An Artificial Intelligence Dataset for Solar Energy Locations in India

Rapid development of renewable energy sources, particularly solar photovoltaics, is critical to mitigate climate change. As a result, India has set ambitious goals to install 300 gigawatts of solar energy capacity by 2030. Given the large footprint projected to meet these renewable energy targets the potential for land use conflicts over environmental and social values is high. To expedite development of solar energy, land use planners will need access to up-to-date and accurate geo-spatial information of PV infrastructure. The majority of recent studies use either predictions of resource suitability or databases that are either developed thru crowdsourcing that often have significant sampling biases or have time lags between when projects are permitted and when location data becomes available. Here, we address this shortcoming by developing a spatially explicit machine learning model to map utility-scale solar projects across India. Using these outputs, we provide a cumulative measure of the solar footprint across India and quantified the degree of land modification associated with land cover types that may cause conflicts. Our analysis indicates that over 74% of solar development In India was built on landcover types that have natural ecosystem preservation, and agricultural values. Thus, with a mean accuracy of 92% this method permits the identification of the factors driving land suitability for solar projects and will be of widespread interest for studies seeking to assess trade-offs associated with the global decarbonization of green-energy systems. In the same way, our model increases the feasibility of remote sensing and long-term monitoring of renewable energy deployment targets.