Coupled observation-operator approximations in outer-loop coupling data assimilation

Philip Browne

doi:10.1002/qj.70266

Return

Coupled observation-operator approximations in outer-loop coupling data assimilation

2026-10-03

DOI：10.1002/qj.70266

Save

View PDF

Abstract

En 中文

Data assimilation provides a way of incorporating information from observations into a model. When an observation of a system is sensitive to multiple components of the model (e.g., atmosphere, ocean, sea ice, land surface, etc.), the data assimilation process should take that into account. One of the key building blocks of data assimilation that can propagate information between model components is the observation operator. However, how to extract information from observations into different model components is challenging. When data assimilation is applied to coupled models, often simplifications are made and the problem is not always tackled in a monolithic way considering all pieces at once. Here we present a methodology for coupled variational data assimilation that allows the effect of coupled observation operators to be approximated whilst maintaining separate minimisations across model components. This allows us to constrain the atmosphere, ocean, and sea-ice components of our coupled model directly from satellite radiances. We show different ways this could be implemented and discuss how it fits within the incremental four-dimensional variational assimilation (4D-Var) framework in use at European Centre for Medium-Range Weather Forecasts (ECMWF). We illustrate the system in action by showing example increments from coupled ocean–atmosphere data assimilation experiments, from both microwave and infrared satellites. The new method provides a way to exploit existing and planned observation systems further, which in turn could help improve numerical weather predictions.

Keywords:

coupled data assimilation

data assimilation

numerical weather prediction

observation operators

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.