EPA/600/A-97/099
9A.14 EFFECTS OF UNCERTAINTY IN METEOROLOGICAL INPUTS ON
03 CONCENTRATION, 03 PRODUCTION EFFICIENCY, AND 03 SENSITIVITY TO
EMISSIONS REDUCTIONS IN THE REGIONAL ACID DEPOSITION MODEL
Yonghong U*. Robin. L Dennis, Gail S. Tonnesen, Jonathan E. Pleim, and Daewon Byun
Air Resources Laboratory, Atmospheric Sciences Modeling Division, NOAA, Research Triangle Park, NC
1. INTRODUCTION
Photochemical models are widely used to
determine the level of VOC and NOx emissions
reductions required to attain the National Ambient Air
Quality Standard for ozone (03). Modeled 03
concentrations, [03], are affected by photochemical
production of 03, by NOx and VOC precursor emissions,
and by meteorological processes that include horizontal
transport, vertical mixing, cloud effects on actinic flux, and
deposition of 03 and its precursors. There is
considerable uncertainty in simulating each of these
processes individually and in simulating the effects of
interactions or feedbacks between chemical and
meteorological processes; thus, it is possible for models
to correctly reproduce observed [03] because of
compensating errors in model inputs or model processes.
This puts into question how much confidence we can
have in model predictions of the effectiveness of VOC
and NOx emissions reductions. Despite the importance
of meteorology, there has been little experience within the
air quality research community in quantifying
meteorological impacts on photochemistry, particularly
with the use of primitive equation meteorological drivers.
In this study, we conduct a series of sensitivity
experiments with the Regional Acid Deposition Model
(RADM) using changes in meteorological inputs, such as
PBL height and cloud effects, to examine their impact on
[03] and on 03 production efficiency. Using each of
these modified meteorology scenarios as a "base case"
we then evaluate the effect of meteorology on the
effectiveness of emissions controls by running additional
model simulations with 15% and 50% reductions in NOx
or VOC emissions.
2. THE BASE RADM EXPERIMENT
The RADM simulations were conducted over the
eastern US for the time period of July 19 to August 12,
1988, at a resolution of 80km on a three-dimensional grid
of 35 by 38 horizontal elements and 21 vertical layers.
Thirteen layers are in the PBL with the lowest layer
extending to approximately 40m above the surface. The
physical and chemical processes included in RADM are
* Corresponding author address. Yonghong Li,
Atmospheric Modeling Division/NOAA, MD-80, Research
Triangle Park, NC 27711; e-mail: yli@hpcc.epa.gov.
advection, vertical eddy mixing, cloud effects, dry
deposition, emissions, and photochemical production and
losses. RADM uses the RADM2 Chemical Mechanism
developed by Stockwell et al. (1990). Anthropogenic
emissions are from the NAPAP 90 emissions inventories.
Biogenic emissions from agriculture and forests are from
BEIS2. Clear sky photolysis rates are calculated by the
delta-Eddington method. The dry deposition velocity is
computed from the resistance-in-series method. The
subgrid-scale turbulent mixing processes in the PBL are
described by using the local eddy diffusivities (Kz), which
are based on boundary layer scaling theory (Byun, 1990
and 1991). For clouds, four primary effects are
represented: (1) vertical redistribution, (2) aqueous
chemical reactions, (3) scavenging effect (loss), and (4)
radiative effect.
3. METEOROLOGICAL INPUT FROM MM5
The meteorological fields used in the RADM
simulations are derived from archives generated by
hydrostatic MM5 simulations, the fifth-generation Penn
State/NCAR Mesoscale Model [Grell et al., 1994], Three-
dimensional fields of horizontal winds (u,v), temperature
(T), specific humidity (Qv) and two-dimensional surface
pressure (p5), column-cumulative precipitation rates,
ground temperature, etc., are archived at hourly intervals
during MM5 runs. We use a meteorology-chemistry
interface program (MCIP) to derive RADM-required fields
from MM5 archives. MCIP estimates various surface and
PBL parameters (such as friction velocity, surface heat
flux, Monin-Obukov length, and PBL mixing height) as
well as dry deposition velocities. These fields are used
as inputs to RADM and are used in RADM algorithms to
determine cloudiness and vertical velocity, and to quantify
various subgrid scale transport processes, dry and wet
deposition rates, and chemical reaction rates.
4. SENSITIVITY EXPERIMENTS
The sensitivity experiments we conducted are
summarized in Table 1. The second column lists the
meteorological parameters or processes to be changed
for the sensitivity experiment. The third and fourth
columns show how the parameters or processes are
changed as compared to the base RADM experiment.
The last column designates a symbol which will be used
in the rest of the paper to refer to each individual
-------
Table 1. Meteorology Sensitivity Experiments,
Exp.
Met Parameter or
Physical Process
PBL height (H)
Their Affected Parameters and Processes, and Designated Symbols
T
2 PBL height (H)
3 PBL mixing scheme for CBL
4 Derivation of surface parameters
and PBL height estimation in MCIP
5 MM4 meteorology
6 All cloud effects
7 radiative effect of clouds
8 all other cloud effects than radiative
Value or Definition for
Sensitivity Experiment
Value or Definition
Base Experiment
Symbol
Designated
50%
1o6%
0.6H
150%
100%
1.5H
Pleim & Chang (1992)
Byun (1990)
ACM
Byun (1991); ground
Byun (1990);
SFC
temperature
Holtslag etal(1995)
MM4
MM5
MM4
all not included
all included
noCId
not included
included
noJcId
not included
included
noWet
sensitivity experiment for convenience of presentation.
The eight experiments listed in Table 1 can be divided
into two categories. Category one covers experiments 1
to 5 that are designed to test PBL mixing schemes and
parameters, especially the PBL height. Category two
includes experiments 6 to 8 designed to test the effect of
clouds processes.
The first two sensitivity experiments in Table 1
apply a simple scaling factor, 50% and 150%
respectively, to the base PBL height (designated as "H"
hereafter) in RADM, while all the other parameters and
processes are unchanged. Their designated symbols are
"0.5H" and "1.5H", respectively. The third experiment,
symbolized as "ACM", utilizes the asymmetric convective
model (ACM) of Pleim and Chang (1992) as an
alternative to describe the eddy mixing process for the
convective boundary layer (CBL) in RADM. The fourth
experiment, symbolized as "SFC", estimates various
surface parameters (such as friction velocity, surface heat
flux, and Monin-Obukov length) in MCIP using the PBL
scaling theory developed by Byun (1991) as opposed to
the surface-layer scaling theory for the base experiment
(Byun, 1990). This SFC configuration has been used in
most of our previous RADM simulations using MM4
meteorology. In order to explain the sensitivity results
presented in section 5, it is crucial to understand the
difference in the details of the estimation algorithms for
convective PBL height among the base, ACM, and SFC
experiments. While the algorithms in all three
experiments are based on vertical profiles of potential
temperature (PT) and the Richardson number, the
difference lies on the initial temperature assumed for a
rising air parcel from near the surface. The base case
uses an effective surface PT which is the near-surface air
PT plus a temperature excess measuring the strength of
convective thermals in CBL (Holtslag et al., 1995). The
same algorithm is employed in MM5 and such computed
H has a maximum value of about 2 km over eastern US.
On the other hand, ACM uses the near-surface air PT,
which results in significantly lower H than the base
experiment. SFC uses the ground PT, which results in
much higher H than the base experiment. Indeed, SFC
has a maximum value of about 3 km over eastern US.
During the daytime when the ground is heated, the
effective surface PT used by the base experiment is
higher than the air PT but lower than the ground PT. A
buoyant thermal with a higher initial temperature will
shoot higher during convection. The fifth sensitivity
experiment, designated as "MM4" in Table 1, uses MM4-
produced meteorology and an older version of MCIP.
This older MCIP has a similar configuration as SFC and
therefore its PBL height is significantly higher than the
base experiment.
The sixth to eighth experiments are conducted to
test how cloud processes affect the model-predicted 03
chemistry. In the sixth experiment, designated as
"noCId", all cloud effects are turned off. In the seventh
experiment, designated as "nojcld", only cloud radiative
effects on photolysis rates are turned off, creating
effectively clear sky photolysis. In the eighth sensitivity
experiment, designated as "noWet", other cloud effects
are turned off while the radiative effect is unchanged. For
each of these meteorology sensitivity experiments, two
runs are conducted. One is with all emissions and the
other is with anthropogenic NOx emissions uniformly
reduced by 15%. For the base experiment, a third run is
also conducted in which anthropogenic NOx emissions
are uniformly reduced by 50%. It will be designated as
".5nox" in the following text. The base run with 15% NOx
emissions reduction will be designated as ".85nox".
5. RESULTS
Results for an urban grid cell in New York City
(NYC) are given in Tables 2 and 4 and for a rural grid
cell in Scotia, Pennsylvania, in Tables 3 and 5. The NYC
cell is characterized by high emissions of NO that titrate
03 causing low [03] near the surface, with higher [03]
levels above the surface layer. The Scotia cell is a rural,
NOx-limited cell in which [03] levels and local
photochemistry are largely determined by the transport of
03 and it's precursors into the cell. The results in Tables
2 through 4 are 7-hour averages from 10am to 4pm local
time during the period of July 20 to August 12. Tables 2
and 3 shows results for the base case scenario, the base
case with 15% and 50% NOx emissions reductions, and
for the eight meteorological sensitivity experiments.
5.1 Ozone Concentration
-------
Table 2. Daytime Average [03] and Local 03 Production Efficiency at NYC for the Base, NOx Emissions Reductions,
and Uncertainty in Meteorology Parameters and Processes. Average Concentrations Are Computed from 10am to 4pm
Local Time for the 11 highest [Q3] days of the base case at New York City.
BASE ,85nox .5nox 0.5H 1.5H ACM SFC MM4 noCId noJcId noWet
03 (> lOOppb) T2S T5S TTE m TT3 T55 TT7 TOS T33 T33 T55~
P(Ox) 160 160 136 189 142 162 144 140 167 161 166
P(Ox) / P(HN03) 7.1 8.4 15.0 6.6 7.3 6.6 7.4 7.2 7.0 7.0 7.1
Table 3. The same as table 2 but for Scotia, Pennsylvania, average values for 4 highest [03] days of the base case
BASE .85nox ,5nox S3F1 iT&R ACM SFC fOT noCId noJcId noWet"
[03] (> 100ppb) T5T n? §5 TSc5 TTS T23 Tft TTi T25 VS. TZT
P(Ox) 58 55 38 72 50 54 50 49 55 54 58
P(Ox) / P(HN03) 47 51 81 52 48 38 45 43 46 46 47
The first row in Tables 2 and 3 shows the average
[03] for days when the Base Case average [03] exceeds
100 ppbv (11 days at NYC and 4 days at Scotia). At
Scotia, average [03] decreases from 121 ppbv to 115
pbbv for the 15% NOx emissions reduction and to 98
ppbv for the 50% NOx emissions reduction, reflecting the
NOx-limited characteristics of Scotia. In NYC, however,
average [03] is unchanged for the 15% NOx reduction
and decreases only 11 ppbv for the 50% NOx reduction
case, which is only half as much as at Scotia. These
results suggest that marginally controlling NOx has little
effect in this NYC cell, although it may benefit downwind
locations.
The PBL-related meteorology sensitivity
experiments (0.5H, 1,5H, ACM, SFC, MM4) show more
significant effects at NYC than at Scotia, with the
difference of [03] from the base experiment ranging from
-20 to 15 ppb at NYC versus -10 to 9 ppb at Scotia. In
the experiments in which we removed cloud processes
(noCId, noJcId, noWet), the NYC [03] increased by 6 to
12 ppbv, and in Scotia [03] increased by only 1 to 4 ppb.
These results demonstrate that variations in the
representation of PBL height, the eddy mixing process,
and cloud processes significantly affect [03] predictions
at NYC, with smaller but still notable effects at Scotia.
For NYC, the [03] reductions due to greater
vertical mixing as in 0.5H, SFC, and MM4 are much
larger than the change due to 15% NOx emission
reduction; in fact, they are comparable to the 11 ppb
reduction achieved by the 50% NOx emissions reduction.
Thus, we estimate that the meteorological uncertainties
in PBL height have comparable impacts on [03] as a 30
to 50% uncertainty in emissions has. This suggests that
it is as important to quantify and reduce meteorological
uncertainty as emissions uncertainty.
5.2 Ozone Production Efficiency
To simulate 03 correctly in regional scale models,
it is necessary to accurately simulate the 03 production
efficiency per HN03 produced, P(03)/P(HN03) or
P(0x)/P(HN03) (Tonnesen and Dennis, 1997). Here Ox
is defined as 03+N02+01D+03P+PAN+2N03+3N205
and P() represents the net chemical production rate. In
Tables 2 and 3 we show the effect of meteorological
inputs on the 03 production efficiency as well as P(Ox),
as integrated over the time period 10 AM to 4 PM for the
eight highest [03] days.
Values of P(0x)/P(HN03) less than 9 suggest
radical-limited conditions (Tonnesen and Dennis, 1997).
For NYC, P(0x)/P(HN03) is less than 8 ppb/ppb for the
base case and all of the meteorology sensitivity
experiments; this indicates that all of these are
radical-limited. P(Ox) increased and 03 production
efficiency decreased in the meteorology experiments with
reduced vertical mixing (0.5H, ACM). This indicates that
reduced vertical mixing causes the system to become
relatively more radical-limited, while increased vertical
mixing causes the system to shift to relatively more
NOx-limited conditions. Results for Scotia are more
ambiguous. Table 3 does show that P(0x)/P(HN03) is
greater than 40 ppb/ppb for all of the meteorology
experiments. This indicates that conditions are very
NOx-limited. In contrast to NYC, the 03 production
efficiency increased in Scotia for the 0.5H experiment
suggesting that reduced vertical mixing caused this rural
cell to become more NOx-limited. It is likely that this
occurs because NOx was converted to HN03 more
rapidly in the upwind urban areas in 0.5H, so less NOx
and PAN were transported to Scotia. These results show
that variations in meteorological inputs do affect 03
production efficiency and 03 sensitivity to NOx.
5.3 Response to NOx Emissions Reduction
This section examines how the meteorological
uncertainties change the system's response to 03
precursors' control. We only present results for NOx
emissions reduction here, in Tables 4 and 5. We
compare the percent change of daytime average 03
concentration due to the 15% NOx emissions reduction
for the base and each sensitivity experiment: ([O3]nox -
[03])/[03]*100, where [03]nox is for the corresponding
15% NOx reduction runs. Some selected results for the
50% NOx reduction are also included in the last three
columns. The results are subdivided, according to the
03 concentration distribution, into an upper third, a
middle third, and a lower third partition, of 8 days each.
-------
K^o7pZn°«eStl SmSlSSS™ R£uc"ons^ncerta'„ty
Class S3 X7*n—?™—i ¦ in
(50% NOx reduction)
75-115
Table 5. The Same As Table 4 but at Scotia. Pennsylvania
(50% NOx reduction)
64-90
•i) -17.0
-16.3 -15.6
13.6 -13.0
-16.4 -15.6 -15.1
For the Base Case at NYC (column 5 in Table 4)
!henaJragePkercent chan9* for the upper third partition
s 0.0 A, and the system on high [03] days is on or near
the ozone ridge line (The ozone ridge line is discussed in
Tonnesen and Dennis, 1997). The percent change
increases to 1.8 /o and 5.8%, respectively, for the middle
and lower partitions, indicating that the system is more
radical-limited on lower [03] days.
The results indicate that lowering the mixing height
moves the system towards being more radical limited.
This is consistent with values of the indicators in Table 2
in which P(0x)/P(HN03) was lower for 0.5H and ACM
(both 6.6) than for the base case (7.1). The shift is
similar for the upper and middle third partitions of the
concentration distribution. For example, the ACM percent
change is 2.1 (2.1 minus 0.0) percentage points larger
than Base for 03 > 115 ppb and 1.9 (3.7 minus 1.8)
percentages points larger for 75 ppb < 03 < 115 ppb.
Figure 1 presents the percent change in the Base Case
for a 15% NOx reduction compared to the percent
A
c
M
change for the ACM case, showing that the shift is
systematic. All of the days are shifted towards greater
radical limitation by approximately the same percentaqe
points, with a slight trend towards a larger shift as the
Base case percent change increases. The intercept in
Figure 1 (1.9) can be compared to the 2.1 and 1 9
percentage point shifts noted above. Sillman et al.(1995)
saw a similar effect for Atlanta when they decreased the
local eddy diffusivity, Kz.
Comparing the 1.5H, SFC, and MM4 results with
the base case indicates that lifting the mixing height
moves the system towards NOx limitation. Figure 2
compares the percent change in the SFC Case to the
percent change for the Base case, showing that the shift
s also systematic. There is a slight trend towards a
arger shift as the Base percent change increases. The
intercept in Figure 2 (-1.0) can be compared to the -1 o
percentage point shift of SFC from BASE for the 03 >
115 ppb partition as shown in Table 4 (-1.0 minus 0 0)
We note that the magnitude of the uncertainty as
We !?au,6
-------
difference between the Base and 1.5H changes the
volume by 50%. We see this asymmetry reflected in the
size of the difference in ozone concentration percent
change due to the 15% NOx emissions reduction
between the 0.5H and 1.5H cases from the Base case.
The results above suggest that meteorology
uncertainties have significant effects on model-predicted
[03] response to NOx and VOC emissions reductions at
NYC. Therefore, it is important to reduce meteorology's
uncertainty to improve confidence in models applications
used to design [03] control strategies. To simply get the
ozone concentration prediction right by tuning the
meteorology may change the system's [03] sensitivity to
emissions which in turn could lead to non-ideal control
strategy decisions.
We also examined the effect of the mixing height
uncertainty on the 50% NOx reduction. As shown in
Table 4 for NYC, the effect of the meteorological
uncertainty scales somewhat less than proportionately to
the size of the emissions reduction. For example, for
[03] > 115ppb, the size of the shift for 0.5H is 6.7 (-2.4
minus -9.1) and 2.5 (2.6 minus 0.0) percentage points for
50% and 15% NOx reductions, respectively. The shift
increases by a factor of 2.6 when the emissions reduction
increased by a factor of 3.3. The size of the shift
averaged over all 24 days for 0.5H is 10 percentage
points (10.5% minus 0.5%). This change is comparable
to the change in emissions reduction effectiveness noted
by Sistla et al. (1996).
At Scotia, PA, a rural site, we see little influence
on the control strategy effectiveness due to the
meteorological uncertainties (Table 5). For the cases
associated with changes in mixing height, we believe this
is due to the opposing effects that happened to cancel in
our particular experiments. We explain this for the 0.5H
case. With 0.5H we would expect the [NOx] to increase
due to a smaller volume and reduce the production
efficiency, changing the control strategy effectiveness.
We also expect the effect of the sensitivity in the urban
areas to be transported into the rural areas. In Table 2
for NYC, we see that P(Ox) increased 18% from the
Base to 0.5H. But the production efficiency,
P(0x)/P(HN03) decreased by 13%. This means that the
P(HN03) increased by 36%. Hence, NOx in the urban
areas is being terminated faster, reducing the NOx
concentrations being transported to the rural areas. The
reduction in transported NOx counters the smaller volume
into which the NOx is going, masking its effect and
resulting in an apparently small sensitivity effect.
Our results suggest that the meteorological
uncertainties in PBL processes and parameters create a
systematic shift in control strategy response that is
independent of the state of the system relative to the
ozone ridge line. On the other hand, the response of
peak ozone to the meteorological uncertainty is
dependent on the state of the system relative to the
ozone ridge line. Thus, one cannot logically deduce the
effect of the meteorological uncertainty on the control
strategy response based on an examination of the [03]
response to the meteorological uncertainty.
6. CONCLUSIONS
These results demonstrate that an accurate
representation of PBL height, eddy mixing, and cloud
processes in the model is extremely important for [03]
predictions and [03] sensitivity. Therefore, meteorological
uncertainty is equally important to be quantified and
improved as emissions uncertainty.
The results suggest that meteorology uncertainty
does convert into uncertainty of model-predicted
effectiveness of emissions control at NYC and even
changes the system's sensitivity regime. This conclusion
holds even on those days when meteorology uncertainty
does not result in differences in [03], In other words, the
effect of meteorology uncertainties on base [03] does not
provide any insight into how the system's response to
emissions control will be changed. Therefore,
meteorology uncertainties and their consequences on
model results must be carefully addressed before we can
rely on numerical model's prediction as a basis in
designing control strategy. It also tells us that merely to
get 03 prediction right does not necessarily get the
system's sensitivity right; independent test and validation
for both must be done, and both are equally important for
model evaluation.
REFERENCE
Byun, D. W., 1990: On the analytical solutions of flux-rofile
relationships for the atmospheric surface layer, J. Appl.
Meteorol., 29, 652-657.
Byun, D. W., 1991: Determination of similarity functions of the
resistance laws for the planetary boundary layer using
surface-layer similarity functions, Boundary Layer Meteorol.,
57, 17-48.
Grell, G.A, J. Dudhia, and D. R. Stauffer, 1994: A Description of
the fifth-Generation Penn State/NCAR Mesoscale Model
(MM5), NCAR/TN-398+STR, National Center for Atmospheric
Research, Boulder, Colorado, 138pp.
Holtslag, A. A. M., E. V. Meijgaard, and W. C. De Rooy, 1995:
A comparison of boundary layer diffusion schemes in
unstable conditions over land, Boundary-Layer Meteor., 76,
69-95.
Pleim, J. E., and J. S. Chang, 1992: A non-local closure model
for vertical mixing In the convective boundary layer, Atmos.
Environ., 26A, 965-981.
Sillman, S., et al., 1995: Photochemistry of ozone formation in
Atlanta, GA - Models and measurements, Atmos. Env., 29,
3055-3066.
Sistla, G., et al., 1996: Effects of uncertainties in meteorological
inputs on urban airshed model predictions and ozone control
strategies, Atmos. Env., 30, 2011-2025.
Tonnesen, G.S., and R.L.Dennis, 1997: An analysis of radical
propagation efficiency to derive combinations of long-lived
species as indicators of 03 sensitivity to NOx and VOC,
submitted to J. Geophys. Res.
DISCLAIMER: This paper has been reviewed in accordance with
the US EPA's peer and administrative review policies and
approved for presentation and publication. Mention of trade
names or commercial products does not constitute endorsement
or recommendation for use.
-------
TECHNICAL REPORT DATA
1. REPORT NO.
EPA/600/A-97/099
2.
3. RE i.
4. TITLE AND SUBTITLE
Effects of uncertainty in meteorological inputs on 03 concentration, 03 production
5.REPORT DATE
efficiency, and 03 sensitivity to emissions reductions in the Regional Acid Deposition
Model
6.PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Y. H. Li
Atmospheric Modeling Division
National Exposure Research Laboratory
Research Triangle Park, NC 27711
8.PERFORMING ORGANIZATION
REPORT NO.
R L Dennis
Atmospheric Modeling Division
National Exposure Research Laboratory
Research Triangle Park, NC 27711
G.S. Tonnesen
National Exposure Research Laboratory
Research Triangle Park, NC 27711
J.E. Plcim
Atmospheric Modeling Division
National Exposure Research Laboratory
Research Triangle Park, NC 27711
D.W. Byun
Atmospheric Modeling Division
National Exposure Research Laboratory
Research Triangle Park, NC 27711
9. PERFORMING ORGANIZATION NAME AND ADDRESS
10.PROGRAM ELEMENT NO.
Same as block 12.
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
National Exposure Research Laboratory
Office of Research and Development
L\ S. Environmental Protection Agency
Research Triangle Park, NC 2771 1
13.TYPE OF REPORT AND PERIOD COVERED
Preprints, 10th Joint Conference on the Applications of Air
Pollution Meteorology with the A&WMA, January' 11-16,
1998, Phoenix, Arizona
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
------- |