the upslope area, metal sheets were inserted into the ground around the plot. The roofs designed for this experiment were made by two layers of V-profile transparent polycarbonate gutters, which conduct the rainfall water to an external gutter connected to a water storage tank, which allow measuring the excluded rain (Figure 1). When placed upside-down, the roof allows the rain to go through and thus works as a roofed control, with similar roof effects (e.g., greenhouse effect) except rainfall exclusion. It is important to highlight that the goal of the treatment was not to completely avoid soil wetting during rainfall but to decrease plant water availability relative to the controls. Indeed, soil wetting during rainfall is possible in the rainfall exclusion roofs when wind is strong and/or there is subsurface runoff.

Figure 1

We established three treatments:

1. Non-roofed controls (hereafter, C_NoR);
2. Rainfall-inclusion roofed control (hereafter, C_R); and
3. Rainfall-exclusion roofs (Drought treatment, hereafter, D).

To address the exceptional climatic variation among sites for the study period, special efforts were made to make the drought treatments comparable between sites. The duration of the experiment was established to guarantee that the rainfall received in the roofed plots during the 12 months previous to the end of the experiment were around or below the 1st percentile of the mean annual rainfall for the long-term reference period (30-50 years). Due to the extreme natural drought experienced in some sites the experiment needed a particularly long period to achieve the 1st percentile value. In all sites except Castelsaraceno, the experiments were extended several additional months (the number varying between sites) after achieving the 1st percentile, with the longest experiment conducted in the driest sites: Santomera (completed by December 2015) and Messara (completed by November 2015).

Soil nutrient availability and labile C, and plant performance were measured in all plots at the beginning of the experiment and at ∼4 months intervals. Measurements on plant reproductive effort were taken during the flowering-fruiting seasons. Using TDR probes, soil moisture was measured on a continuous basis for both the soil underneath plant patches and the upslope bare-soil interpatch.

Plant and soil measurements

Table 1. summarizes the soil quality and plant performance parameters measured in the drought experiment.

Table 1: Vegetation and soil variables (0-5 cm) for repeated measurements at each site for the drought-stress experiment. The sets of parameters measured per site could slightly differ, because of plant/soil/ecosystem properties and available expertise/equipment.

Acronyms list: ● Vegetation growth: TBD: Twig Basal Diameter; TL: Twig length; BBD: Branch Basal Diameter; BL: Branch length. ● Soft eco-physiological traits: SLW: Specific Leaf Weight; RWC: Relative Water Content; SPAD: Chlorophyll content. ● Soil characterization: CEC: Cation-Exchange Capacity SOC: Soil Organic Carbon. ● Nutrient availability: PMN: Potentially Mineralisable Nitrogen. Available P (AvailP) ● Labile pool of soil C: DOC: Dissolved Organic Carbon; HWC: Hot Water Extractable Carbon. ● SWC: Soil Water Content.

Statistical analysis

We analysed all the variables measured for patch (P) and interpatch (IP) microsites using Repeated Measures ANOVA (GLM) with Treatment (D), with three levels, i.e. C_NoR, C_R and R, as between-subject factor and Time (T) as within-subject factor. All data met the normal distribution of residuals and homoscedasticity assumptions. All analyses were carried out using v.23.0 Statistical package (SPSS Inc., Chicago, IL, USA).

 Note: For full references to papers quoted in this article see

» References

1. Effect of drought-stress on soil quality

Figure 1 shows the results for potentially mineralisable nitrogen (PMN) and hot water-extractable carbon (HWC) measured at the Várzea site in Portugal. Both variables showed similar values for patch and interpatch microsites. None of the variables was significantly affected by the drought treatment (Table 1). HWC showed a significant variation in time, peaking in March 2014.

In Castelsaraceno, the dynamics of potentially mineralisable nitrogen (PMN) and hot water-extractable carbon (HWC) did not show any significant effect of the drought treatment (Figure 2), yet both variables showed statistical variation in time. Given the homogeneity of the vegetation in Castelsaraceno, we only considered the patch microsite in this site.

Figure 1
Figure 2

In the Valencia site, patch and interpatch microsites showed similar values (Figure 3) and none of the assessed variables showed any significant effect of the drought treatment (Figure 3 Table 1), with PMN and DOC showing significant variation in time (Table 1).

Figure 4 shows the results from the Messara site in Crete. Likewise the sites reported above, patch and interpatch areas showed similar trends and values. Available phosphorus and potentially mineralisable nitrogen (PMN) did not show any significant effect of the drought treatment (Table 1), while hot water-extractable carbon (HWC) showed significantly different dynamics as a function of the drought treatment, with overall higher values in the open (C_NoR) controls. The three variables showed significant variation in time (Table 1).

Figure 3
Figure 4

Figure 5 shows the results from the Santomera site in Spain. Overall, patch soils showed slightly higher values than interpatch soils. Available P did not show a statistically significant effect of the drought treatment, while PMN in interpatches showed significantly different dynamics and higher values for the drought treatment as compared with both controls (Figure 5; Table 1). Similarly HWC showed a marginally significant increase for the drought treatment in interpatches. All the variables showed significant variation along the study period.

Figure 6 shows the results from the Randi site in Cyprus. Similar to the case in Santomera site, only PMN showed a significant effect of the drought treatment (Table 1), with a higher increase in the patch microsites for the drought plots than for the control plots towards the end of the experiment.

Figure 5
Figure 6

Table 1. Statistics results (F and P-values) of the Repeated Measures Analysis of Variance for the soil variables:

Potentially mineralisable nitrogen (PMN), available phosphorus (Avail P), dissolved organic carbon (DOC) and hot water extractable carbon (HWC) for Várzea, Castelsaraceno, Valencia, Messara, Santomera and Randi sites. Time (T) was used as within-subject factor and Drought Stress (DS) as between-subject factor. T * DS represent the interaction between the two factors. Significant results (p<0.05) and highlighted in bold and marginally significant results (p<0.1) in italics.

2. Effect of drought-stress on plant performance parameters

Figure 7 shows the variation in branch basal diameter (BBD), both as absolute figures and as figures relative to the initial value, for the target plant species as a function of the drought treatment in the Várzea site, Portugal. There was a continuous increase in branch diameter with time, due to the plant regrowth after the wildfire of September 2012, yet there was no statistically significant effect of the drought treatment (Table 2).

Canopy cover (CC) of the target species in Castelsaraseno, Italy, did not show any significant effect of drought, yet there was a significant interaction between Time and Drought factors, with similar variation in canopy for the two controls and drought treatment before the 1st percentile target for the drought treatment was achieved and a higher increase in canopy for the open controls after that moment (Figure 8; Table 2).

Figure 7
Figure 8

Branch length (BL) of the target species in the Valencia site in Spain showed no significant effect of the drought treatment, yet there was a trend towards a lower increase in BL with time for the drought treatment (Figure 9; Table 2).

For the plant biomass (PB) of the target grass species in Messara, Crete, there was no statistically significant effect of the treatments on the absolute or the relative values (Figure 10; Table 2).

Figure 9
Figure 10

Conversely, the change in twig length (TL) of the target shrub species in Santomera, Spain, and the height of the target shrub species in Randi, Cyprus, showed significantly lower values for the drought treatment than for the roofed controls, C_R (Figure 11 and 12, respectively; Table 2).

Figure 11
Figure 12

Table 2. Statistics results (F. and P-values) of the Repeated Measures Analysis of Variance for plant performance variables:

Branch basal diameter (BBD), canopy cover, branch length (BL), plant biomass, twig length (TL) and plant height, and their values relative to the start of the experiment for Várzea, Castelsaraceno, Valencia, Messara, Santomera and Randi. Time (T) was used as within-subject factor and Drought Stress (DS) as between-subject factor. Significant results (p<0.05) and highlighted in bold marginally significant results (p<0.1) in italics.

3. Effect of drought-stress on soil moisture

Figure 13 shows the dynamics of monthly rainfall and soil water content (SWC) for the wettest study sites: Várzea (Portugal) and Castelsaraseno (Italy). in both sites, the control plots without roof showed the highest SWC values, yet only for the patch microsite in Várzea site. Castelsaraseno showed particularly high SWC values as compared to the other sites.

Figure 14 shows rainfall and SWC dynamics for Valencia (Spain) and the Messara (Crete). In general, SWC showed a trend towards lower values for patches than for interpatches, particularly for the drought (D) treatment and in Messara. In Valencia, the highest SWC values were found for the roofed control treatment (R-C), either for patch or interpatch microsites, and the lowest for the drought treatment (D) in soils under plant patches. In Messara, there were no clear differences between control and drought plots.

Soil water content in Randi showed similar dynamics than in Messara, with lower values in patches than in interpatches (Figure 15), while in Santomera, there were no relevant differences between microsites. In Randi, interpatch SWC showed higher values for control than for drought plots, while soils under plant patches did not show any treatment effect. In Santomera, both patches and interpatches showed higher values for control than for drought plots, yet only at the end of the study period, once the natural extreme drought that occurred in the area was over.

Figure 13
Figure 14
Figure 15

4. Discussion

For the driest sites, Randi and Santomera, we found increased PMN and, to a lesser extent, labile C forms (HWC) values with increased drought stress towards the end of the experiment, but none of the other four sites showed any relevant effect of increased drought. Similarly, we found no clear treatment effect on plant performance, except for Randi and Santomera, which showed a decreasing trend in plant growth with increasing drought stress. These results indicate an extraordinary capacity of the plant-soil systems of very dry areas to cope with drought, as only the combined effect of a severe natural drought plus the additional experimentally-induced drought finally resulted in decreased plant growth. The increased PMN and HWC values found in the driest sites in response to increased drought could be explained by an increased amount of litter and dead roots being degraded (López-Poma and Bautista, 2014).

The observed soil moisture dynamics confirmed that the roofs captured rainfall, as both controls gave in most cases higher values for soil water content (SWC) than the drought treatments, but also that the impact of the drought treatment was very different between sites. This probably resulted from the occurrence of exceptionally dry conditions during the experimental period in the dry sites, and exceptionally wet conditions in the wet sites, with both cases leading to small differences between roofed and no roofed plots. Hence, in the dry sites insufficient rainfall is captured to obtain differences between treatments, while in the wet sites the soil beneath the roofs stayed wet through influx of water from outside the roofs. For those ecosystems to drive to tipping points for catastrophic shifts to happen requires probably longer and more severe periods of drought.

Also unexpectedly, the results hinted to a positive side effect of the roofs, e.g. in the form shading or keeping air humidity higher, yet this effect would not be consistent across sites. These results highlight the need for using roofed controls to properly assess the effects of rainfall exclusion. Using examples of plant performance data from the various study sites, Figure 22 further explores and illustrates this finding. By comparing each treatment with the appropriate respective control (i.e., drought versus roofed control and roofed versus non-roofed controls), we found both a negative effect of induced drought and a positive effect of the roof on plant performance, both being higher in drier areas (Figure 16).

Figure 16

 Note: For full references to papers quoted in this article see

» References

• Bestelmeyer, B. T. (2006). Threshold concepts and their use in rangeland management and restoration: the good, the bad, and the insidious. Restoration Ecology 14: 325–329.
• Bremner, J.M., Mulvaney, C.S. (1982). Nitrogen-total. In: Methods of soil analysis, Part 2 “Chemical and Microbiological Properties’’ (A. L. Page, R. H. Miller, and D. R. Keeney, Eds.). American Society of Agronomy, Madison, WI. pp. 595-624.
• Camia A, Amatulli G, 2009. Weather factors and fire danger in the Mediterranean. In: Earth Observation of Wildland Fires in Mediterranean Ecosystems. Chuvieco E. (ed.), Springer, Berlin, pp. 71-82.
• Caon L, Vallejo VR, Ritsema CJ, Geissen V, 2014. Effects of wildfire on soil nutrients in Mediterranean ecosystems. Earth-Science Reviews 139, 47-58.
• Canali S, Benedetti A, 2006. Soil nitrogen mineralization. In: Bloem J., Hopkins, D.W., Benedetti, A. (Eds), Microbiological Methods for Assessing Soil Quality, CABI, Wallingford, UK, pp. 127-135.
• Carvalho A, Monteiro A, Flannigan M, Solman S, Miranda AI, Borrego C, 2011. Forest fires in a changing climate and their impacts on air quality. Atmospheric Environment, 45(31), 5545-5553.
• Carreira JA, Niell, FX, Lajtha K, 1994. Soil nitrogen availability and nitrification in Mediterranean shrublands of varying fire history and successional stage. Biogeochemistry 26: 189-209.
• Delitti W, Ferran A, Trabaud L, Vallejo VR, 2005. Effects of fire recurrence in Quercus coccifera L. shrublands of the Valencia  Region  (Spain):  I.  plant  composition  and  productivity. Plant Ecology 177:57–70
• Bremner JM, Mulvaney CS, 1982. Methods of soil analysis, part 2 chemical and microbiological properties, 595-624.
• Carvalho A, Monteiro A, Flannigan M, Solman S, Miranda AI, Borrego C, 2011. Forest fires in a changing climate and their impacts on air quality. Atmospheric Environment, 45(31), 5545-5553.
• Certini G, 2005. Effects of fire on properties of forest soils: a review. Oecologia 143(1), 1-10.
• Dakos V, Carpenter SR, Brock WA, Ellison AM, Guttal V, Ives AR, Kéfi S, Livina V, Seekell DA, van Nes EH, Scheffer M. 2012. Methods for detecting early warnings of critical transitions in time series illustrated using simulated ecological data. PLoS ONE 7(7): e41010.
• Díaz-Delgado R, Lloret F, Pons X, Terradas J, 2002. Satellite evidence of decreasing resilience in Mediterranean plan communities after recurrent wildfires. Ecology 83, 2293–2303.
• Dury M, Hambuckers A, Warnant P, Henrot A, Favre E, Ouberdous M, François L, 2011. Responses of European forest ecosystems to 21st century climate: assessing changes in interannual variability and fire intensity. IForest, 4, 82-99.
• Eugenio M, Lloret F, 2004. Fire recurrence effects on the structure and composition of Mediterranean Pinus halepensis communities in Catalonia (northeast Iberian Peninsula). Ecoscience 11 (4), 446–454.
• Eugenio M, Verkaik I, Lloret F, Espelta JM, 2006. Recruitment and growth decline in Pinus halepensis populations after recurrent wildfires in Catalonia (NE Iberian Peninsula). Forest Ecology and Management 231(1-3), 47-54.
• Ferran A, Delitti W, Vallejo VR, 2005. Effects of fire recurrence in Quercus coccifera L. shrublands of the Valencia region (Spain): II. Plant and soil nutrients. Plant Ecology 177(1): 71-83.
• Gao Y, Zhong B, Yue H, Wu B, Cao S. 2011. A degradation threshold for irreversible loss of soil productivity: a long-term case study in China. Journal of Applied Ecology 48: 1145-1154.
• Ghani A, Dexter M, Perrott KW, 2003. Hot-water extractable carbon in soils: a sensitive measurement for determining impacts of fertilisation, grazing and cultivation. Soil Biology & Biochemistry 35, 1231-1243.
• Giorgi F., Lionello P., 2008. Climate change projections for the Mediterranean region, Global and Planetary Change, Volume 63, Issues 2–3, September 2008, Pages 90-104, ISSN 0921-8181.
• Giovannini G, 1994. The effect of fire on soil quality. In: Sala, M., Rubio, J.L. (Eds.), Soil Erosion and Degradation as a Consequence of Forest Fires. Geoforma Ediciones, Logroño, pp. 15–27.
• Guenon R, Vennetier M, Dupuy N, Ziarelli F, Gros R, 2001. Soil organic matter quality and microbial catabolic functions along a gradient of wildfire history in a Mediterranean ecosystem. Applied Soil Ecology 48, 81-93.
• Guttal, V., Jayaprakash, C. (2009). Spatial variance and spatial skewness: leading indicators of regime shifts in spatial ecological systems. Theoretical Ecology 2(1): 3–12.
• Haynes RJ, 2005. Labile Organic Matter Fractions as Central Components of the Quality of Agricultural Soils: An Overview.  Advances in Agronomy, 85, 221-268.
• He, Q., Bertness, M. D., Altieri, A. H. (2013). Global shifts towards positive species interactions with increasing environmental stress. Ecology Letters 16:695-706
• Hoinka KP, Carvalho A, Miranda AI, 2009. Regional-scale weather patterns and wildland fires in central Portugal. International Journal of Wildland Fire, 18(1), 36-49.
• Janse, J.H., de Senerpoint Domis, L.N., Scheffer, M., Lijklema, L., van Liere, L., Mooij, W.M. (2008) Critical phosphorus loading of different types of shallow lakes and the consequences for management estimated with the ecosystem model PCLake. Limnologica- Ecology and Management of Inland Waters 38: 203-219.
• Keeney DR, Nelson DW, 1982. Nitrogen-inorganic forms. In ‘‘Methods of Soil Analysis, Part 2. Chemical and Microbiological Properties’’ (A. L. Page, R. H. Miller, and D. R. Keeney, Eds.), pp. 643–698. American Society of Agronomy, Madison, WI.
• Kéfi S., Rietkerk, M., Alados, C.L., Pueyo, Y., Papanastasis, V.P., ElAich, A., de Ruiter, P.C. (2007) Spatial vegetation patterns and imminent desertification in Mediterranean arid ecosystems. Nature 449: 213-217.
• Kéfi S, Guttal V, Brock WA, Carpenter SR, Ellison AM, Livina VN, Seekell DA, Scheffer M, van Nes EH, Dakos V. 2014. Early warning signals of ecological transitions: methods for spatial patterns. PLoS ONE 9(3): e92097.
• Koutsias N, Arianoutsou M, Kallimanis AS, Mallinis G, Halley JM, Dimopoulos P, 2012. Where did the fires burn in Peloponnisos, Greece the summer of 2007? Evidence for a synergy of fuel and weather. Agricultural and Forest Meteorology, 156, 41-53.
• Kuiper, J.J., van Altena, C., de Ruiter, P.C., van Gerven, L.P.A., Janse, J.H., Mooij, W.M. (2015). Food web stability signals critical transitions in temperate shallow lakes. Nature Communications, 6: doi:10.1038/ncomms8727.
• Lavalle C, Micale F, Houston TD, Camia A, Hiederer R, Lazar C, Conte C, Amatulli G, Genovese G, 2009. Climate change in Europe. 3. Impact on agriculture and forestry. A review. Agronomy for Sustainable Development, 29(3), 433-446.
• Lindner M, Maroschek M, Netherer S, Kremer A, Barbati A, Garcia-Gonzalo J, Seidl R, Delzon S, Corona P, Kolström M, Lexer MJ, Marchetti M, 2010. Climate change impacts, adaptive capacity, and vulnerability of European forest ecosystems. Forest Ecology and Management, 259(4), 698-709.
• Lloret F, Pausas JG, Vila M, 2003. Responses of Mediterranean plant species to different fire frequencies in Garraf Natural Park (Catalonia, Spain): field observations and modelling predictions. Plant Ecology 167(2): 223-235.
• López-Poma, R., Bautista, S. (2014). Plant regeneration functional groups modulate the response to fire of soil enzyme activities in a Mediterranean shrubland. Soil Biology & Biochemistry, 79: 5-13
• Mayor, A.G., Bautista, S., Bellot, J. (2009). Factors and interactions controlling infi ltration, runoff, and soil loss at the microscale in a patchy Mediterranean semiarid landscape. Earth Surface Processes and Landforms 34: 1702–1711.
• Mayor, A.G., Goirán. S.B., Vallejo, V.R., Bautista, S. (2016a). Variation in soil enzyme activity as a function of vegetation amount, type, and spatial structure in fire-prone Mediterranean shrublands. Science of the Total Environment 573: 1209–1216.
• Mayor, A.G., Valdecantos, A., Vallejo, V.R., Keizer, J.J., Bloem, J., Baeza, J., González-Pelayo, O., Machado, A.I., de Ruiter, P.C. (2016b). Fire-induced pine woodland to shrubland transitions in Southern Europe may promote shifts in soil fertility. Science of the Total Environment, doi: 10.1016/j.scitotenv.2016.03.243
• Millennium Ecosystem Assessment. 2005. Ecosystems and Human Well-being: Desertification Synthesis. World Resources Institute, Washington, DC.
• Mooij, W.M., Janse, J.H., de Senerpoint Domis, L.N, Hüllsmann, S., Ibelings, B.W. (2007). Predicting the effect of climate change on temperate shallow lakes with the ecosystem model PCLake. Hydrobiologia 584: 443–454.
• Nelson DW, Sommer LE, 1982. Total carbon, organic carbon, and organic matter. p. 539-579. In A.L. Page (ed.) Methods of Soil Analysis. 2nd Ed. ASA Monogr. 9(2). Amer. Soc. Agron. Madison, WI.
• Oesterheld, M., McNaughton, S.J. (1991). Effect of stress and time for recovery on the amount of compensatory growth after grazing. Oecologia 85: 305-313.
• Rietkerk, M., Dekker S.C., de Ruiter P.C., van de Koppel J. (2004). Self-organised Patchiness and Catastrophic Shifts in Ecosystems. Science 305: 1926-1929.
• Riginos, C, Hoffman MT. (2003). Changes in population biology of two succulent shrubs along a grazing gradient. Journal of Applied Ecology 40: 615-625.
• Santana VM, Baeza MJ, Marrs RH, Vallejo VR, 2010. Old-field secondary succession in SE Spain: may fire divert it? Plant Ecology 211: 337–349.
• Scheffer, M. Bascompte, J., Brock, W.A., Brovkin, V., Carpenter, S.R., Dakos, V., Held, H., van Nes, E.H., Rietkerk, M., Sugihara, G. (2009). Early-warning signals for critical transitions. Nature 461, 53-59.
• Pausas JG, Fernández-Muñoz S, 2012. Fire regime changes in the Western Mediterranean Basin: From fuel-limited to drought-driven fire regime. Climatic Change, 110(1-2), 215-226.
• Rietkerk M, Dekker SC, de Ruiter PC, van de Koppel J. 2004. Self-organized patchiness and catastrophic shifts in ecosystems. Science 305: 1926–1929.
• Scheffer M, Carpenter, SR. 2003. Catatrophic regime shifts in ecosystems: linking theory to observation. Trends Ecol. Evol., 18, 648–656.
• Schröder A, Persson L, De Roos AM. 2005. Direct experimental evidence for alternative stable states: a review. Oikos 110: 3–19.
• Serrasolsas I, Khanna PK, 1995. Changes in heated and autoclaved forest soils of SE Australia. II. Phosphorus and phosphatase activity. Biogeochemistry 29, 25-41.
• Trabaud L, 1991. Fire regimes and phytomass growth dynamics in a Quercus coccifera garrigue. Journal of Vegetation Science 2: 307-314.
• Van-Camp L, Bujarrabal B, Gentile AR, Jones RJA, Montanarella L, Olazabal C, Selvaradjou, SK, 2004. Reports of the Technical Working Groups Established under the Thematic Strategy for Soil Protection. EUR 21319 EN/2, 872 pp. Office for Official Publications of the European Communities, Luxembourg.
• Verwijmeren, M., Rietkerk, M., Bautista, S., Mayor, A.G., Wassen, M.J., Smit, C. (2014). Drought and grazing combined: contrasting shifts in plant interactions at species pair and community level. Journal of Arid Environment  111: 53-60.
• Verwijmeren, M., Smit, C., Bautista, S., Wassen, M.J., Rietkerk, M. Waning positive plant-plant interactions under combined drought and grazing stress (in review).
• Vogel, A., Fester, T., Eisenhauer, N., Scherer-Lorenzen, M., Schmid, B., et al. (2013) Separating Drought Effects from Roof Artifacts on Ecosystem Processes in a Grassland Drought Experiment. PLoS ONE 8: e70997.
• Watanabe FS, Olsen SR, 1965. Test of an ascorbic acid method for determining phosphorus in water and NaHCO3 extracts from soil. Soil Science Society of America Proceedings 29: 677–678.