Cloud and shadow in satellite image are obstacles in the land cover classification. In addition, Landsat-7 Enhanced Thematic Mapper Plus (ETM+) image contains scan-gap (striping black areas) due to the failure of the Scan Line Corrector in May 2003. This is also the case with our land cover change analysis in Makassar, Maros, Sungguminasa, and Takalar (Mamminasata), Indonesia. The purpose of this study is to produce an accurate land cover classification map using the composite data of multi- temporal ETM+ data. Three scenes in April, July, and September 2011 were generated by filling the areas of clouds, shadows, and scan-gaps with other data observed in similar seasons. Land cover classification map was generated from 18 bands data by the Support Vector Machine classifier using local-government-produced land use map as training data. The effect of composite on the land cover classification and land cover classification accuracy will be evaluated in the presentation. The significance of this research is the improvements in classification by reducing the obstacles using a simple image composite method.