Spatial and temporal opportunities for forest resilience promoted by burn severity attenuation across a productivity gradient in north western Patagonia

Introduction

Fire regimes in many fuel-rich ecosystems worldwide are dominated by high-severity canopy-consuming (or canopy-killing) fires where low severity fires rarely take place. Although many of these systems have fire-adapted species highly resistant or resilient to severe fires (Jayen et al. 2006; Murphy et al. 2013; Palmer et al. 2018; Owen et al. 2020), others show high vulnerability to these disturbances due to the lack of fire-resistant traits or post-fire regeneration strategies (Holz et al. 2015; Kitzberger et al. 2016b; Landesmann et al. 2021). In these latter systems, it is critical to understand biophysical (bottom–up) and fire weather (top–down) conditions that lead to the attenuation of fire severity. These spatial and temporal drivers could help predict particular locations (e.g. fire refugia) or temporal opportunities (e.g. seasons or periods of particular mild fire weather conditions) where and when high burn severity is attenuated and fire-sensitive ecosystems are less vulnerable to fire, and produce an overall more resilient landscape response. Resilience here may be attained through regeneration either from surviving individuals in fire refugia, or from individuals that may have survived low-severity fires. Thus, it is imperative to understand under which environmental and weather conditions different communities have less severe responses to fire.

Vegetation burn severity, a measure of the immediate impacts of fire on the environment through aerial biomass consumption (Keeley 2009), largely depends on a complex interplay between bottom–up and top–down factors that interact at different scales. Fuel characteristics (amount, continuity and condition to burn) modified by local factors such as radiation, nearness to heat sinks and topography are examples of bottom–up controls of fire severity (Gill 1975; Keeley 2009; Bradstock et al. 2010; Murphy et al. 2013). Worldwide, it is still unclear under which fuel amounts and structures different vegetation types have less severe responses to fire (Parks et al. 2014, 2018; Lindenmayer et al. 2020, 2022; Bowman et al. 2021). In some systems, burn severity is believed to increase with fuel amount up to an asymptote, beyond which it is not possible for fires to burn more severely despite biomass increases (Murphy et al. 2013). However, the vegetation structure and fuel characteristics of highly productive systems such as broadleaf temperate forest or rain forest vegetation may lead to lower burn severity than less productive ecosystems, as happens with fire occurrence (Pausas and Paula 2012). In this case, severity peaks when the combined effects of fuel build-up and drought are maximised, and is reduced where shady environments of highly productive ecosystems generate moister and darker microclimatic conditions (He et al. 2019). Here, burn severity may not be a function of the amount of fuels available, but of the environmental conditions that these fuels experience, such as the structure of the vegetation, topography and weather. However, fuel characteristics are difficult to measure, and these data are not available across different vegetation types and fire seasons. Vegetation productivity is a proxy of fuel characteristics such as amount and condition where high temporal resolution fuel maps are not available. Understanding the relationship between burn severity and productivity across different vegetation types may be the first step towards understanding under which conditions different vegetation types experience less severe fires.

Fire weather is a top–down control of fire severity that can turn less flammable vegetation into highly flammable vegetation. However, it is unclear how variability in fire weather affects burn severity, a critical aspect in the current context of climate change (Hessl 2011). The effect of weather on burn severity depends on the vegetation involved, and the spatial and temporal scale at which the pattern is observed (Cansler and Mckenzie 2014; Abatzoglou et al. 2017; Bowman et al. 2021; Lindenmayer et al. 2021). Warmer and drier conditions can increase burn severity by producing environments conducive to high-intensity fires both between seasons and in individual fires in temperate forests (Abatzoglou et al. 2017; Tran et al. 2020; Bowman et al. 2021; Collins et al. 2022). When looking at the pattern across vegetation types, vegetation characteristics start to be more important than fire weather in driving fire severity patterns (Walker et al. 2020). Thus, the effect of weather on burn severity depends on fuels structure, loading and moisture, and the environmental conditions. In order to understand the role of weather on burn severity patterns, it is important to study several contrasting fire seasons along a representative geographic range of fuel characteristics and vegetation types. We propose that wetter and cooler fire weather conditions would promote lower burn severity in highly productive vegetation types where moister fuel condition and structure conditions act as fire retardants. In contrast, drier and hotter fire seasons would homogenise the landscape, creating conditions for high burn severity across all vegetation types. Finally, sites of intermediate productivity would be less sensitive to weather as fuel amount and conditions create fire-prone systems under any weather scenario.

Patagonia has one of the most striking productivity gradients on Earth, which provides a unique opportunity to test the importance of bottom–up and top–down controls of burn severity across contrasting vegetation types. At the most productive and fuel-rich end of the gradient grow temperate broadleaf forests dominated by fire-sensitive obligate seeders (e.g. Nothofagus dombeyi, N. pumilio) (Veblen et al. 2003; Kitzberger et al. 2005). These forests are most vulnerable to fire the first decades after fire owing to the high amount of continuous fine fuels exposed to desiccation. Once the trees overtop the understorey, vegetation fuel continuity and amount in this lower stratum decrease, increasing the chances of these forests burning with low severity (Paritsis et al. 2015; Tiribelli et al. 2018). After high-severity fires, tree species are unable to regenerate owing to lack of seed sources and/or adverse weather for seedling establishment, and forests shift to shrublands or grasslands (Landesmann et al. 2021). At intermediate portions of the gradient, mixed shrublands coexist in the landscape with dry forests. Shrublands are prone to crown-consuming severe fires, but owing to their resprouting capacity rapidly recover their initial structure (Paritsis et al. 2015; Tiribelli et al. 2018; Landesmann et al. 2021). Finally, dry forests are dominated by the fire-sensitive conifer Austrocedrus chilensis. Despite all this understanding of the interaction between flammability and vegetation, burn severity has only been assessed for individual fires under specific weather conditions (Assal et al. 2018; Landesmann et al. 2021; Franco et al. 2022).

Here, we seek to understand what are the spatial (bottom–up) and temporal (top–down) conditions that allow the fire-sensitive vegetation of Patagonia to burn with lower severity both at the landscape scale and for the main vegetation types in the region. Specifically, we ask: (i) how does the probability of burning with high severity change with climate, topography and distance to water bodies, and between vegetation types; (ii) how does burn severity change across Patagonian productivity gradients and does the pattern hold between vegetation types, and (iii) under different weather conditions? First, we propose that at the landscape scale, changes in productivity largely influence burn severity and we expect maximum burn severity at intermediate portions of the productivity gradient (dry forests and shrublands). Under lower productivity conditions (dry conditions, rocky/sandy sites), burn severity is attenuated owing to sparser fuel loads whereas in highly productive fuel-rich forest ecosystems, burn severity is attenuated owing to the dampening effect of microclimatic conditions on fuel moisture produced by denser tree canopies. Second, we propose that the effects of interannual fire weather variability on burn severity will be more pronounced in high-productivity ecosystems by modifying the attenuation effect of fuel moisture conditions. We expect that under dry/warm weather conditions, highly productive fuel-rich forest ecosystems will increase burn severity compared with wet/cool conditions more than dry forests or shrublands.

To answer these questions, we modelled burn severity as a function of biophysical predictors to find spatial opportunities for lower burn severity in the landscape. To understand burn severity patterns across the productivity gradient and under different vegetation types, we used vegetation cover and NDVI (Normalized difference vegetation index) as proxies for fuel characteristics and productivity. Finally, to find temporal opportunities for lower burn severity, we analysed burn severity–productivity patterns under contrasting fire weather conditions.

Methods

Study area

The study area extends 39–44°S along the Patagonian Andes, Argentina (Fig. 1). In this area, the vegetation varies along a steep precipitation gradient that extends from west to east (3000–800 mm year⁻¹) and less markedly along a temperature gradient associated with elevation and slope exposure (Veblen et al. 1992). Here, we focus on six different community types: (1) wet closed canopy forest dominated by Nothofagus dombeyi with an understorey of the bamboo Chusquea culeou; (2) subalpine forests dominated by the deciduous Nothofagus pumilio; (3) dry forests dominated by Austrocedrus chilensis and small trees such as Lomatia hirsuta and Maytenus boaria; (4) alpine vegetation dominated by cushion shrubs; (5) shrublands of resprouting shrubs and small trees; and (6) forest dominated by exotic conifers (Pinus spp. and Pseudotsuga menziesii).

Fig. 1.

Extent of the study area in north-western Argentine Patagonia. Left: study area and fires scars by year. Right: location of the study area in Patagonia and Argentina.

Nothofagus spp. have thin bark and are easily killed by fire even at medium severity (González et al. 2010; Arroyo-Vargas et al. 2019; Landesmann et al. 2021; Franco et al. 2022), although large trees may survive low-intensity surface fires (Kitzberger et al. 2005). All Nothofagus trees in the study area are non-resprouters, and do not have a persistent seed bank or fire-induced germination and thus can only regenerate from unburnt patches that provide seeds. These trees are also heliophilous, have a short post-fire regeneration window and their seeds are wind-dispersed with short dispersal distances – typically no further than 50–100 m. Trees fail to regenerate after severe large fires and forests are replaced by shrublands of resprouting species that often persist in the forest understorey. The conifer A. chilensis is an intermediate shade-tolerant tree able to regenerate even in dense shrublands (Landesmann et al. 2015) and capable of long-distance seed dispersal, having a better chance to recolonise terrain (Landesmann and Morales 2018). With long periods without fire, Austrocedrus eventually overtops and suppresses shrubs (Landesmann et al. 2016). At relatively drier sites and/or after severe fires, shrublands develop dominated by resprouting short trees (e.g. Nothofagus antarctica, L. hirsuta, M. boaria), tall shrubs (e.g. Embothrium coccineum, Diostea juncea) and the bamboo C. culeou (Veblen et al. 2003). These communities are highly prone to burning owing to site and fuel conditions. Moreover, the decurrent multi-stemmed growth forms provide connected fuel ladders, making them highly flammable. Rapid fuel recovery mediated by vigorous post-fire resprouting results in short fire return intervals (Veblen et al. 1992). Many of these shrublands are early stages of post-fire succession that eventually result in closed-canopy forest, but this is often delayed by repeated burning and heavy livestock pressure (Blackhall et al. 2008, 2017; Paritsis et al. 2015).

Results

General drivers of burn severity

During the 20-fire-season period of this study, 74 fires burned 113 521.1 ha. These fires burned mostly as high fire severity (high 104 667.4 ha (92%) vs low 8853.66 ha (8%). High burn severity probability (HBSP) on average was far above 0.5 for almost every covariate, and in general, their effects were small and uncertain (Fig. 2). HBSP peaked at the intermediate portion of the NDVI gradient, reaching a probability of 0.9 of burning with low severity (0.5–0.9; Fig. 2a). The lowest probability was reached at the lowest portion of the gradient (0.77, NDVI < 0.3). The distance to water bodies, although different from zero (Supplementary Table S8), had a negligible effect (Fig. 2b). Intermediate slopes showed lower HBSP than flat terrain or steep slopes (Fig. 2c), and there was also some edge effect, given that near the fire edge, HBSP was lower than further away from it (Fig. 2d). South-facing mountain slopes had lower HBSP than north-facing slopes, and west-facing slopes lower HBSP than east-facing slopes although the difference was very low (Fig. 2e, f). Annual mean temperature showed a hump-shape with HBSP, where high and low annual mean temperature had the lowest HBSPs (Fig. 2g). Annual precipitation had the inverse pattern to temperature; pixels with mid-high annual precipitations had lower HBSP than the ones with low precipitation (Fig. 2h). Shrublands, exotic species plantations and dry forest had the highest HBSP, while wet forest and subalpine forest had the lowest HBSP (Fig. 3).

Fig. 2.

High burn severity probability (HBSP) relationship with (a) productivity (NDVI), (b) distance to water bodies (lakes and rivers), (c) slope,  (d) distance to the nearest fire edge, (e) northerliness, (f) easterliness, (g) annual mean temperature and (h) annual precipitation. Black lines correspond to the predicted global mean partial effects from a hierarchical regression (fixing the remaining covariates at their means), and grey shading to the 95% highest posterior density interval. This figure relates to the 74 fires that occurred between 1999 and 2019.

Fig. 3.

High burn severity probability (HBSP) for different vegetation types across the productivity gradient. AV, alpine vegetation; DF, dry forest; SH, shrubland; EF, exotic forest; SF, subalpine forest; WF, wet forest. Vegetation types are sorted from low to high by their NDVI range. The box represents 50% of the data, lower and upper boundaries correspond to the first and third quartiles (25th and 75th percentiles), and the lower and upper whiskers extend from the boundary to the largest or lowest value no further than 1.5×IQR n (IQR, inter-quartile range) of the predicted probability (posterior mean). Different letters indicate low overlap between posterior distributions of the severity class (overlap < 5%). This figure relates to the 16 fires that have detailed (fine-scale) vegetation cover maps.

Burn severity patterns in native forests and shrublands in relation to productivity and fire weather

Out of the 20 years analysed, 5 years had above-average summer PDSI conditions and 5 below-average summer temperature, and only 2 years had both conditions. In native forests (dry, subalpine and wet forests) and shrublands, HBSP showed little variation at low or intermediate productivity (NDVI between 0.5 and 0.8), and decreased sharply at high productivity (NDVI > 0.8, Fig. 4a). When fire years were stratified between dry (PDSI below average) and wet (PDSI above average) summers and warm vs. cool summers (temperature above and below average, respectively), divergent patterns were found. HBSP dropped by 0.1–0.2 at NDVIs > ~0.65 during cool summers (Fig. 4b) as a well as during wet summers (Fig. 4c), whereas the HBSP-NDVI variation during warm and dry summers largely resembled that of all fires taken together (Fig. 4).

Fig. 4.

High burn severity probability patterns for pixels of burned native forests and shrublands as a function of productivity (NDVI) during: (a) all fires regardless of the summer weather condition; (b) fires that occurred in summers with below (blue bars) and above (red bars) average temperature; and (c) fires that occurred during summers with below average (green bars, wet) and above average (orange bars, dry) summer Palmer Drought Severity Index (PDSI). Points correspond to median values and the lower and upper whiskers extend to 1.5×IQR n (IQR, inter-quartile range) of the predicted high burn severity probability (posterior mean) for all pixels of the 74 fires that occurred between 1999 and 2019.

Wet forests followed by dry forests showed the largest reductions in HBSP during wet years (Fig. 5b) as well during cool years (Fig. 6). HBSP was reduced during cool or wet summers in wet and dry forests, from 0.9 to 0.53 and from 0.95 to 0.69, respectively, whereas for subalpine forests and shrublands, this probability always remained high (0.9 and ~0.95, Fig. 5a, b). However, when looking at the whole range of productivity within each forest type, subalpine forests reduced HBSP from ~0.95 to ~0.6 during cool summers only at higher productivity, but showed no reduction in HBSP at NDVI < 0.75 (Fig. 6). These lower NDVI locations coincide with north- and west-facing slopes where there is no apparent divergence in HBSP between cool and warm summers (Supplementary Figs S3 and S4). In contrast, wet forests, dry forests and shrublands show a more homogeneous reduction in HBSP during cool summers throughout their respective NDVI and topographic aspect ranges (Fig. 6).

Fig. 5.

High burn severity probability for major vegetation cover classes in different fire weather scenarios. High burn severity probability when (a) summer temperature, and (b) summer Palmer Drought Severity Index (PDSI) were above or below average. Points correspond to median values and the lower and upper whiskers extend to 1.5×IQR n (IQR, inter-quartile range) of the predicted high burn severity probability (posterior mean) for all pixels of the 74 fires that occurred between 1999 and 2019. Bars represent the relative frequency of each fire weather scenario under each vegetation cover class.

Fig. 6.

High burn severity probability as a function of NDVI for different vegetation types for fires that occurred under different summer temperature conditions. Points correspond to median values and the lower and upper whiskers extend to 1.5×IQR n (IQR, inter-quartile range) of the predicted high burn severity probability (posterior mean) for all pixels of the 74 fires that occurred between 1999 and 2019.

Discussion

Biomass-rich and fire-sensitive communities of north-western Patagonia typically burn at high severity. Of the 113 521 ha burned during 20 fire seasons, only 8% was classified as low severity, suggesting that certain very specific fuel, weather and topographic conditions must be met to burn at low severity. Spatially, these conditions are related to low-productivity environments with low fuel loads/continuity (e.g. sandy/rocky outcrops, tree lines), wet environments that generate highly productive wet or subalpine forests, and moist topographic aspects (south and east-facing slopes). In addition, burn severity is also attenuated along fire edges where fires were less intense and eventually stopped burning. Temporally, fires burning in cool/wet summer conditions in productive forest ecosystems tend to reduce the probability of high burn severity. The existence of these microsites where burn severity is attenuated is crucial for the persistence of fire-sensitive taxa in the landscape (in our case, obligate-seeder trees) as they allow for forests’ natural regeneration and persistence in the landscape.

Our results suggest that spatial attenuation of burn severity in systems subject to high burn severity can derive from the interplay between different bottom–up and top–down controls: (1) fuel amount and condition: fuels become sparser when resources for plants are reduced (limiting water, rocky/sandy substrates) and thus fires may burn at lower intensity, leaving unburned patches (Landesmann et al. 2015; Meddens et al. 2018). Fires also burn with lower intensity where fuels are subject to moister environments or where vegetation structure allows for less severe fires (Bradstock et al. 2010; Cawson et al. 2017; Zylstra et al. 2023). (2) Fire breaks: fire intensity decreases near perimeters owing to topographic restrictions like steep downslopes, fire flank areas were fires slowly creep in on rolling debris, dominant winds restrictions (e.g. windward, fire tail areas), or where and when fires encounter man-made or natural suppression conditions (fire lines, heavy rain). (3) Weather: areas that burn under cool/wet summer conditions have a higher probability of burning with low burn severity than the ones burning under hot/dry summer conditions, thus increasing the chances of canopy tree survival, particularly in productive fuel-rich forests (less and sparser surface fuels, wet and dark microenvironments, low connectivity) (Parks et al. 2018).

As predicted, our results show that the relationship between productivity and burn severity followed a hump-shaped relationship across different vegetation types, with dry forests, shrublands and exotic plantations displaying the highest burn severity probability (Fig. 3). In shrublands and dry forests, microclimatic conditions favour fuel desiccation (Paritsis et al. 2015; Barberá et al. 2023), and fuels are more flammable than in broadleaf forests (Blackhall et al. 2015, 2017; Blackhall and Raffaele 2019). Exotic conifer plantations are highly flammable, promoting highly active canopy-consuming fires (Paritsis et al. 2018). In all these vegetation types, dry fine fuel loads are relatively high and fuels have high vertical connectivity (climbers, tall shrubs, untreated plantations) that allow efficient crowning, explaining the pattern we see in our results (Blackhall et al. 2015, 2017; Paritsis et al. 2015, 2018; Tiribelli et al. 2018). Weather does not substantially modify severity when shrublands burn (Figs 5, 6, Supplementary Fig. S2), possibly owing to high levels of fine fuels that do not require long rain-free periods to dry out (Barberá et al. 2023). Despite high burn severity, two of these systems show high levels of resilience. Shrublands have a high proportion of resprouting species that rapidly dominate the post-fire stand (Veblen and Lorenz 1987, 1988) and field measurements show the ability of these species to readily resprout after severe fires and even replace post-fire forest communities (Kitzberger et al. 2016a; Landesmann et al. 2021; Franco et al. 2022). Exotic conifers are highly adapted to fire and rapidly recolonise burned plantations (Raffaele et al. 2016; Franzese et al. 2022). Finally, dry forests (e.g. Austrocedrus-dominated forests) are more fire sensitive, but the ability to grow and survive on dry unproductive rock outcrops (Landesmann et al. 2015) and relative high shade tolerance that enable the species to regenerate under a resprouting shrub canopy (Landesmann et al. 2016) allow long term persistence of the species in the landscape despite overall high levels of burn severity (Gowda et al. 2012).

The lower end of the productivity gradient coincides with rocky areas typical of alpine vegetation as well as some rocky, stony and sandy outcrops at mid elevations (Fig. 3). Here, low burn severity is possibly driven by lower fuel loads and continuity, creating islands of unburned patches promoting some fire resilience. The most productive forest types, such as wet forest and subalpine forests, showed the lowest levels of burn severity (Fig. 3). With high precipitation (>1500 Mean annual precipitation) or south- and east-facing slopes of moister higher elevations (1000–1600 m above sea level), the vegetation structure of dense, highly productive forest allows for darker and moister microclimatic conditions. Here, the discontinuity between understorey and tall canopy fuels (Blackhall et al. 2017; Tiribelli et al. 2018) reduces the probability of fire crowning, particularly during wetter/cooler fire seasons (Figs 5, 6, Supplementary Fig. S2). Under these conditions, and if forests have not burned during the last 100–150 years, larger thick-barked trees can survive and regenerate as a dense new forest cohort (Veblen and Lorenz 1987). If in contrast, fires burn under dry/hot summer conditions (Figs 5, 6, Supplementary Fig. S2), and forests are younger owing to the more frequent occurrence of past fires, trees cannot survive high fire intensity and may transition towards shrublands (Kitzberger et al. 2016b; Tiribelli et al. 2018; Landesmann et al. 2021). This pattern holds along the entire productivity gradient for wet forest but not for subalpine forests. Burn severity in subalpine forests of low productivity was not affected by weather (Fig. 6, Supplementary Fig. S2). These areas are located on north- and west-facing slopes, suggesting a strong topographic control on burn severity in subalpine forests (Supplementary Figs S3, S4).

Although forests may burn with low severity under certain climatic, productivity and topographic conditions, they are vulnerable communities even under these low burn severity scenarios if the first few years post-fire are not masting years for tree species, if cattle are introduced (Blackhall et al. 2008, 2017; Arroyo-Vargas et al. 2019), or if the climatic conditions post-fire do not allow for seed establishment or seedling survival (Kitzberger et al. 2000, 2005). This would lead to increases in burn severity due to vegetation shifts. As in other parts of the world (Karna et al. 2021; Lindenmayer et al. 2021), this positive feedback where severe fires beget severe fires ultimately increases shrubland areas to the detriment of forest area. Thus, the present empirical results underscore the importance of interannual climatic variability in modulating fire severity in northern Patagonia (Holz and Veblen 2011; Veblen et al. 2011) and elsewhere (Abatzoglou et al. 2017; Keyser and Leroy Westerling 2017), and alert us to possible increases in overall fire severity under the ongoing trends and expected scenarios of increasing aridity and temperature in the region (CONAMA 2006; IPCC 2013).

Here, we have shown several mechanisms by which fire regimes that are dominated by high burn severity can, under certain conditions, burn at lower severity, thus increasing forest resilience. These processes have acted in the past, producing the present north-Patagonian forest landscape (Veblen and Lorenz 1987, 1988; Gowda et al. 2012). Ongoing and projected future climate changes for this region suggest that these windows of opportunity for fire-sensitive forest taxa will lessen, thus promoting transitions towards alternative states dominated by more flammable species (Mermoz et al. 2005), and thereby positively reinforcing a regime dominated by frequent and severe fires (Kitzberger et al. 2016b; Tiribelli et al. 2019). Management and mitigation of these projected changes should focus on finding ways to preserve sites in the landscape that promote forest resilience attenuating burn severity. Spatially, we have identified certain site conditions where fires tend to burn at lower severity. Fire crews could take advantage of this information and use arrays of these a priori known locations to concentrate suppression actions as a way to more effectively fight fires and to secure the integrity of key forest patches (e.g. fire refugia on rocky outcrops (Landesmann and Morales 2018)) that will likely function as seed sources for the regeneration of obligate-seeder trees across vast areas burned at higher severity.

Although using temporal windows of wetter/cooler fire seasons seems tempting as opportunities for prescribed burning practices, our data suggest that it is still a risky recommendation as fires spreading under these conditions still have ~50% chance of becoming severe canopy-destroying fires. Nevertheless, identifying temporal windows of severity attenuation can be helpful to guide fire suppression actions (e.g. applying let-burn policies under certain fire weather condition assuming that considerable portions of the landscape will burn at lower severity). Under future climatic scenarios where the probability of fire occurrence will increase (Kitzberger et al. 2022) and opportunities for severity attenuation will become scarcer or more spatially restricted, it is imperative to develop efficient science-based fire management strategies that will help decrease burn severity and favour natural regeneration of closed-canopy broadleaf forests.

Supplementary material

Supplementary material is available online.

Data availability

Data are available on request from the authors.

Conflicts of interest

The authors declare no conflicts of interest.

Declaration of funding

Funding was provided by Ministerio de Ciencia, Tecnología e Innovación Productiva (Argentina), Agencia Nacional de Promoción Científica y Tecnológica, grants PICT-2017-2142 and PICT-2019-406.

Author contributions

F. Tiribelli, T. Kitzberger and J. Paritsis conceived the ideas and designed the methodology; I. Barberá developed the GEE code to calculate severity indexes and calculate weather anomalies; T. Kitzberger, J. Paritsis and F. Tiribelli did the Google Earth pro vegetation classification. F. Tiribelli analysed the data; F. Tiribelli led the writing of the manuscript. All authors interpreted results, contributed critically to the drafts and gave final approval for publication.