Neural Network Approaches to Simulation of Spring Flood in Sites of the Lena River

Abstract

The Republic of Sakha (Yakutia), with an area of 3103.2 thousand km², is one of the largest constituent entities of the Russian Federation with a sharply continental climate and a multitude of large and small rivers and streams. The Lena River is one of the longest rivers of Russia, flows through the whole republic from south to north through the territory with different climatic zones, this leads to a large number of jamming phenomena during the spring floods with a large spatial and temporal range.

Spring-summer floods cause enormous damage to the region, causing flooding of vast territories and objects of the national economy, which determines the relevance and need to develop and improve flood forecasting methods to implement timely measures to prevent and reduce the risk of flooding, especially in densely populated areas and areas with complex infrastructure, such as strategically important facilities, underwater crossings of main pipelines, bridges and power lines. The proposed methods make it possible to estimate water levels during spring floods both on the basis of predicting a time series from previous values and regression dependences on various factors (ice thickness, temperature, etc.) and with sufficient accuracy, which is shown by the results of forecasting maximum water levels for example of two sections of the Lena River, based on historical data.