the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 May 2025
Optimizing physical scheme selection in RegCM5 for improved air–sea fluxes over Southeast Asia
Abstract. This study evaluates the performance of RegCM5 in simulating air–sea fluxes over Southeast Asia through a set of 36 sensitivity experiments testing different physical scheme combinations. Scheme choices vary across five model aspects: radiative transfer, planetary boundary layer, cumulus convection, parameterized microphysics and cloud fraction. A multi-criteria decision-making framework is applied to rank model configurations based on their ability to reproduce spatiotemporal patterns of sea surface wind, latent and sensible heat fluxes, precipitation, and radiative heat fluxes, using mostly satellite-based reference products. No configuration performs consistently best across all criteria: scores assessing latent and shortwave radiative heat fluxes are generally conflicting, with each other and with the scores for precipitation and sea surface wind which instead tend to agree. The choice of cumulus convection scheme drives the performance in simulating the latter two variables, with Tiedtke outperforming and Kain–Fritsch underperforming. In contrast, the best shortwave radiative heat flux simulations are obtained with MIT cumulus convection, in combination with CCM3 radiative transfer. Overall, RRTM/UW-PBL/Tiedtke/SUBEX/Xu–Randall – using the same order of model aspects as listed in the beginning – stands out by maintaining relatively high scores across all assessed variables. Nonetheless, a stronger dissensus in precipitation outputs suggests that reliable rainfall patterns may be a higher priority for decision makers, highlighting CCM3/UW-PBL/Tiedtke/NoTo/Xu–Randall and RRTM/Holtslag/Tiedtke/NoTo/Xu–Randall as the best configurations for this variable. Beyond statistics, further analysis reveals key monsoon-related biases: Indian Summer Monsoon rainfall is generally underestimated, Western North Pacific Summer Monsoon features are overestimated and shifted northward, near-equatorial regions exhibit excessive boreal summer rainfall in most Tiedtke experiments, and austral summer monsoonal sea surface wind and precipitation only impact areas directly north of Australia without inducing the rainfall annual maximum observed in that season over equatorial seas. These findings provide a basis for selecting optimal physics in RegCM5 over Southeast Asia and offer guidance for future applications, including air–sea coupled regional climate modeling.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(15967 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(15967 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-1579', Anonymous Referee #1, 25 Jun 2025
- AC1: 'Reply on RC1', Quentin Desmet, 13 Aug 2025
-
RC2: 'Comment on egusphere-2025-1579', Anonymous Referee #2, 24 Jul 2025
In this manuscript, the authors evaluate the performance of the RegCM5 regional climate model in simulating air–sea fluxes over the Southeast Asian seas. The model was run at a 25 km resolution for the year 2018 using 36 different combinations of physical parameterization schemes, selecting from multiple options for convection, microphysics, planetary boundary layer (PBL), radiation, and cloud fraction. Atmospheric forcing was provided by ERA5 reanalysis at 0.25° resolution, while sea surface temperatures were obtained from the high-resolution SYMPHONIE ocean model running at approximately 0.083° resolution. Model outputs, such as precipitation, surface radiation, latent and sensible heat fluxes, and sea surface wind speed, were evaluated against satellite and reanalysis data. A multi-criteria decision-making framework, incorporating 180 performance metrics across eight oceanic subregions, was used to rank the experiments. The results indicate that the top-performing configuration is a combination of the RRTM radiative transfer scheme, UW-PBL planetary boundary layer, Tiedtke cumulus convection, SUBEX resolved-scale microphysics, and Xu–Randall cloud fraction (identified as 12511, i.e. RRTM/UW-PBL/Tiedtke/SUBEX/Xu–Randall), with the Tiedtke cumulus convection scheme consistently outperforming others, particularly in simulating precipitation and wind. The findings highlight cumulus convection as the primary driver of model performance and suggest that the optimal physical parameterizations may vary depending on the variable of interest (e.g., precipitation vs. shortwave radiation). The manuscript is well written, logically structured, and easy to follow, making it a worthy candidate for publication in Geoscientific Model Development. However, there are some points need to be clarified.
- First, the authors’ use of simulation results from only one neutral year (2018) to evaluate the model’s performance is not sufficiently convincing. A single-year simulation provides only one monthly and annual value per grid point for each variable, which introduces substantial uncertainty into the performance assessment due to the lack of statistical robustness. Furthermore, by excluding years influenced by major climate variability phenomena such as ENSO and IOD, the evaluation overlooks the model’s capacity to simulate responses under extreme conditions, one of the key strengths of dynamical models. As a result, the findings may be overfitted to neutral conditions and may not adequately reflect the model’s robustness or broader applicability across different climate regimes.
- This study is highly valuable for advancing our understanding of air–sea coupling and for supporting the development of coupled models. However, in many practical applications, the accurate simulation of precipitation and temperature over land is even more critical. In fact, coupled models are still relatively uncommon, and most studies continue to rely on standalone RegCM without ocean coupling. Therefore, I suggest that the authors conduct a parallel analysis using the same model configurations over terrestrial subregions where high-quality observational data are available.
- Using ERA5 at the same resolution (0.25°) to force RegCM5 is valid and appropriate for a controlled physics sensitivity study, as done by the authors. However, in this setup, the added value of high-resolution spatial detail from the regional model cannot be fully realized./.
- AC2: 'Reply on RC2', Quentin Desmet, 13 Aug 2025
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-1579', Anonymous Referee #1, 25 Jun 2025
General overview
The authors evaluate the performance of RegCM5 in simulating air-sea fluxes over the Southeast Asian region by testing different physical parametrizations: radiative transfer, planetary boundary layer, cumulus convection, parameterized microphysics, and cloud fraction.
To assess the performance of numerical simulations, the authors use a multicriteria decision-making framework to rank their ability in reproducing sea surface wind, latent and sensible heat fluxes, precipitation, and radiative heat fluxes..
The authors have not found a configuration that properly solves all criteria.
However, they found that using the Tiedtke cumulus convection scheme showed better results for precipitation and sea surface wind.Comments:
The authors mention that the study focuses on the seasonal cycle and that they have chosen the year 2018 to perform the numerical simulations since it offers neutral conditions with respect to large-scale oscillations.
However, the authors remark that it is important to properly take into account the influence of upwellings, eddies, and meanders when forcing atmospheric models. This is the reason why they force RegCM5 with a high-resolution regional SYMPHONIE forced numerical simulation (5 km) instead of with an optimal-interpolation-based SST dataset.
With a 5 km spatial resolution, the ocean numerical simulation will be able to solve eddies of at least 20 km in diameter (effective spatial resolution).
When forcing the atmospheric model, the sea surface data, containing high-spatial resolution features, will be interpolated to the 25 km grid of RegCM5, passing from an effective spatial resolution of 20 km to 100 km.At the end, there is no discussion about the role of mesoscale processes, such as upwelling, eddies, or meanders, in modulating heat fluxes, precipitation, or winds.
Following Frenger et al. (2013) and Villas Bôas et al. (2015), cold and warm eddies are important drivers of latent heat fluxes, and because of that, can modulate the cloud cover and precipitation.
In fact, Villas Bôas et al. (2015) showed that eddies can partially modulate 20% of latent heat fluxes.When the authors compare the monthly SYMPHONIE sea surface temperature results with OSTIA estimations, there are no evident large-scale spatial pattern differences that indicate problems in representing the seasonal cycle.
Instead, the differences, which are larger than 2 oC, seem to be more related to a misaligned occurrence of warm and cold eddies in comparison with the observations.Perhaps it would be easier to directly force RegCM5 with the GLORYS dataset, where mesoscale eddies should be in the proper location, because of the data assimilation.
Regarding the comparison of scores and rankings associated with the 36 experiments, it is difficult to identify which model aspect is more important or promotes more realistic results.
Perhaps it would be better to build a figure that resumes Figure 3, which highlights the model aspects that mostly occur in the top 10 ranks.
In this way, perhaps it would be easier to identify that experiment 12511 is the best performer.Specific comments:
Line 182: Define all acronyms, like EBAF and IMERG.
In line 300, the authors mention that Tiedtke configurations “clearly” outperform the other configurations, but this is not easy to see.
In line 380, the authors highlight the weaker similarity along the equation in Figure 4, but the figure lacks coordinates, which makes it difficult to follow the text.
The authors commonly refer to correlation coefficients, but there is no figure or table to see them, as in line 415.Citation: https://doi.org/10.5194/egusphere-2025-1579-RC1 - AC1: 'Reply on RC1', Quentin Desmet, 13 Aug 2025
-
RC2: 'Comment on egusphere-2025-1579', Anonymous Referee #2, 24 Jul 2025
In this manuscript, the authors evaluate the performance of the RegCM5 regional climate model in simulating air–sea fluxes over the Southeast Asian seas. The model was run at a 25 km resolution for the year 2018 using 36 different combinations of physical parameterization schemes, selecting from multiple options for convection, microphysics, planetary boundary layer (PBL), radiation, and cloud fraction. Atmospheric forcing was provided by ERA5 reanalysis at 0.25° resolution, while sea surface temperatures were obtained from the high-resolution SYMPHONIE ocean model running at approximately 0.083° resolution. Model outputs, such as precipitation, surface radiation, latent and sensible heat fluxes, and sea surface wind speed, were evaluated against satellite and reanalysis data. A multi-criteria decision-making framework, incorporating 180 performance metrics across eight oceanic subregions, was used to rank the experiments. The results indicate that the top-performing configuration is a combination of the RRTM radiative transfer scheme, UW-PBL planetary boundary layer, Tiedtke cumulus convection, SUBEX resolved-scale microphysics, and Xu–Randall cloud fraction (identified as 12511, i.e. RRTM/UW-PBL/Tiedtke/SUBEX/Xu–Randall), with the Tiedtke cumulus convection scheme consistently outperforming others, particularly in simulating precipitation and wind. The findings highlight cumulus convection as the primary driver of model performance and suggest that the optimal physical parameterizations may vary depending on the variable of interest (e.g., precipitation vs. shortwave radiation). The manuscript is well written, logically structured, and easy to follow, making it a worthy candidate for publication in Geoscientific Model Development. However, there are some points need to be clarified.
- First, the authors’ use of simulation results from only one neutral year (2018) to evaluate the model’s performance is not sufficiently convincing. A single-year simulation provides only one monthly and annual value per grid point for each variable, which introduces substantial uncertainty into the performance assessment due to the lack of statistical robustness. Furthermore, by excluding years influenced by major climate variability phenomena such as ENSO and IOD, the evaluation overlooks the model’s capacity to simulate responses under extreme conditions, one of the key strengths of dynamical models. As a result, the findings may be overfitted to neutral conditions and may not adequately reflect the model’s robustness or broader applicability across different climate regimes.
- This study is highly valuable for advancing our understanding of air–sea coupling and for supporting the development of coupled models. However, in many practical applications, the accurate simulation of precipitation and temperature over land is even more critical. In fact, coupled models are still relatively uncommon, and most studies continue to rely on standalone RegCM without ocean coupling. Therefore, I suggest that the authors conduct a parallel analysis using the same model configurations over terrestrial subregions where high-quality observational data are available.
- Using ERA5 at the same resolution (0.25°) to force RegCM5 is valid and appropriate for a controlled physics sensitivity study, as done by the authors. However, in this setup, the added value of high-resolution spatial detail from the regional model cannot be fully realized./.
- AC2: 'Reply on RC2', Quentin Desmet, 13 Aug 2025
Peer review completion
Journal article(s) based on this preprint
Model code and software
RegCM5 code for manuscript "Optimizing physical scheme selection in RegCM5 for improved air–sea fluxes over Southeast Asia" ICTP https://doi.org/10.5281/zenodo.15125814
`scoringtree` Python package for manuscript "Optimizing physical scheme selection in RegCM5 for improved air–sea fluxes over Southeast Asia" Quentin Desmet https://doi.org/10.5281/zenodo.15356967
Interactive computing environment
Scripts and minimal data for manuscript "Optimizing physical scheme selection in RegCM5 for improved air–sea fluxes over Southeast Asia" Quentin Desmet https://doi.org/10.5281/zenodo.15359231
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,966 | 137 | 23 | 2,126 | 27 | 47 |
- HTML: 1,966
- PDF: 137
- XML: 23
- Total: 2,126
- BibTeX: 27
- EndNote: 47
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Quentin Desmet
Thanh Ngo-Duc
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(15967 KB) - Metadata XML
General overview
The authors evaluate the performance of RegCM5 in simulating air-sea fluxes over the Southeast Asian region by testing different physical parametrizations: radiative transfer, planetary boundary layer, cumulus convection, parameterized microphysics, and cloud fraction.
To assess the performance of numerical simulations, the authors use a multicriteria decision-making framework to rank their ability in reproducing sea surface wind, latent and sensible heat fluxes, precipitation, and radiative heat fluxes..
The authors have not found a configuration that properly solves all criteria.
However, they found that using the Tiedtke cumulus convection scheme showed better results for precipitation and sea surface wind.
Comments:
The authors mention that the study focuses on the seasonal cycle and that they have chosen the year 2018 to perform the numerical simulations since it offers neutral conditions with respect to large-scale oscillations.
However, the authors remark that it is important to properly take into account the influence of upwellings, eddies, and meanders when forcing atmospheric models. This is the reason why they force RegCM5 with a high-resolution regional SYMPHONIE forced numerical simulation (5 km) instead of with an optimal-interpolation-based SST dataset.
With a 5 km spatial resolution, the ocean numerical simulation will be able to solve eddies of at least 20 km in diameter (effective spatial resolution).
When forcing the atmospheric model, the sea surface data, containing high-spatial resolution features, will be interpolated to the 25 km grid of RegCM5, passing from an effective spatial resolution of 20 km to 100 km.
At the end, there is no discussion about the role of mesoscale processes, such as upwelling, eddies, or meanders, in modulating heat fluxes, precipitation, or winds.
Following Frenger et al. (2013) and Villas Bôas et al. (2015), cold and warm eddies are important drivers of latent heat fluxes, and because of that, can modulate the cloud cover and precipitation.
In fact, Villas Bôas et al. (2015) showed that eddies can partially modulate 20% of latent heat fluxes.
When the authors compare the monthly SYMPHONIE sea surface temperature results with OSTIA estimations, there are no evident large-scale spatial pattern differences that indicate problems in representing the seasonal cycle.
Instead, the differences, which are larger than 2 oC, seem to be more related to a misaligned occurrence of warm and cold eddies in comparison with the observations.
Perhaps it would be easier to directly force RegCM5 with the GLORYS dataset, where mesoscale eddies should be in the proper location, because of the data assimilation.
Regarding the comparison of scores and rankings associated with the 36 experiments, it is difficult to identify which model aspect is more important or promotes more realistic results.
Perhaps it would be better to build a figure that resumes Figure 3, which highlights the model aspects that mostly occur in the top 10 ranks.
In this way, perhaps it would be easier to identify that experiment 12511 is the best performer.
Specific comments:
Line 182: Define all acronyms, like EBAF and IMERG.
In line 300, the authors mention that Tiedtke configurations “clearly” outperform the other configurations, but this is not easy to see.
In line 380, the authors highlight the weaker similarity along the equation in Figure 4, but the figure lacks coordinates, which makes it difficult to follow the text.
The authors commonly refer to correlation coefficients, but there is no figure or table to see them, as in line 415.