the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Jul 2025
The operational 3DEnVar data assimilation scheme for the Météo-France convective scale model AROME-France
Abstract. Since October 2024 the Météo-France operational convective scale model AROME-France uses a 3DEnVar data assimilation (DA) scheme in order to improve the performances of severe weather prediction. This paper describes the configuration and the evaluation of this 3DEnVar scheme. It summarises the work carried out to configure this scheme in an operational context, with the inherent constraints of numerical robustness, compatible execution times and, of course, correct performances of the forecasts produced. The adjustment of horizontal and vertical localization, inflation, hybridization and the use of Incremental Analysis Update (IAU) are studied in sensitivity experiments, and the impact on the spin-up is investigated. A configuration, and its variation with IAU, are thus defined and evaluated using different scores over a long period and on different severe meteorological situations (winter storm, fog and High Precipitating Events): they largely outperform the operational 3D-Var. The version with IAU has been implemented in the new operational version of the AROME-France assimilation system.
- Preprint
(2016 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-2642', Christoph Schraff, 13 Aug 2025
-
RC2: 'Comment on egusphere-2025-2642', Dominik Jacques, 29 Dec 2025
Review for:
The operational 3DEnVar data assimilation scheme for the Météo-France convective scale model AROME-France
https://doi.org/10.5194/egusphere-2025-2642This manuscript documents the evolution of data assimilation in the hourly cycled AROME prediction system operated and developed at Météo-France. The work presented is serious, and the results appear robust. Documenting the evolution of an operational system is valuable, as it serves as a reference both internally and for teams operating similar prediction systems at other operational centers.
While the scientific quality of the work is good, the manuscript would benefit from a substantial round of revisions to improve clarity and readability. Relatively small improvements, such as the consistent use of acronyms in the text and figures, would significantly enhance readability. For example, 3DVar and EnVar could be used consistently in experiment labels instead of 3dv and 3dev, which are visually similar. Throughout the manuscript, simple reformulations would also help improve the overall flow. It is recommended that proofreading tools be used to identify and correct sentence structures and expressions that are not idiomatic in English.
In particular, Sections 2 and 3.2 should be reworked for clarity. A major difficulty is that many systems with different configurations are described simultaneously. Although the components of the “new” operational system are all presented, relevant information is scattered among descriptions of systems used in previous validation studies. It is suggested that the emphasis be placed more clearly on the new system, while previous studies could be briefly summarized in dedicated subsections.
Figure 1 is also particularly difficult to interpret, even with the accompanying text. Several suggestions for improving this figure are provided in the detailed comments below.
For these reasons, it is recommended that the manuscript undergo a round of major revisions.
Detailed comments in the order they appear in the text
Line 25
“Regarding the first limitation, numerous studies have tried to introduce some time-dependency…”
Would flow-dependent be more appropriate than time-dependency here? The first limitation seems more related to the nature of the error covariances than to temporal aspects. This sentence is also confusing because the second point described just above explicitly discusses the temporal dimension of the problem.
Line 94
“using an original approach”
Please be more specific about what is original in this approach.Section 2
Naming
Line 102: “running a 50-member EDA”
Line 107: “An EDA version is run”
Presumably, these EDAs refer to ARPEGE-EPS and AROME-EPS. These acronyms are self-explanatory and should be preferred over the more obscure “PEARP” and “PEARO”, which are also used in the text. They could also appear in Figure 1.The discussion in lines 125–145 is difficult to follow. It would help to structure this section around the new system being developed, clearly describing the start times of the different runs, and the origin of the initial conditions (ICs) and lateral boundary conditions (LBCs).
Figure 1
This figure is difficult to understand and should be reworked. While wall-clock time is an important constraint in operational forecasting, it may not be the clearest choice for illustrating dependencies between ensemble and deterministic systems. Representing the various components in terms of forecast lead time could make the exchange of ICs, LBCs, and other information clearer.
Additional clarification is also needed in the discussion around line 125 and in Figure 1 itself:
Is “production” synonymous with forecasts?
What do the blue and purple rectangles represent?
Do stacked rectangles indicate ensemble members?
Line 145 refers to green arrows as “perturbed forecasts”, but it is unclear which perturbations are meant. The text discusses perturbed LBCs, stochastic physics (SPP), and observation perturbations via random draws from the observation error covariance matrix for B estimation. Clarifying the different types of perturbations would be helpful.
Consistent use of acronyms between the figure and the text would greatly improve readability. For example, the text refers to AROME-EDA, which does not explicitly appear in the figure, while AEARP, ARO-Fr, and AEARO are not explicitly described in the text.
Consider using the term background instead of guess.
The text refers to H–5, H–4, …, H, H+1, while the figure shows specific hours. Using a concrete example (e.g. 1200 UTC) could make the correspondence clearer and illustrate that the same pattern repeats at other times.Figure 1 legend
“Logical scheme of an extract of …”
Please reformulate this sentence.Lines 170–180
Different pre-conditionings for different systems are discussed here. The link with the new system should be made more explicit. When referring to “the 3D-Var scheme”, it is unclear whether one specific system or multiple assimilation systems are being discussed.
Line 175
Minor comment: the symbol ∘ should be used for the Schur product instead of the lowercase letter “o”.
Line 181
“lead to define the settings”
Please reformulate.Line 210
While precipitation can indeed be viewed as a temporal and vertical integration of model prognostic fields, using 24-hour accumulated precipitation to verify a convection-permitting, hourly cycled assimilation system seems questionable.
Line 220 (and elsewhere)
The expression “strongly smaller” is awkward. Consider alternatives such as much smaller or significantly smaller.
Line 225
“the differences reach only 5% in absolute value”
Is Figure 2 expressed in percent? The legend refers to “relative difference”, which could be interpreted in several ways.Section 3.2
This section would be easier to follow if the experiments were described before their results are discussed.
It is not immediately clear whether the color red or blue indicate better performance.
Using consistent labels between the figures and the text would help.
How long were the experiments run?Figure 4
Please increase the text size for better legibility. A logarithmic scale might be appropriate here.Surface pressure tendencies are strongly influenced by the presence of convection in the domain. Plotting the envelope of minimum and maximum values in the dataset could help put the spin-up values into context.
Line 260
“Spin-up can also be referred as overestimated precipitation …”
Please provide more context, as the connection between spin-up and overestimated precipitation is not self-evident.Line 280
“The results of these experiments are illustrated using the same diagnostics as in section 3.2…”
Please clarify whether the results are shown in Figures 2 and 3, or in additional figures not shown here.Line 320
“Perturbations”
Please provide more detail on the nature of these perturbations.Figure 9 legend
“as a fonction of the forecast range” → function
Figure 10
Ensure that all text is legible; the color bar text is very small. Please also add units (mm) for the maximum values.
Figure 11
Why focus on devi here? While dev and devi are similar, most earlier figures use dev.
Section 5.3.1
It is acknowledged that sample sizes may be limited for intense precipitation events. However, the performance of a single case is largely anecdotal and does not, by itself, demonstrate the superiority of one system over another. In Figure 12, neither model clearly stands out as superior.Citation: https://doi.org/10.5194/egusphere-2025-2642-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,034 | 192 | 33 | 1,259 | 36 | 34 |
- HTML: 1,034
- PDF: 192
- XML: 33
- Total: 1,259
- BibTeX: 36
- EndNote: 34
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
see supplement pdf