1399 / 2024-09-26 18:19:13

Applying Deep Learning To Rendering Extreme Marine Environments

Extreme Weather,ocean,Scene Rendering

In extreme weather events such as hurricanes, floods, and droughts, the ocean is not only an important part of the climate system, but also affects weather patterns through changes in its temperature, salinity, and ocean currents. Changes in the ocean's significant wave height, ocean currents, and sea surface temperature play a key role in the formation and development of extreme weather events. Current numerical forecast models for extreme weather cannot intuitively express their impact on the Earth's environment. To better understand and predict these extreme weather events, we propose a controllable diffusion model for simulating extreme weather phenomena, combining reconstructed scenes with real precipitation data to enrich the dataset and consider multiple interactions between the ocean and the atmosphere. Our proposed method introduces an adaptive weight adjustment mechanism for extreme precipitation events to improve the sensitivity and response speed of the ControlNet model to ocean temperature anomalies (such as El Niño) and atmospheric circulation changes. By adding a permutation self-attention module to the model, the order or arrangement of features is changed, thereby enhancing the model's understanding of the interactive features between the ocean and the atmosphere, especially the impact of ocean current anomalies and ocean heat content distribution on precipitation patterns. In addition, we integrate multi-source data for the regulation and prediction of extreme precipitation, including ocean current data, wave data and precipitation radar data observed by satellite, to form a comprehensive multimodal prediction framework. This framework is not only applicable to flood and precipitation events on land, but also to the prediction of ocean-related disasters. The user interface based on the model can generate accurate extreme precipitation and marine disaster prediction results, providing strong support for relevant decision-making. Qualitative and quantitative studies show that our model can better capture the complex dynamic changes of extreme precipitation events, especially in simulating the impact of the ocean on extreme weather and marine disasters. In all comparative experiments, our simulation results are sota current comparison methods in terms of authenticity and accuracy.