Dear MPAS developers,
I am using MPAS-A with a 60–3 km variable-resolution mesh to simulate the January 2024 winter snowstorm. I ran two experiments with the same dynamics and large-scale forcing: one with Thompson and one with WSM6 microphysics. Both are using the convection_permitting physics suite.
On the native MPAS grid, both experiments show a strong positive relationship between terrain height and simulated precipitation/snow depth. For the 36-h evaluation window:
- 36-h precipitation vs terrain height:
- Thompson: area-weighted Pearson r ≈ 0.42
- WSM6: area-weighted Pearson r ≈ 0.35
- End-of-period snow depth vs terrain height:
- Thompson: area-weighted Pearson r ≈ 0.81
- WSM6: area-weighted Pearson r ≈ 0.74
Area-mean precipitation increases from about 7 mm below 200 m to about 14–15 mm above 800 m. Snow depth increases from near 0 cm in low-elevation areas to about 10 cm above 800 m. The two microphysics schemes therefore have similar terrain-dependent behavior.
The observational issue is that most stations are located at low elevations, so the observational sampling is imperfect. However, this event produced widespread heavy snowfall over low-elevation areas, and many low-elevation stations observed substantial precipitation and snow accumulation. The simulated result appears much more terrain-controlled than the observed event.
I am not assuming the terrain correlation is necessarily wrong, since orographic precipitation and colder mountain temperatures can physically enhance snowfall. My concern is whether some MPAS configuration settings may be amplifying this terrain dependence or creating an unrealistic separation between lowland and mountain precipitation/snowfall.
For the static fields, I used the standard WRF/WPS GEOG static datasets, which are commonly used for WRF preprocessing. The terrain is not from a custom DEM. In the MPAS initialization namelist, I set:
- config_geog_data_path = path to the standard WRF GEOG dataset
- config_topo_data = GMTED2010
- config_supersample_factor = 3
Other relevant static/initialization settings include:
- config_nsmterrain = 1
- config_smooth_surfaces = true
- config_blend_bdy_terrain = false
- config_native_gwd_static = false
- 55 vertical levels
- config_ztop = 30 km
- config_dzmin = 0.3
- config_nsm = 30
- config_tc_vertical_grid = true
My main question is: are there known MPAS settings related to terrain generation, terrain smoothing, static fields, vertical coordinate design, gravity-wave-drag static fields, or the convection-permitting physics suite that could cause precipitation and snowfall to become too strongly correlated with terrain height?
In particular, I would appreciate advice on:
1. Whether the standard WRF/WPS GEOG topography, especially the GMTED2010 terrain source, is appropriate for generating MPAS native-grid terrain at 60–3 km, or whether another terrain dataset/static-field option is usually recommended.
2. Which `init_atmosphere` or static-field settings I should test first if I suspect excessive terrain control.
3. Whether `config_nsmterrain`, `config_smooth_surfaces`, `config_blend_bdy_terrain`, or `config_native_gwd_static` can noticeably affect surface precipitation/snowfall patterns.
4. Whether the 60–3 km mesh is generally sufficient for this type of southern-China winter terrain/precipitation interaction, or whether this behavior may indicate that the effective terrain forcing is still too coarse or too steep.
5. Which MPAS diagnostics would help determine whether the terrain-related precipitation/snow signal is physically reasonable or configuration-related.
6. Are there recommended sensitivity tests for separating realistic orographic enhancement from artificial terrain over-amplification?
Thank you very much for your help.
I am using MPAS-A with a 60–3 km variable-resolution mesh to simulate the January 2024 winter snowstorm. I ran two experiments with the same dynamics and large-scale forcing: one with Thompson and one with WSM6 microphysics. Both are using the convection_permitting physics suite.
On the native MPAS grid, both experiments show a strong positive relationship between terrain height and simulated precipitation/snow depth. For the 36-h evaluation window:
- 36-h precipitation vs terrain height:
- Thompson: area-weighted Pearson r ≈ 0.42
- WSM6: area-weighted Pearson r ≈ 0.35
- End-of-period snow depth vs terrain height:
- Thompson: area-weighted Pearson r ≈ 0.81
- WSM6: area-weighted Pearson r ≈ 0.74
Area-mean precipitation increases from about 7 mm below 200 m to about 14–15 mm above 800 m. Snow depth increases from near 0 cm in low-elevation areas to about 10 cm above 800 m. The two microphysics schemes therefore have similar terrain-dependent behavior.
The observational issue is that most stations are located at low elevations, so the observational sampling is imperfect. However, this event produced widespread heavy snowfall over low-elevation areas, and many low-elevation stations observed substantial precipitation and snow accumulation. The simulated result appears much more terrain-controlled than the observed event.
I am not assuming the terrain correlation is necessarily wrong, since orographic precipitation and colder mountain temperatures can physically enhance snowfall. My concern is whether some MPAS configuration settings may be amplifying this terrain dependence or creating an unrealistic separation between lowland and mountain precipitation/snowfall.
For the static fields, I used the standard WRF/WPS GEOG static datasets, which are commonly used for WRF preprocessing. The terrain is not from a custom DEM. In the MPAS initialization namelist, I set:
- config_geog_data_path = path to the standard WRF GEOG dataset
- config_topo_data = GMTED2010
- config_supersample_factor = 3
Other relevant static/initialization settings include:
- config_nsmterrain = 1
- config_smooth_surfaces = true
- config_blend_bdy_terrain = false
- config_native_gwd_static = false
- 55 vertical levels
- config_ztop = 30 km
- config_dzmin = 0.3
- config_nsm = 30
- config_tc_vertical_grid = true
My main question is: are there known MPAS settings related to terrain generation, terrain smoothing, static fields, vertical coordinate design, gravity-wave-drag static fields, or the convection-permitting physics suite that could cause precipitation and snowfall to become too strongly correlated with terrain height?
In particular, I would appreciate advice on:
1. Whether the standard WRF/WPS GEOG topography, especially the GMTED2010 terrain source, is appropriate for generating MPAS native-grid terrain at 60–3 km, or whether another terrain dataset/static-field option is usually recommended.
2. Which `init_atmosphere` or static-field settings I should test first if I suspect excessive terrain control.
3. Whether `config_nsmterrain`, `config_smooth_surfaces`, `config_blend_bdy_terrain`, or `config_native_gwd_static` can noticeably affect surface precipitation/snowfall patterns.
4. Whether the 60–3 km mesh is generally sufficient for this type of southern-China winter terrain/precipitation interaction, or whether this behavior may indicate that the effective terrain forcing is still too coarse or too steep.
5. Which MPAS diagnostics would help determine whether the terrain-related precipitation/snow signal is physically reasonable or configuration-related.
6. Are there recommended sensitivity tests for separating realistic orographic enhancement from artificial terrain over-amplification?
Thank you very much for your help.