ИСТИНА

The PAIPS (Pulsating Aurora Imaging Photometer System) project — a high-time-resolution UV photometer network on the Kola Peninsula — has revealed a novel class of auroral phenomena: optical microbursts (OMB) . OMBs consist of flash sequences lasting from 10 s to 3 h, with peaks of approximately 100 ms duration. These events are observed in the evening MLT sector. Over three observation seasons (2021–2024), 102 groups of such events have been recorded; the catalog, light curves, and observation conditions are available on the project website: https://uhecr.sinp.msu.ru/paips.html. The formation mechanism of OMBs remains unclear. To investigate the spatial distribution of OMBs, expansion of the PAIPS meridional network is planned and automation of large-volume data processing is required. To this end, neural network-based methods for OMB detection in PAIPS data are under development, which will streamline event identification and enhance selection efficiency in both archival and incoming data from the expanding observational network. Currently, a binary classifier based on a hybrid architecture consisting of convolutional (CNN) and LSTM blocks has been realized. The classification accuracy on the test sample is 94%. As a validation test on real data, two previously manually reviewed datasets were selected. The classifier correctly identified all series previously detected during visual analysis, and also found one additional OMB series. This demonstrates the effectiveness of the developed algorithm. Recall on these test datasets was ≈100%, suggesting a low miss rate under the tested conditions. For a more complete study of optical microbursts, this model was applied to archived data from the PAIP-V, one of the PAIPS family photometer, which was located at the Verkhnetulomsky Observatory. This work was supported by the Russian Science Foundation grant № 22-62-00010. References [1] Klimov P. A., Nikolaeva V. D., Saraev R. E., Shchelkanov K. D., Belov A. A., Kozelov B. V., Murashov A. S., Roldugin A. V., Sharakin S. A. (2024). Advances in Space Research, 74(8), 3556-3568.