Integrating weather observations and local-climate-zone-based landscape patterns for regional hourly air temperature mapping using machine learning

Abstract

Air temperature is a crucial variable of urban meteorology and is essential to many urban environments, urban climate and climate-change-related studies. However, due to the limited observational records of air temperature and the complex urban morphology and environment, it might not be easy to map the hourly air temperature with a fine resolution at the surface level within and around cities via conventional methods. Thus, this study employed machine learning (ML) algorithms and meteorological and landscape data to develop hourly air temperature mapping techniques and methods at the 1-km resolution over a multi-year warm seasons period. Guangdong Province, China was selected for the case study. Random forest algorithm was employed for the hourly air temperature mapping. The validation results showed that the hourly air temperature maps exhibit good accuracy from 2008 to 2019, with mean R2, root mean square error (RMSE) and mean absolute error (MAE) values of 0.8001, 1.4821 °C and 1.0872 °C, respectively. The importance assessment of the driving factors showed that meteorological factors, especially relative humidity, contributed the most to the air temperature mapping. Simultaneously, landscape factors also played a non-negligible role. Further analysis revealed that the maps steadily maintained high accuracy at nighttime (20:00–7:00), which is essential for investigating nighttime urban climate conditions, especially the urban heat island effect. Moreover, a correlation existed between the nighttime air temperature changes and urban morphology represented by the local climate zones. Air temperatures tended to fall more slowly in the core of metropolitan areas than in the urban fringe. Using ML, this study reliably improves the spatial refinement of hourly air temperature mapping and reveals the spatially explicit air temperature patterns in and around cities at different times in a day during the warm seasons. Moreover, it provides a novel valuable and reliable dataset for air-temperature-related implementation and studies.