| 20 May 2026
Impact of ROMEX Radio Occultation Bending Angle Assimilation on Mesoscale Weather Prediction over the Indian Region
Abstract. Data assimilation experiments were conducted for September 2022 at 9-km horizontal resolution using a cyclic three-dimensional variational (3D-Var) assimilation system to assess the impact of a large number of Global Navigation Satellite System (GNSS) radio occultation (RO) bending angle data on mesoscale weather prediction over the Indian region. To enable this, a bending angle observation operator was implemented within the WRF (Weather Research and Forecasting) data assimilation system.
Two experiments were performed: a control experiment (CNTL), in which only conventional observations were assimilated, and a second experiment (RMX-BA), in which RO bending angle observations from the Radio Occultation Modelling Experiment (ROMEX) were assimilated along with conventional data. A 6-h assimilation cycle was performed throughout September 2022. From the 00 and 12 UTC analyses, 72-h forecasts were generated daily, resulting in approximately 60 forecast cases.
Forecasts from both experiments were verified against ERA5 reanalysis for water vapor, temperature, and wind, while rainfall forecasts were compared against Integrated Multi-satellite Retrievals for GPM (IMERG) rainfall estimates. The results show that assimilating RO data improves both the analyses and forecasts of specific humidity, temperature, wind, and rainfall compared to the CNTL experiment. The rainfall forecast skill improved significantly, mainly due to more accurate water vapor in the model's initial conditions. The moist total energy norm (TE), which accounts for forecast errors in water vapor, wind, temperature, and pressure, was reduced by about 20% at the analysis time and by approximately 8.5% in the 72-h forecast.
Overall, the study demonstrated that assimilation of RO bending angle data significantly improved mesoscale weather forecasts over the Indian region.
General comments
This paper assesses the impact of a large number of GNSS-RO real observations made available in the context of ROMEX study on the mesoscale forecast over the Indian Region.
In addition to some classical thermodynamic variables, the article focuses on rainfall forecasting in the Indian Region with verification against ERA-5 reanalyses and IMERG satellite observations. This latter point together with the mesoscale domain contribute to the interest and originality of the study.
Clearly, the study required significant technical work to implement a 2D bending angle observation operator within the WRF data assimilation system and run WRF coupled to ERA5.
The well-written paper presents accessible and understandable results and illustrated with very clear figures. It is mostly very clear and almost ready for publication.
Nevertheless, the manuscript could still be improved by addressing the points discussed below in order of importance from highest to lowest:
Major comments:
- The study covers a period of one month, which I think is the minimum length of time required to draw meaningful and robust conclusions. I guess the main reason is that the study focuses on the summer monsoon which ends in September (the monsoon is mentioned line 82) but this is not explicit written. Is the length of the period really determined by the monsoon forecast, or is it due to the heaviness of the experimentation? If the main objective is to evaluate the impact of ROMEX GNSS-RO data on summer monsoon forecasting, this must be clearly stated (for instance in the first sentence of the abstract).
- Figure 11/Figure 15 (or in a more general way) : are the results statistically significant? This information is important for the robustness of the results.
I suggest you add an indication of a confidence interval (for instance error bars on the curves and profiles).
Minor comments:
- line 8 « Conventional observations » : as there is no official definition for conventional observations, this needs to be explicitly defined here, even if described further in part 2.3.
Here, you mean in situ together with satellite wind-observations.
- line 22 : It seems that the term ‘conventional’ is used to mean ‘in situ’ observations here.
For more clarity, I suggest you use the expression ‘in situ’ instead here.
- line 38 : COSMIC-2 constellation provides obervation between ±40 degrees latitude. This could be mentioned.
- line 95 : the NMC sigle should be spelled out.
- lines 112 to 115 : it is first written that the RO observation errors used in the study are those applied in GFS in operation. But line 115, it seems that GFS has « larger observation error assumptions » in operation. The two sentences seem to contradict each other. I probably misunderstood or this needs to be clarified.
- Use of ERA5 for the boundary conditions and also for verification. Perhaps it could be specified that most of the RMX-BA RO observations were not assimilated in ERA5 reanalysis.
- IMERG estimates are chosen as references for rainfall validation.
In addition to Huffman et al., 2020 the reference to an article on IMERG performances on estimating heavy rainfall might be of interest (for instance https://doi.org/10.1016/j.advwatres.2015.11.008 )
- line 195 : « TE is a vertically integrated index » as TE is (homogeneous to) a norm, perhaps it would be better to change index (which has generally no unit) into norm or metrics.
- line 202 : « as expected TE exhibits a systematic increase with forecast lead-time »
Obviously, this is the general trend but we can see decrease of TE between 0 and 6-h of lead-time for both experiments, also between 12 and 18h, 24 and 30h, 36 and 42h. Just out of curiosity, what’s your interpretation of this?
- line 205 : introduction and definition of the « Impact Parameter » IP. In the field of GNSS-RO, « impact parameter » refers to « nr sin Φ » or a on Figure 3 of the paper (the distance of closest approach for the straight line path).
So, I suggest using another term for the quantities defined here, for instance Impact Index or something else.
- 3.2.2 : « only minor degradation observed near the surface (…) mainly in specific humidity » : near the surface, the degradation in specific humidity is 10-20 % during the first 6 hours (6-8 % for temperature), getting smaller after. But that’s where the forecast error in CNTL is maximum. So, adding ROMEX RO observations does not fix this problem in the initial hours; on the contrary, it makes it worse.
- line 285 : the new IP could be introduced by a sentence like « the impact of RO data assimilation on the rainfall forecast is quantified by... »
- line 301 (and following), figure 15 : ETS (Equitable Threat Score) should be defined or a reference to a paper should be given.
- Figure 15 : the caption should indicate which plot is for 24h accumulated rainfall, which one is for 48h.
- Conclusion line 326 : the biases for specific humidity are often corrected, but not all of them. So, I would nuance the text « The assimilation corrects both moist and dry biases »
Typos:
- line 159 : the name of the variable that represents the refractive-index radius product is missing here or there is an extra comma.
- I suggest you choose one of both writings GNSS-RO (more usual) or GNSS RO, as both are used in the paper.
- caption for Figure 5 : the mathematical symbol for multiplication appears in letters in the size of the grid (5◦ imes5◦ for grid 5◦ x 5◦ grid)
- Figure 6 : the unit for O-B and O-A statistics is missing.
- line 223 : the greek mathematic symbol σ appears as « Std.Dev »
- line 244 : missing « . » at the end of the sentence