Interpolation options for different resolution

[San]

Dear all,

I'm trying to create a geo_em file using the various interpolation options provided by geogrid.
The geo_em file has a resolution of 500m, and the original data (to make LANDUSEF, LU_INDEX and HGT_M field) used to create the geo_em file has a resolution of 100m.

I checked the LU_INDEX variable and found that the interpolation option other than nearest_neighbor did not work.
Of course, LU_INDEX is a categorical variable and it would not be appropriate to use interpolation options other than nearest_neighbor.
However, if use other interpolation options, such as four_pt or sixteen_pt, geogrid must interpolate according to the algorithm of each option, but the execution results are all the same as the nearest_neighbor option.

Meanwhile, when the original data at 1 km resolution were used, each interpolation option appeared to work well.

Is there an algorithm in geogrid/src that compares the original binary data resolution to the model resolution to apply the interpolation option differently?
If not, is there an algorithm that applies appropriate interpolation options when receiving categorical or continuous variables?

Attach the GEOGRID.TBL file and the namelist.wps file that I used to create the geo_em file.

Thank you so much for your help.

 

[Ming Chen]

Please see my answrs below:

(1) Is there an algorithm in geogrid/src that compares the original binary data resolution to the model resolution to apply the interpolation option differently?

No, I don't think so. The interpolation option is always dependent on that specified in GEOGRID.TBL

(2) If not, is there an algorithm that applies appropriate interpolation options when receiving categorical or continuous variables?

For both categorical and continuous variables, the interpolation is conducted based on the option given in GEOGRID.TBL.

 

[San]

Please see my answrs below:

(1) Is there an algorithm in geogrid/src that compares the original binary data resolution to the model resolution to apply the interpolation option differently?

No, I don't think so. The interpolation option is always dependent on that specified in GEOGRID.TBL

(2) If not, is there an algorithm that applies appropriate interpolation options when receiving categorical or continuous variables?

For both categorical and continuous variables, the interpolation is conducted based on the option given in GEOGRID.TBL.

Before writing the question, I referred to the following website:

What resolution of static data should be used?

Hi, I am using three two-way nested domains of resolutions 9km, 3km and 1km. Static data is available upto 30s resolution. How do I decide what resolution data should be used with each of my domains? Is it okay to use static data with resolution much finer than the domain resolution? What is...

forum.mmm.ucar.edu

On the above homepage, I read the following response:
It is totally fine if the static data resolution is higher than the model resolution, because WPS takes care of this issue and gives representative value for each grid box.

When the static data resolution is higher than the model resolution, which algorithms provide representative values for each grid box?.
Does the algorithm that provides the representative value also depend on the GEOGRID.TBL specified option as you mentioned?

The model resolution is 500m resolution, and the static data is 100m resolution.
I leave a question because I wonder why the four point average interpolation method for the categorical LU_INDEX field is the same result as the result of the nearest_neighbor interpolation method.

Thanks very much for your help.

 

[Ming Chen]

Before writing the question, I referred to the following website:

What resolution of static data should be used?

Hi, I am using three two-way nested domains of resolutions 9km, 3km and 1km. Static data is available upto 30s resolution. How do I decide what resolution data should be used with each of my domains? Is it okay to use static data with resolution much finer than the domain resolution? What is...

forum.mmm.ucar.edu

On the above homepage, I read the following response:
It is totally fine if the static data resolution is higher than the model resolution, because WPS takes care of this issue and gives representative value for each grid box.

When the static data resolution is higher than the model resolution, which algorithms provide representative values for each grid box?.

In this case, for land use type, WPS will calculate percentages (or fractions) of various landuse types within a single WRF grid cell, and the one with the largest percentage will be taken as the landuse type in that grid. Details can be found in accum_categorical routine.

Does the algorithm that provides the representative value also depend on the GEOGRID.TBL specified option as you mentioned?

yes that is correct.

The model resolution is 500m resolution, and the static data is 100m resolution.
I leave a question because I wonder why the four point average interpolation method for the categorical LU_INDEX field is the same result as the result of the nearest_neighbor interpolation method.

The only explanation I can find is that, the dominant landuse type of the four surrounding points have the same type as that of the nearest neighbor. Also note that for bilinear interpolation, the nearest point has the largest weight, which may also contribute to the feature you have found, i.e., four point average interpolation method and the nearest_neighbor interpolation method yield same result.

Thanks very much for your help.