Landslide susceptibility modeling by interpretable neural network

Landslides are notoriously difficult to predict because numerous spatially and temporally varying factors contribute to slope stability. Artificial neural networks (ANN) have been shown to improve prediction accuracy but are largely uninterpretable. Here we introduce an additive ANN optimization framework to assess landslide susceptibility, as well as dataset division and outcome interpretation techniques. We refer to our approach, which features full interpretability, high accuracy, high generalizability and low model complexity, as superposable neural network (SNN) optimization. We validate our approach by training models on landslide inventories from three different easternmost Himalaya regions. Our SNN outperformed physically-based and statistical models and achieved similar performance to state-of-the-art deep neural networks. The SNN models found the product of slope and precipitation and hillslope aspect to be important primary contributors to high landslide susceptibility, which highlights the importance of strong slope-climate couplings, along with microclimates, on landslide occurrences. The product of slope and precipitation, along with hillslope aspects, are the main physical factors responsible for landslides in the easternmost Himalayas, according to an interpretable superposable neural network model.

Automated summary

We introduce an interpretable superposable neural network (SNN) optimization method to assess landslide susceptibility, which achieves high accuracy and low model complexity. By training models on landslide inventories from three easternmost Himalaya regions, we confirm the superior performance of our SNN model compared to physically-based and statistical models. We identify slope and precipitation product, as well as hillslope aspect, as important contributors to landslide susceptibility.