Fine-scale classification and mapping of subalpine-alpine vegetation and their environmental correlates in the Himalayan global biodiversity hotspot

Subalpine-alpine vegetation of Himalayan global biodiversity hotspot forms the highest and unique ecosystem of the world. These ecosystems inhabit diverse cold adapted plants, which are currently threatened by global warming. Deciphering vegetation forms and their ecological niches is pre-requisite for evolving conservation strategies. Emerging remote sensing datasets, processing techniques and platforms offer potential to map fine-scale vegetation patterns at ecoregion level. We conceptualised a four-fold classification scheme considering climate, vegetation physiognomy, floristics, and gregarious formations for the subalpine-alpine vegetation of the Western Himalaya spanning over 45,202 km2. Sentinel-2 satellite images were classified using a combination of rule-based and machine learning approach i.e. Random Forest in Google Earth Engine to generate vegetation map at regional scale. Reflectance bands alone provided an overall classification accuracy of 76.46% (kappa 0.78), while, the addition of vegetation indices improved the accuracy to 84.43%. (kappa 0.79). When topographical variables were also considered, the accuracy increased to 91.71% (kappa 0.81). The vegetation map at 10 m resolution represents in total 23 vegetation classes covering subalpine zone (9 coniferous forests, 4 broad-leaved forests, 2 scrubs, bamboo brake and grassland) and alpine zone (2 scrubs and 4 herbaceous). Study enhances knowledge on the coverage, distribution, abundance, diversity of subalpine and alpine vegetation and ecological amplitudes with respect to temperature, precipitation, elevation and aspect. The study outcomes are useful for developing landscape as well as species specific conservation planning and bioresource utilization.

Automated summary

The subalpine-alpine vegetation in the Himalayan global biodiversity hotspot is the highest and unique ecosystem in the world, which is currently threatened by global warming. A classification scheme was developed to map the vegetation patterns in the Western Himalaya region using remote sensing datasets and machine learning approaches. The study provides valuable information for conservation planning and bioresource utilization.