Generalisation of continuous models to estimate soil characteristics into similar delineations of a detailed soil map

The aim of soil mapping is to partition soil bodies using map units, which are more homogenous for specific soil properties than are the soil bodies as a whole. Soil properties are expected to be similar at delineations of a specified soil map unit. Therefore, it is supposed that a model developed to estimate a soil property for one of these delineations could be generalised for the others. This study was conducted to determine the possibility of generalisation (extrapolation) of continuous models of spatial variability to estimate soil physical and chemical properties in similar delineations of a soil map unit. A consociation soil map unit in two different locations of a detailed soil map (1:20000 scale), as similar delineations, was selected in the north-west of Faradonbeh region, Iran. Sixty topsoil samples (0–20cm) were randomly collected in each delineation (totally 120 samples) with 30-m intervals and the samples were GPS-recorded. Laboratory studies consisted of bulk density, pH, calcium-carbonate equivalent, organic matter content, percentage of coarse fragments, and particle-size distribution. First, variography was done according to the soil data of each delineation (named areas A and B) and kriged maps were generated based on their own semivariogram parameters. Then, the kriged map of the soil properties for the second similar delineation (area B) was regenerated based on the corresponding models and their parameters obtained from the first similar delineation (area A). Finally, the regenerated kriged map of each variable was compared with its original kriged map. Visual comparison of the kriged maps of area B obtained from two steps of variography showed very high accordance for all of the soil properties. Quantitative comparison of the kriged maps suggests that the accuracy expected by the users of the soil information should be considered before generalisation of the data for similar units. Lower values of accordance obtained by the Kappa index and, especially, the classification success index than overall accuracy indicate that model generalisation should not be used where high precision of soil information is expected. Discrepancies observed for the kriged maps of the same variables in similar delineations could be due to different soil management practices in the past as a result of different historical developments.