Abstract: It is difficult to reveal spatial distribution patterns of forest fire at a single spatial scale. Here, we explored the spatial variability, correlation, and scale effects of forest burned area in Jiangxi Province between 2003 and 2015 at three spatial scales (city, county, and township levels) using the approaches of variation coefficient, spatial autocorrelation statistics, and scale variance analysis. There were significant differences in the changes of forest burned area among different years, with an overall pattern of increasing first and then decreasing over time. The burned area was the largest in 2008, and the spatial variability was the smallest, showing obvious aggregation characteristics. The distribution pattern of forest burned area differed across spatial scales. With the decreases of spatial scale, the degree of spatial autocorrelation of forest burned area increased, with an increased contribution rate to the whole spatial distribution pattern of forest burned areas. The hotspots of burned area were mainly distributed in the central and southern Jiangxi, which were closely related to local natural and human activities. Understanding distribution patterns of forest burned areas at different spatial scales provide a scientific basis for rational resources allocation of regional forest fire prevention.

Key words: urban green space, urban heat island., cooling effect, field measurement