Test Polygons for Geosystems State Diagnostics and Development of Earth Remote Sensing Data Interpreting Techniques

Abstract

The study presented in the article is aimed at developing methods and algorithms for system analysis of test site data in order to diagnose the state of geosystems and develop methods for interpreting Earth remote sensing data. Materials and methods. The development of methods and algorithms for the analysis and integration of spatial information was based on the analysis of a system of test sites that reveal the features of the interaction between the forest-steppe and forest geosystems of the Volga Upland and the Oka-Don Lowland. The general scheme of the process of compiling a digital map of metageosystems is implemented by solving the following tasks: collecting and preparing a system of thematic maps and databases; systematization of information with the construction of a hierarchy of geosystems; ensemble analysis of multi-zone satellite images with the construction of a synthetic map of geosystems; evaluation of simulation results; obtaining and practical use of spatial information. Resultsand discussion. Combining models into an ensemble based on the proposed architecture of the metaclassifier makes it possible to increase the stability of the analyzing system: the accuracy of decisions made by the ensemble tends to tend to the accuracy of the most efficient monoclassifier of the system. Systematic analysis of territory descriptors integrated on the basis of data from different sources provides a significant increase in the accuracy of metageosystem classification. It is important that the cartograms of the presented descriptors are well interpreted by specialists in the field of data analysis in the geosciences. Conclusion. The use of ensembles built according to the methodology proposed in the article makes it possible to carry out an operational automated analysis of spatial data to solve the problem of thematic mapping of metageosystems and natural processes. The calculation and consolidation of territorial descriptors makes it possible to reduce the dimensionality of the analyzed data, to facilitate the allowable capacity of the machine learning model, to increase its resistance to overfitting, and to prevent a significant decrease in the classification accuracy within the framework of a specific problem being solved.