Citation

Meena, S.R., Ghorbanzadeh, O., van Westen, C.J., Nachappa, T.G., Blaschke, T., Singh, R.P., Sarkar, R. (2021). Rapid mapping of landslides in the Western Ghats (India) triggered by 2018 extreme monsoon rainfall using a deep learning approach. Landslides, 18(5): 1937-1950. Link to paper

Abstract

This study used PlanetScope imagery and deep learning convolutional neural networks (CNNs) to rapidly map rainfall-induced landslides in the Kodagu district of Karnataka state in the Western Ghats of India following the 2018 extreme monsoon event. A fourfold cross-validation approach was employed to select training and testing data, removing potential random effects in model performance. Topographic slope data was incorporated as auxiliary information to enhance model accuracy. The mean accuracies of correctly classified landslide pixels were 65.5% when using only optical data, which increased to 78% with the addition of slope data. The resulting landslide inventory map created using slope data combined with spectral information significantly reduced false positives compared to optical-only approaches. This research demonstrates the practical application of deep learning for emergency response and rapid landslide mapping following extreme weather events. The methodology provides a scalable approach for creating accurate post-disaster landslide inventories that can inform relief efforts and risk assessment in landslide-prone mountainous regions.

AI Visual Interpretation (Beta)