Extended Data Fig. 1: Overview of the two-stage method for global rooftop area estimation.
From: Worldwide rooftop photovoltaic electricity generation may mitigate global warming
The top-down stage involves applying a deep learning model to high-resolution remote sensing imagery to quantify rooftop areas in selected representative regions. The bottom-up stage uses an ensemble of random forest models to model the non-linear relationship between the rooftop areas identified in the first stage and various geospatial variables, enabling the extrapolation of rooftop area estimates at a global scale.
