rivers

This repository contains the code and processing workflows used in the study on predicting the onset of the spring flood rise in Arctic–subarctic rivers of northeastern Russia using machine learning models. The notebooks implement the complete data-processing and modeling pipeline, including parsing of hydrological observations, climate data integration, feature construction, model training, evaluation, and visualization. Hydrological data originate from the gauging-station network operated by the Federal Agency for Water Resources of the Russian Federation (Rosvodresursy). Meteorological forcing data are obtained from the NASA POWER (Prediction Of Worldwide Energy Resources) database. The notebooks provided here reproduce the workflow used to construct the station–year dataset and to train and evaluate the predictive models. The repository is organized into several groups of notebooks corresponding to different stages of the analysis: Data parsing and preprocessing Coord Parsing.ipynb — extraction and cleaning of geographic coordinates of hydrological stations. Info Parsing.ipynb — parsing station metadata including river identifiers, drainage area, and station descriptors. Climate Parsing.ipynb — processing of daily meteorological data obtained from the NASA POWER dataset. Climate Net.ipynb — construction of the spatial climate grid and linking meteorological variables to hydrological stations. River level.ipynb — preparation of daily water-level series and preliminary hydrograph processing. Processing.ipynb — integration of hydrological observations, climate data, and station metadata into a unified dataset. Machine learning modeling Model baseline.ipynb — baseline regression model for predicting the onset of spring flood rise. Model comparison.ipynb — comparison of several ensemble machine learning algorithms. Prediction error vs lead time.ipynb — evaluation of prediction accuracy as a function of prediction lead time. Sensibity analysis.ipynb — sensitivity analysis of predictor variables and model response. Visualization and analysis Visual.ipynb — generation of general figures and exploratory plots. new visual.ipynb — additional visualizations used in the manuscript. Auxiliary scripts compress.ipynb — utilities for file compression and dataset preparation. The notebooks collectively reproduce the data-processing pipeline from raw hydrological and meteorological observations to the final station–year machine-learning dataset and model evaluation results. All notebooks are written in Python and designed to be executed independently for specific stages of the workflow.