Should I use a DTM or DSM as terrain elevation input for high-resolution WRF simulations?

[seti]

​

Hello everyone,

I am preparing a high-resolution WRF simulation over a complex mountainous region in the Italian Alps. My planned nested domains have horizontal grid spacings of approximately:

• d01: 9 km
• d02: 3 km
• d03: 1 km
• d04: 333 m

I would like to update the default WPS terrain elevation dataset with a higher-resolution product. However, I am uncertain whether the terrain input for HGT_M should be based on a Digital Terrain Model (DTM), representing bare-earth elevation, or a Digital Surface Model (DSM), which may include vegetation, buildings, and other surface objects.

My main questions are:

1. For WRF terrain elevation and HGT_M, is a bare-earth DTM generally more physically appropriate than a DSM?
2. Could using a DSM introduce an artificial elevation bias in forested or urban areas?
3. For a 333 m inner domain over steep Alpine terrain, what source resolution would be appropriate before interpolation to the WRF grid—for example 10 m, 30 m, or 100 m?
4. Should a very high-resolution LiDAR DTM be averaged or smoothed before converting it into WPS geogrid format?
5. Are there recommended smoothing criteria or terrain-slope checks to avoid numerical instability at 333 m grid spacing?
6. Is it advisable to use the same terrain dataset for all nested domains, or can a local LiDAR DTM be used only for the innermost domain while a global DEM is used for the outer domains?
7. If local DTM and global DEM datasets are combined, what is the recommended method for blending them near the dataset boundaries?

The study area contains narrow valleys, steep slopes, and high mountain ridges. Therefore, accurate representation of valley-floor and station elevations is important for evaluating near-surface temperature, humidity, wind, and pressure against station observations.

I would appreciate any recommendations regarding suitable datasets, preprocessing steps, WPS configuration, and relevant references or previous experience.

Thank you.

 

[Ming Chen]

Hi, Seti,
Please see my answers below.

Hello everyone,


I am preparing a high-resolution WRF simulation over a complex mountainous region in the Italian Alps. My planned nested domains have horizontal grid spacings of approximately:

• d01: 9 km
• d02: 3 km
• d03: 1 km
• d04: 333 m

I would like to update the default WPS terrain elevation dataset with a higher-resolution product. However, I am uncertain whether the terrain input for HGT_M should be based on a Digital Terrain Model (DTM), representing bare-earth elevation, or a Digital Surface Model (DSM), which may include vegetation, buildings, and other surface objects.

My main questions are:

1. For WRF terrain elevation and HGT_M, is a bare-earth DTM generally more physically appropriate than a DSM?

It is recommended to use DTM, which is strictly consistent with 'terrain height' used in WRF.

1. Could using a DSM introduce an artificial elevation bias in forested or urban areas?

I guess it could, although I haven't done any test to confirm this.

1. For a 333 m inner domain over steep Alpine terrain, what source resolution would be appropriate before interpolation to the WRF grid—for example 10 m, 30 m, or 100 m?

If the input data has higher resolution than the model grid size, WPS will average the data within a single grid to get the value for the grid. In this case, I suppose 10m, 30m, 100m all can be used as input. The only difference is that, 10-m data may have more terrain details than the 100 m data, and thus the average within a 333 m grid may be more realistic.

1. Should a very high-resolution LiDAR DTM be averaged or smoothed before converting it into WPS geogrid format?

I am not sure.

1. Are there recommended smoothing criteria or terrain-slope checks to avoid numerical instability at 333 m grid spacing?

When you run WRF over steep terrain area, the namelist option epssm is used to suppress possible instability.

1. Is it advisable to use the same terrain dataset for all nested domains, or can a local LiDAR DTM be used only for the innermost domain while a global DEM is used for the outer domains?

Personally I like to use the same input data to ensure consistency between inner- and outer domains.

1. If local DTM and global DEM datasets are combined, what is the recommended method for blending them near the dataset boundaries?

I am not sure.

The study area contains narrow valleys, steep slopes, and high mountain ridges. Therefore, accurate representation of valley-floor and station elevations is important for evaluating near-surface temperature, humidity, wind, and pressure against station observations.

I would appreciate any recommendations regarding suitable datasets, preprocessing steps, WPS configuration, and relevant references or previous experience.

Thank you.

 

[seti]

@Ming Chen


Dear Ming,

Thank you very much for your quick response. I really appreciate it.

I would also like to ask which available DTM dataset you would recommend as the most reliable for WRF simulations. Could you please advise me on this?

Thank you in advance for your guidance.

Best regards,