AGROKÉMIA ÉS TALAJTAN 55 (2006) 1 99–108 Correspondence to: GÁBOR ILLÉS, Department of Ecology and Silviculture, Forest Re- search Institute, H-1023 Budapest, Frankel L. út 42–44. Hungary. E-mail: illesg@erti.hu Digital Soil and Landsite Mapping in Forest Management Planning 1 G. ILLÉS, 2 G. KOVÁCS, 2 A. BIDLÓ and 2 B. HEIL 1 Department of Ecology and Silviculture, Forest Research Institute, Budapest and 2 Department of Forest Sites, University of West-Hungary, Sopron (Hungary) Under Hungarian conditions, a randomly based comparison between the data of soil sampling plots and data of existing soil descriptions’ of the same forested area (e.g. in the case of soil maps or rather soil descriptions in management plans, too) would result in a very poor fit if soil properties were regarded. The reason for this is twofold. On the one hand, forested areas have not always been mapped in a system- atic manner. On the other hand, soil properties have not been the first order targets of mapping. They were used as characteristic indicators of taxonomic soil classes. Consequently, the resulting maps contained the soil classes with their typical (or mean) values. In this way, the explored spatial variance of soil attributes within the soil classes and their dependence on the environmental variables were lost (BIDLÓ et al., 2003). This was partly caused by the lack of suitable tools for extending the data of point samples over the whole study area, including the unvisited sites as well. At present we have the tools for building and running complex environmental models and they have been developed and are being used widely in ecological re- searches (AUSTIN, 2002; RECKNAGEL, 2001; LEK & GUEGAN, 1999; LOREK & SONNENSCHEIN, 1999; JORGENSEN, 1997). Naturally, this is valid for applied soil sciences, too (SCULL, 2003; LÁSZLÓ & RAJKAI, 2003). The sound technological basis is provided mainly by the geographic information systems (GISs) and the sophisticated statistical software solutions, which offer effi- cient and fast data mining techniques. Coupling these two provides the means of deriving high resolution and reliable soil property and soil class maps or at least towards maps with known error ranges and confidence levels. The effort for predicting soil properties or soil classes from environmental vari- ables roots in a very simple cause: Preparing accurate soil maps in a traditional way is a very expensive and time consuming activity and the data for the required envi- ronmental variables are generally easier to obtain than data on soils. This approach has already been applied in several studies on soil mapping so far (MCBRATNEY et al., 2000; DOBOS et al., 2000, 2002; ODEH et al., 1992; ZHU, 1997; SINOWSKI & AUSERWALD, 1999).