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

Based on this analysis the puff‐particle and particle‐puff hybrid configurations of the HYSPLIT
system are clearly the best performing, indicating a distinct operational advantage over pure
puff or particle configurations.
6.5 CONCLUSIONS OF THE MODEL PERFORMANCE EVALUATION OF THE LRT DISPERSION
MODELS USING THE ETEX TRACER EXPERIMENT FIELD STUDY DATA
The evaluation of the five LRT dispersion models using a common MM5 dataset and the ETEX
database has provided interesting results about the current capability of LRT models to
reproduced observed tracer concentrations. Four of the five LRT models were able to
reproduce the observed tracer bifurcation at the farther downwind distances. The CALPUFF
model was unable reproduce the observed bifurcation of the tracer cloud and kept the
estimated tracer cloud in a circular Gaussian distribution that was advected too far north.
CALPUFF puff splitting sensitivity tests were performed to determine whether it would help
simulate the bifurcation of the tracer cloud but puff splitting had little effect on the CALPUFF
predictions.
CAMx sensitivity tests were conducted to examine vertical mixing and horizontal advection
solvers and the best performing CAMx model configuration was the one that is most frequently
used in applications, which includes using the CMAQ‐like vertical diffusion coefficients in
MM5CAMx and the PPM advection solver. The vertical diffusion algorithm had a much bigger
effect on CAMx model performance than the choice of horizontal advection solver.
The HYSPLIT sensitivity tests with different particle‐puff variations resulted in a wide range of
model performance with RANK scores that varied from 1.01 to 2.09.
141