Machine Learning Insights into Precipitation: Implicitly Learning Cloud Organization for Improved Predictions
* Register (or log in) to the AI4G Neural Network to add this session to your agenda or watch the replay
In climate modeling, achieving precise prediction of precipitation intensity remains a persistent challenge. While global models yield indispensable data, many consistently exhibit shortcomings in capturing the complexities of precipitation extremes. One plausible explanation lies in the omitted details of subgrid-scale cloud organization, which significantly influences both precipitation intensity and its stochastic behavior. Objectively modeling these patterns has been a challenging task, leaving gaps in our representation of subgrid-scale processes and potential precipitation. Our recent approach, anchored in machine learning methodologies, enables the implicit learning and integration of this subgrid organization from high-resolution precipitable water fields. Notably, we implicitly learn this information using advanced nonlinear dimensionality reduction techniques. The inclusion of the learnt subgrid scale structure significantly improves precipitation prediction. Furthermore, our research highlights the crucial role of memory processes, influenced by subgrid-scale structures, emphasizing their significance in precipitation variability and intensity.
This live event includes a 15-minute networking event hosted on the AI for Good Neural Network. This is your opportunity to ask questions, interact with the panelists and participants and build connections with the AI for Good community.