Work

Reconstruction of Groundwater Level Data Using Temporal Components of Groundwater Fluctuations Based on Wavelet Analysis and Artificial Neural Networks

Year: 2026

Type: article

Abstract: Since the observations of groundwater level (GWL) in Ukraine are not conducted by automated means, the regularity of the data is affected by the human factor as well as social unrest. Since 2022, this has been a full-scale war launched by the russian federation. Continuous long-term GWL observations (to 2011, sometimes until 2017) were used to reconstruct periods with missing measurements, combining autocorrelation analysis, wavelet decomposition, Mann–Kendall trend testing, and artificial neural networks (ANNs). The strongest reconstruction performance was achieved by separating GWL fluctuations into short-, medium-, and long-period components and modeling the dominant medium- and long-period structures. Compared with linear autoregressive baselines, multilayer perceptrons (MLPs) better approximated nonlinear relationships present in the historical record. At the same time, these data-driven models remain sensitive to nonstationarity and should be interpreted as predictive tools rather than causal process models. The data reconstruction study covers the transboundary basin of the Bug River, which is significant for Ukraine and Poland as a water resource.

Source: Water

DOI: 10.3390/w18172105

Authors: Oleksii Shevchenko, Dmytro Charnyi, Ilya Zaslavsky, Vytautas Samalavičius, Yuliia Sovkova

Institutions: Ukrainian Hydrometeorological Institute, The Institute of Environmental Geochemistry of National Academy of Sciences of Ukraine, San Diego Supercomputer Center, University of California San Diego, Vilnius University

Language: English

Topic: Hydrological Forecasting Using AI

Subfield: Environmental Engineering

Field: Environmental Science

Domain: Physical Sciences

SDG: Clean water and sanitation

Open Access status: gold

Estimated APC paid: $3,137