A study was conducted to assess the accuracy of wind wave height forecasts in the Tsemess (Novorossiysk) Bay using the spectral model WAVEWATCH III forced by GFS winds (0.25°). Comparison of model simulations with in situ measurements obtained at the Sheskharis pier during December 2023 – August 2025 showed that the system provides satisfactory forecast quality of wave heights for different lead times up to 72 hours. The root-mean-square error of significant wave height forecasts ranges from 0.17 to 0.22 m, the correlation coefficient is 0.85–0.9, and the systematic bias is negative (–0.09…–0.13 m). The smallest errors were observed for a lead time of 15 hours. Seasonal variability analysis revealed higher forecast accuracy in the autumn–winter period and lower accuracy in the spring–summer season due to the influence of local wind effects. It was found that the model reproduces wave characteristics more accurately during offshore winds (from sea to land) than during onshore winds (from land to sea). The obtained results can be used to improve the accuracy of operational wave forecasting systems and to enhance maritime safety in the Tsemess Bay area.

A possibility of using deep neural networks to estimate the average district-level expected winter wheat yield for the territory of the North Caucasus Administration for Hydrometeorology and Environmental Monitoring is shown. The neural network was trained on a dataset that included satellite indices, meteorological data, and historical series of average regional yields for the period from 2012 to 2023. An experimental search for optimal neural network hyperparameters was carried out and made it possible to achieve the balance between the model’s accuracy and generalizability. A comparative analysis of the accuracy of calculating the expected yield was performed using neural networks, statistical regression models, and machine learning algorithms (decision tree, random forest, linear regression). The results of the analysis showed that the maximum convergence between actual and calculated winter wheat yields is achieved when using the neural network. The results demonstrate the potential of the neural network approach for assessing the expected yield of winter wheat based on the integration of ground-based and satellite data.

The paper investigates main weather anomalies of the 2023/2024 winter and the 2024 summer. It was found that crop formation conditions during this period were significantly worse than in the recent decade. Due to these anomalies, the yield did not meet expectations

The accuracy of seasonal forecasts of sea surface temperature (SST) and sea ice characteristics is assessed. Forecast skill scores are calculated for two versions of the INM RAS Earth system model (INM-CM5 and INM-CM6) differing in the horizontal resolution of the atmospheric and aerosol modules and in some parameterizations of physical processes: cloudiness, indirect aerosol effects, snow cover evolution. It is shown that in most cases the skill scores of SST and ice characteristics forecasts based on the INM-CM6 version are higher than those for the INM-CM5. The scores are comparable with those in foreign centers, confirming the compliance of the forecasting system based on the INM RAS model with the world level of accuracy of seasonal forecasts.

This study presents results of numerical experiments of neutrally-stratified turbulent flows over idealized urban surfaces using a Large-Eddy Simulation (LES) model. It is shown that the turbulent length scales necessary for the formulation of multilayer local one-dimensional Reynolds-Averaged Navier–Stokes (RANS) models of the urban canopy are related to the spatial spectra of turbulence. An algorithm based on the application of Taylor's frozen turbulence hypothesis is proposed to compute an analog of the spatial velocity spectrum inside an urban layer containing objects («buildings»). A qualitative explanation of the dependence of length scales on the morphological characteristics of the urban surface is given.

The paper examines the impact of applying the disturbances of parameters for the schemes describing subgrid-scale processes in the SL-AV atmosphere model on the characteristics of average monthly atmospheric circulation in ensemble subseasonal forecasts. 25-year retrospective forecasts are computed for different seasons with different perturbation sets, the results are compared with forecasts without perturbations. It is shown that perturbing three parameters reduces forecast errors of some variables as compared to undisturbed forecasts without introducing errors into integral characteristics

The analysis of long-term fluctuations in seasonal runoff of the Ural basin rivers is performed, statistical relationships between runoff and the main meteorological factors (air temperature and precipitation) are estimated. Weather stations with the highest correlation between the main meteorological factors and seasonal runoff of the Ural, Sakmara, and Ilek rivers are selected. Empirical dependences of seasonal runoff on air temperature and total precipitation are substantiated for three hydrological stations located on these rivers during the period with minimum anthropogenic load on water resources (slightly disturbed runoff). Their statistical and graph-analytical analysis showed that they have a fairly high correlation for the period of spring flooding and the summer-autumn season, and the dependences for the period of winter low water are much less accurate. Based on the constructed dependencies for a period of significant anthropogenic load on the water resources of the region, a comparative analysis of the calculated (undisturbed) seasonal runoff of the Ural, Sakmara, and Ilek rivers with the observed one is performed.