Learning to Assimilate in Chaotic Dynamical Systems
This work addresses a key bottleneck in simulation-based forecasting for chaotic systems, offering a novel approach that could enhance prediction accuracy in fields like weather modeling.
The paper tackles the problem of estimating initial conditions for forecasting in chaotic dynamical systems by introducing amortized assimilation, a framework that learns from noisy observation sequences without ground truth data, achieving improved effectiveness over existing methods.
The accuracy of simulation-based forecasting in chaotic systems is heavily dependent on high-quality estimates of the system state at the time the forecast is initialized. Data assimilation methods are used to infer these initial conditions by systematically combining noisy, incomplete observations and numerical models of system dynamics to produce effective estimation schemes. We introduce amortized assimilation, a framework for learning to assimilate in dynamical systems from sequences of noisy observations with no need for ground truth data. We motivate the framework by extending powerful results from self-supervised denoising to the dynamical systems setting through the use of differentiable simulation. Experimental results across several benchmark systems highlight the improved effectiveness of our approach over widely-used data assimilation methods.