To isolate the effects of mining, you need to know precise locations of all mines, where economic activity is taking place over time, and the status of environmental quality at the local level.
We have assembled data on the precise location of all operating mines in India, and combined this with high resolution measures of socioeconomic status and environmental change. Using neural networks to identify mine expansions and contractions, we can examine the impact of mining operations on the people who live in the immediate vicinity. This work is ongoing.
Automated scrape of cloud-free imagery for thousands of manually validated mine locations.
Over 60GB of processed training imagery and >1TB test imagery across all of India.
CNN model identifies surface mine locations and size at 5-day temporal resolution.
SANITY CHECK. We expect the classifier to pick up on the same visual characteristics of surface mining as the human eye. The middle image of the "most incorrect mine" is classified as a mine in the training data, but does not appear to be a surface mine from the air; the neural network defined this area as devoid of mining activity, and a human being would have likely done the same.
We have other work on the ecological impacts of growth and the political economy of mining. Explore our paper on the impacts of transportation infrastructure on India's forests forthcoming at the Economic Journal, and our paper on rent-seeking and criminal politicians (under revision at the Review of Economics and Statistics).