Spatial resolution of Formosat-2 data is 8 m for the 4-band multispectral mode and 2 m for the panchromatic mode, respectively. Formosat-2 can revisit the same areas in one-day intervals. The authors used only multispectral data at the 8 m spatial resolution for the purpose of classifying the surface cover types.We used ASTER data acquired in the daytime of March 6, 2001 and Formosat-2 data acquired on July 12, 2004, respectively, to estimate the heat balance in Tainan City.3.2. Meteorological DataThe authors used the ground meteorological data acquired at the meteorological station in Tainan City, managed by the Southern Region Weather Center, Central Weather Bureau of Taiwan. The weather station is located in the center of the city, as shown in Figure 1. Meteorological http://www.selleckchem.com/products/VX-770.html observations come from a standard site but which is surrounded by dense urbanized surfaces and thus it is expected that the observations are representative of canopy layer conditions in the urban area. In the present study, the data used were solar radiation, wind speed, relative humidity, air-pressure and atmospheric temperature acquired at 1100 TST on March 6, 2001, according to the acquisition date of the ASTER data.The meteorological data are summarized in Table 2. The interpolation method is basically the same as that of Kato and Yamaguchi [2]. Because there is only one meteorological observation site and the study area is relatively small, we assumed that atmospheric temperature and air-pressure, at 0m ASL, are the same throughout the study area. Extrapolations of these parameters for each pixel including altitudinal corrections were applied based on the environmental lapse rate with ASTER DEM data. Solar radiation, wind speed and relative humidity were assumed to be constant throughout the study area.Table 2.Summary of the meteorological conditions of the analysis in Tainan City at 1100 TST on March 6, 2001.4.?Surface ClassificationIn order to estimate surface heat fluxes, the authors needed to interpolate the roughness length for sensible and latent heat fluxes, and the minimum stomatal resistance for latent heat flux, respectively, based on surface types as mentioned in Section 2. Because of that, we wanted to classify the surfaces according to vegetation types, and density and height of buildings. In the present study, the authors classified surface types from satellite data by the combined methods of the maximum likelihood, decision tree and manual classification in order to separate surface coverage having similar spectral patterns. First, we classify the surface types in more than 30 categories by the maximum likelihood method. The categories are too numerous for the following analysis because they are based on not only surface coverage but also their spectral pattern. Therefore, they are combined into 8 categories: buildings, roads, water, bare soil, short grass (e.g., lawn), tall grass (e.g., paddy field), bushes, and forests.