This paper presents the simulation results of meteorological and air quality parameters for the Siberian city of Tomsk predicted by mesoscale meteorological and chemical transport models. Changes in the numerically predicted wind velocity fields, temperature, and concentration of major air pollutants were modelled in detail for the selected dates, when anticyclonic weather with cloud free and calm wind conditions was observed in Tomsk. The simulation results have shown that stable or neutral atmospheric stratification with light wind and low ambient air temperature (−30, −20 °C) are the most unfavorable meteorological conditions leading to the near surface pollutants accumulation. The numerical calculation results were compared with observation data from the Joint Use Center (JUC) “Atmosphere” of V.E. Zuev Institute of Atmospheric Optics (IAO) and showed good agreement.

We numerically reproduced the spring riverine thermal bar in the area of the Selenga inflow into Lake Baikal using the nonhydrostatic 2.5D mathematical model. Our model took into account the diurnal variability of the heat fluxes and wind stress on the lake surface based on the atmospheric data from the Babushkin weather station archive during 1–30 May 2015. Propagation of the thermal bar is driven principally by the mechanical and thermal energy of the river inflow together with shortwave radiation while longwave radiation and latent and sensible heat fluxes make smaller contributions. Numerical modeling of the lake hydrodynamics demonstrated that the thermal bar propagation decelerated at night and that strong reverse motion of the thermal bar (toward the shore) was possible within some specific areas due to night cooling and opposing wind. These effects appear following an initial period in which the dynamics are dominated by the river inflow. As the distance from the mouth of the Selenga to the thermal bar increased, the impact of westerly winds on the reverse movement of the thermal bar was reinforced, although the influence of wind does not extend to the full depth of the lake.

The results of mathematical modeling of a nonisothermal turbulent air flow and admixture transport in an idealized street canyon are presented. The simulation is based on the RANS model, numerical algorithm, and application package developed by the authors, which are improved to take into account the effect of the buoyancy force on the aerodynamics and admixture transport. The mathematical model includes Reynolds-averaged stationary three-dimensional Navier–Stokes equations, as well as equations of heat transfer and impurity transport. For closing, a k-eps turbulence model is chosen taking into account the buoyancy forces. The numerical algorithm uses the finite volume method, nonuniform structured grids, and the fictitious domain method. The differential problem is approximated using Van Leer’s monotonized linear upwind scheme and piecewise linear interpolation for dependent quantities. The resulting grid equations are solved sequentially by N.I. Buleev’s iterative method. The SIMPLE algorithm is used to match the velocity and pressure fields. Using the developed package of applied programs, it is found that narrow and high street canyons are the least ventilated at a low wind velocity (~1 m/s), and the higher the street canyon with a constant width the higher the average concentration in the breathing zone. When studying the influence of the degree of heating of the surfaces of a street canyon with the same height and width, the least ventilated case is when the surface temperature of the windward vertical side of the canyon is 15–20°C higher than the ambient temperature. The main reason for this is the formation of a secondary vortex above the road surface, due to which the admixture is poorly carried out of the canyon.

This paper presents a parallel algorithm for numerical solution of equations for the non-hydrostatic mesoscale meteorological model TSUNM3. To justify the choice of the Message Passing Interface technology, computational experiments were performed to compare the effectiveness of the following parallelizing technologies: MPI, OpenMP, OpenACC, and CUDA. 2D decomposition of the gridded domain for the TSUNM3 model has been carried out with MPI technology. It allows for a numerical forecast for the next day in 17 minutes of CPU time on 144 cores of the TSU Cyberia computing cluster.