Strongly Coupled Data Assimilation of Ocean Observations Into an Ocean‐Atmosphere Model

Abstract

A global climate model represents the physical processes of the ocean, sea ice, atmosphere, and land, as well as their interactions by coupling different components in one system. Within a climate model, the ocean and the atmosphere are the two main compartments, which are highly linked to each other. They give/receive their feedback to/from each other in one system thus influencing each other consistently. In contrast to stand-alone models of the ocean or atmosphere, which rely on forcings, a coupled model essentially evolves freely with little influence of forcings. Therefore, the simulation from a coupled model system may be far away from the real state (e.g., Mulholland et al., 2015). Data assimilation (DA) allows constraining the model state by incorporating observations to correct the model evolution in state space (e.g., Zhang & Moore, 2015). With decent initialization, a coupled system is equipped with the potential for real-world prediction.There are two main approaches to DA in a coupled system: weakly coupled DA (WCDA) and strongly coupled DA (SCDA) (Penny & Hamill, 2017). WCDA analyses the state of one or more compartments within a coupled system separately by assimilating their own observations and other components in the system are then influenced indirectly via the coupled model dynamics. WCDA is the most commonly used DA approach for a coupled system (Zhang et al., 2020). One example is our previous study (Tang et al., 2020), who assimilated satellite sea surface temperature (SST) and temperature and salinity profiles into the ocean component for a coupled ocean-atmosphere model. The study found improvements of not only the ocean variables, which were directly updated by the DA, but also the atmosphere variables like the air temperature and wind speed. Most of the previous WCDA experiments assimilated observations of both the ocean and the atmosphere (e.g.,

The ecohydrological response to soil moisture based on the distributed hydrological assimilation model in the mountain region

“…Here is a description of ESTKF (Tang et al, 2021). The state vector in the distributed hydrological assimilation model contains the grid values of soil moisture in the study area.…”

Strongly Versus Weakly Coupled Data Assimilation in Coupled Systems With Various Inter‐Compartment Interactions

“…They tested some configurations of DA and decided not to assimilate oceanic observations into atmosphere to avoid degrading the performance. Tang et al (2021) studied the influence of ocean observations in an atmosphere-ocean CDA system by assimilating sea surface temperature into an atmospheric model. To summarize, sCDA attracts researchers in various disciplines, and its application to various types of realistic coupled models has recently started.…”

Vertical Localization for Strongly Coupled Data Assimilation: Experiments in a Global Coupled Atmosphere‐Ocean Model

“…Goodliff and Penny (2022) used a quasigeostrophic model and variational DA scheme and found that WCDA and SCDA perform similarly in the atmosphere, while WCDA leads to the lowest error in the ocean. Tang et al (2021) assimilated sea surface temperature observations into a global atmosphere-ocean model and found that WCDA and SCDA yield similar errors in the ocean. In the atmosphere, their results were state-variable and location dependent.…”

“…Amidst the varied findings one thing is clear: the choice of coupled DA method depends on the model, model bias, observations, observation density, and underlying DA framework (e.g., ensemble or variational). Even within a single system the optimal method can vary spatially and by state variable (Tang et al., 2021; Yoshida & Kalnay, 2018). Given the value of coupled model initialization in subseasonal to seasonal forecasting and consistent reanalyses, it is important to investigate the potential benefits of SCDA and determine the conditions under which SCDA may outperform WCDA.…”

“…Tang et al. (2021) assimilated sea surface temperature observations into a global atmosphere‐ocean model and found that WCDA and SCDA yield similar errors in the ocean. In the atmosphere, their results were state‐variable and location dependent.…”