Model improves weather forecasts by combining station errors and locations
WxFM-XL: Adapting Univariate Foundation Models to Multi-Station Weather Forecasting
Machine Learning
Summary
Weather forecasts often use models that predict the weather based on data from one spot at a time. The authors found that these models miss two important things when predicting across many weather stations: the distances between stations and how prediction mistakes vary by location. They made a new model called WxFM-XL that connects these ideas by using graphs that show relationships between stations and their typical forecast errors. This new model helps make better weather predictions over multiple locations than previous methods.
What this means in practice
- •For meteorological service teams: Create more accurate regional weather forecasts by combining error patterns and spatial data from multiple weather stations using this model.
- •For energy grid operators: Improve short-term renewable energy output forecasts by integrating multi-station weather predictions enhanced with spatial and error correlations.
Authors
Xiao Wang, Changjian Chen, Zhuo Tang, Rongwen Li, Hongwu Liu, Kenli Li
Abstract
With the rise of univariate time series foundation models (e.g., Sundial, Timer), initial efforts have been made to extend them to multivariate settings. However, these models mainly focus on modeling correlations among variables. When they are applied to multi-station weather forecasting, two important factors are often overlooked: (1) the spatial information of stations, and (2) different error priors of different stations relative to the foundation model. In this paper, we propose WxFM-XL, a model for adapting univariate time series foundation models to multi-station weather forecasting. WxFM-XL introduces a cross-station error correlation prior graph to capture stationwise error priors with respect to the foundation model. Building on this, we further propose a dynamic fusion mechanism that adaptively integrates a spatial correlation graph with the error correlation prior graph. Experiments on multiple datasets demonstrate that our model outperforms state of the art baselines.