Evaluating Microphysics and Boundary Layer Schemes in WRF: Assessment of 36 Scheme Combinations for 17 Major Storms in Saudi Arabia

Sahu, Rajesh Kumar; Bangalath, Hamza Kunhu; Mostamandi, Suleiman; Evans, Jason; Kucera, Paul A.; Beck, Hylke E.

doi:https://doi.org/10.5194/egusphere-2025-912

Preprints

https://doi.org/10.5194/egusphere-2025-912

Preprints

03 Apr 2025

| 03 Apr 2025

Evaluating Microphysics and Boundary Layer Schemes in WRF: Assessment of 36 Scheme Combinations for 17 Major Storms in Saudi Arabia

Rajesh Kumar Sahu, Hamza Kunhu Bangalath, Suleiman Mostamandi, Jason Evans, Paul A. Kucera, and Hylke E. Beck

Abstract. Extreme rainfall events (EREs) and resulting flash floods in Saudi Arabia cause significant risks, including casualties and economic losses. Accurate simulations are crucial for forecasting, climate projections, and disaster management. This study evaluates boundary layer (BL) and cloud microphysics (MP) schemes to simulate EREs in the Arabian Peninsula (AP) using the Weather Research and Forecasting (WRF) model. Thirty-six combinations of four BL and nine MP schemes were tested across 17 EREs at a convective-permitting 3-km resolution, compared with IMERG gridded satellite data for rainfall and station observations for temperature, humidity, and wind speed. Performance was assessed using Kling-Gupta Efficiency (KGE) incorporates correlation, variability, and overall bias. We found good visual agreement between observed and simulated rainfall patterns despite some over- and underestimations. Among BL schemes, the Yonsei University (YSU) scheme stood out as the best performers in terms of rainfall, while Thompson (MP8) ranked the highest among the MP schemes. Goddard (MP7) also delivered strong results. The Thompson-YSU combination yielded the highest mean KGE, performing statistically significantly better than 21 other combinations. Furthermore, performance rankings varied across meteorological variables, suggesting that superior rainfall performance does not necessarily correlate with an overall more accurate simulation. This study highlights the challenges of scheme evaluation and the importance of analyzing many EREs while using reliable reference data. It offers guidance for selecting the most appropriate schemes and lays the foundation for future ERE forecasting and climate modeling improvements in arid regions.

Received: 26 Feb 2025 – Discussion started: 03 Apr 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 10946 KB)

Supplement (6223 KB)

Download & links

Preprint (10946 KB)
Metadata XML
Supplement (6223 KB)
BibTeX
EndNote

Rajesh Kumar Sahu, Hamza Kunhu Bangalath, Suleiman Mostamandi, Jason Evans, Paul A. Kucera, and Hylke E. Beck

Status: final response (author comments only)

RC1:
'Comment on egusphere-2025-912', Anonymous Referee #1, 24 Apr 2025

This study presents interesting insights that could be valuable for the WRF modeling community, particularly in an understudied region. While the data and modeling approach are generally sound, several areas need improvement before publication.
Firstly, the structure and writing style require significant enhancement to meet publication standards. The results should be more detailed, offering specific recommendations—such as which schemes or combinations of schemes to use and which ones to avoid. Additionally, there should be greater emphasis on the processes driving these extreme rainfall events and how these processes are represented in the simulations. Finally, it is essential to identify and highlight key limitations of the study.
Please see the attached comments and also consider the following major points:
- Please use standard terminology (e.g., “Planetary Boundary Layer – PBL” or “convection-permitting”).
- The way of phrasing the aims of the study (i.e., using research questions) is not appropriate for a scientific article. In addition, some of these questions are not so relevant and could be removed (e.g., #10.”How generalizable are our findings?”). The same applies to section or sub-section headings, and the Conclusions, where I would avoid using questions.
- More information should be provided on why the selected processes (PBL and cloud microphysics) were investigated. How do these parameterizations influence the simulation of precipitation? For example, elaborate on what a single and a double moment scheme is. Are there other processes relevant (e.g., convection in the coarser-resolution domain)?
- The selection and definition of extreme cases is problematic, since for many events there is unknown information on the observed precipitation amounts (Table 2). I recommend including an additional column which will show the IMERG nearest grid-cell precipitation.
- It is also unclear if any events lasted more than one day, and if yes, how were these events treated in the analysis? Is one day of spin-up time enough for these runs?
- For extracting the overall statistics, all events were weighted equally. However, in the interpretation of results, it would have been useful to differentiate, for example, between the most and the less extreme events, or between events affecting different parts of the Arabian Peninsula.
- More information on the interpretation of KGE should be provided in Section 3.5. Some references to other studies that use KGE in a spatial context could also be added. Moreover, I strongly recommend using additional evaluation metrics and not relying only on KGE for your conclusions.
- Extensive parts of Section 4 are not results (e.g., L148-161, L217-222, L275-280, L326-332). Please move this and other non-results material to the introduction, data or discussion sections, if relevant.
- The approach described in lines 251-257 should be presented in more detail in the Methods section.
- Figures 6 and 7 should be merged to facilitate the comparison between observations and simulated rainfall. Please be consistent in the date format (panel titles).
- Sections 4.9 and 4.10 are definitely not results material. Please move to other more relevant section(s).

Citation: https://doi.org/10.5194/egusphere-2025-912-RC1
- RC3:
  'Reply on RC3', Abhishek Lodh, 16 May 2025
  Referee Comments for the article titled “Evaluating Microphysics and Boundary Layer Schemes in WRF: Assessment of 36 Scheme Combinations for 17 Major Storms in Saudi Arabia”
  The present article provides a comprehensive literature review in the field of urban meteorology, with a particular focus on the challenges and developments over Saudi-Arabia and middle-east.
  The review work by Sahu Rajesh et al. (2025) is a well-prepared and valuable manuscript that provides a detailed review of evaluating microphysics and boundary layer schemes in WRF over Saudi Arabia. The authors have done an excellent job in addressing extreme rainfall events and its modelling over data sparse region like Saudi Arabia. I think that the manuscript presents a well-structured and comprehensive review with significant improvements in clarity, depth, and organization. Though the manuscript is well structured and presented, I have a very few minor comments related with the manuscript, which I feel that the authors should incorporate. The manuscript has the potential to get acceptance, only after addressing the comments below:
  Major/Minor Comments
  
  Consider making the title slightly more concise and catchier. For example:
  
  “Evaluation of Microphysics and Boundary Layer Schemes for Simulating Extreme Rainfall events over Saudi Arabia using WRF”.
  In Abstract kindly rephrase the line: “Kling-Gupta Efficiency (KGE) incorporates correlation, variability, and overall bias.”
  
  to
  “The Kling-Gupta Efficiency (KGE) metric, which incorporates correlation, variability, and bias, was used for performance evaluation.” Provide necessary (original WMO) citations for the metrics used.
  In section 1 where you structure the ten key questions, consider using letters (a, b, c...) for questions to avoid confusion with numbered sections.
  
  Ensure all acronyms (e.g., MP, BL, KGE, IMERG) are defined on their first use in both the abstract and main text.
  
  Add more region-specific references: While many global references are cited, consider including more recent or specific studies on EREs or WRF performance over Saudi Arabia or the Middle East (e.g., 2022–2024 publications if available).
  
  On page 2, after line number 25 rephrase the line “These events are often linked to the intrusion of intensified subtropical jet stream…” to:
  
  “These events are frequently associated with intrusions of an intensified subtropical jet stream…”
  Make the figure captions of Figure 2, 3 and 4 more self-explanatory by specifying metrics, datasets, and periods used.
  
  In the abstract section after line number 10, “The Thompson-YSU combination yielded the highest mean KGE…”
  
  Rephrase to “Among all 36 combinations, the Thompson-YSU pairing consistently produced the highest mean KGE across the 17 storm events.”
  Ensure consistent use of terms like “EREs,” “events,” “storms” throughout the paper. Stick with one preferred term unless differentiation is needed.
  
  After line number 285, “...models often struggle to replicate the spatial distribution of events precisely.”
  
  Suggestion is to rephrase: “This is expected, as localized convective systems common in the region present challenges for accurately resolving spatial rainfall patterns in mesoscale models.”
  After line number 290: “...the Goddard (MP7) and Thompson (MP8) MP schemes, particularly when paired with the YSU (BL1)…... emerged as superior.”
  
  Rephrased to “...the Goddard (MP7) and Thompson (MP8) schemes, when combined with YSU (BL1), consistently ranked highest across both temporal and spatial KGE assessments.”
  When discussing major findings (e.g., Thompson–YSU being best), consider referencing the figure or table that supports this claim.
  
  The paper can be redrafted to explain the section 4.7 in the beginning i.e. before section 4.1. This is so that readers gets a visual demonstration of the rainfall event in the domain of the study.
  
  In the conclusion section of the study bring out the motivation/conclusion of the study that this is a kind of a verification study for hydrometeorology.
  
  The authors can also verify the 850hPa wind and near surface temperature and provide plots in supplementary section.
  
  Citation: https://doi.org/10.5194/egusphere-2025-912-RC3
  - AC4: 'Reply on RC3', Rajesh Kumar Sahu, 15 Jul 2025
    
    The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-912/egusphere-2025-912-AC4-supplement.pdf
    
    Citation: https://doi.org/10.5194/egusphere-2025-912-AC4
- AC1: 'Reply on RC1', Rajesh Kumar Sahu, 15 Jul 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-912/egusphere-2025-912-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-912-AC1
- AC2: 'Reply on RC1', Rajesh Kumar Sahu, 15 Jul 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-912/egusphere-2025-912-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-912-AC2
RC2:
'Comment on egusphere-2025-912', Anonymous Referee #2, 24 Apr 2025

General Comments
The authors present a valuable study investigating the performance of the Weather Research and Forecasting model (WRF) over the Arabian Peninsula. They did an extensive evaluation of various cloud microphysics (MP) and planetary boundary layer (PBL) with respect to precipitation and standard meteorological variables like near-surface temperature, humidity, and wind speed by means of most recent satellite and in situ data.
In fact, there are not that many extensive studies on the performance of model physics of extreme rainfall events over the Arabian Peninsula (AP). As shown by the authors, different combinations of physics schemes can lead to considerably different results which makes it difficult to use numerical weather prediction (NWP) model simulations for early warning systems.
I clearly see the benefit of such an extensive model evaluation over arid regions, however a major point of concern for me is the experimental setup. While I clearly understand the purpose of applying ERA5 as forcing data, it is critical to apply only 53 model levels on a convection permitting resolution. Several studies, also with other NWP models like COSMO (e.g., Schär, C., and Coauthors, 2020: Kilometer-Scale Climate Models: Prospects and Challenges. Bull. Amer. Meteor. Soc., 101, E567–E587, https://doi.org/10.1175/BAMS-D-18-0167.1.) show that a higher vertical resolution is beneficial on a convection-permitting horizontal resolution.
Also the work of Schwitalla et al. (2020), which was mentioned in Table 1, applied 100 levels at a horizontal resolution of 2700 m. In my opinion it is necessary to have this layer density as, you mentioned this is your manuscript, the AP is usually very dry but you have an extremely strong vertical temperature and moisture gradient, especially near the coast which in case the winds are coming from the sea. This gradient is unlikely to be captured with only 53 levels.
Please make sure that you clearly distinguish between “schemes” and combinations as this is very confusing for the reader.
Please make sure that you do not repeat the explanations of the MP and BL option combinations. Explaining them once and then use “MP8_BL1” throughout the manuscript should be sufficient. Figure captions in the continuous test should start with “Fig.” instead of “Figure”. “Figure” is only used at the beginning of a sentence.
What is your final recommendation regarding the combination of PBL and MP? This should be mentioned as this is an important outcome of your study.
Specific comments:
Manuscript title: Please include the WRF version you are using as different WRF versions can lead to different results.
Line 5: I think “convection-permitting” is more widely used than “convective-permitting”.
Line 11: Where are the “21” combinations are coming from? The abstract suggest you performed 36 combinations.
Line 20: Are there more recent publications which cover the aspect of climate change?
Line 33: “can feed early warning systems”
Line 36: “inform” seems not an appropriate word here.
Line 37: The acronyms “AP” and “WRF” are not explained. Please ensure that all acronyms are explained before they are used in the manuscript. Also add the reference for the WRF model here.
Line 37: “Numerical Weather Prediction (NWP) model”…
Line 41: The microphysics also have an impact on radiation.
Line 53: “to evaluate the best combination”….
Line 56: “using WRF in a two-way nested”… What was your motivation to apply a two-way nesting approach?
Line 66: I do not think that number 9 is a key question of your study.
Line 69: “spans from 16°N to 33°N and 34°E to 56°E”
Starting line 73: The readers may be not familiar with all the different regions of the Arabian Peninsula and Saudi Arabia. I think it would be good to add at least some of the regions and major cities you mentioned to Fig. 1.
Table 1: The study of Schwitalla et al. (2020) is the only experiment in your table with a convection-permitting model resolution. This should be mentioned in the table itself (maybe as a separate column) and/or in the text. Please consider to add the number of model layers of the different studies as this can help the reader to further interpret your findings.
Line 94: Did you use ERA5 pressure or model level data for the initialization of the model? This has influence on the accuracy of the initial conditions as the number of model levels in ERA5 is about 100 up to 30 km while the number of pressure levels is 32 up to approx. 30 km altitude. If you use pressure level data, please explain your decision.
Line 101: Radiosonde stations are plotted in Fig. 1 but were never used or described explicitly throughout the manuscript. Or are they included in the analysis shown in sec. 4.6? If yes, I think it is dangerous to combine them together as you would combine prognostic (3D) variables with diagnostic (2D) variables. Did you consider differences between the station altitude and the model orography in case stations are located in the mountains? If this difference is large, this can alter your results for T, RH, and WS.
Line 108: WRF version 4.4.0?
Line 110: Please add the corresponding number of grid cells and the model top pressure here and to Table 3.
Line 120: The Kessler scheme is an extremely old scheme. Please consider whether it is necessary to apply this in your study on a convection-permitting resolution. It is preferably used in idealized cloud modeling studies.
Line 130: the simulations were run for 84 hours with a 48 hour spin-up followed by a 24h forecast. Why did you run the model for 84 hours? Did you reinitialize the atmosphere after your 48h spin-up?
Figure 1: Please use a different color table as there is too much blue color in the figure. Depending on the printer, it can be difficult to see the red markers on a blue background. Also, add the METAR stations here as this is important for the interpretation of your results.
Table 2: It would be great if you could add the rainfall amount for all cases. Otherwise it is difficult for the reader to judge the level of extreme of the particular event. It really matters if you have 200 mm within a day of more than 40 mm in 30 min.
Table 3: Does the Arakawa-C grid play a role in your study? As far as I know it cannot be changed anyway. Also, the reference for the NOAH LSM is missing here. Regarding the CU scheme: I think you mean that for D02 no CU scheme is used.
Regarding the model setup: Did you use the default data sets for soil texture and land cover? This should be mentioned either in Table 3 or in the text.
Line 140: Please add a short sentence about what a perfect KGE would be. I guess 1 is ideal. What does a KGE of 0.5 tell us (most of the values are below this threshold)? Did you interpolate the simulated precipitation from the WRF model to the IMERG grid? If it was done the other way round, I doubt that this is meaningful. Which interpolation method did you use (bilinear, conservative, nearest neighbor, etc.)? Please also mention the tool or software package used for interpolation.
Line 142: Isn’t γ the ratio of the variances? “Coefficient of variation” sound a bit inappropriate.
Line 145: What was your motivation to use the nearest model grid cell next to the surface observations instead of, e.g., a distance-weighted 3x3 average? I think you cannot expect that the model can simulate your observation 1:1.
Lines 148-161: Please consider to integrate both paragraphs (or parts of them) to the introduction. I think it better fits there rather than in the discussion section.
Line 169-170: you mention that both MYNN schemes and the BouLac scheme rely on the representation of gradients. As you only have coarse 53 model levels, could the explain the detrimental performance of the MYNN and BouLac simulations?
Line 184: What is a “Stratified” BL?
Line 185: This is in contrast to the study of Schwitalla et al. (2020) which you mentioned in Table 1.
Line 201-204: I think this extensive explanation is not necessary here. You may introduce the abbreviations already in Table 3. This allows for saving space.
Line 205: “MP” is used for a microphysics scheme. Please write “cloud microphysics” instead.
Line 217: See my comment to line 201-204.
Line 218: Which advanced microphysical processes?
Line 221: Which sensitivities?
Line 230: I guess you mean “combination” instead of “scheme”.
Lines 231-232: Do the numbers show the absolute value of the mean of “r-1”? This is a bit confusing to the reader. Regarding the KGE: Did you account for a potential “double-penalty” of your model? In case the extreme precipitation is shifted by one grid cell in the model, the KGE may deteriorate.
Figure 3: What do you mean with “long-term bias” in the caption of Figure 3? In the second line of the figure caption, it should be “combination” instead of “scheme”.
Line 247: “(See later Fig. 5)”.
Line 260: You refer to Fig. S4 before Fig. S3 is used. I think the order of the two supplementary figures need to be changed.
Line 279: Which data sets are you referring to? It is unclear.
Line 280: Did you consider the random error provided in the IMERG data set?
Lines 285-292: It may be worth considering to integrate parts of this paragraph to the conclusion section.
Line 295: Can you really conclude this from your study? Precipitation is the end product of a long chain of processes which can have compensating errors. As pointed out earlier, if you mix both 2D and 3D variables here, this can be misleading.
Line 314-315: “WRF model has indicated….”. This sentence sounds a bit awkward.
Line 321: “The WRF model….”
Line 327: Which simplifications? Please elaborate.
Line 337: “the best performing combination in terms of rainfall”.
Figure 6: Showing negative values for accumulated precipitation is not reasonable. Please start with 0 mm or 1 mm.
Lines 340-344: These sentences are confusing. It is not clear to me what you want so say here, especially in relation to the best performing combination.
Figure 7: “… from WRF using the best performing…”
Line 345: The part of the sentence in parentheses can be deleted as this is already explained in line 336.
Line 350: “… for the 17 EREs is 0.20 (not shown)…”. I also would not call the values of “r-1” scores. In my opinion, it is simply a value.
Section 4.9: It this really relevant for your study? My personal feeling is that this is a bit out of scope.
Line 374: What is a “high-performing scheme”?
Section 4.10: Maybe this can be integrated to the conclusion section?
Line 389: Please give an example for arid or semi-arid regions with similar characteristics.
Line 396: Other boundary layer schemes…
Line 402-407: Did you consider investigating cloud properties like integrated cloud water content (e.g., from CMSAF CLAAS: https://doi.org/10.5676/EUM_SAF_CM/CLAAS/V003)? This can give you a hint of what is happening inside the MP schemes with respect to the cloud formation and thus precipitation.
Line 423: “This underlines the complexity…”. Also radiation has an impact on cloud evolution.
Line 425: Only for a particular physics combination.
Line 432: As already mentioned, please reconsider if this part/question is necessary and gives a benefit. This is not a conclusion.
Line 440: “Similar climatic conditions…” like?

Citation: https://doi.org/10.5194/egusphere-2025-912-RC2
- AC3: 'Reply on RC2', Rajesh Kumar Sahu, 15 Jul 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-912/egusphere-2025-912-AC3-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-912-AC3
RC4:
'Comment on egusphere-2025-912', Abhishek Lodh, 16 May 2025

Publisher’s note: this comment is a copy of RC3 and its content was therefore removed on 19 May 2025.

Citation: https://doi.org/10.5194/egusphere-2025-912-RC4
- AC5: 'Reply on RC4', Rajesh Kumar Sahu, 17 Jul 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-912/egusphere-2025-912-AC5-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-912-AC5
RC5:
'Comment on egusphere-2025-912', Abhishek Lodh, 16 May 2025

Publisher’s note: this comment is a copy of RC3 and its content was therefore removed on 19 May 2025.

Citation: https://doi.org/10.5194/egusphere-2025-912-RC5
- AC6: 'Reply on RC5', Rajesh Kumar Sahu, 17 Jul 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-912/egusphere-2025-912-AC6-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-912-AC6

Rajesh Kumar Sahu, Hamza Kunhu Bangalath, Suleiman Mostamandi, Jason Evans, Paul A. Kucera, and Hylke E. Beck

Supplement

https://doi.org/10.5194/egusphere-2025-912-supplement

Rajesh Kumar Sahu, Hamza Kunhu Bangalath, Suleiman Mostamandi, Jason Evans, Paul A. Kucera, and Hylke E. Beck

Viewed

Total article views: 979 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
760	181	38	979	55	21	43

HTML: 760
PDF: 181
XML: 38
Total: 979
Supplement: 55
BibTeX: 21
EndNote: 43

Views and downloads (calculated since 03 Apr 2025)

Month	HTML	PDF	XML	Total
Apr 2025	137	33	8	178
May 2025	120	30	7	157
Jun 2025	28	8	6	42
Jul 2025	96	57	11	164
Aug 2025	116	49	5	170
Sep 2025	263	4	1	268

Cumulative views and downloads (calculated since 03 Apr 2025)

Month	HTML	PDF	XML	Total
Apr 2025	137	33	8	178
May 2025	120	30	7	157
Jun 2025	28	8	6	42
Jul 2025	96	57	11	164
Aug 2025	116	49	5	170
Sep 2025	263	4	1	268

Viewed (geographical distribution)

Total article views: 1,013 (including HTML, PDF, and XML) Thereof 1,013 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 24 Sep 2025

Short summary

This study evaluates 36 microphysics (MP) and boundary layer (BL) scheme combinations in the Weather Research and Forecasting (WRF) model for extreme rainfall over Saudi Arabia. Using Kling-Gupta Efficiency (KGE), results show YSU (BL1) and Thompson (MP8) perform best, while Morrison-MYNN (MP10_BL6) ranks lowest. The mean temporal KGE is 0.37, and the spatial KGE is 0.26, highlighting spatial prediction challenges. Findings aid model evaluation and forecasting in arid regions.


Total:	0
HTML:	0
PDF:	0
XML:	0