TREE SPECIES OF ATLANTIC FOREST AND PAMPA ALLUVIAL FORESTS IN THE CONTEXT OF CLIMATE CHANGE

Although species are continually exposed to variations in climate over time, there is growing concern about the accelerated pace of climate change to which they are currently exposed, as climate is determinant in the geographic distribution of animals and plants. This study evaluated the climatic niche and impact of climate change in the 2061-2080 period on areas of occurrence of indicator tree species of alluvial forests in the south-central portion of the Atlantic Forest and Brazilian Pampa. Nineteen climatic variables were considered for the contemporary and future climates. The species’ climatic niches were modeled using the Maxent algorithm. Climatic adequacy for species in the Parana-Uruguay group was strongly infl uenced by variables related to temperature, while adequacy for species in the Atlantic group was related to rainfall and temperature. A decrease in the area of climatic adequacy is estimated for most species considered, with similar behavior in the two groups assessed. The species tend to occupy a more southern position, using areas of higher altitude, latitude and proximity to the Brazilian coast as climate refuges, highlighting these areas as strategic for environmental conservation.


INTRODUCTION
Considering that climate is determinant in the geographic distribution of species, the expansion and contraction of plant and animal populations in their native ranges have been strongly associated with climate change (Guitérrez and Trejo, 2014). Changes in patterns of species occurrences have been highlighted in several studies (e.g. Thuiller, 2007;Marchioro et at., 2019;Siqueira et al., 2019) as one of the main consequences expected in the global warming scenario. Although species have been continually exposed to variations in the climate throughout their evolutionary history, there is increasing concern about the accelerated rate of recent changes in global climate patterns (Melo et al., 2015;Pecl et al., 2017). Despite several sources of threat to biodiversity, the eff ects of climate change tend to become more prominent than others in the long term (Thuiller, 2007). A better understanding of how species will respond to climate change and the identifi cation of areas of high conservation value based on vulnerability to climate change are therefore highly relevant for impact mitigation (Feeley et al., 2012;Thorne et al., 2018) and the conservation of biodiversity (Costa et al., 2018).
Predictive modeling is an alternative for simulating species distribution in future climate scenarios and assessing the impacts of climate change (Thomas et al., 2004;Costa et al., 2018). The use of appropriate modeling techniques in order to reduce uncertainty by adopting more specifi c criteria in the use of models and carefully selecting geographic and climate data (Medeiros et al., 2013), as well as the use of current knowledge in several fi elds of science (e.g. ecology, paleontology and evolution), can potentially transform the way the vulnerability of species to climate change has been evaluated (Dawnson et al., 2011).
An evaluation of climatic niche and impacts of climate change for the 2061-2080 period in areas of geographic occurrence of tree species that are indicators (indicator species hereafter) of alluvial forests in the central-southern region of the Atlantic Forest and Brazilian Pampa biomes is presented in this paper. Four main questions are answered: i) Will climate change alter the potential distribution range of tree species indicators of alluvial forests? ii) Which fl oristic group of alluvial forests will be potentially more impacted by climate change? iii) Which tree species will undergo more changes in their climatic suitability area? iv) Which climate variables are related to the climatic adequacy of these tree species indicators of alluvial forests?

MATERIAL AND METHODS
The indicator species of alluvial forests identifi ed by Silva et al. (2020) were selected for an impact assessment of climate change on the central-southern region of the Atlantic Forest and Brazilian Pampa biomes. As clarifi ed by Hill et al. (1975), indicator species are those with ecological preferences that identify specifi c environmental conditions. Therefore, indicator species of each group of alluvial forests were used to characterize distinct environmental conditions (p < 0,001):  Silva et al. (2020) observed that the group of alluvial forests in the Parana-Uruguay basin was strongly associated with subtropical (temperate) characteristics, while the group of alluvial forests in the Atlantic basin was strongly associated with areas in tropical conditions, characteristic of the Atlantic basin domain.
Species occurrences were extracted from the Botanical Information and Ecology Network (BIEN) (Enquist and Boyle, 2012;Fegraus, 2012;SpeciesLink, 2012;Anderson-Teixeira et al., 2015;Enquist et al., 2016, GBIF.org, 2018. Data for E. unifl ora were extracted from the Global Biodiversity Information Facility (GBIF.org 2019) because there were no data available from BIEN. All data were spatially fi ltered considering a resolution of 10 minutes to avoid sampling bias (Aiello-Lammens et al., 2015). The geographic occurrence data of the species were rigorously verifi ed and only native range areas were used. Distribution references were obtained from the Flora do Brasil 2020 Database (2017).
Two assumptions usually considered in climate modeling studies on species geographic distribution were adopted to apply the models: i) species are in balance; ii) species do not adapt. The species climatic niches were modeled using the Maximum Entropy algorithm (Maxent) (Philips et al., 2017), using 10,000 pseudo-absences (Lobo and Tognelli, 2011) distributed at random in a 500 km radius around each occurrence (VanDerWal et al., 2009). Five rounds of calibration (70% of data) and test adjustments (30% of data) were conducted. The adequacy of adjustments was verifi ed by TSS statistics (True Skills Statistics), which varies between -1 and 1 (Allouche et al., 2006), where 1 represents the value of the perfect adjustment. The projection of potential spatial occurrence areas was based on the consensus between adjustments to reduce predictive uncertainties inherent to each adjustment,. The mean of diff erent adjustments of climate adequacy estimations was calculated (TSS > 0.4) (Landis and Koch, 1977). Bioclimate variables with the strongest infl uence on the area of climatic adequacy were determined from adjustments with TSS > 0.4. Response curves between the estimate of climatic adequacy and the best explanatory variables were built for each species.
The most pessimistic climate change scenario for the 2061-2080 period (RCP 8.5 -Representative Concentration Pathways), in which humanity will not adopt mitigation strategies, was considered to quantify the impacts of climate change on the area of potential species occurrence using future predictions. These predictions were based on data from the Intergovernmental Panel on Climate Change report (IPCC, 2014) and on the atmospheric circulation model HADGEM2-ES of the MET OFFICE in the United Kingdom.
Consensual projections were produced for indicator species of the two forest groups with TSS adjustments > 0.4 and more than 100 spatially fi ltered occurrences. Five species were considered for the Parana-Uruguay group and 16 species for the Atlantic group. Present and future binary maps (presence/ absence) were produced for these species. Areas were considered adequate for species occurrence when adequacy estimations were higher than 50%. The maps were then superposed to quantify lost pixels, gained pixels, and stable pixels for all species, representing the climatic adequacy area for each one. Possible diff erences in the impact of changes on the groups (Parana-Uruguay and Atlantic basins) were verifi ed through a boxplot and Mann-Whitney test, as the data set did not follow a normal distribution.

RESULTS
After spatial fi ltering, the number of geographic occurrences of tree species in alluvial forests varied from 59 (T. cassinoides) to 984 (A. triplinervia). The analysis of occurrence data showed that the Parana-Uruguay basin (A-E) indicator species were predominantly concentrated in the southern region of Brazil, both in the coastal region and in the continental area.The indicator species of the Atlantic basin (F-U) were generally more distributed regionally but expressively concentrated along the Atlantic coast ( Figure 1). Isothermality (bio 3) and mean temperature in the wettest quarter (bio 8) stood out for species of the Parana-Uruguay group (A-E). While the likelihood of occurrence of A. edulis and E. unifl ora was higher in areas with intermediate isothermal values, G. klotzschiana and C. xanthocarpa were intolerant to areas with high temperatures in the period of the year with the highest rainfall ( Figure 2). On the other hand, rainfall in the driest quarter (bio 14) was the most frequent variable among species in the Atlantic group (F-U), as most species were gradually more intolerant to the month with minimum rainfall under approximately 50mm.
A decrease in area of climatic adequacy for most indicator species in both groups is predicted for the 2061-2080 period (Figure 3). In the Parana-Uruguay group, V. megapotamica (-57.5%) and C. xanthocarpa (-52.4%) were marked by a decrease in the potential area of occurrence. In the Atlantic group, the most relevant losses in area were estimated for C. glaziovii (-86.6%) and G. australis (-83.4%). The species less sensitive to climate change in the Parana-Uruguay and the Atlantic basin were, respectively, E. unifl ora (-34.8%) and M. racemosa (-2.1%). The only species for which the area of climatic adequacy should increase is A. fraxinifolia (+85.2%), in the Atlantic basin, with signifi cant expansion in Central Brazil.
temperature of 20ºC in the wettest quarter (usually in the summer). With the increase of temperature beyond this value, the estimation of adequacy for these species is dramatically reduced. The behavior of V. megapotamica, more infl uenced by temperature in the hottest quarter (also usually in the summer), is similar, but the optimum mean temperature is around 21 o C. Only A. edulis and E. unifl ora in the Parana-Uruguay basin are more infl uenced by isothermality, with climate adequacy optimums between 50 and 60%. Isothermality, which refers to the diff erence between diurnal thermal range and annual thermal range, is related to the tropicality eff ect. These species are therefore less tolerant to higher tropicality conditions.
In the Atlantic group, rainfall infl uences climatic adequacy patterns besides temperature variables. This may be related to the physiological tolerance of species in each group, a determinant of spatial distribution (Lou et al., 2018). Species more tolerant to fl ooding or drought tend to have a wider niche amplitude. The most infl uential variables are rainfall in

DISCUSSION
Distinct geographic distribution patterns were identifi ed for tree species based on the groups of alluvial forests considered in this study. Occurrence in subtropical regions predominates in the Parana-Uruguay group, as in the south of Brazil, while in the Atlantic basin species are more distributed near the coast and are more related to higher tropicality.
When bioclimate variables that infl uence these distribution standards were observed, temperature variables were more infl uential in the Parana-Uruguay group. Analysis of climatic adequacy estimation curves showed that the geographic distribution of the species in the group is associated with milder summer temperatures (subtropical conditions) and medium variations between diurnal and annual thermal ranges (medium isothermality). Estimation of adequacy for the species with the highest indicator value in the Parana-Uruguay group, G. klotzschiana, and for C. xanthocarpa, is favorable up to the approximate mean Red: unstable areas where the climate will stop being favorable for the occurrence of the species; White: unstable areas that will become favorable in the future). the driest month, strongly related to the occurrence of seven species, and isothermality, which signifi cantly infl uences climatic adequacy for four species.
The rainfall seasonality variable is, in turn, strongly associated with the species of highest indicator value in the group, A. fraxinifolia. The interpretation of the response curve may suggest that the regular distribution of rain throughout the year is a limiting factor. On the other hand, the seasonality of high annual rainfall favors its occurrence, with maximum expression under circumstances of high inequality in which the variation of the period with the highest rainfall in relation to the period with less rainfall is comprehended in the interval between 75 to 115%. The majority of species in the group -A. fi rmula, G. australis, L. pisonis, P. contorta, A. dolabripetala, E. edulis and P. grandifl orum -are strongly related to rainfall in the driest month. It is possible to note that the estimation of climatic adequacy for these species is reduced under situations of low rainfall in the driest month (< 50 mm), with an optimum range beyond approximately 75 mm. The sensitivity of the rainfall regime highlighted by modeling suggests that these species are intolerant to extreme drought events.
As stated by Caron et al. (2014), the seasonality of climate elements may change plant physiological behavior. Therefore, climate change, especially high temperatures and low rainfall, may increase water and nutritional stress. High temperatures beyond optimum values is harmful to plant growth and may damage photosynthetic capacity (Kluge et al., 2015;Martinez et al., 2015). Extreme temperatures and water stress may cause stomata closure and reduce carbon intake, reducing photosynthesis and chlorophyll concentration while impacting net carbon gain. Wheeler et al. (2000) stated that increased temperatures may aff ect the most sensitive development phases of plants, such as fl owering, thus reducing seed production due to interrupted pollination or loss of pollen viability, in turn aff ecting other plant patterns and interspecifi c relations (Wheeler and Reynolds, 2013).
The isothermality variable strongly infl uences the climate adequacy of C. glaziovii, N. oppositifolia, C. microcarpa and G. gardneriana, which basically repeat the behavior pattern related to the existence of variations between diurnal and annual thermal ranges observed for species in the Parana-Uruguay group. In addition to tropicality factors related to latitude, isothermality may be related to altitude. Therefore, the species strongly related to this variable are associated with lower altitudes, subject to smaller ranges of diurnal temperature, and intolerant to higher altitudes, in turn related to higher thermal ranges. In like manner, the probability of occurrence of M. racemosa is strongly related to mean diurnal thermal range, with a higher probability of occurrence in lower thermal ranges. Low altitude areas in the absence of extreme droughts are therefore adequate for a large portion of Atlantic species. However, other factors related to temperature also infl uence species in the Atlantic group. P. glabrata and A. triplinervia are more related to mean temperature in the wettest quarter, with temperature optimums around 22 o C and 20 o C, respectively. H. alchorneoides is strongly associated with mean temperature in the driest quarter. Climatic adequacy for this species is higher under mild temperatures, with a climate optimum around 20ºC. The likelihood of occurrence of this species is signifi cantly reduced with an increase in temperature beyond this value.
Estimations for nearly all species assessed in the present study indicate a loss of area of climatic adequacy, with values varying from -2.1% (M. racemosa) to -86.6% (C. glaziovii). The only exception is A. fraxinifolia, as the potential area of occurrence increased. Among 21 species considered, only eight are predicted to lose less than 50% of adequate area, while 13 should incur in area losses higher than 50%. This trend of area loss was also observed by Velazco et al. (2019) upon an evaluation of the vulnerability of Brazilian Savanna (Cerrado) species to climate change which culminated in changes in species distribution patterns.
Stable areas in the Parana-Uruguay group were mostly concentrated in southern Brazil, from the coast to the highest altitudes. The unstable areas where climatic adequacy will decrease occur in diff erent regions in Brazil above the Tropic of Capricorn, with higher expression in the southeastern and part of the Central-Western regions, where temperature increase tends to be more severe. Warming was also estimated in projections of future climate scenarios for the state of Parana (Melo et al., 2015). The authors expressed concern as temperature increase is expected to prompt a reorganization of natural systems. The loss of areas of adequate climate is partially a result of an increase in tropicality, which will generate changes in the rainfall regime and in temperature (Salviano et al., 2016), factors that infl uence species occurrence (Gotelli et al., 2009;Carvalho et al., 2019). Medeiros et al. (2013) corroborated this observation by predicting an increase in annual and seasonal maximum and minimum temperatures for the Southern region. In a study on the Atlantic Forest, Colombo and Joly (2010) identifi ed a pattern of reduction in areas of occurrence of the species considered that varied between 25% and 50%, with migration to more meridional areas in Brazil. Several studies (Thomas et al., 2004;Parmesan, 2006;Lemes and Loyola, 2014) have pointed out that climate change may deeply aff ect environmental dynamics, even leading to species extinctions, overall at local levels.
The areas that tend to remain stable under favorable climate conditions in the Atlantic group are mainly located near the coast in the Northeastern, Southeastern and Southern regions. A reduction in climatic adequacy will occur in unstable areas, mainly concentrated in the Northeast and in the Southeast, both along the coast and in continental regions. Some currently unstable areas will develop an adequate climate for Atlantic species (especially A. fraxinifolia, L. pisonis, H. alchorneoides and G. gardneriana), mostly in Central and Southern Brazil. The displacement geographic occurrence limits to higher altitudes and latitudes in response to temperature increase is a tendency that has been observed for several species (Colombo and Joly, 2010;Chen et al., 2011;Feeley et al., 2012;Moritz and Agudo, 2013) as a consequence of higher tropicality in the future. Besides, the decrease in more continental areas of adequate climate may be explained by a decrease in rainfall in these areas.
The groups assessed include pioneer species such as G. klotzschiana, E. unifl ora, A. fraxinifolia, and C. glaziovii (Aidar et al., 2003;Meyer et al., 2013), usually associated with higher environmental plasticity and higher adaptation capacity to environmental variation than non-pioneers (Valladares and Niinemets, 2008). Pioneer species may thus have better chances of survival from changes in rainfall and temperature. This is in line with Noss (2001), who stated that improved knowledge of functional groups could contribute to the understanding species resistance mechanisms against climate change.
Areas of climatic adequacy are expected to decrease between 30 and 77% for most species in the Atlantic group. However, some species are outside this range and may be less (between 2 and 15%) or more (between 78 and 87%) aff ected. Counteracting the pattern of area loss for species in the group, the area of occurrence of A. fraxinifolia is estimated to increase by more than half, with an expectation of increase in area of climatic adequacy of 85.2%. As mentioned earlier, the occurrence of this species is limited by the homogeneous distribution of rainfall throughout the year. Therefore, as more drought events are expected in the future, the species will benefi t from higher rainfall seasonality and increase its distribution range. A decrease in climate adequacy in the order of 45 to 55% is expected for the majority of the indicator species in the Parana-Uruguay basin, with the exception of E. unifl ora and V. megapotamica, as the values for these species are beyond the limits established for the group. The estimations of reduced climate adequacy for both groups are therefore variable, with relatively similar mean values, but result in no diff erence between groups in terms of decrease in climate adequacy.
Considering that the Atlantic group is subject to higher tropicality, higher adaptation capacity to climate change was expected, with a less signifi cant decrease in areas of climatic adequacy for the indicator species. However, as climate change does not only result in global warming, but also in changes in rainfall patterns, species not adapted to extreme droughts will also undergo a decrease in area of adequate climate. Besides, it is important to note the high variation of the values of decrease in climate adequacy in both groups, which shows that, in both cases, species have diff erent patterns of behavior. This indicates the complexity of forest ecosystems and the diffi culty of establishing patterns for such heterogeneous groups.
The climate in areas of higher altitude and latitude as well as areas nearer the coast are more suitable for the alluvial tree species considered in the present study and may be considered strategic for conservation, as they may function as climate refuges. In a scenario of global warming and altered rainfall patterns, the potential occupation of these climate refuges by diff erent species highlights the need to implement protection measures such as the establishment of protected areas (Pecl et al., 2017).
Although geographic distribution modeling is being widely used, its limitations must be acknowledged. They include the lack of capacity to model biotic interactions, dispersal patterns, and evolutionary changes (Soberón and Peterson, 2005;Torres and Vercillo, 2012). Species adaptation capacity is therefore not considered, as other factors beyond climatic niche (biological, pedological) may create more selective habitats that limit the occurrence of certain species (Pacifi ci et al., 2015).

CONCLUSION
The potential distribution range of species that function as indicators of alluvial forests will be altered due to changes in climate expected in the period 2061-2080. The majority of species considered in the present study tend to undergo signifi cant changes in area of climatic adequacy, with losses above the order of 50%. However, no diff erences were observed between the Parana-Uruguay and Atlantic groups in terms of decrease in area of climatic adequacy. In a global warming scenario, the species assessed will tend to occupy a more southern distribution, occurring in higher altitudes and latitudes, as well as near the coast. These areas will constitute climate refuges, which highlights the relevance of their conservation.