This paper describes the process of classifying a pre-war and post-war Landsat Thematic Mapper image to detect landscape level changes, especially to the forest in Bosnia-Herzegovina. Timber is important in Bosnia for fuel, construction, and as an export product.

Reference data for accuracy assessment were visually interpreted from the 1994 satellite image displayed as a color infrared image. To avoid bias, the visual interpretation was done by an analyst who was not involved in the digital classification. Based on these reference data, the overall classification accuracy was 90 percent.

A maximum likelihood classifier was used in supervised classification of both the pre-war 1987 and post-war 1994 scenes. Each scene was classified to cover types of deciduous forest, coniferous forest, vegetated agriculture, plowed fields, urban, water, and clouds. The area deforested across the TM scene was estimated at 82,500 hectares with 65,000 hectares of conifers and 17,500 hectares of broadleaf forest lost. Most of the larger deforested areas were located to the southwest. Most patches of deforestation were less than 100 hectares in area.

A Normalized Difference Vegetation Index (NDVI) was computed for the 1994 image to locate areas with potentially high timber volume. A threshold value of 0.72 was used to represent potentially high timber stands. These areas totalled 119,000 hectares with 43,000 hectares occurring in patches that exceeded 40 hectares in area. Areas of high NDVI values as potential timber sources were located within 10 km of the cities of Sarajevo, Tuzla, Brcko, and Banjaluka. High NDVI areas were also located within 3km of United Nations designated safe roads.

 email: dfell@merlin.salrm.uaf.edu