Abstract:
Agricultural land evaluation has a key role in the sustainable agriculture. The agricultural land evaluation methodology
is applied to land mapping units for computing a suitability index, based on the value of several soil and environment
indicators, which characterize these land mapping units. The natural land mapping units, are delineated using various
criteria and thematic layers, but most times the approach is subjective. GIS, geomorphometry, remote sensing and
geostatistics bring the possibility to objectively delineate most suitable natural land mapping units for applying the
agricultural land evaluation methodology. The methods for natural land mapping units delineation can be divided in two
classes of methods: supervised and unsupervised. The first, require some knowledge about the area, and can be used to
carry the results for a specific purpose of the land evaluation. The last, related especially to cluster analysis and image
segmentation, depend on the input data and the number of specified classes or the seed points, so require first the
analysis of the input data, to reveal the clusters/seed sampling. Both approaches were used to delineate the natural land
mapping units for a DEM covering a test area, and were used to extrapolate the method settings for a DEM covering 15
villages from Iasi county agricultural area. Because reference data concerning the natural land mapping units is almost
impossible to derive, we analyzed statistically and conceptually the results along a topographic transect, in order to try
to find the most suitable method. Generally, unsupervised segmentation methods gave the best results, and from them
the segmentation procedures, although very intensive from a computational point of view, can depict interesting
patterns of natural aggregation of natural land mapping units.
