Assimilating Observed Surface Pressure into ML Weather Prediction Models

TL;DR

This paper investigates the integration of observed surface pressure data into machine learning weather prediction models using a cycling data assimilation system, highlighting challenges in error growth and covariance estimation.

Contribution

It introduces a cycling data assimilation framework for ML weather models and analyzes their limitations in error growth and covariance estimation.

Findings

ML models diverge due to small-scale noise

Spectral filtering stabilizes but increases errors

ML models poorly estimate short-term error growth

Abstract

There has been a recent surge in development of accurate machine learning (ML) weather prediction models, but evaluation of these models has mainly been focused on medium-range forecasts, not their performance in cycling data assimilation (DA) systems. Cycling DA provides a statistically optimal estimate of model initial conditions, given observations and previous model forecasts. Here, real surface pressure observations are assimilated into several popular ML models using an ensemble Kalman filter, where accurate ensemble covariance estimation is essential to constrain unobserved state variables from sparse observations. In this cycling DA system, deterministic ML models accumulate small-scale noise until they diverge. Mitigating this noise with a spectral filter can stabilize the system, but with larger errors than traditional models. Perturbation experiments illustrate that these models do not accurately represent short-term error growth, leading to poor estimation of cross-variable covariances.