The phenomenon of urban areas becoming warmer than surrounding rural areas, known as urban heat island, is one of the biggest urban issues due to human activities. The study adopted random forests (RF) based downscaling land surface temperature (LST) to assess the surface urban heat island (SUHI) of Fukuoka, Japan. The LST images were first downscaled to 250 m spatial resolution, and then the seasonal and diurnal SUHI was identified. The results implied that the SUHI in summer daytime is the most significant while in winter nighttime is the least significant in the area. Through the bivariate Moran's I test and zonal statistics, the cooling effect of the green-blue spaces were evaluated. The forest has the most substantial urban cooling effect with the average LST change occurs at a distance of about 1 km (maximum 6.32 K). The cooling effect of the sea is significant during daytime, whilst that of the lake/river and the cropland/grassland is not obvious. We concluded the urban population and land use and land cover (LULC) are significantly spatial correlated to UHI, and are important factors of considering the urban heat island (UHI) mitigation strategies.

由于人类活动，城市地区比周围农村地区变暖的现象（称为城市热岛）是最大的城市问题之一。该研究采用基于随机森林（RF）的降尺度地表温度（LST）来评估日本福冈的地表城市热岛（SUHI）。首先将LST图像缩小到250 m的空间分辨率，然后识别季节性和昼夜SUHI。结果表明，该地区夏季SUHI最显着，而冬季夜间SUHI最不显着。通过二元Moran's I检验和区域统计，评估了蓝绿色空间的冷却效果。森林具有最显着的城市降温效果，平均LST变化发生在大约1 km（最大6.32 K）的距离上。白天，海洋的降温效果显着，而湖泊/河流和农田/草地的降温效果不明显。我们得出的结论是，城市人口，土地利用和土地覆盖率（LULC）与UHI具有显着的空间相关性，并且是考虑城市热岛（UHI）缓解策略的重要因素。