Forest soil nutrient spatial prediction method based on artificial neural network Kriging interpolation

Abstract

The invention discloses a forest soil nutrient spatial prediction method based on artificial neural network Kriging interpolation, which comprises the following steps of: obtaining environmental factor grid data; calculating to obtain a forest soil nutrient spatial distribution diagram based on a multi-layer perceptron neural network; carrying out residual calculation between a measured nutrient value and a predicted value; carrying out analysis and verification on the prediction residual of the neural network; carrying out semi-variance calculation of the residual simulating a model determined by a semi-variance function to obtain model types and parameters; carrying out ordinary Kriging interpolation on the parameters of semi-variance model parameters to obtain the spatial distribution of the neural network prediction residual; adding a forest soil nutrient grid and a prediction residual grid based on a multi-layer perceptron neural network to obtain the forest soil nutrient spatialdistribution diagram based on the artificial neural network Kriging interpolation. The prediction precision of the method is obviously improved compared with a method by using only a multi-layer perceptron neural network model or ordinary Kriging interpolation.

Description

technical field [0001] The invention relates to the technical field of soil nutrient spatial prediction, in particular to a method for spatial prediction of forest soil nutrient based on artificial neural network kriging interpolation. Background technique [0002] Forests are the main body of terrestrial ecosystems, and the spatial distribution of soil nutrients plays an important role in the formation and succession of plant community spatial patterns, which are directly related to ecosystem productivity. Therefore, research on spatial prediction of forest soil nutrients is an important support for the realization of sustainable forest use and global change research. In recent years, experts and scholars at home and abroad have conducted research on the spatial prediction methods of soil nutrients in various ecosystems such as cultivated land, forest land, and wetlands. However, due to the high spatial correlation and heterogeneity of forest soil, the spatial distribution ...

AI Technical Summary

Problems solved by technology

However, similar to geographically weighted regression, each prediction unit is represented by the predicted value of soil nutrients at a node, and only the influence of environmental factors at this node is considered, and the spatial autocorrelation of residuals is not considered, so the generated soil nutrient map is not a smooth one. continuous surface, and in the case of fewer nodes, it will lead to soil nutrient mutation phenomenon that is inconsistent with reality

Images