Geospatial Assessment and Modeling of Urban Heat Islands in Quezon City, Philippines Using Ols and Geographically Weighted Regression

Abstract

Abstract. Urbanization has played an important part in the development of the society, yet it is accompanied by environmental concerns including the increase of local temperature compared to its immediate surroundings. The latter is known as Urban Heat Islands (UHI). This research aims to model UHI in Quezon City based on Land Surface Temperature (LST) estimated from Landsat 8 data. Geospatial processing and analyses were performed using Google Earth Engine, ArcGIS, GeoDa, and SAGA GIS. Based on Urban Thermal Field Variance Index (UTFVI) and the normalized mean per barangay (village), areas with strong UHI intensities were mapped and characterized. high intensity UHIs are observed mostly in areas with high Normalized Difference Built-up Index (NDBI) like the residential regions while the weak intensity UHIs are noticed in areas with high Normalized Difference Vegetation Index (NDVI) near the La Mesa Reservoir. In the OLS regression model, around 69% of LST variability is explained by Surface Albedo (SA), Sky View Factor (SVF), Surface Area to Volume Ratio (SVR), Solar Radiation (SR), NDBI and NDVI. OLS yield relatively high residuals (RMSE = 1.67) and the residuals are not normally distributed. Since LST is non-stationary, Geographically Weighted Regression (GWR) regression was conducted, proving normally and randomly distributed residuals (average RMSE = 0.26).

Modelling of Land Surface Temperature Using Gray Level Co-Occurrence Matrix and Random Forest Regression

“…Layers representing different variables related to or known to affect LST are prepared. In this study, the following data layers were from Alcantara et al (2019): LST, NDVI, NDBI, Albedo, Solar Radiation (SR), Surface Area-Volume Ratio (SVR), and Skyview Factor. The generation of these layers were described in detail in Alcantara et al (2019) and will not be presented in in this paper.…”

“…In this study, the following data layers were from Alcantara et al (2019): LST, NDVI, NDBI, Albedo, Solar Radiation (SR), Surface Area-Volume Ratio (SVR), and Skyview Factor. The generation of these layers were described in detail in Alcantara et al (2019) and will not be presented in in this paper. PlanetScope image (acquired on 23 March 2019) is processed to generate soil-adjusted vegetation index (SAVI) and visible green NIR built-up index layers at 3-m resolution.…”

“…The derivation of LST from Landsat image was described in several papers (see Alcantara et al, 2019. Previous studies have used the LST retrieval method developed by Jeevalakshmi et al (2017) and implemented it using Google Earth Engine (GEE).…”

“…These are attempts to understand how LST are affected by the mix of environmental variables and to possibly estimate LST for different future scenarios (e.g., urban expansion, decrease of vegetation cover). Variables describing the distribution of land use land cover types, materials, urban morphology, and solar irradiance, among others, have been used in previous studies to model LST (see Alcantara et al, 2019, Cañete et al, 2019.…”

Thermal and ecological assessment based on land surface temperature and quantifying multivariate controlling factors in Bogura, Bangladesh

“…Urban thermal field variance index (UTFVI) is another widely used indicator derived from LST to evaluate the urban environment's eco-environmental quality or thermal well-being ( Ahmed, 2018 ; Alcantara et al., 2019 ; Alfraihat et al., 2016 ; Kafy et al., 2021 ; Maithani et al., 2020 ; Nguyen et al., 2019 ; Portela et al., 2020 ; Sharma et al., 2021 ; Sobrino and Irakulis, 2020 ). Changes in local wind patterns and humidity, declining air quality and comfort, rising mortality rates, rain, and thunderstorm activity are adverse effects caused by high UTFVI ( Kafy et al., 2021 ; Sejati et al., 2019 ; Singh et al., 2017 ).…”