ИСТИНА

Changes in land use in Russia after 1990 led to the transition of many previously cultivated fields to lay lands. The boundaries between fields were gradually disappearing, and often a few fields are considered now as a single area. The spatial variability of soil properties within such area often is characterized by higher-order non-stationarity. In some respects, non-stationarity could be overcome by using empirical Bayesian kriging, which allows evaluating the probability of those or other variograms in the different parts of the area. Empirical kriging error in this case can be a useful tool for indicating sites with different agricultural practices in the past. Verification of this assumption was conducted on the territory of the Moscow State University Experimental Soil-Ecological Center Chashnikovo, where the majority of the previously cultivated fields were abandoned for about 10-15 years. The study area was situated in Moscow region, about 40 km to the North from Moscow. Soil type was Umbric Albeluvisols (arable sod-podzolic soils under Russian soil classification) formed on limon and moraine loam. Data were processed using empirical Bayesian kriging implemented in the Geostatistical Analyst module of ArcGis package. It was found that the Bayesian kriging error of the values of pH, organic matter content, phosphorus and potassium content within the arable layer were related to the land-use history and clearly divided the area on former fields. Semivariograms for different soil properties of the arable layer were similar within each former field. This regularity was manifested for such labile property as soil moisture and for a very conservative property as soil texture. Maps were done using ordinary and universal kriging. Cross-validation has shown that in most cases the total errors for maps constructed using empirical Bayesian kriging were less, but the decrease was modest.