Search

Data on the distribution of plant species at spatial (grid) scales are required as input for integrative analysis along with related climate, environment, topography and soil data. Although the world's scientific community is increasingly generating data on plant species at various spatial grids and statistically interpolating and extrapolating the available information, data on plant diversity from the Asian continent are scant. Such data are unavailable for India, the mainland of which has part of three of the world's 36 biodiversity hotspots. Although sufficient field sampling is always impossible and impractical, it is essential to utilize fully any available database by adjudging the sampling sufficiency at a given scale. In this work, we used an exhaustive database of the plant species of the Indian mainland that was sufficient in terms of sampling vegetation types. We transformed the data, obtained the distribution at the 1° and 2° spatial grid levels and evaluated the sampling sufficiency at acceptable threshold limits (60% to 80%). The greatest species richness values recorded in the 0.04 ha quadrant, 1° grid and 2° grid were 59, 623 and 1244, respectively. Clench model was significantly (p value < 0.001) fitted using the plant species data at both the grid levels with a very high coefficient of determination (>0.95). At an acceptable threshold limit of 70%, almost all the grids at the 2° level and more than 80% of the grids at the 1° level were found to be sufficiently sampled. Sampling sufficiency was observed to be highly scale-dependent as a greater number of 2° grids attained asymptotic behaviour following the species-area curve. Grid-level sampling insufficiency was attributed to lower numbers of sampling quadrats in forests with poor approachability, which coincided with the world biodiversity hotspots', suggesting that additional sampling was required. We prescribe the use of the 1° and 2° spatial grids with sufficient sampling for any ecological analysis in conjunction with other data and thereby offer grid-level plant species richness data for the Indian mainland for the first time.

India is one of the 12 mega biodiversity countries of the world, which represents 11% of world's flora in about 2.4% of global land mass. Approximately 28% of the total Indian flora and 33% of angiosperms occurring in India are endemic. Higher human population density in biodiversity hotspots in India puts undue pressure on these sensitive eco-regions. In the present study, we predict the future distribution of 637 endemic plant species from three biodiversity hotspots in India; Himalaya, Western Ghats, Indo-Burma, based on A1B scenario for year 2050 and 2080. We develop individual variable based models as well as mixed models in MaxEnt by combining ten least co-related bioclimatic variables, two disturbance variables and one physiography variable as predictor variables. The projected changes suggest that the endemic flora will be adversely impacted, even under such a moderate climate scenario. The future distribution is predicted to shift in northern and north-eastern direction in Himalaya and Indo-Burma, while in southern and south-western direction in Western Ghats, due to cooler climatic conditions in these regions. In the future distribution of endemic plants, we observe a significant shift and reduction in the distribution range compared to the present distribution. The model predicts a 23.99% range reduction and a 7.70% range expansion in future distribution by 2050, while a 41.34% range reduction and a 24.10% range expansion by 2080. Integration of disturbance and physiography variables along with bioclimatic variables in the models improved the prediction accuracy. Mixed models provide most accurate results for most of the combinations of climatic and non-climatic variables as compared to individual variable based models. We conclude that a) regions with cooler climates and higher moisture availability could serve as refugia for endemic plants in future climatic conditions; b) mixed models provide more accurate results, compared to single variable based models. [ABSTRACT FROM AUTHOR]