Abstract: Using the data of land use and land surface temperature (LST) retrieved from Landsat/ETM^+ images, we examined the relationships between the thermal field and landscape pattern in the Nanjing metropolitan region. The correlations between land use types (cropland, forest, water and settlements), mean LST and coefficient of variation (CV) of LST were quantified at three grain sizes (i.e., 1, 2.5 and 5 km). The relative importance of the effects of landscape composition, landscape structure (including landscapelevel and classlevel) and spatial factors on the thermal field pattern was distinguished quantitatively. The results showed that water bodies and forests played a significant role in lowering LST, while this cooling effect produced by croplands was relatively weak. Landscape composition and structure contributed most to explaining the thermal field pattern, while the location factors and the distance from the city center had minor effects. At all three grain sizes studied, considerable variances of mean LST and the CV of LST of each land use type could be explained by the class-level structure (especially at the grain size of 5 km). Landscape composition was the second important factor in explaining the mean and CV of LST, but its unique explanatory power was minimal. Landscape-level structure had a relatively small effect on mean LST but a relatively large effect on the CV of LST; and its unique explanatory power was prominent at the grain size of 2.5 km. Our results suggested that the regulation of land use structure could be a useful approach in order to relieve the urban heat island effect.

Key words: species distribution models, Southwest China, geographic distribution, Gentiana rhodantha, climate change