land Article The Role of Recent (1985–2014) Patterns of Land Abandonment and Environmental Factors in the Establishment and Growth of Secondary Forests in the Iberian Peninsula Marina Palmero-Iniesta 1, *, Josep Maria Espelta 1,2 , Mario Padial-Iglesias 3 , Òscar Gonzàlez-Guerrero 3 , Lluís Pesquer 1,3 , Cristina Domingo-Marimon 1,3 , Miquel Ninyerola 4 , Xavier Pons 3 and Joan Pino 1,2 1 Centre de Recerca Ecològica i Aplicacions Forestals (CREAF), Bellaterra (Cerdanyola del Vallès), 08193 Catalonia, Spain; Josep.Espelta@uab.cat (J.M.E.); l.pesquer@creaf.uab.cat (L.P.); cristina.domingo@uab.cat (C.D.-M.); joan.pino@uab.cat (J.P.) 2 Departament de Biologia Vegetal, Biologia Animal i Ecologia, Universitat Autònoma Barcelona, Bellaterra (Cerdanyola del Vallès), 08193 Catalonia, Spain 3 Grumets Research Group, Departament Geografia, Universitat Autònoma de Barcelona, Bellaterra (Cerdanyola del Vallès), 08193 Catalonia, Spain; mario.padial@uab.cat (M.P.-I.); Oscar.Gonzalez.Guerrero@uab.cat (Ò.G.-G.); xavier.pons@uab.cat (X.P.) 4 Grumets Research Group, Departament Animal and Plant Biology and Ecology, Universitat Autònoma de Barcelona, Bellaterra (Cerdanyola del Vallès), 08193 Catalonia, Spain; miquel.ninyerola@uab.cat * Correspondence: m.palmero@creaf.uab.cat   Abstract: Farmland abandonment has been a widespread land-use change in the Iberian Peninsula Citation: Palmero-Iniesta, M.; since the second half of the 20th century, leading to the establishment of secondary forests across the Espelta, J.M.; Padial-Iglesias, M.; region. In this study, we aimed to address changes in the recent (1985–2014) emergence patterns of Gonzàlez-Guerrero, Ò.; Pesquer, L.; these forests and examine how environmental factors affected their growth by considering differences Domingo-Marimon, C.; Ninyerola, in leaf-habit types. We used a combination of Landsat-derived land-cover maps and aboveground M.; Pons, X.; Pino, J. The Role of biomass (AGB) maps from the European Space Agency to assess the secondary forest establishment Recent (1985–2014) Patterns of Land and growth, respectively, in the study region. We also obtained a set of topographic, climatic and Abandonment and Environmental landscape variables from diverse GIS layers and used them for determining changes over time in Factors in the Establishment and Growth of Secondary Forests in the the environmental drivers of forest establishment and AGB using general linear models. The results Iberian Peninsula. Land 2021, 10, 817. highlight that secondary forest cover was still increasing in the Iberian Peninsula at a rate above the https://doi.org/10.3390/land10080817 European average. Yet, they also indicate a directional change in the emergence of secondary forests towards lower and less steep regions with higher water availability (mean rainfall and SPEI) and less Academic Editor: Ilan Stavi forest cover but are subjected to greater drought events. In addition, these environmental factors differentially affect the growth of forests with different leaf-habit types: i.e., needleleaf secondary Received: 15 June 2021 forests being less favoured by high temperature and precipitation, and broadleaf deciduous forests Accepted: 31 July 2021 being most negatively affected by drought. Finally, these spatial patterns of forest emergence and the Published: 4 August 2021 contrasting responses of forest leaf-habits to environmental factors explained the major development of broadleaf evergreen compared to broadleaf deciduous forests and, especially, needleleaf secondary Publisher’s Note: MDPI stays neutral forests. These results will improve the knowledge of forest dynamics that have occurred in the with regard to jurisdictional claims in Iberian Peninsula in recent decades and provide an essential tool for understanding the potential published maps and institutional affil- effects of climate warming on secondary forest growth. iations. Keywords: drought; forest expansion; land-cover change; land-use change; rural abandonment Copyright: © 2021 by the authors. Licensee MDPI, Basel, Switzerland. 1. Introduction This article is an open access article distributed under the terms and Changes in land use and land cover are one of the main drivers of global environmental conditions of the Creative Commons change given that they affect the properties of land surfaces, the provision of ecosystem Attribution (CC BY) license (https:// services [1] and, ultimately, the Earth’s energy balance and biogeochemical cycles [2,3]. creativecommons.org/licenses/by/ Although deforestation is still one of the main manifestations of changes in land use and 4.0/). land cover, the net rate of forest loss decreased substantially worldwide during 1990–2020 Land 2021, 10, 817. https://doi.org/10.3390/land10080817 https://www.mdpi.com/journal/land Land 2021, 10, 817 2 of 18 due to a reduction in deforestation in certain countries and active and passive afforestation in others [4]. The expansion of secondary forests, i.e., forests that are established in areas with another previous land use, is a phenomenon that has occurred especially in the Northern Hemisphere during the last few centuries [5]. Secondary forests that have been established on former farmland may constitute a large proportion of the total forested area in many regions, e.g., 20–25% in Spain [6,7] and more than 50% in China and Costa Rica [8]. The expansion of secondary forests can have both positive and negative effects on landscape structure and ecological functions. For example, forest cover increase is affecting biodiversity conservation in Europe, with a generalised recovery of forest species, includ- ing threatened species targeted in conservation initiatives [9,10], but also a rarefaction and local extinction of species that are typical of open habitats [10–12]. The expansion of secondary forests may also increase the functional connectivity between forests (e.g., seed dispersal potential) and may facilitate migration and gene flow between tree populations in response to climate change [13], while potentially increasing the propagation of large wildfires, especially in southern Europe [14]. Apart from these environmental changes, the expansion of secondary forests has gained much attention owing to their role in carbon accumulation [15,16]. Fuchs et al. [17] determined that afforestation and cropland abandon- ment made the greatest contribution to carbon sequestration during 1950–2010 in Europe, while Pan et al. [16] reported that the C sink of China’s forests increased by 34% during 1990–2007 primarily due to newly planted forests. Finally, Vila-Cabrera et al. [6] showed that secondary forests that were established from 1956 onwards represent 22% of the total C pool in Iberian forests. Furthermore, it was suggested that forests established on former agricultural land may differ from long-established forests in terms of species composition and their structural and functional characteristics [18]. These forests may benefit from land- use legacies since their soils tend to be richer in nutrients [19] and exhibit greater enzymatic activity [20], which could explain why some secondary forests have higher growth rates than long-established forests: 35% greater plant biomass [21] and 25% higher growth [6], but they may also be more sensitive to climate changes due to differences in functional attributes (lower wood density [22] and finer root morphology [21,23]). This may be of spe- cial relevance under climate change, as drought and heat-induced stress in trees could lead to a reduction in tree growth [24], an increase in mortality rates [25,26] and an increase in forest vulnerability [27,28]. Ultimately, this may cause directional changes in forest compo- sition towards forests with a greater abundance of drought-tolerant species [29,30], e.g., an increase in broadleaf over needleleaf forests [31,32]. On the other hand, the spatial pattern of forest expansion is mostly determined by farmland abandonment, which is supposed to be concentrated in marginal and low productivity areas [33]. Ultimately, the biophysical conditions involving forest expansion may influence the recruitment, growth and mortality of tree species affecting the whole plant community [34–36], which may determine sec- ondary forest productivity and its capacity for carbon accumulation [15,16,34,37]. Although some studies have suggested that the expansion of secondary forests may contribute in a relevant way to carbon sequestration [6,15–17], little is known about the influence of the expansion patterns on the extent of secondary forest growth. The purpose of this study was thus to determine the recent patterns (1985–2014) of secondary forest establishment and growth in the Iberian Peninsula and examine how environmental factors affect them by considering the main forest leaf-habit types. The Iberian Peninsula is an especially appropriate study area given, on the one hand, the expansion of secondary forests that occurred there during the second half of the twentieth century [6,7,33] and, on the other, its diverse topoclimatic conditions and the increasing number of drought episodes in recent decades [38]. Ultimately, this may help to evaluate the potential of recently established forests to climate change mitigation strategies that are aimed at maximising carbon sequestration. We thus established three main hypotheses: (i) the distribution and magnitude of cropland abandonment are closely determined by the environmental characteristics that limit their productivity and, as such, secondary forests will appear first in poorly productive Land 2021, 10, 817 3 of 18 areas (i.e., areas with unfavourable climatic and topographic conditions) since the most productive areas will be the last to be abandoned; (ii) due to the recent increase in drought frequency and severity in recent decades, a greater amount of drought-tolerant broadleaf secondary forests than needleleaf forests will develop; and (iii) drought events will have a greater impact on needleleaf secondary forest growth than on broadleaf secondary forest growth. 2. Material and Methods 2.1. Study Area The study was performed in the Iberian Peninsula (202,067 km2 ; SW Europe), a region with great climatic and topographic diversity (Figure 1) and three biogeographic regions, namely, Mediterranean, Atlantic and Alpine. The mean annual temperatures range from 18 ◦ C (on the southern coast) to 1 ◦ C (in mountainous areas), while the mean annual rainfall varies from 340 mm to over 2400 mm (Topoclimatic Drought Atlas of the Spanish Iberian Peninsula; Domingo-Marimon, 2016) [39]. The large number of mountain ranges (from sea level to more than 2600 m a.s.l.) and steep coastal-inland gradients also contribute to the great environmental heterogeneity of this region. Forests cover 35% of the Iberian Peninsula (EEA Report No. 5/2016). Broadleaf and needleleaf evergreen species are dominant in the study area; broadleaf deciduous forests are less frequent and are mostly found in the Atlantic region and mountainous areas (see Figure 1). The main needleleaf evergreen species in Mediterranean areas are Pinus halepensis and P. pinea, while P. nigra occurs in inland upland areas. Abies alba and P. uncinata are common in Alpine areas in the Pyrenees, while P. sylvestris is found widely from upland Mediterranean to Alpine areas. Broadleaf evergreen forests are dominated by Quercus ilex and Q. suber in lowland sites up to the limit of montane habitats, while Q. coccifera becomes increasingly common as continentality rises. Broadleaf deciduous forests mainly consist of Q. humilis, Q. faginea and Q. pyrenaica in Mediterranean lowlands and uplands, and Fagus sylvatica in the Atlantic region. 2.2. Data Sources To detect the secondary forest emergence, we used Landsat images that were pro- cessed and classified by the Grumets Research Group at the Autonomous University of Barcelona (www.ogc3.uab.cat/acapi/wms/USOS/index.htm; accessed on 10 November 2020, [40,41]). We ensured that the maps were only generated for those years with enough good-quality data and periodicity (i.e., from the mid-1980s). In order to have enough images for map classification (i.e., images that enabled us to correctly identify phenological variation in land cover categories), we worked with quinquennial periods, thus generating maps for every quinquennia from 1985 to 2014. From these maps, we selected a set of available scenes from the Landsat orbits (197, 198, 200, 201 and 202) that were distributed along both latitudinal and longitudinal gradients that were representative of the climatic variability in the Iberian Peninsula (Figure 1). Secondary forest growth was estimated by using the global maps of aboveground biomass from 2017 (100 m resolution) that were generated by the European Space Agency (ESA) Climate Change Initiative Biomass project [42]. Aboveground biomass (AGB) is defined as the amount of living biomass (organic matter) that is stored in vegetation above the soil, including stems, stumps, branches, bark, seeds and foliage, and is expressed as the weight of dry matter per unit ground area (i.e., Mg/ha−1 ) [43]. Land 2021, Land 10, x FOR 2021, PEER REVIEW 10, 817 4 of 18 4 of 20 Figure 1. Secondary Figure forest 1. Secondary distribution forest during distribution during2010–2014 2010–2014 by the themain mainforest forest leaf-habit leaf-habit types types in Iberian in the the Iberian Peninsula. Peninsula. 2.2.2.3. DataDetecting SourcesSecondary Forest Establishment and Growth throughout the Study Period ToWe detected detect a new secondary the secondary forestforest that hadwe emergence, appeared in the study used Landsat period images by were that com- pro- bining maps relating to the different quinquennia. We defined cessed and classified by the Grumets Research Group at the Autonomous University a secondary forest as a of forest that was established on cropland or grassland and is a product of forest expansion, Barcelona (www.ogc3.uab.cat/acapi/wms/USOS/index.htm; accessed on 10 November that is, forests established in areas that for many years previously had been dominated 2020, by [40,41]). a differentWe ensured type of landthat the maps use [44]. We did were only generated not consider for those the transition fromyears with enough shrubland to good-quality data and periodicity (i.e., from the mid-1980s). In order forest as a secondary forest since this was explored by previous studies [45] and because to have enough im- ages thisfor map classification transition (i.e., images mostly corresponds to thethat enabled us regeneration of to correctlyfelled previously identify phenological or disturbed variation in land cover categories), we worked with quinquennial periods, thus generat- forests [7,46]. ing maps Forfor each quinquennium, every quinquennia we from identified 1985pixels belonging to 2014. to forest From these landwe maps, cover that was selected a set of cropland available or grassland scenes from the in the previous Landsat quinquennium orbits (see classes (197, 198, 200, 201 and in 202) the Online Resource that were 1, distributed Table S1). To avoid including pixels that had been falsely classified along both latitudinal and longitudinal gradients that were representative of the climaticas new forests, we only selected forests in a quinquennium that appeared as forests in two subsequent quinquennia variability in the Iberian Peninsula (Figure 1). (i.e., cropland or grassland during 1995–1999 that was classified as forest in both 2000–2004 Secondary forest growth was estimated by using the global maps of aboveground and 2005–2009). To account for potential differences that were related to leaf-habit types, biomass from 2017 we determined (100 m surface resolution) changes that were and annual rates generated of change for bythethemain European Space leaf habit Agency types: (ESA) Climate deciduous, (i) broadleaf Change Initiative Biomass (ii) broadleaf project evergreen and [42]. (iii)Aboveground biomass (AGB) is de- needleleaf evergreen. fined as the amount of living biomass (organic matter) that is stored in vegetation above the soil, including stems, stumps, branches, bark, seeds and foliage, and is expressed as the weight of dry matter per unit ground area (i.e., Mg/ha−1) [43]. Land 2021, 10, 817 5 of 18 Finally, the growth of the selected secondary forests was assessed by estimating their biomass in relation to their age. Biomass was obtained by merging the detected secondary forests during 2010–2014 with the ESA aboveground biomass (AGB) layer. We previously resampled the secondary forest dataset at a pixel size of 100 m (using the modal criterion of the most represented value) to fit with the resolution of the biomass dataset. Forest age was assessed in years as the difference between the final quinquennium (2010–2014) and the quinquennium of forest establishment. 2.4. Environmental Drivers of Secondary Forest Establishment and Growth To assess the environmental context of the secondary forests in the study area, we selected a set of potential drivers of forest expansion and growth [45,47–50]. Based on previous work (e.g., [45,50–52]), we chose a series of climatic, topographic and landscape factors that are known to affect forest composition, structure and growth. 2.4.1. Climatic Variables To characterise the main climatic conditions, we used the Topoclimatic Drought Atlas of the Spanish Iberian Peninsula [39], which includes monthly aggregates from 1950–2015 at a spatial resolution of 100 m of rainfall, mean temperature and the Standardised Precipi- tation Evapotranspiration Index (SPEI) as an indicator of water deficit. We calculated the mean temperature (◦ C) and rainfall (mm) annually and seasonally for 1950–2015 (hereafter, the historic climate) and for the time that the secondary forest grew (hereafter, the recent climate). We used SPEI values that were calculated at a timescale of 12 months, which is an optimal and appropriate scale for studying the long-lasting dry periods that characterise hydrological droughts [53,54]. We defined a drought event as a period of consecutive months in which the SPEI values were equal to or lower than −1, as suggested in the literature [53,55,56]. For a statistical quantification of drought episodes, we calculated the commonly used drought parameters [55,57] for both the historic and recent climatic periods, i.e., drought frequency (number of episodes during a period divided by period; events/year), mean drought duration (average duration of drought events in the period; months), mean drought intensity (average of the mean SPEI values during the period of drought events) and mean drought severity (average of the accumulated SPEI values during the period of drought events). 2.4.2. Topography We used the digital elevation model of the Iberian Peninsula at a spatial resolution of 90 m generated by the Kraken group from the University of Extremadura [58] to characterise the mean elevations (m) and slopes (degree) of emerging secondary forests. 2.4.3. Forest Cover Previous studies have observed that the amount of forest cover in the landscape has a positive influence on forest expansion and determines the distribution patterns of secondary forests [45,50]. Thus, we determined the percentage of forest land cover at the beginning of the study period by counting the number of forest pixels in a radius of 1 km of each secondary forest pixel. Finally, we resampled the spatial resolution of the abovementioned land-cover raster datasets from 30 m to 100 m to coincide with the resolution that is employed in most environmental datasets generated using the majority criterion (assigning the value of each pixel based on the most abundant value). We overlapped multiple environmental maps, after resampling and changing the coordinate systems wherever necessary, to determine the environmental values in all cells classified as secondary forests. 2.5. Statistical Analyses As for the secondary forest establishment, we first performed ANOVA tests to de- termine whether the environmental conditions of established secondary forests differed between forest leaf-habit types (broadleaf deciduous, broadleaf evergreen and needleleaf Land 2021, 10, 817 6 of 18 evergreen). Data were log-transformed when necessary to meet the assumptions of nor- mality. Where significant differences occurred (p < 0.05), comparisons between means were performed using Tukey’s HSD multiple comparison test. We then performed a general linear model (GLM) to test the association of the time (quinquennium) of the establishment of secondary forests with the environmental variables described above. For this analysis, we took the historic climatic variables from the period 1950–2015 as being representative of the climatic conditions in the area. We decided to work with the absolute value of the severity and intensity of drought factors (all values were negative) so that the higher the value, the greater the severity or intensity of the drought, which facilitated the interpretation of the effect in the model. To prevent multicollinearity, we produced a correlation matrix with a Spearman rank test (Electronic Supplementary Materials 2, Figure S1) to avoid using highly correlated variables (r < |0.7|). Thus, when variables were correlated, we chose the best variable by selecting the one with the strongest effect on the response variable. The simplest general linear models were selected following a dredge procedure using the MuMIn R package [59], which removed non-significant vari- ables from the general model, to assess significant changes in model predictions using the Bayesian information criterion (BIC). From the models with a difference in BIC relative to BICmin < 2, we chose the most parsimonious model by selecting the model with the fewest predictor variables following the procedure described in Crawley [60]. Then, we tested for spatial autocorrelation of the model residuals using Moran’s I Autocorrelation Index [61] to ensure sampling independence (Electronic Supplementary Materials 1, Table S2; [62]). As for forest growth, we first performed ANOVA tests for each forest quinquennium to determine whether AGB differed between forest leaf-habit types and performed pair-wise comparisons of means using Tukey’s HSD tests wherever significant differences occurred (p < 0.05). We also performed a GLM to test the association of AGB with environmental variables and age (midpoint of each quinquennium). For this analysis, we used the recent climatic variables calculated from the year the secondary forest became established to the end of the study period (2015) to describe the climatic conditions in which secondary forests grew. In this case, we also used absolute values of the drought severity and intensity factors. We employed the same criteria to prevent multicollinearity and avoid highly correlated vari- ables used in the abovementioned model (Electronic Supplementary Materials 2, Figure S2). Furthermore, we considered forest leaf-habit as a factor and the environmental interactions with the forest leaf-habit. The simplest general linear model was selected using the same criterion as described above. Then, as for the model analysing forest establishment, we tested for spatial autocorrelation of the model residuals using Moran’s I Autocorrelation Index (Electronic Supplementary Materials 1, Table S3). 3. Results 3.1. Secondary Forest Establishment The total forest area (i.e., the balance between gains and losses) increased by 7.7% during 1985–2014, which represents a mean annual increase of 0.31%. Broadleaf deciduous (BD) forests showed the highest increase from their initial area (30.3% during the study period, 1.21% annually). Broadleaf evergreen (BE) forests increased by 5.2% (annual increase 0.21%), while needleleaf evergreen (NE) forests showed no changes in their area. Secondary forest establishment occurred at an annual rate of 0.60 ± 0.15% (mean ± SE of quinquennia). BD forests had the highest annual establishment rate (1.09 ± 0.21%), followed by BE (0.81 ± 0.32%) and NE (0.24 ± 0.11%) forests. By the end of the study period, 14.8% of the forest area corresponded to secondary forests that had emerged during the 1985–2014 period (54.2% of BD, 31.1% of BE, and 14.7% of NE forests; see Figure 2 for the spatial distribution of secondary forests by forest leaf-habit. Land2021, Land 2021,10, 10,817 x FOR PEER REVIEW 78of of18 20 Figure 2. Secondary forest distribution and area (percentage of 10 km pixels) during 1985–2014 for the main forest leaf-habit Figure types in2.the Secondary forest distribution Iberian Peninsula: broadleafand area (percentage deciduous, broadleaf of 10 km pixels) evergreen during 1985–2014 and needleleaf evergreen.for the main forest leaf- habit types in the Iberian Peninsula: broadleaf deciduous, broadleaf evergreen and needleleaf evergreen. Secondary forest establishment occurred in areas with a mean (± SE) elevation of 667 ±Secondary 0.8 m, a slope 6.2 ± 0.0% and forestofestablishment occurred in areas a previous treewith cover of 30.3(±±SE) a mean elevation 0.05%. of 667 The mean ± 0.8 m, aconditions climatic slope of 6.2of± these 0.0% andareasa previous tree cover were: annual 13.6 ±The of 30.3 ± 0.05%. temperature: ◦ C, climatic mean 0.01 annual precipitation: conditions of these710 ±areas 0.49 were: mm, 0.46 annual± 0.00 droughts/year temperature: with°C, 13.6 ± 0.01 4.74 ± 0.00 annual months of precipitation: duration, 710 ± 0.49intensity mm, 0.46of± SPEI values: −1.30 ± 0.00 droughts/year 0.004.74 with and±severity of SPEI 0.00 months values: −6.91 of duration, ± 0.00. intensity of BD SPEIsecondary forests values: −1.30 became ± 0.00 andestablished severity ofinSPEI areasvalues: with higher −6.91 ±average 0.00. BDannual precipitation secondary forests rates, became steeper slopes and established more in areas forest with coveraverage higher but with a lower annual mean annual precipitation temperature rates, than steeper slopes BE andand moreNEforest forests (Table cover but1). They with also became a lower established mean annual in areasthan temperature withBE lower andelevations, NE forests on less 1). (Table steep They slopes also and lower became average annual established in areasprecipitation, but with higher with lower elevations, on lessmean steepannual slopes temperatures and greater drought intensity and severity than BD and lower average annual precipitation, but with higher mean annual temperatures and and NE forests. NE forests became established in areas of greater environmental variability greater drought intensity and severity than BD and NE forests. NE forests became estab- with less drought duration, intensity lished in areas and severity of greater than either environmental BD or BEwith variability forests. less drought duration, intensity and severity than either BD or BE forests. Land 2021, 10, 817 8 of 18 Table 1. Mean (±SE) values of environmental variables for sites with secondary forests and differ- ences between secondary forest leaf-habit: broadleaf deciduous (BD), broadleaf evergreen (BE) and needleleaf evergreen (NE). Letters show significantly different means at p < 0.05. Secondary Forest Environment Elevation (m) Slope (%) Forest cover (%) All 667 ± 0.8 6.17 ± 0.01 30.3 ± 0.05 BD 681 ± 1.5 a 8.26 ± 0.02 a 36.7 ± 0.10 a BE 666 ± 1.1 b 5.25 ± 0.02 b 27.6 ± 0.06 b NE 684 ± 1.8 a 6.05 ± 0.03 c 29.2 ± 0.11 b Mean annual Annual precipitation Drought frequency temperature (◦ C) (mm) (event/year) All 13.6 ± 0.01 710 ± 0.49 0.46 ± 0.00 BD 12.5 ± 0.01 a 895 ± 0.87 a 0.46 ± 0.00 a BE 14.0 ± 0.01 b 654 ± 0.59 b 0.46 ± 0.00 a NE 13.7 ± 0.01 c 623 ± 1.00 c 0.48 ± 0.00 b Drought duration Drought severity Drought intensity (months) All 4.74 ± 0.00 −6.91 ± 0.00 −1.30 ± 0.00 BD 4.72 ± 0.00 a −6.88 ± 0.00 a −1.29 ± 0.00 a BE 4.79 ± 0.00 b −6.98 ± 0.00 b −1.31 ± 0.00 b NE 4.60 ± 0.00 c −6.74 ± 0.00 c −1.30 ± 0.00 c SPEI Longitude Latitude All 0.004 ± 0.000 435,901 ± 440 4,453,824 ± 369 0 BD 0.005 ± 0.000 a 552,225 ± 786 a 4,604,454 ± 642 a BE 0.005 ± 0.000 b 349,954 ± 535 b 4,401,588 ± 436 b NE 0.004 ± 0.000 c 526,596 ± 899 c 4,404,518 ± 733 c The best model describing the environmental factors that were associated with the time of establishment of the secondary forests during 1985–2014 is shown in Figure 3. The time of establishment of the secondary forests (quinquennial) was positively associated with the mean summer precipitation, mean SPEI and mean drought intensity, but negatively associated with slope and forest cover. Significant interactions of forest leaf-habit with some environmental factors in the model indicated that, over time, both BE and NE were established in areas with lower elevation, lower forest cover areas and lower drought intensity compared with BD forests, while BE forests were established in less steep areas than both NE and BD forests. An autocorrelation test (Electronic Supplementary Materials 1, Table S2) detected very low yet significant values of the Moran’s Index on the study sample and its model residuals (0.071 and 0.027, respectively). 3.2. Secondary Forest Growth The aboveground biomass (AGB) of the secondary BD forests had mean values (±SE) ranging from 61.3 ± 1.1 Mg ha−1 in the youngest forests (5 years) to 101.5 ± 0.9 Mg ha−1 in the oldest (25 years). These values were higher than those of BE and NE forests for most age groups (Figure 4). By contrast, secondary BE forests had the lowest AGB values for most age groups, ranging from 35.0 ± 0.7 Mg ha−1 to 58.3 ± 0.6 Mg ha−1 . Finally, NE secondary forests had mean AGB values ranging from 37.3 ± 1.4 Mg ha−1 to 77.9 ± 0.9 Mg ha−1 . The GLM model for the effect of local environmental factors on the growth of sec- ondary forests explained 42% of the variance (Figure 5). The growth of secondary forests was positively associated with age, slope, forest cover, summer precipitation and mean annual temperature, but negatively associated with drought frequency and intensity. Addi- tionally, BE forests had the lowest AGB, while NE forests had the highest AGB. In addition, the GLM revealed significant interactions between the forest leaf-habit and climatic factors, Land 2021, 10, 817 9 of 18 that is, the positive effect of summer precipitation was higher for BE forests than for NE and BD forests, the positive effect of temperature was lower in NE forests than in BE and BD forests, while the negative effect of drought frequency and intensity was lower in BE and NE forests than in BD forests. Finally, both slope and summer precipitation had a negative interaction, which indicates that the greater the slope, the lower the positive effect of summer precipitation. The autocorrelation test (Electronic Supplementary Materials 1, Table S3) showed large decreases in the Moran’s I values from the study sample (0.252) to Land 2021, 10, x FOR PEER REVIEW 10 of 20 the residuals of the GLM model with the explanatory variables (0.005), with both values being significant. Figure 3. General linear model (GLM) parameter estimates of the relationships between the time Figure 3. General linear model (GLM) parameter estimates of the relationships between the time (quinquennium) of secondary forest establishment and elevation (m), slope (◦ ), annual precipitation (quinquennium) of secondary forest establishment and elevation (m), slope (°), annual precipita- (mm), tion mean (mm), annual mean SPEI annual values, SPEI drought values, intensity, drought forest intensity, cover forest in a in cover 1 km a 1 radius (%) and km radius interaction (%) and in- with secondary forest leaf-habit broadleaf evergreen (BE) and needleleaf evergreen teraction with secondary forest leaf-habit broadleaf evergreen (BE) and needleleaf evergreen (NE) (NE) forests compared forests to broadleaf compared deciduous to broadleaf (BD) forests. deciduous Climatic (BD) forests. values values Climatic were calculated from thefrom were calculated meantheof the mean of 1950–2015. period the period 1950–2015. SignificanceSignificance is indicatedisusing indicated using thep-values: the following following*: p-values: 0.05, ***: *: 0.05, ***: 0.001. 0.001. 3.2. Secondary Forest Growth The aboveground biomass (AGB) of the secondary BD forests had mean values (±SE) ranging from 61.3 ± 1.1 Mg ha−1 in the youngest forests (5 years) to 101.5 ± 0.9 Mg ha−1 in the oldest (25 years). These values were higher than those of BE and NE forests for most age groups (Figure 4). By contrast, secondary BE forests had the lowest AGB values for most age groups, ranging from 35.0 ± 0.7 Mg ha−1 to 58.3 ± 0.6 Mg ha−1. Finally, NE sec- ondary forests had mean AGB values ranging from 37.3 ± 1.4 Mg ha−1 to 77.9 ± 0.9 Mg ha−1. Land 2021, 10, 817 10 of 18 Land 2021, 10, x FOR PEER REVIEW 11 of 20 Figure Figure 4. Average 4. aboveground biomass (AGB)biomass Average aboveground of secondary forests (AGB) ofby age for the different secondary forests byforest ageleaf-habit for thetypes. Differ-forest different ent letters indicate significant differences in AGB between forest leaf-habit types for the same age according to Tukey’s test (p