Abstract: Land surface temperature (LST) is a key variable in climatic and ecological studies. However, accurate measurements of LST over continents are not yet available for the whole globe. This thesis first reviews the state of the science of land surface temperature (LST) estimates from remote sensing platforms, models, and in situ approaches. Considering the suspicions, we review the current Land Surface Temperature confirmation and estimation method. Then the requirements for LST products are specified, from the different user communities. In this paper analysis a physics-based method to retrieve LST from the MODIS daytime MIR data in channels 22 (centered at 3.97 µm) and 23 (centered at 4.06 µm). In this method to separate the reflected solar direct irradiance and the radiances emitted by the surface and atmosphere. The MIR spectral region (3–5 µm) has many advantages with respect to the TIR spectral region. MIR using the multispectral thermal imager and found that LST retrieved from MIR is only half as sensitive to errors in LSE as those retrieved from TIR. Consequently, it seems to be more appropriate to retrieve LST from MIR rather than TIR data. However, measurements in the MIR region at satellite altitudes during the daytime consist of a combination of both reflected radiance due to solar irradiance and emitted radiance from both the surface and the atmosphere. In this paper proposed clustering method is implemented to process subsequences of time series data and detects land cover change temperature measured as a function of time. Land cover change temperature measured is declared when consecutive subsequences that are extracted from one MODIS time series transitions from one cluster to another cluster and remains in the newly assigned cluster for the rest of the time series. The temporal sliding window designed to operate on a subsequence of the time series to extract information from two spectral bands from the MODIS product.
Keywords: Land Surface Temperature (LST), Mid-Infrared (MIR), Modis, Image Segmentation, Fuzzy c-means clustering.