This study presents an integrated approach for predicting the spatial distribution of gold, copper, and lithium concentrations using machine learning, geostatistical methods, and multivariate geospatial data. The problem is formulated as a spatially dependent multivariate regression task, distinguishing it from traditional classification-based mineral prospectivity approaches. A unified database was developed, incorporating geochemical indicators, geomorphometric terrain parameters, remote sensing data, and spatial coordinates. Correlation analysis with an adaptive threshold was applied to optimize the feature set and improve model robustness. The results show that linear methods are limited in capturing nonlinear relationships, while ensemble methods provide significantly higher predictive accuracy. In some cases, geostatistical methods achieve the best performance, emphasizing the importance of spatial structure. Feature importance analysis indicates that gold prediction is primarily driven by geochemical indicators, spatial coordinates, and terrain characteristics. Results for copper and lithium confirm the general applicability of the proposed approach. Overall, the study demonstrates the effectiveness of combining machine learning and geostatistics for modeling geochemical processes.

This work presents the corporate geoinformation system Mineral AI, developed to support spatial-analytical decision-making for subsoil use in the East Kazakhstan region. The system is implemented on the ArcGIS Enterprise platform and integrated with the ArcGIS Pro desktop application, providing a unified technological workflow from data acquisition to publication of analytical results. Mineral AI supports the full lifecycle of geodata management, including field acquisition and spatial referencing of geological, lithogeochemical, and geophysical observations; ingestion into the corporate environment; data normalization, cleaning, and classification; anomaly detection; and harmonization of indicators for subsequent analysis. Following preprocessing, the data are subjected to spatial analysis, geostatistical interpolation, 3D modeling, and predictive assessment of prospective mineralized zones. Particular attention is given to constructing probabilistic prospectivity layers based on the integration of geological, lithogeochemical, and structural factors. The analytical outputs are published as managed geoservices with controlled access via the ArcGIS Enterprise portal, ensuring data consistency and version control across expert units. The results demonstrate that this architecture transforms the assessment of mineral resource potential from a fragmented expert-driven activity into a formalized, reproducible, institutionally supported procedure suitable for exploration planning, optimization of geological survey resource allocation, and justification of subsoil-use management decisions.

This study focuses on the coefficient inverse problem arising in magnetotelluric (MT) sounding, which plays a crucial role in geophysical exploration and subsurface characterization. The main objective is twofold: first, to construct a reliable forward numerical model based on the Helmholtz equation with a complex-valued conductivity coefficient, and second, to develop a stable inversion procedure for reconstructing the conductivity distribution from boundary measurements. The forward problem is discretized using a finite-difference approximation, ensuring numerical stability and accuracy for both the direct and adjoint formulations. To address the ill-posed nature of the inverse problem, a misfit functional is introduced, measuring the discrepancy between simulated and observed boundary data. This functional is minimized using the iterative Landweber method, which provides a simple yet robust tool for stabilizing reconstructions. Numerical experiments are carried out for a synthetic conductivity model consisting of a smooth background medium with an embedded localized anomaly. The obtained results demonstrate the ability of the proposed method to recover key structural features of the anomaly. The presented framework offers a promising foundation for the development of practical inversion algorithms applicable to real geophysical MT data. Key words: Helmholtz equation, magnetotelluric sounding, inverse problem, Landweber method, numerical solution.