Historical Fires Induced Deforestation in Relict Scots Pine Forests during the Late 19th Century

Mountain forests are subjected to several pressures including historical land-use changes and climate warming which may lead to shifts in wildfire severity negatively impacting tree species with low post-fire growth resilience. This is the case of relict Mediterranean Scots pine (Pinus sylvestris) forests in the Sierra de Gredos mountains (central Spain). We reconstructed the historical fire regime of these forests since 1700 by using paleoecology, historical ecology and dendroecology. We detected an increase in charcoal accumulation rate and coprophilous fungi in peat bogs during the late 19th century when the pine pollen percentage sharply decreased, historical records of fire peaked and many trees showed growth suppressions. We inferred an increased wildfire incidence during the late 19th century, which could have shaped the current distribution of Scots pine forests. This shift in fire-forest interactions can be explained by the uncontrolled use of mountain forests and grasslands due to the dissolution of “Mesta”, one of the major and lasting transhumance livestock associations in Europe. Integrating historical human and climate influences on fire regimes allows decomposing the resilience and conservation components of relict forests.


Introduction
Shifts in forest wildfire regimes are mainly triggered by changes in land use and climate, interacting with other causes [1,2]. Bridging paleoecological and historical reconstructions of fire recurrence and severity allow detecting those shifts and the interactive effects of land-use changes and climate and characterizing the historical range of variability of fire-prone forest ecosystems [3][4][5]. Reconstructing fire regimes is necessary to assess forest resilience through time in response to combinations of climatic, ecological and human drivers [6].
Mediterranean mountain forests have shown resilience to fire in terms of regeneration and growth dynamics [7]. However, rapid socioeconomic and land-use changes during the 20th and 21st centuries, including rural depopulation and abandonment of traditional landscape management, have enhanced forest expansion increasing the amount and continuity of fuel and leading to more frequent, large drought-driven fires [2,8]. It is unknown if Mediterranean forests will keep being resilient under future scenarios with abundant fuel loads and increased warming and aridification [7]. Detecting fire regime shifts through paleoecological and historical data allows determining the relative role of human-caused, historical burning in Mediterranean forests and assessing post-fire recovery [9][10][11].
According to several reconstructions, fire has contributed to rapid structural and compositional changes in Mediterranean forests at least for the past 2000 years [9][10][11]. For instance, the local demise of Pinus nigra Arn. in northern and central Spain has been attributed to recurrent human-caused fires occurring at a high frequency [12][13][14], and the low post-fire regeneration of this thin-barked species. Some of these Mediterranean pine forests are relict or isolated populations such as those formed by Scots pine (Pinus sylvestris L.) in the Sierra de Gredos mountains, central Spain (Figure 1a). The relict status of Scots pine in that region may be related to shifts in the historical fire regime [15][16][17][18][19], with more recurrent fires than under "natural" regimes, since Scots pine is one of the most vulnerable European pine species to fire damage [20].  [21]). Dots indicate isolated Scots pine populations. (b) Location of four Scots pine forests (white circles) investigated in the Sierra de Gredos mountains, central Spain (Gu: Guisando; Pp: Puerto del Pico; Hy: Hoyos del Espino; Nv: Navarredonda) and pollen/charcoal records (red stars: Pp, Puerto del Pico; Se, Serranillos; Aaf, Arroyo de Aguas Frías). Elevation is given in m a.s.l.
Here we reconstruct fire regimes during the past 300 years in the Sierra de Gredos to assess their role as historical factors leading to the current relic status of Scots pine in this region. We reconstruct fire incidence by combining palaeoecological (charcoal) and dendroecological (sharp growth suppressions) data and comparing them with historical fire records (archival sources). We hypothesize that a high fire frequency reduced the cover and density of Scots pine forests in the Sierra de Gredos.

Study Area and Tree Species
The study area is located in the Sierra de Gredos, central Spain (Table 1, Figure 1). The climate of this mountainous region is Mediterranean with continental influence, characterized by cold-wet winters, wet springs and warm-dry summers. Mean annual temperature is 14.5 • C, mean maximum and minimum temperatures are 30.1 • C and 0.0 • C, respectively, and total annual precipitation is 1483 mm (data from "Arenas de San Pedro" meteorological station; 5 • 5 28 W, 40 • 12 31 N, 510 m a.s.l.). Drought may last from May to September. Geological substrates are dominated by granites, and soils are moderately deep at mid-elevation but rocky and thin in high-elevation Scots pine stands [22]. The vegetation of the study area includes forests, shrublands and also sparse pine woodlands on rocky slopes, and includes oaks (Quercus ilex subsp. ballota (Desf.) Samp., Quercus pyrenaica Willd.) from 600 to 1600 m, and shrubs (Cytisus oromediterraneus Rivas Mart.) above 1600 m and pines. The resin tapping industry has favored planting Pinus pinaster Ait. at low to mid elevations, whereas scattered Scots pine stands appear at high elevations from ca. 1200 up to 1800 m [17]. In this area, pine woodlands have burnt ( Figure 2) disproportionately more than other vegetation types since the 1970s after the abandonment of traditional land use and grazing practices associated with small fires affecting high-elevation scrublands [23]. Scots pine is a long-living, thin-barked, non-serotinous and shade-intolerant species that can survive low-severity surface fires, but is less tolerant to mid-severity fires than the thick-barked, fast-growing P. pinaster [20,24].

Field Sampling of Scot Pine Forests and Dendroecology
We selected relict Scots pine stands situated near the sites of paleoecological records and in areas historically burnt (Figures 1b and 2) but subjected to different climate conditions. We sampled two Scots pine stands located at high (Puerto del Pico) and at low (Guisando) elevation (Table 1). In each stand, we sampled living Scots pine trees with diameter at breast height (Dbh) above 10 cm and located in a ca. 0.5 ha large area. Due to the conservation measures of these relict Scots pine forests, we discarded obtaining partial cross sections from living trees, and we took cores at 1.3 m using 5-mm Pressler increment borers perpendicular to the maximum slope.
Dendrochronology was applied to cross-date the tree-ring width series [25]. Cores were air-dried, glued onto wooden supports and polished with sandpaper until rings were clearly visible. Then, samples were visually cross-dated and tree-ring widths were measured to the nearest 0.01 mm using a binocular scope and a LINTAB measuring device (Rinntech, Heidelberg, Germany). Cross-dating was checked using the COFECHA program [26].
Individual tree-ring width series were detrended to remove non-climatic, size-related growth trends [25]. We applied a power transformation and then a cubic smoothing spline with 50% frequency-response cut-off was fitted to obtain ring-width indices. These indices were subjected to autoregressive modeling to remove most first-order autocorrelation. Finally, site chronologies were obtained by averaging the residual ring-width indices on a yearly basis using a bi-weight robust mean. We obtained two types of chronologies for each of the two study sites: (1) standard, non-pre-whitened or preserving serial autocorrelation; and (2) residual, pre-whitened chronologies. For the chronology construction, we used the ARSTAN software [27]. To characterize each tree-ring series, we calculated the mean and first-order autocorrelation, a measure or similarity between consecutive rings, of raw ring-width data, and the mean sensitivity, a measure of relative change in width between consecutive rings and correlation of trees with site mean chronology of standard chronologies [25]. The two study sites' chronologies were compared with tree-ring data of the same species and from two nearby sites (Navarredonda, Hoyos del Espino; Table 1) gathered from the International Tree-Ring Data Bank (ITRDB) webpage [28] and measured by Klaus Richter in 1985. We also considered previous chronologies developed in the study area for Scots pine and other pine species to check our cross-dating [13,14,29,30].
We calculated the differences between the standard and residual chronologies following [14] to assess short-term changes in growth. According to these authors, post-fire growth declines should be characterized in stands historically burnt, and these growth suppressions should not be related to other stress factors such as pests, diseases or droughts which differently affect tree growth [4,14]. We considered the standard chronology of the observed post-fire growth values, and the residual chronology of the predicted values after removing the temporal autocorrelation influence. We assumed that an abrupt decrease in growth following a fire (growth suppression) was related to a loss of photosynthetic biomass or deterioration of cambium because of high temperatures as has been observed in boreal Scots pine forests where post-fire growth suppressions lasted from 1 to 5 years [31].
Finally, we reconstructed pine recruitment in 20-year classes based on estimated age at 1.3 m. Age was estimated by counting rings in cores with pith or by calculating pith-offset estimates in cores without pith. This was achieved by fitting a geometric pith locator to the innermost rings and converting the distance to the theoretical pith into the number of missing rings [32]. We estimated tree age as the maximum number of tree-rings in each individual summed to the number of years from the base to the coring height (1.3 m). The time to grow from the base to 1.3 m was estimated to be 6 years based on the equation fitted to trees from the Guisando site: height = 0.347 age 0.736 , R 2 = 0.85.

Paleoecological Methods
To reconstruct the regional forest history during the period 1700-2000, three pollen and charcoal records from the study area were selected corresponding to three cored mountain peat bogs ( Figure 1): one situated at mid elevation (Arroyo de Aguas Frías, 1120 m a.s.l.) and two located at high elevation (Puerto del Pico, 1395 m; Serranillos, 1700 m). The processing and dating of samples have been described elsewhere [16][17][18][33][34][35][36]. These procedures are explicitly documented for the Arroyo de Aguas Frías site in the Supplementary Material. Here we only consider the percentage values of the main taxa identified in the pollen records related to fire (Scots pine, pyrophytic broom communities of Genisteae shrubs), the relative abundance of coprophilous fungi associated with grazing activity, as well as fire events related to peaks in charcoal accumulation rates.

Historical Fire Records Obtained from Archives
Historical fire records were obtained through systematic research in national (National Historical Archive, General Archives of the Administration, Archives of the former Ministry of Agriculture, Spanish Military Police Archive, Spanish National Library), regional or provincial (Province Historical Archives, Forestry Administration Archives, private archives) and municipal archives. This allowed reconstruction of the fire history for the Sierra de Gredos mountains (n = 1094 records) from the year 1497 until 2013. Three types of archival sources were considered: administrative documents from administrations with fire use regulation and land management power since the 16th century, judicial and police sources including court registers and police reports since the 17th century and printed press (official journals, newspapers, books) [10,37]. Historical fire records were georeferenced with three different levels of increasing accuracy (municipality, site or area without specified boundaries and forest or plot with accurate limits of property) depending on the historical source. The historical fire database includes comprehensive information on fire occurrence with data on: date, location, land ownership, land cover/use, burnt area, fire duration, fire cause, suppression resources, losses, etc. [1]. We acknowledge the statistical limitations of the archival sources [1,10], but they still provide relevant qualitative information on historical fire regimes.

Paleoecology: Late-19th-Century Shift in Fire Regime and Reduction of Scots Pine Forests
Several proxies indicate a reduction of Scots pine forests during the past three centuries in the study sites with an abrupt shift in the late 19th century when the fire regime changed. For instance, in the Arroyo de Aguas Frías site, there is an abrupt drop in Scots pine pollen during the late 19th century which coincided with an increase in Genisteae pollen and coprophilous fungi indicating an increase in pastoral pressure (Figure 3). This change is also documented between the second half of the 19th century and the early 20th century in the Serranillos site, where the abundance of coprophilous fungi increased.
The changes in Scots pine pollen from the range of 50-70% during the 18th century and most of the 19th century to 10-30% values after the late 19th century and afterward indicate a shift from more extensive pine forests to scattered, fragmented stands. In the Puerto del Pico site such a decrease in pine pollen was not observed, but coprophilous fungi increased in the mid-19th century and in the 20th century.
Charcoal analyses indicate that the abrupt deforestation of pine forests observed in the late 19th century in the Arroyo de Aguas Frías is related to fire events in the late 19th century (1886 and 1893; Figure 3

Tree Growth
On average, radial growth was 1.97 mm (Table 2), with sites such as Guisando showing higher growth rates because of the presence of younger trees (Figure 4). The mean autocorrelation was 0.79 and the mean sensitivity was 0.26 indicating a high similarity between consecutive rings but also a high relative year-to-year variability in width. The mean correlation of trees with the site chronology was 0.61 confirming a high common signal, probably controlled by climate variability.  During the best-replicated period common to all tree-ring series , the two ITRDB series were highly correlated with them (Pearson correlation, r = 0.76, p < 0.001) and also with the Puerto del Pico series (r = 0.25-0.37, p < 0.05) but not with the Guisando series which showed lower correlations with the ITRDB series (r = 0.23, p = 0.07). The series of the two sampled sites were positively correlated (r = 0.34, p = 0.005). These results indicate a slightly different climate-growth response in Guisando as compared to the other sites in agreement with its lower autocorrelation and lower correlation with the site chronology ( Table 2).

Fire Proxies in Tree-Ring Records: Recruitment Patterns and Growth Suppressions
None of the sampled Scots pine trees were recruited from 1880 to 1920 when the frequency of historical fire records peaked according to archive sources (Figure 4). In the Guisando site, at least three age classes could be distinguished, with young trees (age at 1.3 m between 20 and 100 years, 51% of trees), middle-aged trees (age between 150 and 200 years, 45% of trees) and one old tree of 284 years (age between 280 and 300 years, which represent 4% of all sampled trees). In the Puerto del Pico site, middle-aged trees were also abundant (45% of individuals), and 68% of trees were established to be from 1800 to 1880. In this site, there was a negative but not significant correlation between the log-transformed number of fire events and the frequency of recruited trees (Spearman correlation, r s = −0.44, p = 0.10). Overall, we found a dominant cohort of middle-aged pines recruited in the first half of the 19th century.
In the Guisando site, we detected very low values of the differences between standard and residual indices in the mid (1861) and late 19th century (1888, 1891, 1895) and also in the early 20th century (1922,1944,1947,2006) (Figure 5). The growth suppressions observed during the 1880s and 1890s in this site coincided with the peak of fire records. Similar years of low difference values were also observed in the Puerto del Pico site, but they did not surpass the 0.05 lower threshold (dashed line in Figure 5). The series of differences between the standard and the residual chronologies of the Guisando site was negatively related to the number of fires (r s = −0.21, p = 0.007, period 1850-2013).

Discussion
Our findings confirm our hypothesis supporting that a high fire frequency reduced the abundance of Scots pine in the Sierra de Gredos mountains. This corresponded to an abrupt shift in the fire regime cascading on feedbacks between wildfire, vegetation and human use of mountain forests, shrublands and pasture. We found a peak in wildfires during the late 19th century according to several proxies (charcoal accumulation rates, archive sources; Figures 3 and 4), which coincided with a reduction in Scots pine pollen in several sites (Arroyo de Agua Frías, Serranillos), a lack of tree establishment ( Figure 4) and growth suppressions in the Guisando site ( Figure 5). Nevertheless, there could be lagged responses of pine forests to the change in fire regime or dating uncertainties of fire events, which could explain why after several fires in the late 19th century, pine pollen steadily increased during the 20th century in the Arroyo de Agua Frías site. This trend could also reflect the recovery of pine forests in the study area after the reduction of the number of fire events during the mid-20th century (Figure 4).
In the study region, the last 300 years are mainly characterized by the deforestation of high-elevation Scots pine forests [18] related to an increase of human pressure, that is, clearance of high-mountain pine forests by means of anthropogenic fires to obtain pastures related to transhumant pasture uses [17]. This is confirmed by the increase in post-fire shrubland (Genisteae) communities above the disturbed treeline and coprophilous fungi after the 19th-century fire regime shift (Figure 3), as illustrated by the Arroyo de Agua Frías site. This deforestation process was not related to any major timber harvesting during the late 19th century according to historical management plans [10], and the presented tree-ring data do not show growth releases suggesting the removal of trees and the reduction of competition in the sampled stands ( Figure 5).
The data from Serranillos reflect the presence of more extensive pine forests from the mid-18th to the mid-19th centuries as inferred from the changes in pollen percentage; while a reduction of pollen record there and in the Puerto del Pico along its chronological trajectory would denote the existence of isolated high-mountain pines or scattered stands [38]. These data seem to suggest that the increase of human-mediated fires and livestock grazing was connected to the decline of Scots pine in high-elevation sites [9,20]. The lower percentage of Scots pine pollen in the Puerto del Pico record is explained because this area has been the main passage for transhumant cattle since the Middle Age, making it an open woodland with scattered stands [35].
The beginning of the Late Modern Period (1800 to the present) in the Sierra de Gredos mountains coincides with prominent socioeconomic and political changes such as the dissolution of the "La Mesta" system, one of the main transhumance livestock associations in European history [39]. The termination of "La Mesta" in the 19th century far from lightening the livestock pressure on mountain areas as the Sierra de Gredos, triggered the intensive use of pastures for sheep and cattle for local farmers [20,35]. This manifested as a further local expansion of high-mountain shrublands (e.g., Genisteae) and grasslands with the increase of proxies of pastoral pressure such as coprophilous fungi and minimum percentages of Scots pine pollen [18]. These changes suggest an increase in local grazed areas and grazing pressure in response to the abrupt increase in fire frequency and severity during the late 19th century [13,14]. Such a shift in the fire regime was the most plausible cause of reduced establishment (Figure 4), either because of low recruitment or high mortality rates or both and severe growth suppressions ( Figure 5). We noted that both forest responses were site contingent since the recruitment signal was more evident in the high-elevation Puerto del Pico site, where severe fires are frequent (Figure 2), whilst the growth signal was recorded in the low-elevation Guisando site. It is also remarkable that some growth suppressions (e.g., 1960-1963) and fire scars (e.g., 1975) were dated in trees sampled in the Puerto del Pico site [40], during two decades with a high frequency of fires ( Figure 5). This may be explained by the fact that the late-19th-century wildfires drastically reduced pine cover whereas recent fires occurred under the present fire exclusion policy when the landscape was more open. It is also interesting to note that other reconstructions of fire history found a high incidence of fires during the mid to late 19th century in several Mediterranean countries such as Greece [41,42], Spain [43] and Algeria [44]. Further research could investigate if the 19th-century shift in fire regime was linked to local socioeconomic changes or also influenced by regional changes in climate conditions such as increased aridification.
We consider that the 1890s growth suppression can be explained by widespread surface fires which differently affected trees with moderate to high diameter (we estimate Dbh between 15 and 25 cm in the 1890s of the largest trees). This fire regime primarily depended on human burning and secondarily on climate variability since warm spring conditions occurred during those decades [13]. Scots pine growth in the study area is enhanced by wet spring and summer conditions [29,45], and reconstructed summer precipitation in the 1890s showed the smallest range of variability since 1800 [46]. Surface fires of different intensity usually affect Scots pine forests, albeit lethal crown fires could occur in dense stands under drought conditions [20]. In the study case, a high frequency of mid-to high-severity surface fires could have triggered the decline of Scots pine forests.
The late-19th-century shift in the fire regime contributed to shaping the current distribution and structure of relict Scots pine forests in the Sierra de Gredos mountains. Such episodes of rapid socio-economic and ecological changes and loss of forest resilience may inform management policies based on prescribed burning to recreate pre-19th century disturbance regimes associated with smaller, frequent fires and a greater abundance of Scots pine mountain forests. In addition, these forests are located near the species' southernmost distribution limit and should be monitored, better investigated and preserved since climate warming could further threaten their persistence. Plantations with other Scots pine provenances or with the more flammable P. pinaster should be avoided near these relict stands to preserve their genetic uniqueness and promote their long-term post-fire recovery. From a scientific point of view, combining paleoecological, historical, dendroecological and ecological (tree census data, forest inventory plots) sources of information represents a unique opportunity to disentangle the roles played by humans and climate as long-term drivers of forest dynamics [5,30].

Conclusions
We documented a regime shift in fire-vegetation dynamics during the late 19th century. In this study, human-caused burning was the main driver of the reduction in cover and abundance of Scots pine, which currently forms relict scattered stands in the Sierra de Gredos mountains. Paleoecological (increase in charcoal, decrease in pine pollen, increase in Genisteae pollen and coprophilous fungi), historical (increase in archive records of fires) and dendroecological (decrease in recruitment, growth suppressions) proxies point out to the high frequency and severity of fires during the late 19th century as triggers of deforestation related to shifts in land use of the mountain landscape (dissolution of the "Mesta" transhumant livestock system, extensive anthropogenic wildfires to create woodlands more intensively grazed). The disturbance regime shifts in the late 19th century illustrate a severe loss of resilience of Scots pine forests with legacy effects, including the current structure and distribution of relict forests of the species in the Sierra de Gredos mountains.