Assessing the Impact of Land Cover Changes on Surface Urban Heat Islands with High-Spatial-Resolution Imagery on a Local Scale: Workflow and Case Study

Peng Ren,Xinxin Zhang,Qinglin Meng,Haoyan Liang

doi:10.3390/su11195188

Abstract

Low-altitude remote sensing platform has been increasingly applied to observing local thermal environments due to its obvious advantage in spatial resolution and apparent flexibility in data acquisition. However, there is a general lack of systematic analysis for land cover (LC) classification, surface urban heat island (SUHI), and their spatial and temporal change patterns. In this study, a workflow is presented to assess the LC’s impact on SUHI, based on the visible and thermal infrared images with high spatial resolution captured by an unmanned airship in the central area of the Sino-Singapore Guangzhou Knowledge City in 2012 and 2015. Then, the accuracy assessment of LC classification and land surface temperature (LST) retrieval are performed. Finally, the commonly-used indexes in the field of satellites are applied to analyzing the spatial and temporal changes in the SUHI pattern on a local scale. The results show that the supervised maximum likelihood algorithm can deliver satisfactory overall accuracy and Kappa coefficient for LC classification; the root mean square error of the retrieved LST can reach 1.87 °C. Moreover, the LST demonstrates greater consistency with land cover type (LCT) and more fluctuation within an LCT on a local scale than on an urban scale. The normalized LST classified by the mean and standard deviation (STD) is suitable for the high-spatial situation; however, the thermal field level and the corresponded STD multiple need to be judiciously selected. This study exhibits an effective pathway to assess SUHI pattern and its changes using high-spatial-resolution images on a local scale. It is also indicated that proper landscape composition, spatial configuration and materials on a local scale exert greater impacts on SUHI.

Highlights

Against the background of extensive urbanization, cities and conurbations experience rapid expansions, especially in developing countries
The results show that the supervised maximum likelihood algorithm can deliver satisfactory overall accuracy and Kappa coefficient for land cover (LC) classification; the root mean square error of the retrieved land surface temperature (LST) can reach 1.87 ◦C
We examined the workflow and assessed the accuracy of LC classification, LST retrieval, and applicability of some indexes for evaluating surface urban heat island (SUHI) on a local scale

Summary

Introduction

Against the background of extensive urbanization, cities and conurbations experience rapid expansions, especially in developing countries. Vast areas of rural land are changed into impervious surfaces such as buildings and roads, which generally absorb and re-radiate solar radiations effectively [1,2,3,4,5] These -stored solar radiations and artificial heat generated from productions, transportations, and civilian life lead to severe urban heat islands (UHI) [6,7,8,9]. Research has been proliferating on the analysis of the patterns of UHIs, their influences, and corresponding adaptive strategies [14,15,16,17] These studies are usually based on the relationship between land surface temperature (LST) and land cover (LC) [18,19,20], which play crucial roles in describing the urban thermal environment. Pongracz et la. [27] found that the surface urban heat island intensity (SUHII) in nine central Europe cities demonstrated appreciable monthly variations

Objectives

Methods

Results

Discussion

Conclusion