BUILDING DAMAGE EXTRACTION TRIGGERED BY EARTHQUAKE USING THE UAV IMAGERY

Large-scale maps of building rooftop area (BRA) are crucial for addressing policy decisions and sustainable development. In this paper, we propose a deep-learning method for high-resolution BRA mapping (2.5 m) from Sentinel-2 imagery (10 m). The resulting China building rooftop area dataset (CBRA) is the first multi-annual (2016–2021) and high-resolution (2.5 m) BRA dataset in China. Cross-comparisons show that the CBRA achieves the best performance in capturing the spatiotemporal information.

The spatial and temporal distribution characteristics of aftershocks around the fault are analyzed according to the stress changes after the main earthquake. The model can be used to predict the multi-timescale anisotropy distribution of aftershocks fairly. The finite fault model of the main earthquake is used in the construction of the prediction model. The model is a deep neural network; the inputs are the stress components of each point; and the output is the probability of an aftershock.