The pattern of bird distribution along the elevation gradient of the Sutlej River basin, western Himalaya, India

We examined the species richness of birds along the elevation gradient of the Sutlej River basin in Himachal Pradesh in the western Himalaya of India. Birds were sampled at 318 sites categorized into 16 elevation bands ranging from 498 to 3700 m between June 2012 and April 2013. A total of 203 bird species were recorded. Species richness showed a monotonic decline with increasing elevation, with 27% of species recorded within a narrow elevation range. We tested the roles of explanatory variables such as environment (temperature, precipitation, area, & Mid-domain Effect (MDE) richness) and habitat (Normalized Differential Vegetation Index (NDVI): July, November & March) on the observed distribution pattern. The observed species richness pattern was strongly correlated with temperature, while three other variables—precipitation, area, and MDE richness—contributed negligibly to the observed pattern. The present study indicates that climatic conditions and vegetation are the major contributors for determining species richness along the Sutlej River basin. Thus, a customized approach is crucial for conservation of species in the elevation range.

Understanding the association between species richness and elevation gradients is essential as it provides insights into the observed patterns and processes responsible for the relation, which in turn supports conservation efforts (Stevens 1992;Raman et al. 2005;Acharya et al. 2011). In view of this, the present study was carried out to document and describe the distribution pattern of bird communities along the elevation gradient in Sutlej River basin of Himachal Pradesh belonging to the western Himalayan part of India. The western part of Himalaya is an important area of regional endemism and is a priority region for conservation (Rahmani & Islam 2004). The study aims to provide comprehensive information on the species richness pattern from yet another important region of the western Himalaya, thereby contributing more information to the debatable topic of species distribution patterns along an elevation gradient.

Study area
The Sutlej River basin in the western Himalayan mountain range (30.85-32.91 0 N & 76.26-79.00 0 E) is situated in the state of Himachal Pradesh (Fig. 1). The entire region is characterized by numerous mountain ranges, hills, rivers, and forests (Rahmani & Islam 2004;Sharma & Kumar 2012). The Sutlej River is one of the major physical features of Himachal Pradesh and flows in the southwesterly direction, bisecting the state. The Sutlej River basin covers wide-ranging variations in elevation (498-6685 m), climate (tropical to temperate), and vegetation types (tropical forest to alpine pastures). The temperature varies from a minimum of -3.8°C in February during winter to a maximum of 31.9°C in May during summer. The average annual rainfall (June-September) is 1,035.1mm (Indian Meteorological Department 2012) and precipitation in the form of snow is recorded during winter from December to February. We stratified the study area into 16 elevation bands of 200m interval. Within each band, point counts were fixed in proportion to the availability of the area. ASTER Global Digital Elevation Model (GDEM) was used to obtain the area available in each elevation band. A total of 318 sampling points were surveyed at different elevation bands (Table 1).

Bird sampling
The present study was conducted from June 2012 to April 2013. To assess the difference in the richness and abundance of birds along the elevation gradient, point count method (Gaston 1975) was followed. A minimum of 1km distance (calculated using QGIS 2.12.1) was maintained between the consecutive points. In each point count, a single observer recorded every bird seen inat 10min intervals (Ralph et al. 1995;Raman 2003) within the fixed distance of 50m radius. The number of individuals and species were recorded based on both visual and vocal estimation. The sampling was avoided in inaccessible areas and at point count stations with evidence of human settlement. The surveys were carried out 30min after dawn and 30min before sunset on days with suitable weather conditions. The point counts were conducted from 0600 to 1100 hr in the morning and 1500 to 1800 hr in the evening. The

12718
probability of twice counts was minimized by excluding flocks that appeared in the same direction. To enhance the probability of detecting elusive or rare species, point counts varied between one to four replications in each season (summer, winter, & rainy season). Thus, a total of 1,239 point counts were conducted throughout the study. The average species richness of sampling replications has been used for analyses. The migratory status and taxonomic system used in this study followed Grimmett et al. (2011).

Habitat variables
We extracted temperature, precipitation, and other 19 bioclimatic variables from the WORLDCLIM database for the analysis. The database presents climatic data from 1950-2000 (50 years) at a spatial resolution of 1km (Hijmans et al. 2005). We used normalized differential vegetation index (NDVI) values from 2012 and 2013 to measure primary productivity, downloaded from Google Earth Engine (LANDSAT 7 with 32 days interval: https:// earthengine.google.com/datasets/).

Analysis
We used non-parametric estimator (Chao 1) to calculate the estimated species richness using the statistical software program EstimateS 9.1 (Colwell & Coddington 1994;Colwell 2016). A precise estimate of observed and estimated species richness for a sampling interval was assumed if the species accumulation curve approached a plateau (Chao et al. 2005). Linear regression analysis (Sokal & Rohlf 1995) was performed against observed species richness with elevation to assess the species distribution pattern.
We performed Monte Carlo simulation using the software Mid-Domain Null model (McCain 2004) for testing geometric phenomena or mid-domain patterns of species ranges. The computed action simulates species richness curves based on empirical ranges or range midpoints within a bounded domain. It is based on the analytical--stochastic models (Colwell & Hurtt 1994). The species ranges are randomly shuffled within a bounded geographical domain and resulting peak species richness at intermediate elevation (Colwell & Lees 2000). To test the geometric boundary constraints in relation to the distribution of species richness, 95% prediction curves were raised based on 50,000 simulations without replacement using empirical range sizes (with 200m elevation increases). Linear regression was applied between empirical species richness and the average of the predicted number of species to assess the model fit.
We used simple linear regression model to explore the effect of each explanatory variable that could support the bird species richness along an elevation gradient. The explanatory variables such as mean annual temperature, mean annual precipitation, area, NDVI for July, October, November, February, March & April, and MDE predicted richness were regressed with estimated species richness (Chao 1). Among the environmental variables, temperature, precipitation, and area were highly correlated, similar to the habitat variables NDVI for July, November, and March. These highly correlated variables were selected for stepwise multiple regression analysis to identify the best explanatory variables that predicted the estimated species richness along elevation. In the set of variables, we selected estimated species richness as dependent variable and others as independent variables. In each successive step, the model with significant results was dropped and tried with other variables to find the next significant one.
We examined the elevation range profile for each species occurring at all elevations, whether the species utilized all intermediate elevation or was stuck between elevation maximum and minimum limits (Patterson et al. 1998). We then performed Pearson correlation to test the relationship between the range size of each species against the lower and upper limits of its elevation ranges. Regression and correlation analysis were performed in statistics software package SPSS 17.0 (Chicago, IL, USA).

Species richness
A total of 203 bird species, including 147 residents and 56 non-residents, were recorded across 16 elevation bands during the study period. The number of species detected in each elevation band varied from 20 to 140. The bird species richness was highest at around 700m elevation in the Sutlej River basin. The species accumulation curves across all the bands almost reached an asymptote, except in the elevation gradient above 2100-2500 m and 3500-3700 m (Fig. 2). The observed species richness corresponded well with the estimated Chao 1 (r = 0.99, p < 0.000).

12719
pattern in the Sutlej River basin showed a monotonic decrease in species richness with increasing elevation. A similar pattern was observed in the MDE, where the data revealed poor fit to the MDE predictions. The range size of resident bird species showed that only 12.5% of the empirical points occurred inside the predicted range of MDE null model (Fig. 4). The regression relationship between empirical species richness and the mean of the predicted richness is insignificant (F (1.14) = 0.026, p = 0.87, R 2 = 0.002). MDE for all the bird species was also insignificant (F (1.14) = 0.001; p = 0.97; R 2 = 0.0001) and 81% (13/16) of the empirical points fell out of the predicted limits (Fig. 5).
In simple linear regression, estimated species richness (Chao 1) was significantly correlated with mean temperature (F (1.14) = 79.10, p = 0.000, R 2 = 0.85). Other environment factors such as MDE predicted richness and habitat variables such as NDVI for October, February, and April were insignificant. In the stepwise regression first model, the annual mean temperature was significantly correlated with estimated richness (Table 2). After removing the temperature, the second model habitat variable, NDVI March, was significantly correlated and the third model precipitation and area together showed a significant correlation.
Variation in empirical species richness in different elevation bands matched closely with the observed species richness. Abrupt turndown in species richness at elevation ranges 900-1,100 m and 1,300-1,700 m, however, was closely associated with predicted range boundaries (Fig. 5). It shows that the deviation of MDE prediction between empirical and observed species richness slightly goes wrong in Sutlej River basin.

Elevation range profile
The elevation range profile for birds in Sutlej River basin recorded a higher number of species (72% of species) occurring below 2500m elevation gradients.
A total of 55 bird species was restricted to very narrow elevation range size (200m), 10 species were restricted above 3,000m (Table 3), and 17 species were spread all over the gradients (Fig. 6). A total of 77 (38%) bird species were below 600m elevation range and 19 species occurred above 3,000m elevation range. The elevation range size of low elevation species, those occurring below 2,000m (upper limit: r = 0.825, p < 0.0001; lower limit: r = -0.383, p <0.0001), and high elevation species, those occurring above 2000m (upper limit: r = 0.282, p < 0.01; lower limit, r = -0.813, p < 0.0001), was positively correlated with upper limit and negatively correlated with lower limit. The range size of low elevation species, however, was inclined to decrease with elevation and that of high elevation species to increase with elevation.

DISCUSSION
In Sutlej River basin, we found a clear association between bird species richness and the factor (temperature) influencing the species distribution across the elevation bands. The MDE analysis also showed the decreasing trend of species richness along the alelevation gradients. In addition, species range size revealed that most of the species have a narrow elevation range within the sampled elevation gradients.

Species richness
Previous studies along Sutlej River basin examined a section of the elevation gradient for assessing the avian communities (Gaston 1993;Thakur et al. 2006;Miller et al. 2008;Jayapal & Ramesh 2009;Kulkarni & Goswami 2012). This is the first study

Mid-domain analysis
The bird species richness along the elevation gradient was greater in three distinct peaks: 500-700 m, 1500-1700 m, and 3300-3500 m. This pattern reflects that the species richness of birds in Sutlej River basin is highly patchy. The present study, however, revealed that species richness of birds declined monotonically with increasing elevation. The monotonic decline in species richness along the elevation gradient has commonly been reported in various taxa and regions (Rahbek 1995(Rahbek , 2005Patterson et al. 1998;Lessig 2008). In contrast, a few studies in the Himalaya observed high species richness at middle elevation than higher and lower elevations (Raza 2006;Acharya et al. 2011;Joshi et al. 2012;Joshi & Rautela 2014;Joshi & Bhatt 2015), although they found very little support for the MDE (Acharya et al. 2011). Lack of mid-elevation peak in species richness indicates that geometric constraints have a relatively low influence on the bird species richness pattern in the Sutlej River basin and western Himalaya.
The effect of climatic and productivity variables on the distribution pattern of birds in this river basin was tested. The monotonic decline in species richness was influenced by temperature as evidenced by the stepwise multiple regression model (Table 2). This observed result corresponds with Acharya et al. (2011) andStevens (1992), where the species richness along the elevation gradient was influenced by climatic factors followed by habitat variables. In contrast, other studies in the Himalaya observed that the vegetation composition (Raza 2006;Joshi et al. 2012) and structure (Joshi & Rautela 2014;Joshi & Bhatt 2015) determined the species richness in a gradient, rather than the elevation and other climatic factors. These studies covered a part of the elevation range and did not consider a wideelevation range as that followed in the present study and by Acharaya et al. (2011). This might be the reason for the absence of the influence of climatic variables on bird species distribution in other studies (Joshi et al. 2012).
In resident birds, 19% of observed species richness fells within 95% null model, which might be influenced by the geometric constraints than in other bands. For overall bird species, geometric constraints influenced between 1600 and 1800 m and >3500m elevation, whereas observed species richness fell between 900 and 1700 m and >3500m elevation of null model. The sudden decline in both resident and overall bird species richness above 2000m is in concurrence with increasing elevation and is due to sparse forest cover and harsh climatic conditions. These changes in climatic conditions could be a cause for the decline of species richness above the transition zone (>2000m) in Sutlej River basin.

Elevation range profile
Species richness of birds in Sutlej River basin was greater within 2000m elevation (116 species, 57%) and rest were found above 2000m. The elevation range limits of each species varied in this study area and showed very narrow ranges. Out of the 203 species, 17 species occurred in all sampled elevation bands while 55 species were found only in few elevation bands showing that their range sizes are narrow. Several factors such as habitat structure, inter-specific competition, and environmental tolerance (Jankowski et al. 2010;Laurance et al. 2011) would have driven the species into narrow elevation bands. Species with broader elevation ranges were insectivorous or granivorous. The insectivorous birds such as Indian Roller Coracias benghalensis, Indian Pitta Pitta brachyura, Whiskered Yuhina Yuhina flavicollis & Red-billed Leiothrix Leiothrix lutea and granivorous birds such as Jungle Bush-quail Perdicula asiatica, Laughing Dove Spilopelia senegalensis, and Emerald Dove Chalcophaps indica were observed only at low elevation bands (<2000m). Stevens (1992) stated that the relationship between latitude and elevation differences experienced by the organisms along geographic gradients is due to the breadth of climatic conditions. Species that have narrow elevation ranges tend to have less climatic tolerance and thus are vulnerable to extinction due to global warming (Colwell et al. 2008;Laurance et al. 2011). For instance, broader elevation range species can migrate upward or downward in response to climatic change and could maintain their viable population sizes (Laurance et al. 2011).
This is the first study in the western Himalaya to assess the distribution pattern of birds along a broad elevation gradient (500-3500 m). Our observation revealed that the bird species richness was significantly greater at a lower elevation than at mid-and high elevation, showing a monotonic decline in species richness. This observed pattern is strongly correlated with the climatic variable, atmospheric temperature, in Sutlej River basin. A large proportion of resident birds occurs in a narrow elevation range, indicating the necessity of sustainable conservation efforts. One of the major threats to the birds in Sutlej River basin is the hydropower projects that are reducing the green cover of the mountains, which perhaps is driving away species from their native elevation ranges (especially 600-1600 m) (Pandit & Grumbine 2012). Hence, a highly customized approach is crucial for the conservation of this entire elevation range.

Grey-capped Emerald
Dove

Great White Egret
Ardea alba +