John - Thanks for this helpful feedback. I'll pass your suggestion about using "NA" vice NaN" along to the developer and provide his response in another comment. I will also mention your proposed bugfixes to enable future releases of MET to handle "NaN" values.

John - the developer thinks that he can change the "NaN" to "NA", so we are pursuing that option which takes advantage of the MET's internal handling of bad or missing data. I used the grep command you suggested which worked exactly as you said. It removed the following 7 lines:
1 MSATWR MSATWR01 20220705_161500 32.59018526 -106.7417173 1321 11 NaN 2 NA NaN
2 MSATWR MSATWR01 20220705_161500 32.59018526 -106.7417173 1321 52 NaN 2 NA NaN
3 MSATWR MSATWR01 20220705_161500 32.59018526 -106.7417173 1321 33 NaN 10 NA -1.62301324149007
4 MSATWR MSATWR01 20220705_161500 32.59018526 -106.7417173 1321 34 NaN 10 NA 1.18073001225701
5 ADPUPA MSATWR01 20220705_161500 32.59018526 -106.7417173 1321 11 NaN 10 NA NaN
6 ADPUPA MSATWR01 20220705_161500 32.59018526 -106.7417173 1321 33 NaN 2 NA -1.24151906693094
7 ADPUPA MSATWR01 20220705_161500 32.59018526 -106.7417173 1321 34 NaN 2 NA 0.862982163853926

I need to run the ASCII2NC to convert the file containing the remaining obs to NetCDF and then run Point-Stat to confirm everything is working as expected.

My assessment of the impact of removing these lines is as follows:
Lines 1,2,5 -->no impact as there is no usable data when the obs value is missing and the obs pres level is missing.
Lines 3,4 -->Appears to be possible adverse impact because the obs for U and V winds at 10m AGL (standard level) should be processed by MET as values at 10 AGL and the ob pressure level is not needed for processing on AGL levels.
Lines 6.7 -->no impact as MET will not be able to process the U and V winds at 2m AGL on non-standard levels without a pressure value. The non-standard obs are verified against model output on a pressure level instead of an AGL level.

Does the above correctly state the impacts?

Thanks.

John, at first I didn't realize that the NaN lines really do contain legitimate observations in them. So instead of running grep to remove them from the file, I'd recommend running sed to just replace all instances of NaN with NA. And that's consistent with the change you've already discussed with the developer.

The good news is that it's a simple one-liner:

sed -i 's/NaN/NA/g' qc_obs_used_earth_relative.2022_07_05_161500.txt

This sed command modifies the file in place. If you'd rather keep the original file untouched and write the modified output to a new file, don't use the -i option, like this:

sed 's/NaN/NA/g' qc_obs_used_earth_relative.2022_07_05_161500.txt > qc_obs_used_earth_relative.2022_07_05_161500_patch.txt

As for your line-by-line assessment, there's a lot of details there. Put simply, if the pressure value (column 8) is bad data, then you can't (or shouldn't) use it to verify forecasts at pressure levels. But big picture, you should be able to configure Point-Stat to access/verify against any of the U/V vector observations in this file. Rather than thinking about all possible permutations, I'd recommend considering your forecast and observation data sources and decide which comparisons make logical sense and are scientifically useful. And then configure Point-Stat to make those comparisons.

And if you have trouble getting Point-Stat to compare what you'd like (e.g. fcst/obs name, fcst/obs level, and message type), I can probably help with those details.

Great suggestion, John. So using the sed command you suggested actually accomplishes the replacement which is consistent with the modification to the MATLAB script to change NaN to NA.

Appreciate your offer to assist with determining the most appropriate comparisons. I would like to review the possibilities with you for processing the observations in the file I uploaded. Originally, my plan was to process the observations at non-standard levels such as UGRD and VGRD at 2m AGL and TMP at 10m AGL by assigning the message type ADPUPA to those observations so I could verify them on a pressure level instead of on an AGL level. I have forecast output on pressure levels which are reasonable to be used with observations at 2m and 10m AGL. Is this the way to go, or can I match up those observations at non-standard levels some other way?

@jwraby, like I said, there's any number of possibilities. There's no right or wrong answer here. Its your decision to configure MET how you'd like... to make the fcst/obs comparisons that you'd like... to answer the scientific questions that you'd like to investigate.

It certainly is fine to verify forecasts pressures levels against ADPUPA or MSATWR observations on pressure levels.

Let me point you to this section about vertical interpolation in Point-Stat. You can decide whether you want to just match point observations to a single model level (e.g. NO VERTICAL INTERPOLATION: verify TMP/P850 forecast using any observations of temperature between P825 and P875) or interpolate the forecast pressure level data in the vertical to the pressure of the observation (e.g. WITH VERTICAL INTERPOLATION: verify all forecast pressure levels TMP/P800-1000 using any observations of temperature in that same range).