Assessing the Effect of the Stochastic Structure of Inflow into Lake Baikal on the Long-Term Level Regime of the Lake

Abstract

The statistical homogeneity of the total inflow into Lake Baikal have been analyzed. The periods of 1901–1933 and 1933–2014 have been found not only to show statistically significant difference between the coefficients of autocorrelation of the total inflow, but also changes in its stochastic structure. For the first period, the autocorrelation function of the total inflow into Lake Baikal corresponds to AR(1)-process, that for the second period corresponds to a single-component AR(2)-process or the most adequate two-component AR-process with components in the form of AR(2)-processes. The use of AR(1)-process to describe the total inflow into Lake Baikal in 1933–2014 has been shown to lead to underestimation of the dispersion and autocorrelation coefficient of water level variations in Lake Baikal. The stochastic structure of the long-term variations of the total water inflow into the lake that has formed after 1934 has to be taken into account to obtain more accurate estimates of these characteristics. With the stochastic properties of water inflow into the lake taken into account properly, extremely high water levels in Lake Baikal show higher exceedance probability than those derived from the model based on AR(1)-process used to describe the inflow. For example, the level of 1% exceedance probability that corresponds to AR(1)-model of inflow will become a level with 2% exceedance probability if the inflow is correctly modeled as an AR-process with two components in the form of AR(2)-processes.