Methodology
Fig-1 shows the data prosessing flow in this study. The first step of the study is to make the land cover classification. TERRA /MODIS has 3 different modes of spatial resolution - 250m, 500m and 1000m - with 36 spectral bands in total. In the study, the land cover characterization is made using 250m resolution optical images with categories of built up, forest, grass, water body and bare soil. The classification result is used to provide the emissivities of the land cover categories7). Emissivity is then used to estimate the land surface temperature from the brightness temepature value in the thermal band image with 500m resolution data1,2,3,4). Finally the heat island impact was analyzed. However, the land surface temperature is influended by the geographic conditions such as the climate and weather at the time in the area, so that it is difficult to compare the temperature itself to indicate the heat island intensity. In order to compare the heat island impact among the cities, the normalized surface
temperature images were prepared by equalizing the temperature values of forested region(Veg-2 in the classification category in this study) in suburbs where the heat island effect is supposed to be unrelated. Fig-2 shows the method of calculating the normalized land surface temperature. In this study the temperature at Vegetation-2 category in Seoul image was used as the parameter for normalizing the land surface temperature for other cities.

Fig. 1 Processing Flow

Fig. 2: Relationship between observed temperature and normalized temperature