Forecasting land surface drought in urban environments based on machine learning model

Research output: Journal Publications and ReviewsRGC 21 - Publication in refereed journalpeer-review

View graph of relations

Author(s)

Detail(s)

Original languageEnglish
Article number106048
Journal / PublicationSustainable Cities and Society
Volume118
Online published16 Dec 2024
Publication statusPublished - Jan 2025

Abstract

Urban drought, a subtype of socio-economic drought, has received limited attention compared to other types. Given the shifts in water supply patterns due to global climate change and ongoing urbanization, understanding and predicting how urban design affects drought is crucial for sustainable human settlements. Previous research has primarily focused on large-scale predictive modeling, making it difficult for architects and urban designers to address potential drought risks. This study addresses that issue by proposing an urban planning prediction model based on generative adversarial networks that integrates temperature vegetation dryness index (TVDI) maps. The model is trained on relevant datasets by using Guangzhou as a case study, including land cover, land surface temperature, and normalized difference vegetation index data. TVDI maps are generated from untrained image pairs based on urban planning parameters. Model validation, including accuracy analysis and scenario simulations, is conducted to assess the model’s ability to predict urban land surface drought resulting from changes in urban planning. This approach highlights its proactive capacity to anticipate and reveal long-term or gradually unfolding drought trends, offering valid tools for urban design and planning. © 2024 Elsevier Ltd.

Research Area(s)

  • Urban drought, Temperature Vegetation Dryness Index (TVDI), Generative Adversarial Networks (GAN), Machine learning