Towards Strongly-coupled Ensemble Data Assimilation with Additional Improvements from Machine Learning
Abstract. We assessed different coupled data assimilation strategies with a hierarchy of coupled models, ranging from the simple coupled Lorenz model to the state-of-the-art coupled general circulation model CFSv2. With the coupled Lorenz model, we assessed the analysis accuracy by strongly-coupled Ensemble Kalman Filter (EnKF) and 4D-Variational (4D-Var) methods with varying assimilation window lengths. The analysis accuracy of the strongly-coupled EnKF with a short assimilation window is comparable to that of 4D-Var with a long assimilation window. For 4D-Var, the strongly-coupled approach with the coupled model produces more accurate ocean analysis than the ECCO-like approach using the uncoupled ocean model. Experiments with the coupled quasi-geostrophic model conclude that the strongly-coupled approach outperforms the weakly-coupled and uncoupled approaches for both the full-rank EnKF and 4D-Var, with the strongly-coupled EnKF and 4D-Var showing a similar level of accuracy higher than other coupled data assimilation approaches such as the outer loop coupling. A strongly-coupled EnKF software framework is developed and applied to the intermediate-complexity coupled model SPEEDY-NEMO and the state-of-the-art operational coupled model CFSv2. Experiments assimilating synthetic or real atmospheric observations into the ocean through strongly-coupled EnKF show that the strongly-coupled approach improves the analysis of the atmosphere and upper oceans, but degrades observation fits in the deep ocean, probably due to the unreliable error correlation estimated by a small ensemble. The correlation-cutoff method is developed to reduce the unreliable error correlations between physically irrelevant model states and observations. Experiments with the coupled Lorenz model demonstrate that strongly-coupled EnKF informed by the correlation-cutoff method produces more accurate coupled analyses than the weakly-coupled and plain strongly-coupled EnKF regardless of the ensemble size. To extend the correlation-cutoff method to operational coupled models, a neural network approach is proposed to systematically acquire the observation localization functions for all pairs between the model state and observation types. The following strongly-coupled EnKF experiments with an intermediate-complexity coupled model show promising results with this method.
Eugenia Kalnay et al.
Eugenia Kalnay et al.
Eugenia Kalnay et al.
Viewed (geographical distribution)
Review for “ Towards Strongly-coupled Ensemble Data Assimilation with Additional Improvements from Machine Learning” by Kalnay et al.
This manuscript reviews the coupled data assimilation and strongly coupled data assimilation research conducted by Dr. Kalnay’s group. The manuscript is well-written, with appropriate literature references. The content is very beneficial for the data assimilation community as well as the general readers who are interested in coupled data assimilation. I think it is well fit for NPG. In terms of the manuscript, I only have a few minor comments.
1 I can understand the abbr. SC and WC represent strongly-coupled and weakly-coupled, respectively. Please define them in the manuscript.
The same comment also applies to abbr. “NMC”
Ln 11-12 the simple coupled Lorenz model —> a simple coupled Lorenz model
2 Ln 19 I am confused about the “full-rank” EnKF
3 Ln 24 55 upper oceans —> upper ocean
4 Ln 125-126, The smallest RMSE shows at an assimilation interval of 8 time-steps that is your smallest assimilation interval. I think it is worth pointing out here.
5 Fig. 2 From my understanding, ECCO only updates the boundary forcing and parameters, not ocean state variables. From the figure, the initial conditions of model states are updated by DA. Please clarify it.
6. Comparisons of 3/4D-Var and EnKF in a coupled QG Model
There are UC_clim, UC_3days and UC_1day. it is worth providing details on their adjustment. Which is equivalent to the regular UC applying in the atmosphere and ocean?
The improved ocean does not enhance the RMSE in the atmosphere throng dynamic coupling, which needs some discussion.
Ln 225-226 Fig. 4(a-b) only demonstrates two CDAs (WC, SC), not three methods, From fig. 4b, I can not conclude SC is better than WC.
7 Ln 255 “lower than SC 3D-Var” should be “higher than SC 3D-Var”
8 Ln 257 “For 4D-Var, applying more outer loops and longer assimilation window lengths further reduces the analysis error”. The state mentioned here has no support.
9 “accuracies smaller than SC 3D-Var” should be “higher than”
10 Fig.6 shows that the different RMSE between SC and WC has not reached to equivalent, especially the surface T/S, which needs to point out.
11 Ln 340 it is confusing for the statement “due to the missing vertical localization not used in the Ocean-LETKF.” Please rephrase.
12 LN 356 “states are assimilated by the SC EnKF”. The mean of SC EnKF here indicates the cross-model update, which is different from the other places indicating the whole DA algorithm.