Using deep neural networks for predictive modelling of informal settlements in the context of flood risk

Yue Zhu,Emily So,Christian Geiß

doi:10.1088/1742-6596/1343/1/012032

Yue Zhu, Emily So + Show 1 more

Open Access

https://doi.org/10.1088/1742-6596/1343/1/012032

Copy DOI

Abstract

Global climate change has substantially increased the risks of cities being adversely affected by natural hazards such as floods. Among the inhabitants of cities at risk, residents dwelling in informal settlements are the most vulnerable group. To identify the future exposure of informal settlements, we adopt a data-driven model from the machine learning domain to anticipate the growth patterns of formal and informal settlements in flood-prone areas. The potential emergence of informal settlements in Shenzhen, China, is predicted by the proposed method. Then, through an analysis of the flood susceptibility of the predicted informal settlement areas, the emerging vulnerability of Shenzhen towards flooding is revealed.

Highlights

IntroductionCities are becoming increasingly exposed due to exponential urban growth
Convolutional neural network for classification and prediction Comparing with conventional land use modelling tools, Machine Learning (ML) techniques excel at processing complex variables and predicting dynamic patterns
A Convolutional Neural Network (CNN) model is used to identify past growth patterns of informal settlements among various land use types, which are extracted from satellite images

Summary

Introduction

Cities are becoming increasingly exposed due to exponential urban growth Among these growing settlement areas, the emergence of informal settlements has raised significant concerns due to their poor quality of construction (Scovronick, Lloyd, & Kovats, 2015), limited access to urban facilities, and the restricted choice of site selection. These informal settlements have become one of the most vulnerable parts of a city.

Objectives

Methods

Results

Conclusion