Simulating and mitigating extreme urban heat island effects in a factory area based on machine learning

Abstract

Abstract Urban heat islands (UHIs) have caused radical changes in urban climates. However, the extreme UHI (E-UHI) formed in factory areas deserves more attention. To mitigate the E-UHI, machine learning is used for simulating and quantifying the marginal utility of the scale, shape, type, stage, and structure of the factory on the land surface temperature (LST), factory LST ( LST f ), surrounding LST ( LST s ) and increase value ( ΔLST ) level. The results show that the scale of all types of factories affects LST f and LST s , and the shape of steel factories affects LST s and ΔLST . The LST in factories that require high-temperature environments (e.g., smelters) is significantly higher than that in other factories (e.g., sales plants). The ΔLST of green space (GS), staff activity ground (SG), material transfer ground (MG), material storage area (MA), factory building (FB), smelting area (SA) and casting building (CB) are 3.95\xa0°C, 4.01\xa0°C, 5.08\xa0°C, 5.15\xa0°C, 5.24\xa0°C, 5.49\xa0°C and 7.32\xa0°C, and their optimal ranges are 8.84%–15.09%, 16.65%–25.52%, 3.91%–35.91%, 0.00%–8.70%, 5.06%–13.60%, 23.33%–48.02%, and 0.00%–5.73%, respectively. Appropriately standardizing the scale and shape, controlling the temperature of the high-temperature generation stage, reducing the proportion of CB, MG and MA, and increasing the proportion of GS and SG are effective ways to alleviate the E-UHI. The findings provide theoretical guidance for resource-based cities to mitigate E-UHIs.