A comprehensive transferability evaluation of U-Net and ResU-Net for landslide detection from Sentinel-2 data (case study areas from Taiwan, China, and Japan)

Omid Ghorbanzadeh,Thomas Blaschke,Hejar Shahabi,Pedram Ghamisi,Alessandro Crivellari

doi:10.1038/s41598-021-94190-9

Abstract

Earthquakes and heavy rainfalls are the two leading causes of landslides around the world. Since they often occur across large areas, landslide detection requires rapid and reliable automatic detection approaches. Currently, deep learning (DL) approaches, especially different convolutional neural network and fully convolutional network (FCN) algorithms, are reliably achieving cutting-edge accuracies in automatic landslide detection. However, these successful applications of various DL approaches have thus far been based on very high resolution satellite images (e.g., GeoEye and WorldView), making it easier to achieve such high detection performances. In this study, we use freely available Sentinel-2 data and ALOS digital elevation model to investigate the application of two well-known FCN algorithms, namely the U-Net and residual U-Net (or so-called ResU-Net), for landslide detection. To our knowledge, this is the first application of FCN for landslide detection only from freely available data. We adapt the algorithms to the specific aim of landslide detection, then train and test with data from three different case study areas located in Western Taitung County (Taiwan), Shuzheng Valley (China), and Eastern Iburi (Japan). We characterize three different window size sample patches to train the algorithms. Our results also contain a comprehensive transferability assessment achieved through different training and testing scenarios in the three case studies. The highest f1-score value of 73.32% was obtained by ResU-Net, trained with a dataset from Japan, and tested on China’s holdout testing area using the sample patch size of 64 × 64 pixels.

Highlights

Earthquakes and heavy rainfalls are the two leading causes of landslides around the world
Landslide inventory maps are usually prepared by extracting the landslide information from Remote Sensing (RS) images, including optical satellite images and synthetic aperture radar (SAR) data, because of the relatively low cost associated with obtaining RS images and their wide coverage a rea[13,14]
Chen et al.[31] used a Convolutional Neural Network (CNN) algorithm based on the Gaofen-1 High-Resolution (HR) RS image and slope information derived from a 5 m spatial resolution digital elevation model (DEM) for landslide extraction in three different cities in China

Summary

Introduction

Earthquakes and heavy rainfalls are the two leading causes of landslides around the world. Chen et al.[31] used a CNN algorithm based on the Gaofen-1 High-Resolution (HR) RS image and slope information derived from a 5 m spatial resolution digital elevation model (DEM) for landslide extraction in three different cities in China. They used 28 × 28 pixels sample patches and achieved a quality percentage of 61%. Ghorbanzadeh et al 2019 evaluated two different CNN algorithms trained by different sample patches with a range of window sizes from 12 × 12 to 48 × 48 pixels and compared the landslide detection results with those of state-of-art Machine Learning (ML) models. Using a post-processing method for mask operations and screening, they were able to achieve the highest accuracy of more than 80%

Objectives

Methods

Results

Discussion

Conclusion