Improvement of spatial-temporal urban heat island study based on local climate zone framework: A case study of Hangzhou, China

Abstract

Large-scale long-period urban heat island (UHI) intensity (UHII) prediction with high spatiotemporal resolutions, satisfactory accuracy, and calculation efficiency is crucial and challenging for UHI mitigation studies. This study proposes a framework combining an urban weather generator (UWG), local climate zone (LCZ), deep-learning method, and Python automatic cyclical calculation to obtain hourly UHII throughout one year and over a total of 1920 blocks in Hangzhou City. The spatial-averaged hourly UHII are between −2 °C and 6 °C in more than 96 % time, and those at 0 °C–1 °C contribute to 43.80 % of the total time. Spring gives the most intense nocturnal UHII, while winter has the weakest one. A significant diurnal UCI phenomenon could accompany the strong nocturnal UHI phenomenon. From the synthetic (based on seasonal data) 24-h curves, mean UHII and UCII (average of the positive and negative values, respectively) drop by approximately 25 % in winter compared to spring. Spatially, UHII is higher in the central regions with compact buildings but significantly lower in the high-vegetation-coverage regions. Based on LCZ framework, LCZ 1 (compact high-rise configurations) has the highest UHII, independent of examined periods. UHII relations to building coverage, vegetation ratio, and building height are individually quantified. The building coverage has the highest influence on annual UHII, with a correlation coefficient as high as 0.76. Results indicate that, for UHI mitigation purposes, the percentage of compact high-rise configurations (LCZ 1) in the urban area shall be limited, and the vegetation ratio is better to be greater than 20 %.