The decline in soil fertility in recent decades has led to a number of problems in agriculture. The spread of degradation processes forces the use of modern methods of modeling and forecasting erosion. The purpose of this work was to make a long-term forecast of the degradation processes of ordinary chernozems on sloping lands in crop rotations of various designs and using various agricultural technologies using automated neural networks. The research was conducted in a multifactorial long-term experiment, which is located on the slope of the Bolshoy Log beam, in 1990-2022. The experience was laid down in 1986 in the system of contour and landscape organization of the slope area with a steepness of up to 3.5-4°. In the experiment, crop rotations with a different ratio of the proportion of pure steam and perennial grasses in the structure of crops, as well as two tillage systems, were studied. As a result of the research, a forecast was made for a hundred years ahead using automated neural networks, which revealed a general trend of attenuation of erosion processes on slopes, but with varying degrees of intensity, both in crop rotations of various designs and when using various agricultural technologies for cultivating crops. Cultivation of crops in crop rotations with a different proportion from 20 to 40 % of perennial grasses in the structure of sown areas reduces the runoff of meltwater and stormwater by 22.9-24.9 %, washing away the soil by 26.4-28.6 %, and in some cases completely, under exist-ing conditions. The use of soil protection treatment will reduce the runoff of meltwater and stormwater by 39.2-72.0 %, and flushing by 41.5 %.