Analyzing environmental determinants of tree species distributions and regeneration pattern in Western Himalaya, India


 Proper documentation of tree distribution across the globe has been considered crucial to assess the relationships between species occurrence and changing climate, and accordingly in designing the conservation action plans. Himalaya is one of the regions in the world where the temperature is gradually increasing at a rate higher than the global average. Therefore, it becomes imperative to understand the most influential parameters for major tree species distribution and its regeneration pattern across their habitat in Western Himalaya (WH) via direct gradient analysis. We used National Forest Inventory (NFI) data which has a robust statistical design with data collected in a consistent manner across the timeline in systematic order. This allowed us carrying out detail analysis to reveal the aforesaid relationship and pattern. Three topographical factors (altitude, aspect, slope), four major climatic variables (temperature, temperature seasonality, precipitation and its seasonality) and four edaphic factors (soil depth, soil humus, soil consistency and soil coarse fragments) were considered as defining variables. The results indicated that the altitude and temperature were the most determinant factors followed by precipitation in shaping the distribution of major tree species in WH. The analysis also indicated the upward shift of regenerating individuals of these tree species towards higher altitude. These relevant information about the extent of current tree distribution and their regeneration pattern over the last one and half decade might help in systematic conservation planning and monitoring range dynamics under future climate change conditions.

data which has a robust statistical design with data collected in a consistent manner across the timeline in systematic order. This allowed us carrying out detail analysis to reveal the aforesaid relationship and pattern. Three topographical factors (altitude, aspect, slope), four major climatic variables (temperature, temperature seasonality, precipitation and its seasonality) and four edaphic factors (soil depth, soil humus, soil consistency and soil coarse fragments) were considered as de ning variables. The results indicated that the altitude and temperature were the most determinant factors followed by precipitation in shaping the distribution of major tree species in WH. The analysis also indicated the upward shift of regenerating individuals of these tree species towards higher altitude. These relevant information about the extent of current tree distribution and their regeneration pattern over the last one and half decade might help in systematic conservation planning and monitoring range dynamics under future climate change conditions. Background Under ongoing rapid climate change scenario, understanding the ecological requirements of tree species distribution is a major research agenda in plant ecology (Enquist et al. 2016;Franklin et al. 2016). This would help in better understanding of tree diversity, a key driver of forest functioning (Paquette and Messier 2011;Pichancourt et al. 2014) and forest ecosystem service provisioning (Gamfeldt et al. 2013; Thompson et al. 2014).
The information related to plant species distribution patterns along complex environmental gradients are ecologically important from management perspectives because it can be used (i) for woodland conservation management plans (Roces-Díaz 2014), (ii) for successful ecological restoration plans (Onaindia et al. 2013, Pei 2018, (iii) to identify biodiversity and productivity patterns (Fischer et al. 2014) and biodiversity protection practices such as forest degraded areas (Zhang et al. 2016) and (iv) for identifying the ecological niches of plant species to predict their potential response to climate change (Zonneveld et al. 2009). Since studies based on species range shift indicates that all species have climatic limitations beyond which they cannot survive , it becomes important to understand and monitor the in uencing factors of distribution patterns of plant species (Bai et al. 2004). However, for many regions in the world including Himalayas, historical data on species distribution are lacking (Hamid et al. 2018).
Studies have shown that the distribution of vegetation types and oristic patterns are most associated with environmental factors, including local variables/topographic factors (elevation, slope aspect, slope degree), soil factors (a soil's physical and chemical properties) (Huang 2002;Jabeen and Ahmad 2009;Zhang et al. 2012), and human impact factors (Enright et al. 2005;Hoang et al. 2011). Soil type and topographic variables were the most signi cant factors affecting species diversity and woody vegetation of a locality (Hejcmanova-Nezerková and Hejcman 2006;Zhang et al. 2012). Climatic factors were central in controlling the distribution and species richness patterns of various tree species ). The edaphic factors were reported to be more important than climatic one in shaping distribution of twelve temperate tree species in Switzerland (Walthert and Meier 2017). Since the environmental factors such as temperature, precipitation, atmospheric pressure, solar radiation, and wind velocity change systematically with elevation, thus elevational gradients are powerful natural experiments for testing the ecological and evolutionary responses of forests to environmental changes (Cui et al. 2005;K€orner 2007). Such rapid change in altitudinal gradient at small distances found in Himalaya makes it noteworthy for ecological studies (Singh and Singh 1992;Chandra 2010).
The Himalaya is recognized as among the 36 global biodiversity hotspots (http://www.cnservation.org/priorities/biodiversity-hotspots). Due to diverse ecological, biogeographical and evolutionary factors that favor high endemic biodiversity (Joshi and Pant 2012), the region is highly sensitive to climate change and have been a data-de cient region (IPCC, 2007) and is under-represented in scienti c literature on climate change-induced species range shifts in comparison to other mountains of the world (Schickhoff 2005;Dutta et al. 2014;Schickhoff et al. 2015). Since Himalayan ecosystems have been referred to be extremely sensitive under future climate ) and heavily in uenced by climate change (Polanski et al. 2014;Chakraborty et al. 2018), any change in temperature or rainfall pattern will adversely impact the entire ecosystem already under stress due to anthropogenic activities (Ravindranath et al. 2006). Evidences indicate that the Himalayan region is warming at a higher rate than the global average rate Jianchu et al. 2007;Shrestha & Aryal, 2011;Xu et al., 2009). Dash et al. (2007) reported that in the WH a 0.9 °C average increase was observed during the 102-year period 1901102-year period -2003102-year period . Shrestha et al. (1999 reported that temperature increments are greater in the uplands than that in the lowlands regions in the Himalaya. However, the precipitation doesn't show de nite trend, but a distinct shift from snow to rain has been was described in literature (Singh et al. 2011).
The temperature and precipitation-related vegetation studies in Himalaya are limited and concentrated to localized places (Xu & Liu, 2007;Liang et al., 2015). The changes at the regional level are yet to be documented  and therefore the landscape level primary data on distribution and abundance could be critical in understanding key aspects of macro-ecological patterns (Oommen and Shanker 2005).
The species and environmental relationship have been explored in Himalaya in recent past (Acharya et al. 2011;Gaire et al. 2014). The energy variables (potential evapotranspiration and temperature seasonality) were reported to be major determinants of species richness in WH (Panda et al. 2017). Kharkwal et al. (2005) found the altitude, temperature and rainfall as major determinants of species richness in Kumaon Himalaya whereas many acclaims that it is the forest structure and composition which are mainly driven by elevation and climate (Vetaas 2000;Sharma et al. 2016b;Sharma et al. 2017Sharma et al. , 2018. Climate and soil properties have been reported to strongly constrain the species distribution (Dubuis et al. 2013). The distribution pattern of a plant species indicates its adaptability to various environments (Wang et al. 2004). Under the climate change regime (Tewari et al. 2017;Upgupta et al. 2015) various environmental parameters may respond differently (degree and rate of change). The vegetation distributions are likely to undergo change in near projected climate change (Gao et al. 2017).
Regeneration is another critical aspect as it determines future species composition and stocking. If it gets con ned to a particular range of habitat conditions, that extent becomes crucial in determining species's geographical distribution (Grubb 1977). It also decides the existence of a community under varied environmental conditions, and therefore becomes more important to study the regeneration pattern, which may be in uenced by both natural (Behera et al. 2012;) and anthropogenic (Chaturvedi et al. 2017) factors. The lack of adequate forest regeneration is an important issue recognized by both foresters and ecologists (Ceccon et al. 2004;Mishra and Singh 2017). It may be due to lack of viable seed production, insect and animal predation, unfavorable microclimatic conditions, overgrazing, habitat changes, and biological invasions. The regeneration patterns may also act as early indicator of the impact of climate change (Bulletin E-3221, 2014). Regeneration in himalayan forests was quoted to be adversely affected by the uncontrolled grazing by domestic livestock by removing young seedlings and saplings, which also causes soil loss due to trampling (Saberwal 1995). De cit regeneration will affect the entire ecosystem structure and functioning. From the very past, literature reports it to be of utmost importance to have baseline scienti c data on regeneration characteristics of dominant tree species on landscape level to understand ecological implications in distribution of trees (Forman and Godron 1986). The regeneration pattern predicts future growth of the forest, and hence important for forest management (Khanna 1996).
There is a serious lack of systematic studies and empirical observations about species-level impacts of climate change in the Himalayas (Gautam et al. 2013). The literature reports that major/widespread species with large population and high fecundity are more likely to persist and adapt but species with small populations, fragmented ranges, or low fecundity, may face population collapse and need facilitated migration (Serra-Diaz et al. 2017). However we believe that if major species distribution is driven by very few environmental parameters with narrow regeneration zone, then they may not be able to persist and adapt without facilitated migration and therefore it becomes important to monitor reproduction and regeneration.
The WH contributes with reportedly 490 tree species inventoried out of which 372 species found in wild (Bhatt et al. 2016). Despite apparent similarity to temperate forests in terms of tree species (oak and pinedominated systems) and cool average temperatures, these forests possess distinct ecological traits. High nutrient turnover rates and productivity and a phenology adapted to the summer drought, make them more similar to tropical ones (Thadani et al. 2014).
This study targeted 15 major tree species identi ed in NFI forming representative major forest types of WH. We included major abiotic components of environmental heterogeneity i.e. climatic parameters like temperature and precipitation, soil variables like soil depth, humus, soil consistency, coarse fragments and topographic variables like slope, aspect and altitude. We also used one very signi cant NFI variable i.e. most commonly found tree species under regeneration observed in an area of 2 ha around the center of the inventory plot. Such sites can be considered promising for the growth and perpetuation of the corresponding tree species. With the huge amount and wide coverage of data availability in NFI, this study explores the physiographic zone level of environmental in uence on the distribution of tree species in WH. Our main objective of current analysis is to nd whether such a scenario exist for major tree species of WH? We test for the null hypothesis i.e. all the environmental variables play equal role in determining the distribution of selected tree species in WH. As our study focuses on the relative importance of climate variables, soil and topography, we do not include biotic interactions and disturbances. In addition, we sought to see the most determining variable for the 15 tree species. Since mature trees are long lived and may not be good indicators of climate change tree regeneration pattern of these 15 species are analyzed. The range of species regeneration in comparison to their natural course of occurrence would help us to understand their overall status in the recorded forest area.

Materials And Method Study Area
Study area is WH and comprises 19 districts belonging to three different states of India -Jammu and Kashmir, Himachal Pradesh and Uttarakhand ( Fig. 1) These three states possess largest number of forest types (35 in each) as mentioned by Champion & Seth (1968) in India and collectively has more than 28% area under forest (FSI 2019). The WH exhibit semi-arid and cold arid climates, respectively, the far eastern ranges represent some of the wettest places on earth with > 4000mm annual precipitation. The effect of the monsoon becomes less pronounced in the WH. Unlike the central and eastern parts, the WH receives higher precipitation during winter in the form of snow. The WH is structurally complex with elevations ranging from 300 to over 8126 m, and the drainage system is composed of rivers, lakes and glaciers. The WH forest vegetation ranges from tropical dry deciduous forests in the foothills to timberline followed by alpine scrub and alpine meadows. The forest types greatly varies from dry and moist deciduous forest in tropical zone dominated by Shorea robusta, Sub-tropical pine forest ruled by Pinus roxburghii, moist and dry temperate forest consisting of broad leaved species of evergreen Oak species (Q. oblongata, Q. dilatata and Q. semecarpifolia) and Rhododendron spp. and by deciduous species like Carpinus viminea, Juglans regia, Aesculus indica, Alnus nepalensis etc., then sub-alpine forest of A.pindrow, A. spectabilis, Pinus walliciana, Picea smithiana and Betula, followed by alpine scrub comprising of Juniperus species (FSI 2011).
The mountains rise abruptly, resulting in a diversity of ecosystems that range from alluvial grasslands, subtropical broadleaved forests to dominance of conifers in the temperate zone to alpine meadows above the tree line. High-altitude cold desert ecosystems encompass a signi cant area of the region. The region supports a number of glaciers that are key source of water for sustaining downstream ecosystems and the world's largest human population.

Methodology and Data Collection
Strati ed two stage sampling was employed, with districts as the primary sampling units, within which secondary units are selected by strati ed systematic sampling ( Fig. 2a and 2b). Different sampling approach were used at the second stage for (1) forests, (2) urban Trees outside Forest (TOF) and (3) rural TOF. Over the selected districts, systematic square grids based on Survey of India toposheets were laid out that de ne the sample locations for the secondary sampling units.
Two 1 ¼′ x 1 ¼′minute sub grid within each grid of 2 ½′x 2 ½′minutes forms the basic sampling units. The intersection of diagonals of such 1 ¼′x 1 ¼′minute sub grid marked as the center of plot on the map where a plot of 0.1 ha in size (Fig. 2c) was laid out to inventory all the trees with diameter at breast height (dbh) ≥10 cm. Tewari and Kleinn (2015) describeed the developmental history of NFI in India in detail and its biodiversity relevance has been mentioned by Thakur et al. (2018). Since the NFI data covers a signi cant part of WH and in an in systematic manner, it allows the assessment of the distribution pattern of tree species encountered in NFI. The NFI data collected during the period 2002-2015 was used for the study. The data afterwards was not used since the methodology got changed for the new NFI in the year 2016. There were one thousand eight hundred and thirty two (1832) plots found with at least one of the 15 selected tree species across the large environmental gradients. These selected 15 tree species were found in more than 10 plots.
The literature recommends few major environmental variables characterizing macroscale hypotheses for biodiversity patterns (Mouchet et al. 2015) out of which we selected 4 climatic variables (Table 1): monthly climatic variables and bioclimatic variables were obtained from WorldClim database (http://www.worldclim.org). These variables were selected to re ect the averages as well as the climatic boundaries of the species range. These data are a set of global climate layers with a spatial resolution of approximately 1 km 2 . The coordinates of the plots were used to get the value of various climatic data used in the study.
Various soil parameters used in the study have been de ned in the NFI manual: "Manual of instructions for eld inventory-2002" (http:/fsi.nic.in/documents/manualforest_invetory_2.pdf) speci cally designed for eld inventory. Humus is the decomposed organic matter (leaves, twigs, branches etc.) forming the upper most soil layer and distinguished from the undecomposed leaf litter. Soil consistency comprises the nature of soil material that is expressed by the degree and kind of cohesion or resistance to deformation or rupture. Texture of soil refers to relative occurrence of clay, silt and sand particles. Coarse fragments like gravel, boulders, loose stones present in the soil mass. Depth of soil will be estimated from the soil sample plots and guessing the remaining depth. Apart from this soil erosion, texture (clay, sand and loamy soil) and color were noted down in NFI, however since slope is considered soil erosion is not considered, soil texture corresponds to different size of soil particles -it might be useful indicator for unique species but may not collectively indicate biodiversity, soil color that cannot be linked to biodiversity have not been considered in this study. The relationship between structure and microenvironment was estimated by simple correlation and Canonical Correspondence Analysis (CCA). We used CCA to establish the importance of environmental drivers for species distribution using PCORD software. CCA is an ordination technique, which shows nonlinear relations between species and environmental variables and prefers the best weights for environmental variables. Abundance data for trees were used for this. In CCA, environmental variables were standardized to Z score and squared root transformation of species data were carried out. Monte Carlo permutation test by using 999 permutations was carried out by PCORD to test the signi cance of the eigenvalues of all the canonical axes. In the ordination diagrams, species names was abbreviated using ve letters from the scienti c name of each plant by combining the single initial letters for the genus and four letters of species.
The six soil variables (Table 1) were analyzed through literature. The soil erosion was most important because it reduces forest productivity by decreasing the soil water availability and removing plantavailable nutrients (Elliiot et al. 1999). The other variables were subjected to factor analysis to see the most correlated variable so as to reduce the number of variables nally to be subjected to CCA. The soil depth shows some correlation to rockiness and coarse fragments individually and a correlation was also found between rockiness and coarse fragments while weaker correlation was observed among others ( Table 2). The KMO value (Table 3) shows adequacy of the sample. Therefore, out of listed soil variables in Table 1, two variables soil depth and soil erosion was considered to represent the edaphic components in CCA. A main matrix composed of 1832 plots and 11 environmental factors (included climatic/ topographic/ edaphic variables) and second matrix composed of 1832 plots and 15 species (containing tree abundance per species/ per sample unit) was constructed to analyze the relationships between the species distribution and the environmental factors.

Results
The CCA showed that the eigenvalues for the rst three axes were 0.805, 0.402, 0.149 and speciesenvironmental correlations for these axes were 92%, 67% and 42.6% respectively (Table 2a). These values along with the results of Monte Carlo test (p < 0.001), indicated a highly signi cant correlation between the environmental variables and tree species distribution. The gradual decrease of eigenvalues of rst three axes indicated a well-structured dataset assuring that the CCA analysis performed well in describing relationships between vegetation and environmental variables (Table 2a). The CCA biplot diagram (Fig. 3), clearly shows that the species are distantly dispersed suggesting different environmental variables in uences the distribution pattern to varying extent. The rst axis was signi cantly (p > 0.001) responsible for the species environment relationship (Table 2c).
Cumulatively the rst three axes of CCA ordination clari ed 14.4% of the variance in the species data (table 4a). The higher value of species-environmental relation (92.3%) indicated that species data were strongly related to the measured environmental variables (Table 4c). The Monte Carlo test was attempted to test the signi cance of an observed test statistica (Table 4b).
The Fig. 3 and Table 5 clearly shows that the rst CCA axis negatively correlates with altitude while positively correlates with annual mean temperature (i.e. with increasing altitude the annual mean temperature decreases.) and annual precipitation. The second CCA axis correlates with temperature seasonality. Majority of these 15 tree species are strongly in uenced by altitude and temperature seasonality. The table 4a, 4b and 4c shows the overall variance in the data, species species correlations and especies -environemnt correlations as obtained in the results. Table 4a Number p is not reported for axes 2 and 3 because using a simple randomization test for these axes may bias the p values.  p is not reported for axes 2 and 3 because using a simple randomization test for these axes may bias the p values.  As can be seen biplot ordination diagram (Fig. 3), Shorea robusta (Srobu), a tropical tree species falls in entirely different zone of environment. P. roxburghi is found in sub-tropical zone and shows positively distributed along b12 (annual precipitation) which is supported by literature where the growth of chir pine has been found positively correlated with the precipitation (Sigdel et al., 2018;Chaunhan et al.;, Aryal et al., 2018. The monsoonal and non-monsoonal regimes are known to in uence the tree distribution . P. roxburghi, S. robusta and Q. oblongatum are seems to be driven by precipitation and its seasonality thus may be in uenced strongly by monsoon which forms extensive forest in WH. Rise in temperature and resultant enhanced evapo-transpiration desiccates the seeds of Q. oblongata more rapidly in the soil seed banks due to global temperature rise has been hypothesized in literature (Bhatt et al. 2015). P gerardiana, P. wallichiana, P. smithiana have been reported to be non monsoonal and are falling in the region of low precipitation seasonality. Enough moisture and temperature during favorable growth season are key determinant for growth and development in S. robusta. (Kumar and Chopra, 2018) Four of the fteen tree species -Betula utilis (Butil), Pinus gerardiana (Pgera), Abies spectabilis (Aspec) and Abies pindrow (Apind) falls in the high elevation zone quite separated from other tree species. The literature also reports B.utilis and A. spectabilis forming associations in around 4000 m in treeline in WH (Singh and Singh, 2002). A. spectabilis shows good regeneration between 2268-3000m as evident in literature (Sharma et al.2017). Phytosociologically, P. gerardiana shows high endemism in its natural zone and reported to occur in more than 70% of the area wherever it exist in north-west Himalaya (Kumar et al. 2016).
Picea smithiana and Cedrus deodara are distributed above 2000 m whereas Pinus wallichiana is found along the entire length of WH (1200-3800m). Literature reports these species at very high altitude usually in dominant form or codominant form (Liang et al, 2014;Gairola and Todaria, 2008). B.utilis and P. gerardiana shows no association with any other tree spp. A.pindrow, B.utilis and Acer caesium have been found to occur in equal importance value indices (IVI) in Kumaun region of WH (Gairola et al., 2008). The oak forest is distributed along the altitudinal gradient in WH with, Q. leucotrichophora (1200-2200 m), Q. oribunda (2201-2700 m) and Q. semecarpifolia (2701-3300 m) .
We didn't nd B. alnoides as most commonly found tree species under regeneration in any of the NFI plots under study. P. smithiana showed its presence in 3.82% of total plots wherever it was found while B. utilis and R. arboreum were found on less than 15% of total plots wherever it occurred. All the rest were present in more than 22% of their total plot of occurrences. We found six tree species reported in less than 33% (of their total occurrence) with reported as most commonly found tree species under regeneration. Their regeneration along the altitude (Fig. 4) clearly shows their presence throughout the altitude. We also analyzed (for all the fteen tree species) the presence of trees (dbh > = 10 cm) in parallel to their regeneration along the altitude as observed in the inventory plots and found that in except for A. spectabilis, P. wallichiana and S. robusta, the minimum altitude at which regenerating individuals are found are higher than the altitude where parent trees are reported from. The regenerations of B. utilis, A. spectabilis, A. pindrow, and A. acuminatum have been reported to be evidently expanding into alpine zone (Sharma et al, 2018).

Conclusion
The study proves that the distribution of major forest forming tree species in WH is governed by the altitude, annual mean temperature and precipitation seasonality. Considering the altitude to be a stable parameter, under the changing climatic regime, the annual mean temperature and precipitation seasonality are likely to be the major driving factors for any alteration in the distribution of forest forming tree species in WH. Since eleven tree species reportedly showing dominant regeneration at the altitude above the one where parent trees are growing, this pose a question whether forest forming tree species are migrating upwards in general.
The manner in which increasing established sapling at higher elevation of Pinus wallichaiana (added by growth ring study) indicated upward shift in WH (Dubey et al, 2003), can this phenomenon of nding regenerating tree individuals in visibly dominant form in large area of 2 ha at higher elevation be considered as sign of upward shift in WH. If so, at least this study is giving a signal towards this trend.
With increasing altitude availability of space will decrease and consequently only few of many species will be able to survive and perpetuate.
Since the data belongs to 2002-15, nding B. alnoides to be totally absent from important species under regeneration in NFI plots could be an alarming sign since there may be a regenerating individuals but may not be found in visibly dominant condition.
Thus it becomes very important that any change or unusual nding in the distribution pattern/ range of these tree species must be reported to understand the in uence of environmental determinants on tree species constituting a particular forest ecosystem and further on ecosystem services provided by that ecosystem. Especially the species like Rhododendron arboreum, Betula utilis, Pinus gerardiana and Quercus semecarpifolia seemingly shows dominant regeneration at an altitude where parent trees are found indicating healthy ecosystem. An intensive regeneration based study of forest forming tree species through permanent observation plot studies may provide insight into micro climatically operating drivers. Such studies based on huge datasets from the ground and their linking with the global data helps to understand the observed trend and strengthen the consensus on scienti c ndings. Availability of data and materials: Data will not be shared as it belongs to NFI of India and government policy does not allow putting the raw data on any sort of public domain.
Competing interest: The authors declare that there is no con ict of interest.
Funding: This research did not receive any speci c grant from funding agencies in the public, commercial, not -for-pro t sectors.
Author's contribution: Arun Kumar Thakur carried out the study and drafted the manuscript.
Rajesh Kumar conceived and designed the study.
Raj Kumar Verma coordinated and helped to draft the manuscript.
Pankaj Kumar helped in the statistical analysis and helped in framing the draft.  Regeneration of six tree species (found in less than 33% of their occurrence as major species under regeneration) along altitude