7. System-Analytical Modeling of Water Quality for Mountain River Runoff

The method of system-analytical modeling of complex natural systems is proposed to construct the simulation balance models for water and hydrochemical runoff of mountain rivers. An integral set of universal models for seasonal and long-term runoff dynamics was developed based on the example of 34 mid-size and small rivers of the Altai-Sayan mountain country as a regional case study. The set includes (a) regional climate model, (b) water runoff balance model, and (c) seven hydrochemical runoff balance models. The latter characterizes the hydrochemical composition of the river runoff: three nitrogen mineral forms (\({\mathrm{NO}}_{2}^{-}, {\mathrm{NO}}_{3}^{-}, {\mathrm{and\,NH}}_{4}^{+}\)), phosphates (\({\mathrm{PO}}_{4}^{3-}\)), ions, total dissolved iron, and suspended matter. To calculate the seasonal runoff, four hydrological periods/seasons were specified: winter low water, spring–summer flood, summer low water, and autumn low water. A total of 13 typological geosystem groups (landscapes) were selected to account for a landscape structure of river basins. In “b” and “c” models, the hydrological and hydrochemical regimes of river basins are divided into 13 standard types that correspond to selected hydrological seasons and landscapes. Each type depends on spatially generalized monthly dynamics of precipitation and air temperature. These meteorological characteristics are calculated in “a” model and expressed in percent of specified long-term mean values to be the same throughout the Altai-Sayan mountain country. GIS data on the relief and landscape structures of mountain river basins represent the input information for the “b” and “c” models. These data include the area and average altitude of the basins, the area and elevation of landscapes, the altitude of the outlet, the length of river channels (between the river head and the outlet), and the area of arable land. The spatially generalized for the Altai-Sayan mountain country normalized monthly precipitation and mean monthly air temperature as well as water runoff estimated for individual landscapes in river basins with “a” and “b” models serve as input factors for seven “c” models. The sensitivity of models to variations of input factors was evaluated. The sensitivity is expressed as a contribution of a particular factor to the variance of the observed values of the output variable (water or hydrochemical runoff). A quantitative assessment of the water and hydrochemical runoff sensitivity was obtained for the following factors: landscape structure of river basins, basin lateral slope, precipitation, temperature, arable land area. Both RSR < 0.60 (RMSE-standard deviation ratio) and NSE > 0.65 (Nash–Sutcliffe model efficiency coefficient) estimated for each of the developed runoff models represent their good or very good performance. The most probable water and hydrochemical runoff can be forecasted for 3–4 months ahead with a twice reduced variance as compared with the similar forecast by the observed mean runoff. The elaborated balance models with anew selected landscapes and the updated values of parameters can be applied to any mountainous area and allow to estimate and manage the seasonal and long-term dynamics of water quality.