Research of methods and processing of databases on the biomass of Eurasian forests as neural networks. Part 2. New opportunities for artificial intelligence in predicting climate-driven changes

Abstract

The assessment of the carbon storage capacity of forests has reached the global level, and the assessment of greenhouse gas absorption at carbon landfills is relevant. The authors have developed and published three author's databases on the biological productivity of Eurasian forests. It is shown that for databases, correct algorithms of alternative methods give close results, and an incorrect algorithm gives a significant shift in the result in relation to the model of the same ideology, but built according to the correct algorithm. The resulting models are used to predict changes in these indicators over time based on the principle of spatio-temporal substitution. It has been established that the climatic conditionality of the studied bioproduction indicators is of a general nature for both quantitative and qualimetric indicators of the biomass of trees and forest stands. The resulting models are applied in the construction of a neural network to predict changes in these indicators over time based on the principle of space-time substitution. In the process of machine learning and solution, it was found that the climatic conditionality of the studied bioproduction indicators is of a general nature for both quantitative and qualimetric indicators of the biomass of trees and forest stands.