The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Publications Copernicus
Articles | Volume XLII-4/W19
23 Dec 2019
 | 23 Dec 2019


M. M. Tinoy, A. U. Novero, K. P. Landicho, A. B. Baloloy, and A. C. Blanco

Keywords: urban heat island, hot spots, cold spots, multiple regression, quantile regression, modelling

Abstract. This study produced spatiotemporal hot and cold spot occurrence maps for Davao City for the period 1994-2019 using land surface temperature (LST) images. Urban heat is theorized to have been affected by some, if not all, of the following impact factors: air pollutant concentrations/particulate matter (PM10), vegetation “abundance” (using EVI), building “density” (NDBI), albedo, topography, and population density. A mobile traverse sampling was performed in the morning and afternoon of 15 April 2019 to measure PM10 in the city’s identified hot spots. The remaining factors were generated from imagery (i.e., Landsat 8, Synthetic Aperture Radar) and obtained from the Philippine Statistics Authority. These factors were analyzed against the LST which was obtained through Project GUHeat’s methodology. The relationships between the factors and LST were studied through multiple and quantile regression models (MRM & QRM). Results showed that variable PM10 does not have any significance in the MRM. Meanwhile, QRM were fitted to different quantile values, namely: 10th, 25th, 50th, 75th, and 90th. It is only at the 90th quantile where all the independent variables are good predictors for the LST. Albedo is the most important predictor for the LST at 10th quantile whereas Elev for the 25th quantile. However, when LST is at the 50th, 75th, and 90th quantiles NDBI is the most significant variable at predicting LST. Reliable spatiotemporal assessment and modelling of surface temperature are essential for urban planning and management to formulate sustainable strategies for the welfare of people and environment.