Multi-Scale Application of Advanced ANN-MLP Model for Increasing the Large-Scale Improvement of Digital Data Visualisation Due to Anomalous Lithogenic and Anthropogenic Elements Distribution

Abstract

The main objective of this paper is to compare and improve spatial distributions models for Pb and Cu in air and soil using the universal kriging and ANN-MLP at the macro regional scale. For this purpose, both models have been applied for visualization of a spatial distribution of lead (Pb) and copper (Cu) in a morphologically and geologically complex area. Two river basins in the eastern part of North Macedonia, have been selected as the main research region due to the extensive anthropogenic impact of long-lasting mining activities, with emphasis on the specific geochemistry of the area. Two environmental media (soil and moss) have been selected as they are much more available as space from biospheres submitted for destruction processes globally. Surface soil and moss as bio-indicator element measurements were submitted in correlation with geospatial data obtained from DEM, land cover data, and remote sensing, and are incorporated into spatial distribution mapping using an advanced prediction modeling technique, ANN-MPL. Both methods have been further compared and evaluated. The comparative data outputs have led to the general conclusion that ANN-MPL gives more realistic, reliable, and comprehensive results than the universal kriging method for the reconstruction of main distribution pathways. The more the factors influencing the process of distribution of the elements increase, the more the use of ANN-MPL improves.

Keywords

• trace elements
• soil
• moss
• spatial distribution
• artificial neural network–multi-layer perceptron
• universal kriging