The semi-arid savanna grasslands (SSG) of peninsular India are important habitats with a unique assemblage of endemic species. They are also critical to millions of pastoralists and agro-pastoralists for whom these are the major grazing areas for their livestock. Yet, the forest-centric bias towards vegetation classification in India has failed to properly recognise this biome, and as a result, it has been neglected and subject to large-scale landuse change. Most efforts at mapping the extent of this biome using remote sensing data have tended to underestimate the extent due to difficulties in differentiating between grass cover and dryland agriculture. We used a novel approach of using multi-date MODIS NDVI data and an unsupervised classification to create a probabilistic output of SSG occurrence. We also used ancillary data to predict occurrence of SSG using NDVI and Bioclimatic layers with a regression tree rule-set classification, which identifies the bioclimatic envelope of SSG, and the predicted extent. To determine the current protection status of SSG, we conducted a GAP analysis with the current protected area network. The results show that the SSG biome is primarily spread over eleven states of India with 1.2 to 9.1% coverage of the geographic area for the high probability of SSG occurrence class. However, the overall protection status of SSG in these states is low, with only 0.1 to 8.7% under the PA network. The states with the highest area of SSG include Maharashtra, Gujarat and Rajasthan. To conserve and manage the last remaining areas of SSG in the country, we suggest a sentinel landscape approach, with a systematic conservation prioritisation exercise combining both biodiversity value as well as human-use to ensure a sustainable and equitable use of these threatened ecosystems.