Generating multi-temporal landslide inventories through a general deep transfer learning strategy using HR EO data

Kushanav Bhuyan; Hakan Tanyaş; Lorenzo Nava; Silvia Puliero; Sansar Raj Meena; Mario Floris; Cees van Westen; Filippo Catani

doi:10.1038/s41598-022-27352-y

Generating multi-temporal landslide inventories through a general deep transfer learning strategy using HR EO data

Sci Rep. 2023 Jan 4;13(1):162. doi: 10.1038/s41598-022-27352-y.

Authors

Kushanav Bhuyan^{1

2}, Hakan Tanyaş³, Lorenzo Nava⁴, Silvia Puliero⁴, Sansar Raj Meena^{4

3}, Mario Floris⁴, Cees van Westen³, Filippo Catani⁴

Affiliations

¹ Machine Intelligence and Slope Stability Laboratory, Department of Geosciences, University of Padova, 35131, Padua, Italy. kushanav.bhuyan@studenti.unipd.it.
² Centre for Disaster Resilience, Department of Applied Earth Sciences, Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, 7514 AE, Enschede, The Netherlands. kushanav.bhuyan@studenti.unipd.it.
³ Centre for Disaster Resilience, Department of Applied Earth Sciences, Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, 7514 AE, Enschede, The Netherlands.
⁴ Machine Intelligence and Slope Stability Laboratory, Department of Geosciences, University of Padova, 35131, Padua, Italy.

Abstract

Mapping of landslides over space has seen an increasing attention and good results in the last decade. While current methods are chiefly applied to generate event-inventories, whereas multi-temporal (MT) inventories are rare, even using manual landslide mapping. Here, we present an innovative deep learning strategy which employs transfer learning that allows for the Attention Deep Supervision Multi-Scale U-Net model to be adapted for landslide detection tasks in new areas. The method also provides the flexibility of re-training a pretrained model to detect both rainfall- and earthquake-triggered landslides on new target areas. For the mapping, we used archived Planet Lab remote sensing images spanning a period between 2009 till 2021 with spatial resolution of 3-5 m to systematically generate MT landslide inventories. When we examined all cases, our approach provided an average F1 score of 0.8 indicating that we successfully identified the spatiotemporal occurrences of landslides. To examine the size distribution of mapped landslides we compared the frequency-area distributions of predicted co-seismic landslides with manually mapped products from the literature. Results showed a good match between calculated power-law exponents where the difference ranges between 0.04 and 0.21. Overall, this study showed that the proposed algorithm could be applied to large areas to generate polygon-based MT landslide inventories.