WRF-Chem MOZCART Performance

[Zongrun Li]

Dear all,

I have ran a one-year simulation for CONUS. Here is our settings:

We conducted year-long WRF-Chem (version 4.2) simulations for the year 2017, focusing on the CONUS region. The simulation setup included the following configurations and input sources:

1. Chemical Mechanism: We used the MOZCART chemical mechanism, combining MOZART for gases and GOCART for aerosols.

2. Model Resolution and Domain: The model resolution was set to 36 km, covering the entire CONUS region.

3. Meteorological Conditions:

   • Time-varying meteorological initial and boundary conditions were derived from the 12 km North American Mesoscale Forecast System (NAM).

4. Chemical Initial and Boundary Conditions:

   • Chemical species' initial and boundary conditions were provided by CAM-Chem.

5. Emissions:

   • Anthropogenic Emissions: Derived from the U.S. National Emission Inventory (NEI2017) and converted from SMOKE outputs to be compatible with WRF-Chem.
   • Biogenic Emissions: Calculated online using the Model of Emissions of Gases and Aerosols from Nature (MEGAN).

6. Additional Inputs and Adjustments:

   • Background GOCART Data: Created using the PREP-Chem-SRC program.
   • Seasonal Dry Deposition: Provided by the season_wesely program.
   • Climatological O₃ and O₂ Overhead Columns: Required for the FTUV photolysis option, generated using the exocoldens program.

Here is the model performance:

I followed the instruction MOZCART provided. Is there any steps I made wrong so that the performance is not really well?

Thanks,

Zongrun

[Zongrun Li]

I evaluate the model performance by comparing it with the EPA AQS monitoring sites.

[Gabriele Pfister]

Hello Zongrun

I would say that your performance looks actually pretty good from the plots you sent. You have to keep in mind that you compare a 36km model grid to very localized observations and cannot expect perfect agreement. 

Gabi

--

================================
Gabriele Pfister

Atmospheric Chemistry Observations & Modeling (ACOM) Lab

NSF National Center for Atmospheric Research (NCAR)

Email: pfi...@ucar.edu

Work Phone: +1 303 497 2915

Web: https://staff.ucar.edu/users/pfister

[Zongrun Li]

Thank you! I’ve just noticed that the performance is comparable to what I obtained from GEOS-Chem (0.5° × 0.625°). Initially, I was focusing on comparing it with a 12 km CMAQ simulation, and I found that WRF-Chem tended to underestimate ozone and NO2 levels. I think that the resolution might be a contributing factor as you suggested. Since this is my first time running a year-long simulation with WRF-Chem, I want to ensure that my settings are configured correctly.

[MarwaClimate2024]

Hello Mr. "zongr." How are you? I am doing research on simulating the atmosphere with aerosols to try to reduce solar radiation but I am facing a problem with the simulation program W R F - c h em can you help me with this?

[MarwaClimate2024]

Hello  Mrs.  Gabriele Pfister  ." How are you? I am doing research on simulating the atmosphere with aerosols to try to reduce solar radiation but I am facing a problem with the simulation program W R F - c hem can you help me with this?

[Gabriele Pfister]

Hello Zongrun

I am not sure what your request is. 

Also, I just noticed from your previous email that you use GOCART background fields for your runs? These are not needed when you run with MOZCART since the mechanism has full gas phase chemistry. Make sure you carefully go through the examples and Instructions we provide on our website (https://www2.acom.ucar.edu/wrf-chem) to ensure you have all namelist settings correct. 

Gabi 

[Zongrun Li]

Hi Gabi,

Thanks, it is a thing that good to know since creating the GOCART background file is tedious and seems have less impacts on WRF-Chem simulations. 

Could I ask an additional question for the PM2.5 simulations using MOZCART? I have no questions for the NO2 and O3 simulation yet and the results looks reasonable. However, I recently checked the PM2.5 and the daily PM2.5 performance is not really well. I used the NEI2017 emissions provided by NCAR (the one provided along with EPA_ANTHRO) and I find the PM2.5 performance is poor (see my attached file). It seems the WRF-Chem did well predictions on PM2.5 below 30 but worse for the extreme events. Is it normal or I did something wrong for simulations? I did not find ptfire in the provided NEI2017. Could it be a reason?

Thanks,

Zongrun 

[Gabriele Pfister]

Hello Zongrun

why your model is not performing well can have many reasons and it would require a detailed analysis to identify shortcomings. Just a few comments that might help guide you: 

- I am not sure what resolution you are running the model at but when you compare to surface sites then keep in mind that the observations are very localized and the model with a coarser resolution might not be able to represent these scales. Specifically if the observations are taken in urban areas. You might also want to look at different regions separately since aerosol characteristics can vary significantly between different parts of the CONUS. 

- I am not quite sure what your plots show but you might also want to compare diurnal cycles. Model performance can often vary between e.g., day and night.  Similarly, if you have longer simulations, you might also want to see if your model performance is better in one season than another since dominant sources might also vary with season. 

- The PBL can have a strong influence on surface PM2.5 concentrations and I suggest you also look into that. 

- GOCART does not simulate SOA, so if SOA is an important component you might have to switch to a more detailed aerosol scheme

-Do I understand correctly that you do not include any fire emissions at all? If so, then I strongly suggest adding them (you can, e.g. use FINN fire emissions as described on our website). Also make sure you have online dust and sea salt emissions active. 

Gabi

[Zongrun Li]

Hi Gabi,

Thanks for your suggestions. I found I have some issues on simulating the correct PM2.5 at the ocean region in my previous one year simulations (2017). 

I revised my namelist based on your suggestions (adding FINN and remove GOCART background) and I attached the namelist here. I did a one day simulation and I can still found the high PM2.5 at the ocean. 

I checked the wrfchemi file and it looks fine. I checked SEAS variable in wrfinput (zeros as initial values) and wrfbdy and it looks with low values. I suspicious that the dry deposition at ocean had some problem and here is the message in the rsl.error. Is the high PM at ocean due to the dry deposition problem? Do you have any suggestions?

d01 2017-01-01_20:40:00 calling dry_deposition_driver

d01 2017-01-01_20:40:00 DOING DRY DEP VELOCITIES WITH WESELY METHOD

d01 2017-01-01_20:40:00 in dry_dep_wesely

d01 2017-01-01_20:40:00 AEROSOL DRY DEP VELOCITIES  = 0.0

Thanks,

Zongrun

[Zongrun Li]

I could also provide the previous namelist for running this one year simulation which shows high PM over the ocean.

[Mary Barth]

Hi Zongrun,

I agree with your suspicion about dry deposition. Yet your namelist option for sea salt emissions seems correct to me. If it were me, I would run again with a higher wrf_debug level to see if there is more information about why drydep velocity = 0. 

I would also look into the model code to find where the code is writing that statement. That is, 

grep "AEROSOL DRY DEP VELOCITIES" *F

in the chem/ directory. 

I also noticed that you do not have wet scavenging turned on. I'm not sure it would reduce the PM concentrations over the ocean but it would be a better configuration.  Set:

wetscav_onoff                       = 1,

 chem_conv_tr                        = 1,  
 conv_tr_wetscav                     = 1,   
 conv_tr_aqchem                      = 0,   

This covers both resolved and convective storms.

- Mary

^--^--^--^--^--^--^--^--^--^--^--^--^--^--^-^--^--^--^-^--^

Mary Barth 

Senior Scientist   

phone: 303-497-8186    email: bar...@ucar.edu

NSF - National Center for Atmospheric Research

P.O. Box 3000

Boulder, CO 80307

Staff web page

SIMA Web Site

MUSICA Web Site

DC3 Web Site 

^--^--^--^--^--^--^--^--^--^--^--^--^--^--^-^--^--^--^-^--^

[Zongrun Li]

Hi Mary,

Thanks. I checked the code

The aer_mech_id is set for zero for MOZART-GOCART and I think that's the reason I find this message in the log.

I revised my settings as you suggested, enabling wet scavenging and switching to MOZART-MOSAIC (attached is the namelist). Recently, I ran another one-month case (July 1 to August 16, 2017) and analyzed the average PM over this period. The PM deposition over the ocean seems reasonable, as the eastern coastal region shows lower PM levels. However, I still observe high PM concentrations in the western coastal region. I'm unsure whether this is realistic or if I made an error. I would appreciate any comments you might have.

Thanks,

Zongrun

[Gabriele Pfister]

Hi Zongrun

I think that the high PM2.5 just off the West Coast does not have to be unreasonable. It could be transport from the West Coast that is trapped in the shallow marine boundary layer or seasalt aerosols or even sometimes plumes from Asia. I cannot tell though how high the values really are because your scale only goes to just over 10 ug/m3. Have you looked at different aerosol components to understand which ones most contribute to the signal? 

We can see these high aerosol plumes also in our models on some days. For example, you can check out our global model forecasts here and our WRF-Chem forecasts here.

Gabi