Melnyk et al. (2026) Time Series Analysis and Periodicity Analysis and Forecasting of the Dniester River Flow Using Spectral, SSA, and Hybrid Models

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Water
• Year: 2026
• Date: 2026-01-22
• Authors: Serhii Melnyk, Kateryna Vasiutynska, O. Butenko, Iryna Korduba, Роман Трач, A. M. Pryshchepa, Yuliia Trach, Vitalii Protsiuk
• DOI: 10.3390/w18020291

Research Groups

Not specified in the provided text.

Short Summary

This study applied spectral and singular spectrum analysis to Dniester River runoff (1950–2024) to identify dominant periodicities (approximately 30, 11, 3–5.8, and 2 years) linked to large-scale natural drivers, and developed ensemble forecasts predicting reduced runoff for 2025–2028 followed by recovery.

Objective

• To identify the dominant periodic components governing the hydrological regime of the Dniester River basin and assess their links to large-scale natural drivers.
• To develop data-driven runoff models and medium-term forecasts for the Dniester River using a combined spectral–singular spectrum analysis framework, complemented by CNN–LSTM and hybrid ensemble approaches.

Study Configuration

• Spatial Scale: Dniester River basin, a transboundary basin shared by Ukraine and Moldova, utilizing data from three gauging stations.
• Temporal Scale: Mean annual runoff time series from 1950 to 2024 for analysis; medium-term forecasts generated for 2025–2034.

Methodology and Data

• Models used: Spectral analysis, Singular Spectrum Analysis (SSA), Cross-spectral analysis, Coherence analysis, Convolutional Neural Network – Long Short-Term Memory (CNN–LSTM) models, Hybrid ensemble approach.
• Data sources: Time series of mean annual runoff from three gauging stations on the Dniester River.

Main Results

• Four stable groups of periodic variability were identified in the Dniester River's mean annual runoff, with characteristic timescales of approximately 30 years, 11 years, 3–5.8 years, and 2 years.
• These periodicities correspond to known major atmospheric–oceanic oscillations (e.g., AMO, NAO, PDO, ENSO, QBO) and the 11-year solar cycle.
• Cross-spectral and coherence analyses revealed a statistically significant relationship between solar activity and river discharge, with an estimated lag of approximately 2 years.
• SSA reconstructions were highly reliable, explaining more than 80% of the discharge variance.
• Forecast comparisons indicated that spectral methods tend to amplify long-term trends, CNN–LSTM models produce conservative trajectories, and a hybrid ensemble approach provides the most balanced and physically interpretable projections.
• Ensemble forecasts predict reduced runoff during the period 2025–2028, followed by a recovery in runoff from 2029 to 2034.

Contributions

• First systematic application of a combined spectral–Singular Spectrum Analysis (SSA) framework to the Dniester River basin for hydrological regime characterization.
• Basin-specific integration of this novel framework to consistently characterize runoff variability and assess large-scale natural drivers.
• Identification and quantification of dominant periodic components in Dniester River runoff and their statistically significant links to major atmospheric–oceanic oscillations and solar activity.
• Development and comparative assessment of data-driven runoff models (spectral, CNN–LSTM, hybrid ensemble) for medium-term forecasting in the Dniester basin.
• Provision of actionable, physically interpretable ensemble forecasts for future runoff, supporting long-term water-resources planning and climate adaptation strategies in the region.

Funding

Not specified in the provided text.

Citation

@article{Melnyk2026Time,
  author = {Melnyk, Serhii and Vasiutynska, Kateryna and Butenko, O. and Korduba, Iryna and Трач, Роман and Pryshchepa, A. M. and Trach, Yuliia and Protsiuk, Vitalii},
  title = {Time Series Analysis and Periodicity Analysis and Forecasting of the Dniester River Flow Using Spectral, SSA, and Hybrid Models},
  journal = {Water},
  year = {2026},
  doi = {10.3390/w18020291},
  url = {https://doi.org/10.3390/w18020291}
}