Estimating Maximum Water Levels During Winter Flooding at Some Segments of the Lena River

Abstract

The Sakha Republic (Yakutia) has a large territory that covers various climatic zones and a network of water bodies and thus is exposed to a wide range of natural emergencies. The most typical of them is spring-summer floods that cause flooding of vast territories, facilities and infrastructure, thus causing enormous damage to the economy; it determines relevance of developing and perfecting flood prediction methods to reduce the hazard level and possible damage. This research presents application of multiparametric models and neuron networks for development of a predictive model that allows forecasting the spring flooding hazard from statistical data accumulated through 44 years of observation and regressive modeling. The proposed methods allow evaluating the spring flood water levels as a function of various factors (thickness of ice, temperature, etc.) with sufficient accuracy, as confirmed with the results of predicting maximum water levels for two segments of the Lena River. Selection of the river course segments was determined by nearby location of potentially hazardous facilities whose flooding may cause significant property loss. Factors influencing spring flood levels have been determined.

Modeling water level using downstream river water level observations and machine learning methods

“…Traditionally, such calculations are based on hydraulic models [1,2] (if reliable data on the configuration of water surface slopes are available) or statistical models (if an archive of data on level observations at both sections is available) [3]. Some authors discuss using time series predictions [4], which makes it possible to obtain estimates of the water level during spring floods.…”