Algorithm for Using Artificial Intelligence to Monitor Agrogenic and Anthropogenically Disturbed Soils

Introduction

In 2019, by Presidential Decree No. 490, the National Strategy for the Development of Artificial Intelligence for the period until 2030 was approved in the Russian Federation.” (Decree of the President..., 2019). It is aimed at bringing the country to a leading position in the world. Interest in AI is driven by the emergence of new technologies and graphics processors, the growth in the computing power of modern computers, cloud computing, and the explosive growth of big data. The high speed and accuracy of models obtained through automated machine learning (Kataev et al., 2023a) has opened up new opportunities for process automation and digital transformation in different areas of human life.

The emergence of unmanned aerial vehicles, digital cameras and display sensors made it possible to continuously collect multi-temporal and multispectral data with high resolution (Kataev et al., 2023b; Tokareva et al., 2018, 2020, 2021; Pasko et al., 2019a,b, 2020a, b). Their analysis using GIS and AI represents a huge potential for solving modern problems facing the natural sciences, the agro-industrial complex, etc. The use of AI in the field of agriculture, soil science and agrochemistry is a relatively new, but promising task, since it allows to improve management efficiency, implement optimal preventive measures based on forecasts and prompts of AI systems, automate routine operations and identify hidden factors affecting the effectiveness of measures and offer information available for decision-making by ordinary users.

The synergistic use of artificial intelligence, remote sensing of the Earth and geoinformation systems opens up great opportunities for solving scientific and applied problems in the field of soil ecology monitoring and agricultural production management (Agroecological state, 2018). It allows you to quickly obtain and analyze objective and reliable data on the state of soil and vegetation according to unified methods, at a time – from sites significantly distant from each other, or information about their condition in previous years. The accumulated experience will form a database for machine learning of neural networks and the selection of significant criteria. The compiled algorithm will become the basis for modeling the state of the agroecosystem and forecasting its development in regulated and/or natural conditions.

As part of the implementation of the project on the use of artificial intelligence for monitoring agrogenic and disturbed soils, the following are provided:

• 1.

  Preliminary processing and analysis of spatial data obtained in the field and from remote sensing data;

• 2.

  Automated classification without training;

• 3.

  Interpretation and verification of results;

• 4.

  Field survey of the territory;

• 5.

  Selection and description of key areas for creating training samples. Creating training standards;

• 6.

  Evaluation of the quality of standards;

• 7.

  Choosing the decisive classification rule;

• 8.

  Classification;

• 9.

  Post-classification processing using the smoothing method and generalization.

The scientific literature available to us does not contain information about the use of intelligent systems for studying soils. The purpose of the chapter is to apply the capabilities of intelligent systems to solve urgent problems of agrochemistry and soil science, as well as to model crop yields. The issues of anthropogenic soil degradation and the use of agrogenic soils of agricultural lands are selected as urgent problems.

Anthropogenically transformed soils receive a lot of attention in modern science (Agroecological state..., 2018).