The purpose of this workshop was to to produce a community assessment of approaches and methodologies used for chemistry modeling in cloud and mesoscale models. Modeling issues that arise in coupled chemistry-dynamical models were examined for the purpose of providing guidance for the development of the dynamical and chemistry components of the WRF model. Listed below are brief summaries of the presentations and discussions. As a result of the workshop, a working group for the Weather Research Forecast (WRF) chemistry model is now being formed. Peter Hess is the head of the working group, and people interested in joining the working group should contact Peter Hess or Bill Skamarock (hess@ucar.edu, skamaroc@ucar.edu). Information on the WRF model and its development can be found at http://wrf-model.org
1. Dynamics
Two issues were the focus of discussions related to model dynamics: Mass conservation and advection schemes.
The candidate dynamics kernels for the WRF model (using both Eulerian and semi-Lagrangian techniques) are fully nonhydrostatic, fully compressible, and are formally mass-conserving.
Eulerian prototypes for the WRF model dynamics solver have been constructed within the WRF software framework using upwind and centered advection schemes from 2nd to 6th order accuracy. Features of advection schemes that are desirable for chemistry applications are high order accuracy, positive definiteness, and monotonicity. The advection schemes in the prototypes are not positive definite nor monotonic, and as such may not be ideal for some chemistry applications.
It was suggested that one or two schemes be supported that are well
suited to advection of chemical species. Experts in the atmospheric
chemistry community who may be able to help guide this effort are Richard
Easter (Pacific Northwest National Laboratory), Stuart McKeen (NOAA/AL),
Daewon Byun (NOAA/EPA), Talat Odman (Georgia Institute of Technology),
Piotr Smolarkiewicz (NCAR), Ricky Rood (NASA/GSFC), Michael Prather (UC-Irvine),
and Philip Rasch (NCAR). Current advection schemes that are
being used by the atmospheric chemistry community are the modified Bott
scheme, the piecewise parabolic method (PPM), and others.
Plans: Evaluate candidate advection schemes and implement/test/support 1-3 of the most promising schemes.
2. PBL Parameterizations
Several general types of PBL schemes were discussed; 1) Richardson-number dependent K models, 2) K-profile models, 3) Single-point (Higher-order) closure models, 4) Mixed layer models, and 5) Multi-stream exchange models (including mass-flux approaches and transilience theory).
PBL scheme capabilities important for chemistry applications include accurate prediction of the PBL height, inclusion of counter-gradient and possibly non-local transport, and the ability to predict the covariance between two species. It was also recognized that the parameterization should depict the interaction between the PBL and shallow convective and cumulus clouds, which is not included in existing atmospheric models, and should be coupled to the biogenic emissions and dry deposition. The PBL parameterizations that might be appropriate for chemistry are the multi-stream schemes and the higher order closure schemes, although the ability of any of these schemes to fulfill the atmospheric chemistry needs is an open question. The PBL scheme that is being ported to the early WRF prototypes is the MRF scheme (Hong and Pan, 1996 MWR, 2322-2339, used in the NCEP Medium Range Forecast (MRF) model; it is a K-profile model with nonlocal diffusion and countergradient effects). Other schemes (yet to be determined) will be ported at a future time.
Plans: Work with the WRF physics working group to identify and port PBL schemes that will address the chemistry modeling needs, and implement species transport within the appropriate PBL schemes.
3. Convective Parameterizations (deep and shallow convection)
The needs of the atmospheric chemistry community are (1) to have an estimate of updraft and downdraft mass fluxes and associated entrainment and detrainment, (2) to have an estimate of the amount of condensate and its evaporation, and (3) to have an estimate of wet deposition. It is also desirable to have schemes that are easily modified and easily swapped. Examples of schemes that possess most of these properties and that are already being used in coupled meteorology/chemistry models are versions of the Arakawa-Schubert scheme, the Kain Fritsch scheme, and the Grell scheme. Another scheme to consider is Bechtold's scheme (Bechtold's scheme can be obtained from ftp://ftp.aero.obs-mip.fr/pub/salsa/becp/convect/, and is used in the MESONH model: http://www.aero.obs-mip.fr). There was also discussion of the need for test cases to evaluate tracer transport within the convective parameterizations.
The convective parameterization that is being ported to WRF is the the Kain-Fritsch scheme. Other schemes (likely the Grell scheme and others yet to be determined) may be ported in the future.
Plans: Work with the WRF physics working group to identify and port convective parameterization schemes that will address the chemistry modeling needs, and implement species transport within appropriate schemes.
4. Cloud Microphysics Parameterizations
It was suggested that a 3 class ice microphysics parameterization be used for model simulations that resolve convection (horizontal resolutions of a few kilometers or less). However, if the convection is parameterized, a 2 class ice microphysics parameterization should be sufficient. It is clear that ignoring ice will produce errors in the spread of clouds and in precipitation rates.
An explicit microphysics parameterization or a double-moment bulk water microphysics parameterization should be used for aerosol-cloud interactions.
The WRF model prototypes currently have two microphysics schemes implemented, a Kessler-type microphysics (warm rain only) and a 3 class ice microphysics (Lin et al). Other microphysics will be ported to WRF in the future, including double-moment type schemes and other schemes that are more sophisticated than the 3 class ice microphysics schemes. This is an active area of research within the dynamics community, so a variety of schemes should become available.
Plans: Work with the WRF physics working group to identify and port
microphysics schemes that will address the chemistry modeling needs, and
implement species transport (and possibly aqueous chemistry; see topic
5 below) within some suitable subset of schemes.
5. Cloud Effects on Chemistry
It was agreed that NOx production from lightning should be parameterized in WRF with chemistry. For simulations that do not resolve convection, this parameterization could be combined with the convectiveparameterization. For simulations that do resolve convection, the parameterization of NOx production from lightning should be consistent with the microphysics parameterization.
Wet deposition should be included with the resolved clouds to represent the flux of chemical species in the precipitating hydrometeors. The depiction of wet deposition in parameterized clouds should be estimated in the convective parameterization scheme using the flux method.
Some form of aqueous chemistry should be included in WRF with chemistry. At a minimum, aqueous oxidation of SO2 by H2O2 and O3 should be included. Other aqueous chemistry mechanisms that could be included are those outlined in RADM (Chang et al, 1986), Walcek et al. (1997), Liu et al. (1997), or the user's choice.
The effect of ice on chemistry, at this stage, should be considered as a research project and no mechanisms should be included in WRF with chemistry at this point.
Currently, there is no chemistry (of any kind) in the WRF model
Plans: (1) Identify appropriate lightning-NOx parameterizations and port them to the WRF model. (2) Evaluate and implement appropriate methods for wet deposition in the WRF model in the resolved precipitation and in the convective parameterizations.
6. Chemical Solvers and Mechanisms
It was suggested to have one complex technique and a few simpler numerical techniques for solving the chemistry. Solvers suggested for the model are SMVGear, RODAS, the Euler Backward Iterative, and themethod of Hertel et al. (1993). The efficiency of some of the solvers might be improved by the use of some techniques employed in SMVGear.
It was also suggested to have 3 base mechanisms, Carbon Bond IV (regionalized version from PNNL), SAPRC mechanism - 2000 (detailed), and RACTM version 2 which was developed for the regional scale (Stockwell). A very simple mechanism could be used for non-chemistry studies. In the long-term, it would be preferable to use detailed explicit (master) mechanisms.
The implementation of chemical compilers was also discussed. Chemistry preprocessors could be contributed from several groups (P. Hess, NCAR; W. Stockwell, DRI; C. Mari, CNRS; J. Young, EPA; and H. Jeffries, UNC).
Currently, there is no chemistry in the WRF model
Plans: The WRF chemistry working group should identify, a) 1-3
chemistry solvers that could be implemented, b) 1-3 chemistry mechanisms,
and c) a preprocessor suitable for the model.
7. Emissions
The biggest hurdle for emission inventories is retrieving source level data bases. The North Carolina Supercomputing Center has developed a tool (Sparse Matrix Operator Kernel Emissions, SMOKE which can be found at http://envpro.ncsc.org/products/smoke/) to digest the data bases and produce appropriate model input.
It would be desirable to develop a detailed geographic information system (GIS) data base. This data base could be updated every few years to capture the changing use of the land.
As a default, it is suggested to have a global data base obtained from GEIA or EDGAR. However, the details, such as the location of the point sources, the plume heights, and the speciation of the chemical species, of the emissions are needed. This information is fairly well developed for North America, but is poorly characterized in other regions of the world. More funding is needed to improve our emissions databases.
Biogenic emissions should be based on meteorology and should interact with the dry deposition parameterization. Episodic events, e.g., fires, should be included.
Currently there are no emissions of chemical species in the WRF model.
Plans:
8. Dry Deposition
The land surface and vegetation description in WRF will be fairly sophisticated (NCEP, FSL, and others each have a scheme). There needs to be consistency between the biogenic emissions, dry deposition, surface moisture, turbulence, and chemistry parameterizations in the description of the planetary boundary layer. Biogenic emissions from both vegetation and soil should be considered. Likewise, dry deposition of chemical species onto different types of vegetation needs to be parameterized. In forest canopies, the extinction of radiation will limit the amount of biogenic emissions and dry deposition because these processes are linked to photosynthesis. Similarly the photochemistry in a forest canopy is altered by decreasing photodissociation rates in the lower parts of the canopy. Turbulence in forest canopies is important to depict as it controls the intermittent exchange of chemical species between the canopy and planetary boundary layer, in that these species are released in bursts to the PBL. To accurately describe dry deposition and other processes in the boundary layer, there should be a depiction of the interaction between the turbulence of the planetary boundary layer (section 2), the biogenic emissions (section 7), the radiation treatment in forest canopies, and the dry deposition. Also, a crucial factor in the whole representation of the atmosphere-biosphere trace gas exchange is the characterization of the land cover and land use in terms of the fractions of different surface cover types and its characteristics, e.g., of Leaf Area Index, canopy height, surface roughness. A consistent use of these land cover and landuse properties for trace gas exchange and other micrometeorogical processes is important because of the strong link between the atmospheric chemistry and physical and dynamical processes in the surface layer and the PBL.
Plans:
9. Radiation
The look-up table method of calculating photolysis frequencies is sufficient for clear-sky conditions. However, when clouds and aerosols are considered, the calculation of actinic fluxes and photolysis frequencies should be performed in the WRF and chemistry model.
It was also emphasized that there needs to be consistency with the actinic flux (and therefore photolysis frequency) calculation and the heating rate calculation.
Calculation of the cloud fraction is very important for photolysis rates, heating rates, and aqueous chemistry.
Plans: Communicate need for cloud fractions to the WRF physics group. Evaluate possibilities for producing consistency between the actinic flux calculations and the heating rates.
10. Aerosols
Several methods are used to depict the size and composition of aerosols in models. The modal approach assumes a mathematical function for the size distribution and is a ``simple" method. The quadrature methodof moments and the sectional approach are more complex methods. The method used may be a choice of the user.
As a default, is it helpful to have aerosol mass, size, number, and composition determined from a data file (derived from another model, or measurements).
Plans: As needed, aerosol models can be implemented within WRF. This is very much an area of research, and multiple aerosol models may be implemented for different needs and for evaluation. The coupling of aerosols in radiation and microphysics within the WRF model should be explored.
11. Near-term Activities
Prioritize and move forward with the plans outlined in thesummaries above.
Determine what topics need funding and who is interested in participatingin implementing parameterizations and WRF modifications.
A working group for WRF chemistry is now being formed. Peter Hess isthe head of the working group, and people interested in joining theworking group should contact Peter Hess or Bill Skamarock (hess@ucar.edu, skamaroc@ucar.edu).
It was suggested that a single column version of WRF be developed for
testing of new codes and for other studies. This need was also expressed
in the recent WRF planning workshop (held on 29-30 March 2000) independent
of the chemistry needs, and one will likely be developed after decisions
concerning which WRF prototype to use have been made.