Documentation of the Evaluation of CALPUFF and Other Long ...

2000a). SCIPUFF also has its own simplified mass‐consistent wind field processor referred to as
MC‐SCIPUFF (Sykes et al., 1998). Gridded meteorological fields are normally supplied to
HYSPLIT and FLEXPART using software that converts prognostic meteorological data into
formats that are directly ingested into the respective dispersion models. The CAMx model also
uses software to reformat output from a prognostic meteorological model into the variables
and formats used by CAMx.
Use of a diagnostic wind field model (DWM) as the primary method to supply meteorological
data to the dispersion models under review creates additional uncertainty in the
intercomparison of the five dispersion models. DWM’s, such as CALMET, have the ability to
ingest prognostic data from models such as the PSU/NCAR MM5 (Grell et al.,1995) or the
Advanced Research Weather Research and Forecasting (WRF‐ARW) (Skamarock et al. 2008) as
its first guess wind field. However, this method of using the prognostic meteorological data as
the first guess field for the DWM does not preserve the integrity of the original meteorological
field. For example, the CALMET DWM adjusts the wind fields for kinematic and thermodynamic
effects of terrain and also rediagnoses key meteorological parameters such as planetary
boundary layer heights. Thus, to conduct a proper evaluation of the dispersion models on the
same basis, each of the models should be operated with the same meteorological dataset. In
order to maintain consistency with this study objective, it would not have been appropriate to
use either MC‐SCIPUFF or CALMET to produce three‐dimensional meteorological fields for their
respective dispersion model.
In order to facilitate direct intercomparison of models using a common prognostic
meteorological dataset, it is necessary to supply meteorological fields to CALPUFF and SCIPUFF
in the same manner as the particle models and grid model included in this study. SCIPUFF has
the ability to ingest prognostic data sets directly in either MEDOC (Multiscale Environmental
Dispersion Over Complex terrain) (Sykes et al., 1998) or HPAC formats. The Pennsylvania State
University developed the MM5SCIPUFF utility program (A. Deng, pers. comm.) to convert MM5
fields into the MEDOC format which is directly ingested into the SCIPUFF. Similarly, the US EPA
developed the Mesoscale Model Interface (MMIF) software to convert MM5 fields into the
CALPUFF meteorological input format (Emery and Brashers, 2009). With these two utility
programs, it was now possible to evaluate the five LRT models using a consistent set of
meteorological inputs.
Due to the inherent differences that exist between each of the five LRT models, it was not
possible to standardize dispersion model options. Rather, options selected for each class of
models were similar to the extent possible. For example, more advanced model features
(turbulence dispersion, puff splitting) were used for CALPUFF simulations as these represent
the state‐of‐the‐practice for puff dispersion models and are most consistent with the
capabilities of the SCIPUFF modeling system, helping to facilitate greater inter‐model
consistency for this evaluation.
CALPUFF is typically only recommended to distances of about 300 km or less (EPA, 2003). This
would effectively limit the useful range of CALPUFF to the first 24‐36 hours of ETEX simulation.
However, recent enhancements to the CALPUFF modeling system include both horizontal and
vertical puff splitting, incorporating the effects of wind shear on puff growth, potentially
allowing for use of CALPUFF at distances greater than the nominal recommended limit of about
300 km, and allowing for more direct intercomparison with the two particle models and one
grid model used in this study which are free of this restriction. The default method for CALPUFF
vertical puff splitting is to allow for splitting to occur once per day by turning on the puff
105