Jump to main content or area navigation.

Contact Us

Atmospheric Modeling and Analysis Research

Multiscale Meteorological Modeling for Air Quality

Air quality models require accurate representations of air flow and dispersion, cloud properties, radiative fluxes, temperature and humidity fields, boundary layer evolution and mixing, and surface fluxes of both meteorological quantities (heat, moisture, and momentum) and chemical species (dry deposition and evasion). Thus, meteorological models are critical components of the air quality modeling systems that evolve with the state of science.

Because of this evolution, EPA must frequently challenge established models and configurations. This includes examining not only new physics schemes, but also data assimilation strategies which lower uncertainty in model output.EPA must also develop and refine physical process components in the models to address new and emerging research issues. Each of these research objectives has the overarching goal to improve meteorological model simulations to ultimately reduce uncertainty in air quality simulations.

EPA's meteorology modeling research program involves several key projects that have led to improved meteorological fields. The first is the transition from the MM5 mesoscale model system to the Weather Research and Forecasting (WRF) model that represents the current state of science. Part of this effort was to implement in WRF the land-surface (Pleim-Xiu; PX), surface-layer (Pleim), and planetary boundary layer (Asymmetric Convective Model version 2; ACM2) schemes that had been used in MM5 and are designed for retrospective air quality simulations.

Transitioning the mesocale modeling ssytem to WRF included improving the PX land-surface physics that included a deep soil nudging algorithm and snow cover physics that dramatically improved temperature estimations in the winter simulations and areas with less vegetation coverage. Additionally, EPA worked to implement the nudging-based four-dimensional data assimilation (FDDA) capability in WRF that had been available in MM5.

Another effort has been a re-examination of FDDA techniques, including use of an objective re-analysis package for WRF (“OBSGRID”) to lower the error of analyses used to nudge the model toward the observed state.

Current results of the implementation of new physics in WRF show that our configuration is comparable to or exceeds the level of MM5 in terms of uncertainty or error in near-surface variables like 2-m temperature, 2-m moisture, and 10-m wind as indicated in Table 1 below. This is true only when the new analysis package is used to improve analyses used for FDDA and soil moisture and temperature nudging in WRF. Figure 1 below shows RMSE differences between WRF and MM5 where both models were configured as similar as possible (i.e, PX LSM, ACM2 PBL, etc). The large number of dark blue and purple areas indicate WRF has a much lower temperature error than MM5.

A new evaluation method that utilizes both wind profiler and aircraft profile measurements provides a routine method to examine not only the uncertainty of simulated wind in the planetary boundary layer, but also the less examined temperature structure. The WRF model has low error in temperature (median absolute error of 1.0 to 1.5 K or less), wind speed (< 2.0 m/s) and wind direction (< 30 deg) in the planetary boundary layer, which is generally less than the error near the surface (Figure 2 below). The model also simulates the evolution of the wind structure, including features like nocturnal jets and the convective mixed layer (See Figure 3 below), with low error (<2.0 m s-1). Our current configuration of WRF has met the requirements for the transition from MM5.

Table 1. Summary of surface-based model performance statistics for each simulation. Also provided is the RMSE (2-m temperature only) of analysis dataset used for indirect soil moisture and temperature nudging of the PX LSM.
Summary of surface-based model performance statistics for each simulation. Also provided is the RMSE (2-m temperature only) of analysis dataset that was used for the indirect soil moisture and temperature nudging of the PX L

Spatially distributed RMSE difference (2-m temperature) between the WRF and MM5 for August 2006. Negative values indicate WRF has a lower error and positive values indicate MM5 has a lower error
Figure 1. Spatially distributed RMSE difference (2-m temperature) between the WRF and MM5 for August 2006. Negative values indicate WRF has a lower error and positive values indicate MM5 has a lower error.

Mean absolute error (MAE) profiles of model simulated temperature, wind speed and wind direction for August 2006. The observations used to compute MAE include 19 NOAA wind profilers located in the central United States
Figure 2. Mean absolute error (MAE) profiles of model simulated temperature, wind speed and wind direction for August 2006. The observations used to compute MAE include 19 NOAA wind profilers located in the central United States.

Diurnal mean wind speed profiles (height above ground level) for January and August 2006. The left column represents the mean observed wind speed computed using 19 NOAA wind profilers located in the central United States
Figure 3. Diurnal mean wind speed profiles (height above ground level) for January and August 2006. The left column represents the mean observed wind speed computed using 19 NOAA wind profilers located in the central United States.

Recommendations for using the Pleim-Xiu LSM, Pleim Surface-layer and ACM2 PBL in WRF-ARW

EPA has worked many years to develop meteorological model physics schemes designed for simulating weather in retrospect for use by air quality models. This effort began with MM5, but the focus is now the WRF-ARW model. We don't recommend using the PX LSM for forecast applications because it uses 2-m temperature and mixing ratio analyses to employ an indirect soil moisture and temperature nudging algorithm. In a forecast application, these fields would be a forecast rather than observational-based analysis or re-analysis. It is possible to run the PX LSM with the nudging scheme disabled, but very little testing has been done, so at this time it is not recommended.

EPA has done sensitivity experiments and helped develop a few tools that lower the error of PX LSM and ACM2 based model simulations. The following steps are recommended:

  1. In WRF Version 3.1 a surface analysis nudging scheme is implemented. This option requires the generation of a WRF surface FDDA file with the naming convention wrfsfdda_d01. This file contains near surface fields like 10-m wind, 2-m temperature, 2-m mixing ratio, snow and a few other fields. The PX LSM requires this file to perform the indirect soil moisture and temperature nudging. In the namelist.input file that specifies the model configuration options, the fields four-dimensional data assimilation (FDDA) should be set to one. The start and end time and analysis interval should also be specified. This configuration will pass the proper analyses to the PX LSM module, which automatically enables the soil nudging.
  2. At this time wrfsfdda_d01 file can only be generated using a tool called Obsgrid, which takes the base analysis and observations on the WRF grid and generates the reanalysis fields. Once the surface FDDA file is generated, the next step is to run real.exe to generate the 3-D FDDA file (wrffdda_d01), WRF input file (wrfinput_d01) and lateral boundary conditions (wrfbdy_d01). The namelist variable num_soil_levels should be set to two for the PX LSM.
  3. Our typical meteorological simulations cover months, seasons and in many cases an entire year. We spin up the PX soil model by starting 5-10 days before the period of interest. We run the model in 5.5 day run segments, so the model is completely re-initialized every five days. For the first run segment of the spin-up period we used a utility called IPXWRF (IPXWRF for WRFV3.0, WRFV3.1 and newer) to initialize the deep soil temperature as the average 2-m temperature from the wrfsfdda_d01 surface analysis file. We also set a namelist option (pxlsm_smois_init) in the wrf namelist (namelist.input) to one. This sets the deep soil moisture as a function of soil moisture availability and soil type.
  4. After the first run segment, the namelist option pxlsm_smois_init should be reset to zero. Also, IPXWRF should be reconfigured so both layers of soil moisture and temperature are re-initialized to values from the previous run segment. This keeps soil moisture, temperature and the nudging of these soil variables consistent throughout long simulations, which reduces the overall error of near surface meteorology.

Contacts: Jonathan Pleim, Tanya Otte, Robert Gilliam

Related Publications:
  • Gilliam, R. C. and J. E. Pleim, 2010, Performance assessment of new land-surface and planetary boundary layer physics in the WRF-ARW, J. Appl. Meteor. and Clim. in press.
  • Gilliam, R.C., C. Hogrefe, And S.T. Rao, 2006, New Methods For Evaluating Meteorological Models Used In Air Quality Applications, Atmospheric Environment, 40(26), 5073-5086
  • Otte, T. L., 2008: The impact of nudging in the meteorological model for retrospective air quality simulations. Part I: Evaluation against national observation networks. J. Appl. Meteor. Climatol., 47, 1853-1867.
  • Otte, T. L., 2008: The impact of nudging in the meteorological model for retrospective air quality simulations. Part II: Evaluating collocated meteorological and air quality observations. J. Appl. Meteor. Climatol., 47, 1868-1887.
  • Otte, T.L., A. Lacser, S. Dupont, and J.K.S. Ching, 2004: Implementation of an Urban Canopy Parameterization in a Mesoscale Meteorological Model. J. Appl. Meteor., 43, 1648–1665.
  • Pleim J. E., R. Gilliam, 2009: An indirect data assimilation scheme for deep soil temperature in the Pleim-Xiu land surface model. J. Appl. Meteor. Clim., 48, 1362-1376.
  • Pleim, J. E., 2007: A combined local and non-local closure model for the atmospheric boundary layer. Part 1: Model description and testing. J. Appl. Meteor. Climatol., 46, 1383-1395.
  • Pleim, J. E., 2007: A combined local and non-local closure model for the atmospheric boundary layer. Part 2: Application and evaluation in a mesoscale model. J. Appl. Meteor. Climatol., 46, 1396-1409.
  • Pleim, J. E., 2006: A simple, efficient solution of flux-profile relationships in the atmospheric surface layer, J. Appl. Meteor. Climatol., 45, 341-347.

Jump to main content.