The aim of this project was to develop information criteria to characterize environmental mapping data, particularly as applied to heterogenous solonetzic grassland, where the wide variability of the distribution of the water soluble salts poses great difficulty for mapping. The proposed methodology facilitates building geographic databases with optimizing accuracy on cost-benefit functions and is therefore applicable for baseline studies as well as for areas where detailed information is available. First, the project elaborated an image tiling method guided by the local heterogeneity. Using simulated images, researchers found that, based on tiling, one could design sampling schemes which are more effective than regular or random or image classification based on sampling design. Researches then carried out a detailed field sampling on a heterogeneous solonetzic grassland including botanical description and the taking of soil samples. After the laboratory analyses, researchers studied the soil-vegetation correlation and proved its applicability in estimating both quantitative and qualitative soil properties. Based on the early results, researchers confirmed the applicability of the theoretical findings, i.e., the mapping process can be guided by the local heterogeneity of the mapped properties. The project has pointed out that the spatial pattern of the remotely sensed image used is similar to the pattern of the relevant soil properties, though there is poor correlation between sampled variables and satellite imagery. Consequently, the applicability of remotely sensed data at the spatial scale of the mapping proved limited. The project has produced interpolated maps of soil properties and vegetation cover of the study area using geostatistical methods. Recently, the researchers have been looking for opportunities to utilize information on spatial structure of images to produce vegetation and soil maps, and have elaborated a wavelet decomposition method providing multi-scale pattern description which they intend to use to produce soil and vegetation maps applying conditional simulation. The data and procedures used have been integrated into a geographic information system (GIS). (Author abstract, modified)
