Analysis of the water level regime of the Taz River for optimization of waterway operation

Abstract

. Improving the safety and efficiency of shipping routes is a pressing issue for water transport in the Arctic zone, which depends on the accuracy of hydrological regime forecasting. This study aimed to analyze the water level regime to develop a methodology for forecasting maximum water levels at the outlet of the Taz River (Krasnoselkup village), which is essential for navigation in this region. The study utilized a comprehensive approach, including exploratory statistical analysis, a time series stationarity assessment, a probabilistic assessment of extreme levels, and a comparative analysis of the effectiveness of various forecast models. The study was based on a 50-year observation series (1976–2025). The results demonstrated stationarity of the series in the absence of a statistically significant trend. A comparative evaluation of the classical ARIMA model and a number of machine learning algorithms (linear regression, random forest, gradient boosting, and neural network) revealed the superiority of the ARIMA(0,1,1) model. The Random Forest machine learning model demonstrated the best performance among ML methods, but its predictive power (R² = -0.0098) was unsatisfactory. The practical significance of this work lies in the fact that the developed model and the calculated flood levels of varying frequency (e.g., a 10-year flood of 1051.2 cm) can be integrated into a decision support system for optimizing navigation planning, risk assessment, and the design of hydraulic structures on the Taz River. The results confirm the feasibility of using classical statistical methods for forecasting relatively short stationary hydrological series in data-limited conditions.