Hi all,
I need to use WRF-CMAQ to simulate airpoluution for year 2030. So I wonder what scenario data I should use for both climate and airpoulation data ?
I want to run the model with spacial resolution of 1Km, and for 1 year (2030).
Please kindly advise me in these regards.
Thanks.
There is no simple answer as to what scenario you “should” run.
Many studies have been conducted using future climate scenarios to drive WRF and CMAQ, as well as WRF-Chem. For example, see Nolte et al., Atmos Chem. Phys. 2018, Nolte et al., J. Air Waste Manage. Assoc. 2021 (using WRF and CMAQ),Yahya et al., Atmos. Environ. 2017, Pfister et al., JGR 2014 (using WRF-Chem). There are many other such studies, including review articles (Fiore et al., JAWMA 2015; Fu and Tian, Curr. Poll. Rep. 2019). However, to my knowledge these studies are all at much coarser spatial resolution than 1 km.
You don’t say what the goal is of your application, but if you are trying to assess the impact of climate change on air quality, I would caution against relying on a single year simulation. There is substantial interannual variability in meteorology that affects air pollution concentrations. Just simulating one future year and one recent year, and attributing the difference to climate change, is not valid.
2 Likes
So many thanks for your explanations and for sharing the papers. Regarding the goal of my study, I need to mention that I am interested in predicting what AQ might look like in different future scenarios compared with each other and with the present day – typically simulating a year of air pollution (e.g. it might be 2030) for comparison with (e.g.) 2023 or 2024.
However, I need to know some issues to make sure this approach is practical.
-
If I want to run the WRF model for 12 months, is it better the run the model for short periods with 1-month intervals, or it is better to run the model continuously for 12 months? Which one provides more reliable results for meteorological variables?
-
What approach should I consider to model air pollution linked to future policy change?
-
You mentioned “I would caution against relying on a single-year simulation. There is substantial interannual variability in meteorology that affects air pollution concentrations”, so what is your solution in this regard? Would you please explain more?
-
Is it possible to run the CMAQ for 12 months continuously?
Please kindly advise me in these regards.
Best wishes,
You are asking some good questions here, but they do not all have definitive answers.
The effect of changes in air pollutant emissions is generally larger than the effect of climate change, especially over the near term. When comparing 2023 or 2024 to 2030, most practitioners, at least at EPA, would keep the meteorology unchanged in the future year and model the effect of the future emissions scenario.
Different researchers have employed different approaches for downscaling global climate model simulations with WRF. Our group has chosen to use long-term continuous simulations with nudging (Bowden et al., J Clim 2012; Otte et al., J Clim 2012; Spero et al., JAMC 2015). Others have not used nudging but have instead used frequent reinitializations. There is by now a fairly large and growing body of research on regional climate downscaling. Several review articles have been published, and there are model intercomparison activities (e.g., CORDEX, NARCCAP).
I use CMAQ driven by WRF after it has been processed by MCIP. In other words, the meteorology is “offline” or “uncoupled” from the simulated air pollution. That would not be the case with WRF-Chem. I have simulated continuous decadal (actually 11-year) periods, subset from the multidecadal WRF simulations our group has conducted. These are at relatively coarse (36-km) spatial scale. Others have chosen to sample specific days for modeling, e.g., Mahmud and Kleeman, ACP 2012).
How to account for interannual variability (e.g., Fiore et al., JGR 2022) while keeping the modeling computationally tractable is a challenge.
It is theoretically possible to run CMAQ for long periods continuously. In practice, most researchers at EPA (including myself) execute the model for 24 hours at a time, for convenience in scripting and input data preparation. In these runs, the model state at the end of one day is the initial condition for the next, so we consider these to be continuous simulations.
2 Likes
So many thanks for your explanations. I need to read more about the issue, and then I will come back again to discuss the issue further.
Best wishes,