The article compares two ensemble forecasting systems, S1 and S2, using the SLAV072L96 model, in terms of forecasts with a lead time of up to 6 weeks. The first system, S1, uses an ensemble of 61 members, and the ensemble of initial states is generated with the help of the breeding method. The newer system, S2, has 41 members, and the ensemble of initial states is prepared using the data assimilation technique based on the local ensemble Kalman filter. It is shown that despite a smaller number of ensemble members, S2 is not inferior to S1; instead, it has some advantages, which are especially remarkable over long integration intervals (up to 46 days). At the same time, both systems are characterized by inadequate spread of the ensemble and asymmetric distribution of forecasted values and hence require additional adjustment and improvement. It is concluded that the use and further development of the S2 system is preferable, since the same or slightly higher quality of ensemble forecasts is achieved with lower computational costs.
Based on long series (for the period 1958–2024), a study was conducted on the statistical structure of the field of available moisture content in the one-meter layer of soil under crops in European Russia, and the impact of observed climate changes on the soil moisture regime was assessed. It has been shown that in the 21st century, the soil moisture content is higher than that in the period 1958–1999, and the currently observed climatic changes are positive for the agricultural industry of Russia. The curves constructed for seasonal variations of available moisture content according to modern data can be applied in operational agrometeorological practice. The main reasons for the changes were analyzed: an increase in the recurrence of warm winters and the cyclical nature of the climate system.
The paper discusses results of analyzing the correspondence of the empirical probability distribution of daily precipitation’s absolute maxima to the theoretical Pearson and Frechet curves. An application of the method of constructing probability curves for the largest members of the samples has shown that the laws of maximum daily precipitation distribution are satisfactorily described using the Pearson distribution. The use of the Frechet distribution leads to a significant overestimation of distribution quantiles in the region of rare events.
The paper presents the results of developing the Lavinex Russian system for dynamic impact on the snow cover at avalanche sites. The system for artificial avalanche triggering using the gas-air mixture explosions consists of a detonating device installed on a concrete foundation in the upper part of an avalanche site (2–5 m below the top of the avalanche initiation zone) as well as an autonomous gas (oxygen and propane) component storage and supply unit (“shelter”) installed out of the area of possible impact of falling stones or snowslides. The components of the system are connected by gas supply pipelines. The device is controlled remotely. The tests have shown that an explosion of gas mixture consisting of 1/6 of propane and 5/6 of oxygen makes it possible to generate an excessive pressure up to 30 kPa (relative to atmospheric pressure) onto the snow surface at the avalanche site. This significantly exceeds the pressure of 8 kPa generated by the foreign analogue, the GAZ.EX system used in Russia and abroad.
This study employs a machine learning approach based on the Support Vector Machine (SVM) for water body classification from remote sensing imagery. To enhance the model’s performance, generalization ability, and robustness, the Particle Swarm Optimization (PSO) algorithm was used to optimize the SVM parameters, constructing the PSO-SVM model. Using the Gaofen-1 (GF-1) satellite imagery from Poyang Lake, the classification results before and after optimization were compared. The PSO-SVM achieved a 7.3332
