Mapping the rapid development of photovoltaic power stations in northwestern China using remote sensing

Many leading countries are boosting renewables, especially solar energy, as a major way to mitigate future energy crises and climate change. Particularly, in China, the number and scale of photovoltaic (PV) power stations have grown unprecedentedly in the last decade. There is an urgent need to monitor the PV power development in order to accurately estimate national renewable potentials and understand the ecological impacts. However, there are few efforts towards providing spatially explicit and time-series datasets of PV development at the regional and national scales. To fill the gap, this study proposes an integrated remote sensing approach for PV power stations mapping by combining image segmentation and object-based classification (ISOC) techniques. We took five northwestern provinces of China as an illustration and produced 30-m medium-resolution PV power station distribution maps from 2007 to 2019. Our analysis shows that the total area of PV power stations in the five provinces increased to 722 km 2 in 2019, with producer, user and overall accuracies of 86%, 100% and 93%. Of the 309 PV station clusters (hereafter, PV parks), the top 7% largest ones account for 61% of the total area of PV power stations, indicating that PV power stations in the Northwest tend to be developed in the form of large-scale centralized PV parks. The land used for PV power stations was mainly converted from four land cover types: Gobi Desert, sandy land, sparse grassland, and moderate grassland. The central government policy on facilitating clean energy played a major role in driving the rapid expansion of PV parks across the country. The methodology and results of this study will help policymakers, researchers, and practitioners to develop corresponding industrial standards and environmental regulations to mitigate the potential environmental impacts, and promote more sustainable development of solar energy in China and beyond.