EXERCISE 5

Exercise 5: Examining the WRF-Chem output for the South America domain

***************************************************************************

***************************************************************************

The purpose of this exercise is to provide a basic guide to new users in regards to the process of analyzing the model output and which tools to use. The more experienced WRF-Chem users, or those that have conducted mesoscale modeling studies in the past, might find this exercise to be very fundamental.

1. Now that your WRF-Chem simulation has finished, the time has come to examine the output. The first goal would be to verify that the simulation is good representation of the actual observed atmosphere. The second would then be to analyze the simulation data to determine what processes are taking place.

   A wide variety of programs can be used to post-process the WRF model data. As stated on the linked web page, these programs include ARWpost, NCL, etc., but one can use other programs as well such as GrADS. In fact, if one decides to use netCDF formatted output, there are many other programs that can be used to visualize and analyze the simulation results. The point being that a user is not limited to a single analysis/plotting application in order to view the simulation results. Use the tool(s) that you are most comfortable with in order to view the simulation results.

   
   

2. As the first goal indicates, before examining the chemistry solution, one needs to look at just the simulated meteorology. Does the simulation reproduce the observed behavior of the atmosphere for your simulation dates? Look at the different meteorology fields - (potential) temperature, pressure and winds for example. Make plots of the many different fields like surface (potential) temperature and mean sea level pressure. Compare these plots with the observations taken on the same day and in locations contained in your domain. The simmulation needs to reproduce the observations as closely as possible.

   The following image of surface temperature (K) and mixing ratio (kg per kg) at 00 UTC on 12 Sept. 2012 was made using the NCL program provided at the link at the bottom of this exercise.

   

   
   

3. Also look at the vertical structure of the atmosphere by plotting soundings (e.g., skew-T logP diagram). Compare the 00 and 12 UTC soundings with the simulation results. Focus on the lower troposphere and the boundary layer evolution. Capturing the evolution and vertical extent of the planetary boundary layer is very important as this impacts the mixing of chemical species as well as the transport.

   Often the boundary layer is not evolving correctly in the simulation domain. It could be the choice of boundary and layer parameterizations. It is well known that the YSU PBL parameterization can produce a PBL structure that is too warm and dry while the MYJ is the opposite. The HRRR operational model (3 km grid spacing for the 48 contiguous states of the USA) uses the more advanced MYNN PBL parameterization in the run. But some times one needs to focus on other fields - like soil moisture or snow cover - to more accurately capture the diurnal evolution of the atmosphere. For a multiple simulations one can look into cycling the soil related fields and see if that improves the boundary layer evolution.

   The following sounding is taken from the grid point near Sao Paulo, Brazil. The NCL program that produced these images is provided at the link at the bottom of this exercise.

   

   
   

4. So far the surface temperautre and diurnal structure of the atmosphere appears reasonable so far, then look at any of the resolved density driven circulations in the atmosphere (e.g., sea breezes, mountain flows). For this exercise, the 100 km grid spacing does not resolve features like the sea breeze so these fields will not be plotted.

   
   

5. With the meteorology in the simulation looking good, it is now time to examine the chemical species. One of the first items to examine is the emissions fields. For this, make a plot of CO (Carbon Monoxide). The background mixing ratio values for CO is roughly 0.08 ppmv in the WRF-Chem simulations. Urban locations, major highways and large fires should show up as locations of higher values ranging from 1 to .1 ppmv.

   The two images of Carbon Monoxide (co) at 00 UTC on 12 Sept. 2012 are shown below. The image on the left is for a simulation that does not include the MOZART initial and lateral boundary conditions. The image on the right is for a simulation that used MOZART data to provide the initial and lateral boundary conditions for the chemistry fields. Notice the difference of the peak values, the overall lower mixing ratio values and the strong gradient at the lateral boundaries when the MOZART GCM data is not included.

   

   
   

6. Other chemical species to examine, if they are included would be ozone and pm 2.5. In this exercise, it is assumed that you will eventually run with full chemistry and so the ozone and PM fields will exist.

   In your simulation the background ozone values vary with height with lower values in the troposphere (roughly .02 to .08 ppmv) at the surface and high values (1 to 10 ppmv) in the stratosphere. The values for ozone should vary during the day with higher values in the troposphere occuring in the afternoon and lower values over night and in the early morning.

   In the following two images the ozone values during the morning (03 UTC on 11 Sept. 2012) and the afternoon (15 UTC on 11 Sept. 2012) are shown. Notice that, despite the color scales being different in the two figures, the ozone values are higher during the afternoon hours.

   

   If ozone is too low, examine the emissions fields and verify that emissions for NO and NO2 are correct. Also examine the photolysis arrays (e.g., PHOTR4) and verify that photolysis is occuring in the simulation. If there is no photolysis, then check your namelist settings. If they are correct - photolysis is turned on - then examine the clouds as the meteorology for the simulation might have an issue and generated too many clouds. If PM is too small, then it is likely the emissions are not being included - especially if there is little PM10. For high values, it is possible there is a blow-up in one of the chemistry arrays. Start looking for a problem by examining the various aerosol fields in the Aitken and accumulation modes (e.g, sulfate is so4ai and so4aj, nitrogen is no3ai and no3aj, elemental carbon is eci and ecj)

   
   

7. For aerosols, one probably should examine the pm 2.5 (PM2_5_DRY)and pm 10 (PM10)diagnosed fields first. These will be the sum of all aerosol species in the Aitken and accumulation modes. Concentrations in the troposphere typically range from 0 to 100 micrograms per meter cubed. There can be much larger concentrations in locations of large fires, or during a significant pollution event.
   

   If the values for PM are too small, then it is likely the surface emissions are not being included - especially if there is little PM10. For high values where PM is well over 1000 micrograms per meter cubed, it is possible there is a blow-up in one of the aerosol chemistry arrays. Begin to diagnose the reason for the high values by looking for a problem by examining the various aerosol fields in the Aitken and accumulation modes (e.g, sulfate is so4ai and so4aj, nitrogen is no3ai and no3aj, elemental carbon is eci and ecj). Don't forget that some times there is a reason for the PM 2.5 values to be over 1000 micrograms per meter cubed. For example, a very large fire, or a large complex of fires can generate these high reading.

   
   

8. Now that one has examined a series of plots for the simulation days and confirmed the simulation appears to be generating reasonable results, one needs to do the same for observations. One can either generate plots overlaying the observations and compute fields like correlations, bias, etc., or instead use other utilities like the Model Evaluation Tools (MET). MET is designed to be a highly-configurable, state-of-the-art suite of verification tools. It was developed using output from the Weather Research and Forecasting (WRF) modeling system but may be applied to the output of other modeling systems as well.

This concludes WRF-Chem South America domain tutorial exercise 5

Return to WRF-Chem South America Domain Tutorial Page