Using a time series of Landsat TM data for digital mapping to fill information gaps in topsoil texture central Tunisia

In arid and semi-arid areas, bare soils occupy a larger area than the vegetation cover. The vegetation covers only 10% to 30% of the soil surface with seasonal chlorophyll activity. The soil surface should thus be directly detectable by remote sensing. Concerning our study site in central Tunisia, existing soil maps are neither exhaustive nor sufficiently precise for environmental modeling or thematic mapping. The main purpose of our study was to produce topsoil texture map at fine spatial resolution over our area by combination of Landsat Thematic Mapper data. Landsat images were acquired in summer and in the plowing and sowing period in fall. A maximum of one image was selected per year. Vegetation areas were masked using a sill of the normalized difference vegetation index for each image. Relationships between textural indices (MID-Infrared) and particle size analysis were studied and were used to produce clay map at a spatial resolution of 30 m. Ordinary kriging and cokriging, by combining more than one image, were used to fill in the gaps created by the vegetation mask and to predict clay content of each pixel of the image at 100 m grid spatial resolution. Results showed that ordinary kriging can identify certain linear structures such as the wadi bed with low estimated clay content. Cokriging using more than one date improved the prediction of the soil fraction over the masked area.