impact of meteorological inputs on surface o 3 prediction
DESCRIPTION
Impact of Meteorological Inputs on Surface O 3 Prediction. Jianping Huang 9 th CMAS Annual Conference Oct. 12, 2010, Chapel, NC. Co-Authors. - PowerPoint PPT PresentationTRANSCRIPT
Impact of Meteorological Inputs on Surface O3 Prediction
Jianping Huang
9th CMAS Annual Conference
Oct. 12, 2010, Chapel, NC
Co-Authors Jeff McQueen1, Youhua Tang1,2, Binbin Zhou1,2,
Marina Tsidulko1,2, Ho-Chun Huang1,2, Sarah Lu1,2, Brad Ferrier1,2, Bill Lapenta1, Geoff DiMego1 (1: NOAA/NCEP/EMC, 2: IMSG)
Daewon Byun3, Pius Lee3, Daniel Tong3,4 (3: NOAA/ARL, 4: ERT)
Ivanka Stajner (NOAA/NWS/OST)
Motivation and objectives Motivation
- O3 over-predicted especially by CB05 and in coastal regions
Objectives
- to evaluate meteorological inputs
- to reduce O3 over-prediction
Outline National Air Quality Forecasting
Capability Current issue of O3 forecasting Verification of meteorological inputs Sensitivity ofO3 prediction to cloud
parameters Summary
National Air Quality Forecasting CapabilityNational Air Quality Forecasting Capability
Emission model: SMOKE - NEI 2005
- BEIS V3 Met model: WRF/NMM (NAM,
12 km/L60)
- T, RH, Wind, etc.
- Cloud, PBL (re-calculated by PreMAQ)
AQ model: CMAQ (12km/L22) - Oper: CONUS(CB04),
AK/HI(CB05/Aero-4)
- Exper/Dev: CONUS(CB05/ Aero-4)
http: www.weather.gov/aqhttp: www.weather.gov/aq
Current issue of O3 forecasting
6
8-hr max O3 significantly over-predicted in NE coastal region as compared to AIRNOW
5x (Exp.) 8-hr max O3 Aug-31-10 5x (Exp.) 8-hr max O3 Aug-31-10
ppb
ppb
Current issue of O3 forecasting (cont.)
7
Daily 8-hr max O3 (exp.) over-predicted (CONUS)Time period: July 1st to August 31st, 2010
Date (12 UTC Cycle) Date (12 UTC Cycle)
obs
fcst
bias
rmse
Emissions - NEI 2005 Meteorological inputs - wind, etc. - cloud, PBL height (re-diagnosed in PreMAQ) CMAQ - deposition velocity, etc. - CB05 mechanism Lateral boundary condition - static
Causes of O3 over-prediction
Verification tool and data Verification tool: Forecast Verification System (FVS)
- Grid2obs
- Grid2grid
- Statistics (e.g., rmse, bias) and FHO (e.g., csi, ets, far) Met observational data
- T, RH, Wind: ANYSFC, ADPUPA, ONLYSF, VADWND
- Cloud: AFWA (global, 10 x 1o, 1-hr), CLAVR-x (global, 0.5o x 0.5o, 6-hr) O3 data
- AIRNOW Studied time period
- O3 and met verification: Jul. 1 to Aug. 31, 2010
- Sensitivity testing: Aug. 5 – 31, 2010
FVS statistics parameters
FVS Statistics variables: F.H.O. F = grid fraction of forecasted > threshold O = grid faction of observed > threshold H = grid fraction of both forecasted and
observed > threshold Basic statistics scores Bias=F/O=(a+b)/(a+c) Critical Success Index CSI=H/(F+O-H)=a/(a+b+c) Probability of Detection POD=H/O=a/(a+c) False Alarm Ratio FAR =1-H/F=b/(a+b) Thresholds: O3: > 55, 65, 75, 85, 105, 125, 150 (ppb)
N=a+b+c+d
F=a+b
O=a+c
H=a
b
c
d
a
Verification of met inputs
Date
Date
Date
black: rmsered: bias
Date
T (o
C)
rmse
, b
ias
of
T (o
C)
Relative humidity (RH)Temperature (T) black: obs meanred: fcst mean
RH
(%
)rm
se ,
bia
s o
f R
H (
%)
Domain: CONUS
Verification of met inputs (cont.)
WS
(m
/s)
rmse
, b
ias
of
WS
(m
/s)
Date
Date
black: rmsered: bias
Wind speed (WS) black: obs meanred: fcst mean
Cloud cover (%)
Date
Date
rmse
, b
ias
of
TC
LD
(%
)T
CL
D (
%)
Domain: CONUS
How does cloud impact O3 prediction?
Photolysis rate
Jcld=J0[1+Cf(1.6trcos()-1] below cloud,
Jcld=J0[1+Cfi(1-tr)cos()] above cloud,
where J0 is the clear sky photolysis rate, Cf is cloud cover, is the zenith angle, αi is a reaction dependent coefficient, and tr is cloud transmissivity, which is a function of cloud water content and cloud thickness.
Cloud parameterization in PreMAQ
- Cloud cover: Geleyn et al. (1982) (below PBL); Schumann (1989), Wyngaard and Brost (1984) (above PBL)
- Liquid water content: Welcek and Taylor (1986), Change et al. (1987, 1990).
NAM Cloud: more complicated cloud parameterization schemes (Ferrier et al. 2002)
13
Cloud cover FHO statistics
Against AFWA for CONUS, Aug 05-31, 2010
Total cloud cover threshold Total cloud cover threshold
black: defaultred: modified
% %
Cloud cover FHO statistics (cont.)
Total cloud cover threshold
black: defaultred: modified
Total cloud cover threshold
Against CLAVR-x for CONUS, Aug 05-31, 2010
% %
Sensitivity run: default vs. Modified PreMAQ
08-31-2010: 13 UTC 08-31-2010: 19 UTC
Hourly-mean O3 difference (modified-default)
ppb ppb
8-hr max O3 verification: CONUS
black dash: default-fcstred dash: modified-fcst solid: obs
obs
fcstrmse
bias
Date (12 UTC Cycle) Date (12 UTC Cycle)
black: defaultred: modifiedsolid: rmsedash: bias
8-hr max O3 verification: NEUS
black: defaultred: modifiedsolid: rmsedash: bias
black dash: default-fcstred dash: modified-fcst solid: obs
Date (12 UTC Cycle) Date (12 UTC Cycle)
8-hr max O3 FHO comparison: CONUS
8-hr max O3 threshold
black: defaultred: modified
8-hr max O3 threshold (ppb)
ppb
ppb
8-hr max O3 FHO comparison: NEUS
8-hr max O3 threshold 8-hr max O3 threshold
black: defaultred: modified
ppb
ppb
Summary O3 over-prediction is often observed especially near
North-eastern coastal region. Met verification results present that while temperature,
relative humidity, and total cloud cover simulated by NAM show very good agreement with observations, NAM does not capture the time variability of the observed wind well.
The sensitivity study indicates that direct taking cloud parameters (cloud cover, liquid water content, cloud base and top) from NAM outputs may slightly improve surface O3 prediction especially over the NE coastal region.