Meteorological Modeling for the August 2000 Houston-Galveston Ozone Episode: PBL Characteristics, Nudging Procedure, and Performance Evaluation
Abstract
This report describes evaluations of the performance of various configurations of
the MM5 modeling system, as compared to planetary boundary layer (PBL) structure and
profiler winds. Soundings from the three sounding sites are grouped by time of day and
by regime. Systematic differences between the different model runs are found; the
differences between the models and observations vary from site to site. Higher vertical
resolution did not produce improved boundary layer structure. The MCNC runs had
well-mixed PBL’s, even at night, but were too shallow during the day. Most of the runs
with the MRF PBL were similar and performed fairly well. One area of possible concern
is the systematic underestimate of the strength of the sea breeze inversion, an error which
may lead to too much diffusion of constituents into and out of the advancing marine air.
The Gayno-Seaman PBL scheme appeared to be more realistic, but its sea breeze
inversion was too strong.
Wind errors were computed at a variety of heights, grouped by weather regime, and
with 24-hour running means and departures from running means. Most of the model
error was associated with the departures from the running means. The MRF PBL
schemes tended to perform best overall. All model runs except MCNC developed large
biases at heights above 1 km. The MCNC run was worst during Regime 1 but was best
during Regime 2 when other model runs produced only a small fraction of the observed
low-level jet speeds.
Based on these and previously-reported evaluations of various model
configurations, a particular configuration was chosen. This configuration uses the MRF
PBL with 43 vertical levels and one-way nesting. The soil moisture availability is
specified to decrease during the model integration, to simulate evaporation of rain that
fell just prior to the ozone episode. A new subroutine was added to the MM5 to permit
model restarts with updated soil moisture.
The nudging strategy is then outlined. The approach followed here uses a large
time window for nudging so as to effectively average out possibly erroneous hour-to-hour
variations in low-level winds that were introduced during the quality assurance process.
The default value for nudging strength is used. No nudging is performed prior to August
25 because the convection on the previous days are not resolved by the profiler network
and any attempt at nudging would cause aliasing in the model fields.
The final model run, called the “driver” run, also utilizes lower-tropospheric
nudging of water vapor on the 12 km grid. This nudging is designed to suppress a robust
outflow boundary which sweeps through Houston on August 31 in the model forecasts
but not in the observations. The nudging successfully prevents convection from
developing on the 4 km grid and reinforcing the outflow, but a weak wind surge does
reach Houston during the evening.
The thermodynamic performance of the driver run is very similar to its predecessor
runs, since no nudging is applied to the temperature field. The wind field is dramatically
improved, at least by comparison to the profiler data. Since this data set was used to
nudge the model in the first place, further objective verification of the improvement due
to nudging is necessary.
The wind and temperature fields during the high ozone days of the episode are
examined in detail. On two of the days, August 30 and 31, the model wind and
3
temperature fields have realistic large-scale and small scale features and further
improvement is unlikely. Three other days, August 26, August 29, and September 1, are
generally accurate but have wind errors which are likely to lead to position errors in the
simulation of ozone by a photochemical model. The remaining day, August 25, had
erroneous or too-light surface winds. On this day, high values of ozone are likely to be
simulated by a photochemical model, but it is possible that the mixture of ozone
precursors that leads to the high ozone will be fundamentally different due to the
transport errors.
In summary, the driver model run produces generally accurate daytime lowertropospheric
temperatures and winds. On most days of the episode, the meteorological
fields appear to be adequate for driving the particular combination of mixing and
chemical processes that lead to high ozone on each of those days.
Department
Atmospheric SciencesCollections
Citation
The following license files are associated with this item: