MoWE : A Mixture of Weather Experts

2025-09-10

AI Summary

Recent data-driven weather forecasting models have exhibited performance stagnation. To address this, we propose a novel forecasting paradigm based on multi-expert ensemble—without designing new base models—by dynamically weighting the outputs of multiple existing models via a vision Transformer-based gating network. Our key innovation lies in a spatiotemporal adaptive dynamic weighting mechanism that optimizes fusion weights according to forecast lead time and geographic location. Coupled with a lightweight training strategy, the framework enables efficient, scalable expert mixture modeling. Evaluated on a 2-day forecasting task, our method reduces RMSE by 10% relative to the best-performing AI-based meteorological model, significantly outperforming naïve averaging ensembles while incurring lower computational overhead.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Computer Vision: Multi-modal Vision
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web search models and ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Data-driven weather models have recently achieved state-of-the-art performance, yet progress has plateaued in recent years. This paper introduces a Mixture of Experts (MoWE) approach as a novel paradigm to overcome these limitations, not by creating a new forecaster, but by optimally combining the outputs of existing models. The MoWE model is trained with significantly lower computational resources than the individual experts. Our model employs a Vision Transformer-based gating network that dynamically learns to weight the contributions of multiple "expert" models at each grid point, conditioned on forecast lead time. This approach creates a synthesized deterministic forecast that is more accurate than any individual component in terms of Root Mean Squared Error (RMSE). Our results demonstrate the effectiveness of this method, achieving up to a 10% lower RMSE than the best-performing AI weather model on a 2-day forecast horizon, significantly outperforming individual experts as well as a simple average across experts. This work presents a computationally efficient and scalable strategy to push the state of the art in data-driven weather prediction by making the most out of leading high-quality forecast models.

Problem

• Combining existing weather models optimally for better forecasts
• Reducing computational resources compared to individual expert models
• Improving forecast accuracy beyond state-of-the-art AI weather models

Innovation

• Mixture of Experts combination approach
• Vision Transformer-based gating network
• Dynamic weighting of expert models