Scaling is essential issue in understanding and predicting landscape pattern and ecological process in ecological research. As a significant indicator and ecological parameter affected by urbanization in urban ecosystem, landscape pattern and ecological process of urban heat island (UHI) and its relationship between others ecological factors are scale-dependent. The latest studies demonstrated that landscape metrics provides an effective method in quantifying UHI pattern. However, issues concerning UHI are highly complex and variable. The study of impacts of changing scale on UHI should be strengthened. A common objective of scale related research in UHI is to find the underlying mechanisms by use of the selected scales. The purpose of this study is to identify the scale characteristics and scale domain of UHI pattern, and provide basic information for pattern analysis and scaling in UHI research.
In this paper, taking Pearl River Delta as an example region, we obtained land surface temperature (LST) from Landsat TM by mono-window algorithm. Then mean-standard deviation method was employed to transform LST into thematic map of thermal categories. The basic spatial unit is 30m. By the scale with pixels on side of the grid cell, 2、3、4、5、6、7、8、9、10、15、20、25、30、35、40、45、50、60 and 70 basic cells were assembled using software ArcGIS. Several landscape metrics, including patch density (PD), Percentage of Landscape (PLAND), Largest Patch Index (LPI) and Shannon's Diversity Index (SHDI), were computed to detect the UHI pattern in different spatial scales. The scale effects of spatial grain sizes were analyzed under class and landscape levels.
The results demonstrated that as the spatial grain size increased, dominance of high thermal class decreased and transformed dramatically into neighbor thermal class. Scale effects appear significantly in class and landscape levels. The responding characteristics were varied by specific thermal class and landscape metrics. PD and FRAC_MN in both class and landscape levels were easier to predicted, indicating these two metrics could be treated as good methods of scaling in UHI pattern. Thresholds of grain size were obvious, such as grain size 150m of PD, PLAND, CLUMPY and CONTAG in the region. The appropriate grain extent (30-150m) was detected for calculating landscape metrics of UHI pattern. According to scale inflexions and response curves discussed in the paper, scale domains of landscape matrices were confirmed. Generally, from 30m to 150m was the scale domain to UHI pattern. That means that related ecological model of UHI can be scaled across this scale extent by ordinary transformation method. This study may offer useful information of UHI mechanism and scaling of UHI spatial model in future research.