Influences of the urban heat island (UHI) became more and more serious in the industrial areas of Pearl River Delta. Spatial patterns and dynamic characteristics of the UHI were not well understood because of limitation of research methods and techniques used for this type of study. The spatial patterns of meteorological conditions were important for the formation of the UHI. Spatial patterns of urban heat environment were simulated by the fifth generation Penn State/NCAR Mesoscale Model (MM5) and studied using data retrieved from remote sensing (RS) images. Thermal landscape was described by using classic landscape indices.
The non-hydrostatic mesoscale model (MM5) simulations were conducted using four nested domains, and used to simulate the thermal environment. The finest mesh grid increment of the nesting was 1 km. The land surface temperature (LST) was retrieved by single sixth thermal infrared data of ETM+ at 11:00 am (local time) on Sep 14, 2000. The distributions of LST and urban heat island intensity (UHII) were similar between model simulations and RS data. It showed that there were several UHI centers in Pearl River Delta. The differences of temperature between simulated by MM5 and RS images and observed data from automatic weather stations (AWS) were roughly 1 ℃. LST distributions retrieved by ETM+ showed more fine-scale details than MM5 simulations. Wetland landscape, urban green-land, and river could break up UHI into segments. It revealed that these three types of land-use were important to define urban heat environment. Some small urban cold islands appeared near large parks, lakes, and reservoirs.
The intensity of UHI was divided into six categories by temperature differences. The heterogeneity of thermal environment could be revealed by using the categories of UHI intensity. Therefore the thermal landscape types were defined with six types. By the methods of landscape ecology and GIS, the spatial pattern indices of thermal landscape from MM5 and RS were analyzed. Spatial autocorrelation indices (Moran I and Geary c indices) of LST obtained from MM5 and RS were relatively approximate and similar, and the dynamic curve of autocorrelation was nearly cosinoidal. Some landscape indices were very closely correlated such as Percent of Landscape (PLAND), Number of Patches (NP), Mean Fractal Dimension index (FRAC-MN), Landscape Shape index (LSI), Perimeter-Area Fractal Dimension (PAFRAC), Simpson’s Diversity index (SIDI), Simpson’s Evenness index (SIEI) and Landscape Division index (DIVISION). So the dynamic characteristics of spatial pattern of thermal landscape from MM5 simulations could be described by these indices. The results showed that LSI and NP were higher during daytime than during night time, and that all the indices were more fluctuant during daytime than during night time. Indices of DIVISION and PAFRAC had small wave peak values at 5:00 am (local time) and had inflexions at 21:00 pm (local time) because of land-sea breezes, urban-rural breezes and the surface long-wave radiation divergence.
Based on the methods of landscape ecology, the integration of MM5 simulation and RS was found to be effective approach to analyze the spatial patterns of UHI. Dynamic characteristics, especially diurnal variation of spatial patterns of thermal landscapes and other meteorological conditions could be studied with the mesoscale model MM5 simulations.