Comparison of Two-Way and Offline Coupled WRF-CMAQ

Dear all,

I am using the wrf-cmaq twoway model in CMAQ version 5.5.

I thought that the results from the two-way model with the direct shortwave feedback turned off should be identical to those from the offline model, but they are not. (with same wrf_input, icon, bcon, emission files)

I ran the model for January, and the PM2.5 concentration in the two-way model is significantly lower than that in the offline model.

Why are the two results different? Is there other chemistry-meteorology feedbacks in the two-way model besides the aerosol direct effect?

Additionally, I have confirmed that changes in clouds lead to a reduction in radiation. As far as I know, the two-way model does not account for aerosol indirect effects. How can we understand the changes in clouds?

@ycyeong –

I am not an expert on the WRF-CMAQ two-way model, but you should not expect to see the same results in the two-way model as you do in the sequential model because the meteorology in the two-way model is not a temporal interpolation between the hours (as occurs in the sequential model).

If you have not already done so, please see Wong et al., Geosci. Model Dev., 2012 (GMD - WRF-CMAQ two-way coupled system with aerosol feedback: software development and preliminary results). Figure 9b gets to your PM issue, even though older versions of WRF and CMAQ were used in the article than what you used.

–Tanya

1 Like

That hourly-interpolated meteorology is inadequate has long been documented,even before the existence of the coupled WRF-CMAQ.

A 2002 AWMA conference presentation on work sponsored by CARB showed that for a 4KM California domain, the model output differences between using interpolated hourly-timestep meteorology and interpolated ten-minute-timestep meteorology were quite significant: as much as 30ppb difference locally in modeled ozone. Differences were largest at the sunrise and sunset transitions: hourly sampling of meteorology is inadequate to simulate the non-linear changes in mixing behavior at these times.

Another high-temporal-resolution phenomenon modeled by the met-models is gravity waves; an hourly-resolution “snapshot” of gravity waves leads to a failure to model the meteorological dynamics correctly.

All of this is governed by the Shannon-Nyquist Theorem from Information Theory.

There is a considerable literature about the relationship between spatial resolution and the size of features the meteorology models can resolve; it is known that MM5 resolves features of about size 7*DX, whereas WRF does better, resolving features of size 5*DX (both of which are considerably inferior to the information-theory limit if 2*DX).

You can think of wind speed as a conversion factor between spatial resolution and temporal resolution, so a 10M/S top-of-PBL wind speed means that a 4KM spatial-resolution met model requires output temporal resolution of 400 S = (4000M / (10M/S) ) in order to have a hope of resolving the features modeled by the met-model. The effective 5*DX resolution of WRF means that you could probably get away with a (400S*(5*DX/(2*DX)) = 1000S temporal resolution to model most of the meteorological effects.

The better solution, of course, is to put the chemistry-transport model directly into the met model…

Hi ycyeong,

I assume when you ran the WRF-CMAQ and CMAQ offline model, they both used the same PE configuration (WRF or WRF-CMAQ sets it up automatically but on CMAQ offline side, you can choose your only. If the PE configurations are different, it could cause differences.
Another thing is temporal resolution as Tanya and Dr. Coats have both mentioned. WRF-CMAQ coupled model is an online model with a design to let user to decide how frequent the met information is being sent to CMAQ (typically in the provided script, it is set to 5:1, i.e. met information is sent to CMAQ every other 5 wrf time steps). Recently I have done a study to determine the impact of temporal resolution in WRF-CMAQ coupled model on air quality (O3 and PM25), but I haven’t had a chance to analyze the result.
Another point you need to realize is there is one wrf time step difference in the met information that sends to wrfout and to CMAQ.
Hopefully this gives you some insights on what your have observed.

Cheers,
David