Bulgarian Journal of Agricultural Science
Array ( [session_started] => 1718994914 [LANGUAGE] => EN [LEPTON_SESSION] => 1 )
Help
 
Register

Login:


Forgot Details? Sign-up



Space-time modeling and forecasting steppe soil fertility using geo-information systems and neuro-technologies
V. Pichura, L. Potravka, N. Stratichuk and A. Drobitko
Abstract: The study presents the results of using geo-information systems and neuro-technologies for modeling spatial heterogeneity and forecasting changes in agro-chemical properties of steppe soil fertility exemplified by Kherson region of Ukraine. Modeling allowed determining general regularities of impacts of the current agricultural practices on changes in the content of macronutrients over the past 50 years that have caused the ongoing process of a gradual decrease in the content of humus, nitrogen, phosphorus and potassium in steppe soils. A lack of balanced crop rotations, regular, uniform and necessary supply of fertilizers, occurrence of water erosion, including irrigation erosion and deflation, and also long-term irrigation led to a drop in the content of macronutrients in 1970–2020: the content of humus – by 0.36% (from 2.56% to 2.20%) or by 14.1% in statistical relation; mobile phosphorus – by 34.2% (from 62.0 mg·kg-1 to 40.8 mg·kg-1); exchangeable potassium – by 17.8% (from 442.4 mg·kg-1 to 363.8 mg·kg-1); a decrease in the content of nitrifiable nitrogen – by 17.0% (from 23.0 mg·kg-1 to 19.1 mg·kg-1) on the average in 2013-2020. Geo-statistical analysis made it possible to determine spatial regularities of changes in the content of macronutrients in steppe soils. The method of autocorrelation analysis was used to measure the minimum and maximum radii of typicality of the formation of agro-chemical properties of steppe soils being from 2.5 to 12.5 km, respectively. It indicates considerable spatial heterogeneity in distribution of macronutrients within the contours of different soil types. The neuro-technological modeling resulted in the creation of three-layer artificial neural networks for space-time modeling of the content of macronutrients in steppe soils. The reliability of approximation of neuro-models on test samples equaled 92.4–94.8%. An irreversible process of gradual depletion of steppe soils is forecasted under the current agricultural practices: a drop in the content of humus – by 0.01% per year on non-irrigated lands, by 0.03% per year on the average on irrigated lands; nitrifiable nitrogen – by 0.04 mg·kg-1 of soil on non-irrigated lands, by 0.06 mg·kg-1 of soil per year on the average on irrigated lands; mobile phosphorus – by 0.16 mg·kg-1 of soil per year on non-irrigated lands, by 0.18 mg·kg-1 of soil per year on the average on irrigated lands; exchangeable potassium – by 1.9 mg·kg-1 of soil per year on non-irrigated lands, by 3.1 mg·kg-1 of soil per year on the average on irrigated lands. The obtained result determines territorial priorities of the regional policies, making it possible to apply differential effectiveness of the soil- protecting block of agricultural systems.
Keywords: forecasting; GIS-technologies; humus; modelling; neuro-technologies; nitrogen; phosphorus, potassium; steppe soil fertility
Date published: 2023-02-24
Download full text