The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLI-B2
https://doi.org/10.5194/isprs-archives-XLI-B2-55-2016
https://doi.org/10.5194/isprs-archives-XLI-B2-55-2016
07 Jun 2016
 | 07 Jun 2016

HISTORICAL GIS DATA AND CHANGES IN URBAN MORPHOLOGICAL PARAMETERS FOR THE ANALYSIS OF URBAN HEAT ISLANDS IN HONG KONG

F. Peng, M. S. Wong, J. E. Nichol, and P. W. Chan

Keywords: 3D building model, aerial photogrammetry, digital surface model, land surface temperature, urban morphological parameter

Abstract. Rapid urban development between the 1960 and 2010 decades have changed the urban landscape and pattern in the Kowloon Peninsula of Hong Kong. This paper aims to study the changes of urban morphological parameters between the 1985 and 2010 and explore their influences on the urban heat island (UHI) effect. This study applied a mono-window algorithm to retrieve the land surface temperature (LST) using Landsat Thematic Mapper (TM) images from 1987 to 2009. In order to estimate the effects of local urban morphological parameters to LST, the global surface temperature anomaly was analysed. Historical 3D building model was developed based on aerial photogrammetry technique using aerial photographs from 1964 to 2010, in which the urban digital surface models (DSMs) including elevations of infrastructures and buildings have been generated. Then, urban morphological parameters (i.e. frontal area index (FAI), sky view factor (SVF)), vegetation fractional cover (VFC), global solar radiation (GSR), Normalized Difference Built-Up Index (NDBI), wind speed were derived. Finally, a linear regression method in Waikato Environment for Knowledge Analysis (WEKA) was used to build prediction model for revealing LST spatial patterns. Results show that the final apparent surface temperature have uncertainties less than 1 degree Celsius. The comparison between the simulated and actual spatial pattern of LST in 2009 showed that the correlation coefficient is 0.65, mean absolute error (MAE) is 1.24 degree Celsius, and root mean square error (RMSE) is 1.51 degree Celsius of 22,429 pixels.