Abstract:

Traditional numerical weather prediction (NWP) models have made remarkable progress over the past few decades. Yet, as we inch closer to the limits of deterministic forecasting, there remains a growing need for innovative techniques to enhance forecast accuracy. Enter machine learning (ML) — a subset of artificial intelligence that is causing a paradigm shift in meteorological forecasting. This presentation offers a comprehensive overview of the latest advancements in the application of machine learning at Tomorrow.io across the weather forecasting value chain. There will be a focus on a core algorithm that post-processes NWP data with a multi-task neural network with a Continuous Ranked Probability Score (CRPS) custom loss function to generate probabilistic forecasts, but will also cover ML techniques used in short-term nowcasting and generative AI to summarize weather impact for customers.