Figure 1: Emission sources (LA - Los Angeles, SD - San Diego, SF - San Francisco, LV - Las Vegas, PH - Phoenix, SL - Salt Lake City) and receptors (SPMO - Spirit Mountain, MEAD - Meadview, LOME - Long Mesa) discussed in the paper.
September 1997
Colorado State University RAMS (Regional Atmospheric Modeling System)
( Pielke et al.,1992) was used to simulate meteorological fields in
southwestern United States over the entire year of 1992 in connection with
the MOHAVE project
( Uliasz et al.,1996).
These daily simulations were
performed on two nested grids with a horizontal grid spacing of 60 and 12 km
respectively. The second grid covers the vicinity of the Grand Canyon National
Park and most sources and receptors considered in the MOHAVE project and
discussed in this paper (Fig.1). The daily RAMS simulations did not take
advantage of special meteorological measurements available during the MOHAVE project.
Unfortunately, the daily RAMS simulations were not followed by higher
resolution simulations which could better resolve the extremely complex
terrain topography of the region.
The RAMS meteorological fields were used as input for the Lagrangian Particle Dispersion (LPD) model (Uliasz,1993, Uliasz,1994) to run a variety of passive tracer simulations in both source- and receptor-oriented modes ( Uliasz et al.,1996). The LPD model was also coupled with another meteorological models used in the MOHAVE project. This paper describes the results of validation of the RAMS/LPD modeling approach applied to atmospheric transport from the Los Angeles Basin to the Grand Canyon region with the aid of methylchloroform as a tracer of opportunity.
Methylchloroform is entirely anthropogenic and is released primarily in the metal fabrication, electronics and aerospace industries. Most emission sources of methylchloroform in southwestern US are concentrated in the Los Angeles area. White et al. (1990) examined methylchloroform data at Spirit Mountain (SPMO), Meadview (MEAD), and Cajon Pass on the edge of the Los Angeles Basin during June-August 1985-1987. They found that summertime concentrations at SPMO and MEAD show strong weekly cycles with the lowest values typically from late Sunday to Tuesday and lag similar cycles observed in the Los Angeles Basin by 1 to 2 days. This observed pattern is caused by a nearly complete shutdown of methylchloroform emission during weekends. These findings imply that methylchloroform may be used as an endemic tracer of the Los Angeles area. Contributions from other urban areas are expected to be much lower. Therefore, several studies of long range transport in southwestern United States have used methylchloroform as a tracer of opportunity Miller et al.,1990, Pryor and Hoffer, 1992, Pryor, 1995.
Our validation study was performed for a period of six months from May to
October, 1992 during which air pollution transport from the Los Angeles Basin
to the Grand Canyon area was frequently expected. Hourly concentrations of
methylchloroform were measured during this period at
three locations in the Grand Canyon area: Spirit Mountain, Meadview
and Long Mesa (LOME). Particle simulations were carried out backward
in time for these three receptors and the corresponding influence functions
were calculated in one hour sampling intervals. No attempt was made to estimate
the emission rate of methylchloroform from the Los Angeles area or other potential
sources. A contribution from a given source was calculated in terms of concentration
normalized by emission rate. A complete shutdown of emission during weekends
was assumed. Each source was approximated by a 50x50x0.5 km box.
Figure 2: Methylchloroform concentrations above background by hour of day and day of the week at three receptors during May-October, 1992: observed values (red) and the simulated contribution from the Los Angeles Basin with (green) and without (blue) weekend emission. Simulated normalized concentration is multiplied by the factor of 0.25x10^12.
Figure 2 presents observed methylchloroform concentration above
background by hour of the day and day of the week during the analyzed six month period.
The background concentration was estimated using the 25th percentile concentration
Pryor, 1995.
The regular weekly cycles with the lowest values on Tuesdays are evident in
the observed data. The amplitude of these cycles is the highest for the SPMO
receptor which is the closest one to the Los Angeles Basin and is
significantly less for the MEAD and LOME receptors located farther east.
The simulated Los Angeles contribution with the emission shut down during weekends
reproduce the weekly cycles of the observed data quite well.
For a reference, the Los Angeles contribution
simulated for the constant emission is also presented.
It is interesting to note that the effect of zero emission during Saturdays and
Sundays is evident for a large part of the week. This suggests that the transport
time between Los Angeles and the considered receptors is much longer than two days
due to the effect of mesoscale circulations.
Figure 3: Time series of the observed methylchloroform concentration at SPMO (red) and time series of corresponding percentiles (blue)
In order to directly compare the observed and simulated methylchloroform
episodes, the time series of one hour values were smoothed using a 24-hour
running average. Next, frequency distributions were calculated for each time
series for the 6 month period and, finally, the concentration values in time
series were replaced by their corresponding percentiles with values varied
between 0 and 1. Figure 3 illustrates this procedure by
comparing the methylchloroform concentration time series observed at SPMO with its
corresponding percentile time series. This procedure enhances the appearance of
the concentration episodes.
Under the assumption that the Los Angeles Basin is the only significant
source of methylchloroform, most of the observed methylchloroform episodes are well
reproduced by the model at all three receptors although the model performance
is less accurate at MEAD
( Uliasz et al.,1996).
Correlation coefficients between the observed and simulated percentiles of
concentration at all three receptors for each month and for the entire period
are presented in Figure 4 . The correlation coefficients are evidently lower for
MEAD than for the two other receptors, probably due to the fact that MEAD
is located in more complicated terrain at the mouth of the Grand Canyon.
The correlation coefficient is not the best measure to characterize agreement
between model predictions and observations. Several small time lags between
simulated and observed methylchloroform episodes significantly contribute to
a low correlation. An interesting alternative to
compare model results with observations is offered by spectral analysis (e.g.,
Sirois et al.,1995). Coherence derived from a cospectrum
calculated for two time series varies between 0 and 1 and may be interpreted
as a correlation between these time series expressed as a function of
frequency or wavelength. The coherence between the simulated and observed time
series of methylchloroform concentration varies significantly with wavelength
(time scale) but has similar features for each receptor site (Fig.5).
All cospectra show that simulated time series are to some extent coherent with the
observed ones for the time scale longer than about three days. For the shorter
time scale, the cospectra become very noisy. The maximum coherence (between
0.6 and 0.8) appears at time scales between 150 and 180 hours at SPMO and
MEAD. This corresponds to the weekly cycles presented in Figure 2.
For LOME this maximum is not so well marked and is shifted to longer time scales.
There is also another maximum in all three coherence spectra corresponding to a time
scale of about four days.
Figure 6 presents coherence spectra calculated for the concentration
time series at two receptor sites: SPMO and MEAD. The coherence between simulated
time series is much higher than between the observed values. It indicates that
the model does not take into account some features, like local topography,
which distinguish MEAD from SPMO. A higher model resolution is required to
correctly simulate concentration affecting MEAD than is necessary for the
other receptor sites.
In the above analysis, the Los Angeles Basin was treated as the only significant source of methylchloroform. In order to examine the validity of this assumption, potential contributions from five other urban areas were determined ( Fig.7). Due to transport patterns, some of these areas, e.g., Phoenix, may have high potential contributions to concentrations at the receptors. However, concentration predictions for these sources show no correlation with observations. On the other hand, the predicted contribution from San Francisco shows a positive correlation with measurements, but the potential impact of this source is an order of magnitude lower than the contribution from Los Angeles. Due to lack of transport patterns from the north there was practically no impact from the Salt Lake City area. Therefore, one can assume with a good approximation that methylchloroform can be used during the analyzed period as an endemic tracer for the Los Angeles Basin. However, it should be noted that the potential contribution from Las Vegas shows some correlation with the observations at MEAD. Because of its proximity to the receptors of interest, the role of this source area should be more carefully examined with the aid of higher resolution modeling. It is also interesting to see that the model can distinguish between contributions from closely located source areas like Los Angeles and San Diego.
Despite necessary compromises in the design of long term RAMS simulations, the results from validation of the RAMS/LPD modeling with the aid of methylchloroform observations are encouraging. We were able to correctly reproduce the regional scale transport between the Los Angeles Basin and the Grand Canyon region. Performance of the RAMS/LPD modeling is evidently better than the performance of simpler modeling approaches applied to the same region ( Uliasz et al.,1996). However, it is also obvious that the numerical resolution of RAMS in the daily simulations was not sufficient to resolve all important mesoscale circulations related to complex terrain features.