CMAQ evaluation tools

With CMAQ running semi-successfully on System X (THH "waiting too long for completion" hangups crop up occasionally but are system- and not code-related), I am now using the evaluation tools to compare model output to observations. Here is the order of reformatting:

1) AQS AIRS files, including STN/IMPROVE and PAMS speciation in AIRS formats, downloaded from http://www.epa.gov/ttn/airs/airsaqs/detaildata/downloadaqsdata.htm
2) Use the rd_airs program to convert these to an intermediate format
3) Use the airs2ext program to convert the intermediate format to CASTNET format
4) Use sitecmp to compare the CASTNET-formatted data to your model output, the ACONC files

On compiling the evaluation tools:
1) The Makefile assumes the Portland Fortran90 compiler, so I had to recompile a Portland Fortran version of IOAPI.
2) I had to specify the locations of the libioapi.a and libnetcdf.a files in the Makefile.
3) The correctly formatted EXT files are in $IOAPI/ioapi/fixed_src.
4) For the sitecmp program, XSTAT3.EXT does not exist, but the version of this file that came with the unified version of CMAQ is covered by definitions in PARMS3.EXT, so I commented out the line for XSTAT3.EXT.
5) To get the program combine to compile, I had to modify the Makefile and put module_specdef.o ahead of all the other .o files. We will be calculating PM2.5 as

PM25T25,ug/m3,
(ASO4I[2]+ANO3I[2]+ANH4I[2]+AORGAI[2]+1.167*AORGPAI[2]+AORGBI[2]+AECI[2])*PM25AT[1]+
(ASO4J[2]+ANO3J[2]+ANH4J[2]+AORGAJ[2]+1.167*AORGPAJ[2]+AORGBJ[2]+AECJ[2]+ANAJ[2]+ACLJ[2]+A25J[2])*PM25AC[1]+
(ASO4K[2]+ANAK[2]+ACLK[2]+ACORS[2]+ASOIL[2])*PM25CO[1]

as obtained from Prakash Bhave at NOAA/EPA.

CMAQ rebuild

The CMAQ project is back underway! Unfortunately, the storage changes on System X have changed the home directory path from /nfs/home to /home. CMAQ stores absolute pathnames in the binaries, so it's rebuild time!

Progress report

Just a progress report:

1) After almost 18 hours of coding, CMAQ 4.6 runs on System X! I had to completely reinstrument the parallel I/O and stencil exchange libraries to get it working, but I've managed to run the CMAQ tutorial on System X (the results check out, so I didn't break anything ;) ). There's still a problem with file permissions on the compute nodes, so I haven't been able to try a VISTAS run yet, but the LASCA staff are working on it (so they say). I'll probably need some help identifying the most appropriate data files for input, and I'm not convinced that bcon and icon only need to be run once. I think you need to re-run them for different domains, so that means finding the correct inputs. Is this true?

2) I'm rebuilding CMAQ 4.5 on ene-411-lmarr to try to get our data to match the benchmark. I have questions about which input files are the most useful, but there are so many choices I think I'm just going to have to sit down with Linsey one morning and sort this out. E-mail!

3) I've found a detailed description of the entire Virginia ambiant air monitoring network in the Virginia Ambiant Air Monitoring 2002 Data report (http://www.deq.state.va.us/airmon/pdf/AnnualRpt02.pdf). This includes station numbers, pollutants, locations, airs numbers, and lat/long. I'm working on finding data for each of these sites.

4) I've found complete data for two IMPROVE sites: JARl1 and SHEN1. I'm still trying to make sense of what I found. Data is in /home/jlinford/ambiant_data as OpenOffice.org spreadsheets.

5) Does anybody know where I can simply "look up" data given a monitoring site? I mean, the EPA must keep this stuff somewhere, and what good are these goofy site id numbers if you can't query a database with them....

It's .... still .... running ....

Adding the OCEAN file to CMAQ has made it even slower. We've been running for several hours now with no end in sight. Another thing, I'm seeing these messages in the output log:

grep -A2 WARNING nohup.out


>>--->> WARNING in subroutine RDEMIS_AE:INTERPX
Variable "SURF" not in file OCEAN_1
M3WARN: DTBUF 0:03:20 Jan. 1, 2002
--
>>--->> WARNING in subroutine RDEMIS_AE
Could not interpolate SURF from OCEAN_1 . Surf-zone emissions set to zero.
M3WARN: DTBUF 0:03:20 Jan. 1, 2002
--
>>--->> WARNING in subroutine RDEMIS_AE:INTERPX
Variable "SURF" not in file OCEAN_1
M3WARN: DTBUF 0:10:00 Jan. 1, 2002
--
>>--->> WARNING in subroutine RDEMIS_AE
Could not interpolate SURF from OCEAN_1 . Surf-zone emissions set to zero.
M3WARN: DTBUF 0:10:00 Jan. 1, 2002
--
>>--->> WARNING in subroutine RDEMIS_AE:INTERPX
Variable "SURF" not in file OCEAN_1
M3WARN: DTBUF 0:16:40 Jan. 1, 2002
--
>>--->> WARNING in subroutine RDEMIS_AE
Could not interpolate SURF from OCEAN_1 . Surf-zone emissions set to zero.
M3WARN: DTBUF 0:16:40 Jan. 1, 2002
--
>>--->> WARNING in subroutine RDEMIS_AE:INTERPX
Variable "SURF" not in file OCEAN_1
M3WARN: DTBUF 0:23:20 Jan. 1, 2002
--
>>--->> WARNING in subroutine RDEMIS_AE
Could not interpolate SURF from OCEAN_1 . Surf-zone emissions set to zero.
M3WARN: DTBUF 0:23:20 Jan. 1, 2002


We'll just have to see how things turn out...

The search begins

Ok, so now we need to...

1. Build a spreadsheet of stations in VA, data available at station, lat/long for stations
2. Gather ambient data for stations in VA
3. Convert lat/long to x/y and x/y into gridcell
4. Compare real data with model data
   

The yellow brick road is marked by these webpages:

• Ambient Data for Model Evaluation
  http://pah.cert.ucr.edu/vistas/ambient.shtml
  
• Ambient Datalist
  http://pah.cert.ucr.edu/vistas/docs/Ambient_datalist.doc
• Selected AQS Code Descriptions
  http://www.epa.gov/ttn/airs/airsaqs/manuals/codedescs.htm
• VIEWS Website
  http://vista.cira.colostate.edu/views/Web/IMPROVE/improve_data_prelim.htm
  

Let the hunt begin!

New PGroup license installed, CMAQ rebuilt

I've installed the new PGroup license file and rebuilt CMAQ. pgf90 is FAST, much faster than ifort or g95. (And before you say anything, I "make clean"'ed and deleted all .o and .mod before building). CMAQ is now ready to go. We just need the OCEAN file. From the output log:

>>--->> WARNING in subroutine OPEN3
File not available.


*** ERROR ABORT in subroutine RDEMIS_AE
Could not open OCEAN_1 file

Next steps

So, here's what we do next:

1) Benchmarkmark our results to make sure the data is good. Here's the plan:
a) Rebuild the model with commercial pgf90. Still need to get the license....
b) Rebuild with aero4. Need sea-salt OCEAN file. Linsey is finding this.
c) Edit the list of time-averaged output species (AVG_CONC_SPCS) in the run script.
d) Run the model with these species: NO2, NO, O3, HO, HO2, N2O5, HNO3, CO, ISOP, ETH, TOL, SO2, ASO4J, ASO4I, ANH4J, ANH4I, ANO3J, ANO3I, AORGAJ, AORGAI, AECJ, AECI
e) Compare our results to demo results.

2) Gather ambiant data for Virginia so we can compare model to real measurements. This could take some time, but here's a starting point: http://pah.cert.ucr.edu/vistas/ambient.shtml

3) Compare our results to real data. Here's some resources (not links, just files I'm supposed to already have somewhere) :
a) EVALUATION_TOOLS.txt
b) Model_Performance_Evaluation.pdf

(1)a-c shouldn't take more than a day. (1)d will take another 10+ hours of runtime. (1)e could take 1-2 days, depending on how complete we want to be.

(2) Might be a week or so, depending. It's just a lot of digging.

(3) ???? It's going to take some time to match real stations with grid cells and then extract the data. I'll know better when I get closer.

Did I miss anything?