Please use this identifier to cite or link to this item:
Scopus Web of Science® Altmetric
Type: Journal article
Title: Innovative use of spatial regression models to predict the effects of green infrastructure on land surface temperatures
Author: Bartesaghi-Koc, C.
Osmond, P.
Peters, A.
Citation: Energy and Buildings, 2022; 254:111564-1-111564-18
Publisher: Elsevier BV
Issue Date: 2022
ISSN: 0378-7788
Statement of
Carlos Bartesaghi-Koc, Paul Osmond, Alan Peters
Abstract: Understanding the complex and dynamic interplay and cumulative effects of green infrastructure (GI) and urban form on land surface temperatures (LST) is important to design and implement heat mitigation strategies. Past research has mostly employed two-dimensional (2D) indicators, simple correlations and conventional regression models using coarse-level analytical approaches that obviate spatial autocorrelation effects. For the first time, this study applies a holistic approach to evaluate GI and urban settings as complex dynamic systems. The objectives of this paper are to: (1) develop novel ‘spatially-based’ predictive models that account for spatial dependencies; (2) implement a fine-scale analytical unit (<50 m) for a more precise and accurate analysis; (3) incorporate the ‘multi-temporal’ diurnal and seasonal variations into predictions; and (4) propose the novel combination of 2D and 3D morphological, compositional and configurational parameters of GI and urban form derived from very high resolution (VHR) remotely-sensed data (<2m), using Sydney metropolitan region as case study. Results show a strong spatial association of LST at fine scale (<50 m) and spatial autocorrelation among residuals in traditional models. Spatial error model (SEM) exhibits a superior performance over conventional multivariate regression, however, results presented significant heteroscedasticity caused by the large temperature variability in certain areas, although this problem was partially solved. Future studies should incorporate unmeasured factors related to material-specific properties (i.e. albedo, emissivity), and capture the thermal variation within urban areas by segmenting datasets into zones with relatively homogenous thermal and physical properties. Overall, ground imperviousness mostly defines the LST profile of a place, with a relative warming effect of 0.23 °C and 0.61 °C during day; and 0.18 °C and 0.41 °C at night per 10% of area increment in winter and summer, respectively. The same increment in the proportion of water and trees contributes to a maximum LST reduction of 0.42–0.85 °C in summer, and 0.25–1.17 °C in winter; however, this causes an increase of nocturnal LST between 0.12 °C and 0.30 °C throughout the year. In general, the cooling effects from GI do not outweigh the warming effects from man-made surfaces. Compared to abundance, the spatial configuration of trees is less influential on LST. Ground sky view factor (GSVF), altitude and distance to coast are of relative importance in defining LST profiles. These results used to numerically simulate different greening scenarios at neighbourhood scale for Sydney; illustrating the potential of spatial models to define heat mitigation scenarios to inform urban design and planning policies.
Keywords: Surface urban heat island; Predictive modelling; Mitigation Strategies; Land Surface temperature; Spatial Error Model; Multivariate Regression; Ordinary Least Square Regressions
Description: Available online 24 October 2021
Rights: © 2021 Elsevier B.V. All rights reserved.
DOI: 10.1016/j.enbuild.2021.111564
Published version:
Appears in Collections:Architecture publications

Files in This Item:
File Description SizeFormat 
  Restricted Access
Embargo ends January 20241.96 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.