To:		Mike McCorison, Dee Morse

From:		Howard Gebhart

Subject:	Review of Revised CALPUFF Modeling – Big West of California
LLC

Date:		February 6, 2007

	This reports provides a review of the updated Class I modeling impact
analysis for the proposed Big West of California LLC Clean Fuels Project
(CFP).  The CFP is designed to increase the production of gasoline and
diesel fuel that meets California fuel specifications at the Big West
refinery, located 2.5 northwest of Bakersfield, CA (formerly Shell Oil
Company).  Although there will be new process units associated with the
CFP, the project does not increase the crude oil throughput for the
refinery.

	The Class I modeling analysis addresses potential impacts at Dome Land
and San Rafael Wilderness (managed by USDA – Forest Service) as well
as Sequoia National Park (managed by National Park Service).  The
closest approach of each area is as follows:  Dome Land 82 km, San
Rafael 85 km, and Sequoia 107 km.

	The major changes from the March 2006 modeling analysis were to
increase nitrogen oxides (NOx) and sulfur dioxide (SO2) emissions from
the fluid catalytic cracking unit (FCCU).  The FCCU NOx emissions
increased from 202 to 404 lb/day 

(20 ppmv to 40 ppmv).  The FCCU SO2 emissions increased from 281 to 703
lb/day 

(20 ppmdv to 50 ppmdv).  The increased emissions reflect a higher
short-term emission rate, whereas the March 2006 CALPUFF modeling used
the proposed best available control technology (BACT) emissions rate,
which was based on the annual mean.  The FCCU is one of the larger
emission sources associated with the Clean Fuels Project.

	Other emission changes associated with the CFP included a small
increase in emissions at the ground flare, plus decreases in emissions
at the MHC Heater 14-H2 and the diesel fire water pump engine.  Also,
two small sources (utility boiler and elevated alky flare) were removed
from the modeling as these sources are no longer included with the
project.  

 	Other than the changes in emissions described above, the modeling
procedures appear for the most part to be unchanged from the March 2006
modeling analysis.  The Class I modeling followed the CALPUFF-Lite
screening procedure, based on five years of meteorological inputs from
the Bakersfield airport.  The new modeling used 1999, 2000, 2001, 2003,
and 2004.  2001 was omitted due to extensive “missing” data for wind
speed in the Bakersfield airport data.  The March 2006 modeling used the
five year period 2000-2004 (including 2001 with the apparent data
recovery problems.)  

All CALPUFF modeling inputs were consistent with the recommended default
values and Federal Land Manager (FLM) guidance.  The emissions inputs
included particle speciation and 10 ppb was selected for the background
ammonia concentration.  Visibility calculations used MVISBK = 6 (normal
for CALPUFF-Lite screening) along with the FLAG “natural background”
aerosol concentration values.  Particle speciation was incorporated by
using a 2.46 extinction coefficient for the PM mass, which represents
the weighted mean extinction for the various PM constituents.  Overall,
the modeling appears to have been done correctly and follows FLM
guidance.

	Modeling results are based on information submitted in a revised PSD
application from the applicant’s consultant (Ashworth Leininger Group)
dated December 2006.

	Based on the reported CALPUFF results, ambient concentrations of
modeled pollutants (NOx, PM-10, and SO2), were below the applicable
“significant impact levels” for all Class I areas modeled.  For the
SO2 modeling, all sulfur input to the system were modeled as SO2.

	I noticed some minor problems with the PM-10 modeling, but these errors
do not appear to have affected the modeling results and conclusions. 
First, the PM-10 inventory was modeled as six separate size classes in
order to allow for deposition of the PM-10 mass.  While this is
permissible, the deposition effect in the Big West CALPUFF study is
probably small, given that 82% of the PM-10 mass was modeled as 0.625
microns or smaller.  The deposition error occurred in that the largest
size was chosen as the “mass mean diameter” for each of the six size
classes (e.g., the PM mass for less than 

0.625 microns was all modeled at 0.625 microns diameter).  This would
tend to overestimate PM-10 deposition and as such, underestimate PM-10
concentration impacts.  The other PM-10 modeling error was that the
secondary sulfate and nitrate formation were not counted in the PM-10
mass for the increment modeling.  Nevertheless, the 

PM-10 concentrations were reported to be at most 25% of the applicable
SIL, so these minor errors were unlikely to change to conclusions
regarding PM-10 emissions and their impact on Class I PSD increment
consumption.  

The total nitrogen deposition values were less than the FLM deposition
analysis threshold (DAT) of 0.005 kg/ha-yr at all Class I areas modeled.
 Nitrogen deposition included all of the relevant pollutant species,
including deposition of ammonium sulfate.  

For sulfur deposition, the modeled impacts slightly exceeded the DAT,
with maximum annual predictions of 0.0056 kg/ha-yr for Dome Land and
0.0052 kg/ha-yr at San Rafael.  Dome Land showed sulfur deposition above
the DAT for three of the five years modeled and San Rafael was over the
sulfur deposition DAT for only one of the five years modeled.  For the
five-year average, the Dome Land sulfur deposition was 0.005054 kg/ha-yr
and the San Rafael sulfur deposition was 0.00473 kg/ha-yr.  The 

five-year average calculations shown above assume that the worst-case
receptor in each Class I area is the same for each year, which is not
necessarily the case.  The sulfur deposition was less than 0.005
kg/ha-yr at Sequoia in all five years.  

	The visibility modeling predicted a small number of days in excess of
the 

5% threshold (based on change in extinction coefficient) at each Class I
area modeled.  The frequency of impact was 3.2 days per year at Dome
Land, with a maximum change in extinction of 6.76%.  San Rafael modeled
2.4 days/yr above 5% change in visibility, with a maximum change in
extinction of 6.59%.  At Sequoia, the frequency of visibility impact
above 5% was less than 1 day per year, with a maximum change of 6.23%. 
The visibility modeling included emissions of primary sulfate, with the
SO2 emissions reduced slightly to account for the percentage of total
sulfur modeled as sulfate.  This adjustment is appropriate as it
maintains a sulfur balance in the emissions inventory.     

	Although the CALPUFF-Lite visibility modeling predicted a few days
above the 5% threshold for change in extinction, the frequency of these
exceedances is small 

(3 days per year of less) with a magnitude only marginally above the 5%
visibility impact threshold.  Likewise the sulfur deposition is
marginally above the 0.005 kg/ha-yr SIL. Although any decision regarding
the significance of such impacts is reserved to the FLM, other cases
where modeled impacts were slightly above the applicable SIL have been
judged to be insignificant.  In addition, the modeled impacts were all
computed using the CALPUFF-Lite screening procedure.  In all
probability, if a refined CALPUFF modeling study were performed, the
impacts would be lower and would fall within the established SILs. 
However, the FLM may still request a refined CALPUFF modeling study for
this project at its discretion.

	Please contact me at 970/484-7941 or by email at   HYPERLINK
"mailto:hgebhart@air-resource.com"  hgebhart@air-resource.com  if there
are any questions regarding this review. 

Howard  

