Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evolution of Data-driven Single- and Multi-Hazard Susceptibility Mapping and Emergence of Deep Learning Methods

Jaya Sreevalsan-Nair, Aswathi Mundayatt

TL;DR

The paper addresses how data-driven methods for single-hazard susceptibility mapping have evolved and how these approaches extend to multi-hazard susceptibility mapping (MHSM). It reviews SHSM workflows, highlights DL as a rising method with strong performance in FSM/LSM, and surveys state-of-the-art MHSM strategies that largely rely on late fusion of SHSMs. A key contribution is the proposed vision to treat MHSM as a multimodal DL problem, leveraging early, late, or hybrid fusion to exploit inter-factor and inter-hazard relationships through semantic segmentation. The work outlines a generalized algorithm for MHSM and argues that deep multimodal learning offers a promising, yet largely untapped, pathway to improve multi-hazard risk assessment in practice.

Abstract

Data-driven susceptibility mapping of natural hazards has harnessed the advances in classification methods used on heterogeneous sources represented as raster images. Susceptibility mapping is an important step towards risk assessment for any natural hazard. Increasingly, multiple hazards co-occur spatially, temporally, or both, which calls for an in-depth study on multi-hazard susceptibility mapping. In recent years, single-hazard susceptibility mapping algorithms have become well-established and have been extended to multi-hazard susceptibility mapping. Deep learning is also emerging as a promising method for single-hazard susceptibility mapping. Here, we discuss the evolution of methods for a single hazard, their extensions to multi-hazard maps as a late fusion of decisions, and the use of deep learning methods in susceptibility mapping. We finally propose a vision for adapting data fusion strategies in multimodal deep learning to multi-hazard susceptibility mapping. From the background study of susceptibility methods, we demonstrate that deep learning models are promising, untapped methods for multi-hazard susceptibility mapping. Data fusion strategies provide a larger space of deep learning models applicable to multi-hazard susceptibility mapping.

Evolution of Data-driven Single- and Multi-Hazard Susceptibility Mapping and Emergence of Deep Learning Methods

TL;DR

The paper addresses how data-driven methods for single-hazard susceptibility mapping have evolved and how these approaches extend to multi-hazard susceptibility mapping (MHSM). It reviews SHSM workflows, highlights DL as a rising method with strong performance in FSM/LSM, and surveys state-of-the-art MHSM strategies that largely rely on late fusion of SHSMs. A key contribution is the proposed vision to treat MHSM as a multimodal DL problem, leveraging early, late, or hybrid fusion to exploit inter-factor and inter-hazard relationships through semantic segmentation. The work outlines a generalized algorithm for MHSM and argues that deep multimodal learning offers a promising, yet largely untapped, pathway to improve multi-hazard risk assessment in practice.

Abstract

Data-driven susceptibility mapping of natural hazards has harnessed the advances in classification methods used on heterogeneous sources represented as raster images. Susceptibility mapping is an important step towards risk assessment for any natural hazard. Increasingly, multiple hazards co-occur spatially, temporally, or both, which calls for an in-depth study on multi-hazard susceptibility mapping. In recent years, single-hazard susceptibility mapping algorithms have become well-established and have been extended to multi-hazard susceptibility mapping. Deep learning is also emerging as a promising method for single-hazard susceptibility mapping. Here, we discuss the evolution of methods for a single hazard, their extensions to multi-hazard maps as a late fusion of decisions, and the use of deep learning methods in susceptibility mapping. We finally propose a vision for adapting data fusion strategies in multimodal deep learning to multi-hazard susceptibility mapping. From the background study of susceptibility methods, we demonstrate that deep learning models are promising, untapped methods for multi-hazard susceptibility mapping. Data fusion strategies provide a larger space of deep learning models applicable to multi-hazard susceptibility mapping.

Paper Structure

This paper contains 13 sections, 2 figures, 2 algorithms.

Table of Contents

  1. Introduction
  2. Background on Single-Hazard Susceptibility Mapping
  3. Flood Susceptibility Maps
  4. Landslide Susceptibility Maps
  5. Background on Multi-Hazard Susceptibility Mapping
  6. Other Relevant Single-Hazard Maps
  7. State-of-the-Art MHSM Computation Methods
  8. A Generalized Algorithm
  9. Background on Deep Learning Models for Susceptibility Mapping
  10. Single-Hazard Susceptibility Mapping
  11. Multi-hazard Susceptibility Mapping
  12. Vision for DL in MHSM and Discussion
  13. Concluding Remarks

Figures (2)

  • Figure 1: Comparative results of ML/DL models used for single-hazard susceptibility mapping in terms of the percentage of area in different levels of severity of the natural hazard and the area under the ROC curve (AUC) metric for (a) FSMs costache2020novelnachappa2020flood and (b) LSMs fang2021comparative.
  • Figure 3: MHSM computation using different fusion strategies in deep multimodal learning when using conditioning factors (CFs) of multiple single-hazard susceptibility maps (SHSMs). Here, CF(SHSM-i) refers to a set of raster images of CFs for the $i$th SHSM. (Adapted from zhang2021deep)

Theorems & Definitions (1)

  • Definition 1