Capabilities of Landsat 5 Thermatic Mapper data in regional mapping and characterization of inland valley agroecosystems in West Africa

This study demonstrated the capabilities of Landsat-5 Thematic Mapper (TM) data in regional mapping and characterization of inland valley (valley bottoms plus valley slopes or fringes) agroecosystems. A methodology was developed involving image enhancement, display and digitizing, as well as image and Geographical Information System (GIS) manipulation techniques for rapid studies of inland valley characteristics over large areas. A case study was conducted using Landsat-5 TM data over the regions of Save and Parakou in the Republic of Benin (path: 192, row:54 of Landsat TM World Reference System) encompassing about 3.12 × 106ha (study area) to highlight the strengths of the methodology. The results showed the ability of TM data, using the espoused methodology, to map inland valley characteristics: (a) as well as or better than 1 : 200000 topographic maps; and (b) to a significantly similar level to that of 1 : 50000 topographic maps. The technique involving a ratio image of TM4/TM7 (red), TM4/TM3 (green) and TM4/TM2 (blue) when displayed in a magnifying factor of one or greater provided the single best enhancement and display technique for mapping inland valley bottoms (see illustrations in figures 4, 5 and 11). A combination of this technique with two other image enhancement techniques, involving a false colour composite (FCC) of TM4 (red), TM3 (green) and TM5 (blue) and a ratio RGB (red-green-blue) image of TM4/TM5, TM4/TM3 and TM4/TM1, improved the results significantly. Relative to 1 : 200000 topographic maps TM data mapped up to 118 per cent stream frequencies (number of streams km−2) and 99 per cent stream densities (length of streams km−2). However, relative to 1 : 50000 topographic maps TM data mapped only 74 per cent stream frequencies and 77 per cent stream densities. The ability of TM data to provide the shape, size, area, spatial distribution and land use characteristics of inland valleys (see figure 11, for example), unlike other data sources, such as topographic maps, have been highlighted in this paper. TM-derived valley bottom widths and vegetation characteristics (e.g., Ratio Vegetation Index; RVI) of valley bottoms and valley fringes provided highly positive correlations or formed statistically significant relationships with the corresponding ground-measured information. In the total study area of 3.12 × 106ha, covered by a full Landsat-5 TM scene (image), there were 9 per cent valley bottoms (see figure 11) and 192 per cent valley fringes as mapped using the espoused methodology. Of this, 7.9 per cent of the valley bottoms and 159 per cent of the valley fringes were cultivated. The percentage of inland valleys and their cultivation intensities relative to their distance from the road network and settlements were established. The major road network and a significant proportion of the settlements were also mapped directly from the TM data. These results indicate the strength of the methodology to rapidly map and characterize inland valley agroecosystems at regional level using high-resolution remotely sensed data.