27 november 2009
DESCRIPTION
Introduction of REAS and adjoint inverse modeling of NO x emissions over eastern China using satellite observations. 27 November 2009. Jun-ichi Kurokawa National Institute for Environmental Studies, (NIES), Tsukuba, Japan. Brief review and future plans for REAS - PowerPoint PPT PresentationTRANSCRIPT
27 November 2009
Introduction of REASand
adjoint inverse modeling of NOx
emissions over eastern Chinausing satellite observations
Jun-ichi KurokawaNational Institute for Environmental Studies,
(NIES), Tsukuba, Japan
Brief review and future plans forREAS
(Regional emission inventoryin Asia)
J. Kurokawa1,*, T. Ohara1, H. Akimoto2, N. Horii3,K. Yamaji4, X. Yan5, and T. Hayasaka6. 1. National Institute for Environmental Studies, Tsukuba, Japan 2. Acid Deposition and Oxidant Research Center, Niigata, Japan 3. Faculty of Economics, Kyushu University, Fukuoka, Japan 4. Frontier Research Center for Global Change, Yokohama, Japan 5. State Key Laboratory of Soil and Sustainable Agriculture, Institute of Soil Science, Chinese Academy of Science, Nanjing, China 6. Research Institute for Humanity and Nature, Kyoto, Japan.
Summary of REAS version 1.1
SO2 NOx CO BC OC NMV NH3 CH4 CO2
Combusiton Fossil FuelBiofuel ● ● ● ● ● ● ● ● ●
Non-Combustion
IndustrialProcess ● ● ● ● ●
Others ● ● ●
AgricultureAgriculturalField (Soil) ● ● ●
Livestock ● ●
Ohara et al.,ACP,7,4419-4444,2007; http://w3.jamstec.go.jp/frcgc/research/p3/emission.htm
Item Description
Years1980-2003
2010, 2020High/Middle/Low for China
Scenario ERI, China; IEA W.E.O.Spatial Resolution 0.5 degree × 0.5 degree
Temporal Resolution Annual (Monthly for Soil NOx)
Stationary Sources (Combustion) E = Σ (A/NCV)×S×( 1- SR)×( 1- R) SO2
E = Σ A×EF× ( 1- R) NOx, CO, BC, OC, …E EmissionsA Energy ConsumptionNCV Net Calorific ValueS Sulfur Contents of FuelSR Sulfur Retention in AshR Removal EfficiencyEF Emission FactorΣ Fuel and Sector Type
Stationary Sources (Non-Combustion) E = A×EF× ( 1- R)
A Production/Population/… Road Transport E = Σ A × EF × FE / (SG×NCV)
A Energy ConsumptionEF Emission FactorFE Fuel EconomySG Specific GravityNCV Net Calorific ValueΣ Vehicle Type
Basic methodology for REAS 1.1
Grid Allocation:
・ Large Power Plants → Large Point Sources → Location of LPS・ Other Sources → Area Sources → Population (rural/urban/total), Road Network
・ Averaged emission factors for each fuel and sector type were used.・ Temporal variations of EFs by emission regulation measures etc. were not considered in detail.・ Rcently, many new studies of Asian emissions have been reported. But REAS has not utilized them, yet.
Historical emissions between 1980 and 2003
・ Total energy consumptions in Asia more than doubled between 1980 and 2003.・ Asian emissions show rapid growth rates:
SO2:119%, NOx:176%, CO:64%, NMVOC:108%・ NOx emissions in China show a marked increase of 280% over 1980 levels and in particular, the growth after 2000 is high.
SO2 [Mt/yr] NOx [Mt/yr]
CO [Mt/yr] NMVOC [Mt/yr]
0.0
0.5
1.0
1.5
2.0
2.5
3.0
1980 1985 1990 1995 2000 2005 2010 2020年
NOx
(199
6)
年値
に対
する
比率
REASREAS_PSCREAS_REFREAS_PFCZhang et al. (2007,2009)Hao et al. (2002), Tian et al. (2005)
NO2 (Richter et al., 2005)衛星
PFC
PSC
REF
● REAS ◆ Zhang et al. (2007, 2009) ▲ Hao et al. (2002), Tian et al. (2005) ■ Richter et al. (2005) NO衛星観測 2
Irie et al.(2009)
Increasing rates of REAS (00 → 05) 11.8% (SO2) 11.2% (NOx) 7.7% (CO)
NO2 VCDs
NO
x (A
ll da
ta a
re n
orm
aliz
ed to
199
6)
High
Low
Middle
Zhang et al., ACP,9, 5131-5153,2009
Time sereis of NOx emissions in China
Coal-fired Power Plants
Gasoline Vehicles
・ Main focus is on China and, if possible, other East Asia, India and SE Asia.・ New information from recent literatures and databases will be collected. ・ Emission factors will be updated considering the detailed technologies and their implementation reates for each year.・ Information about LPS including newly constructed or closed will be updated, and if we can get good data, large industrial plants will be treated as LPS.・ GIS database, Monthly statistics , Meteorologica data, Satellite observations.・ Results of inverse modeling will be utilized to update the data sets.・ Future scenarios will be generated by AIM (Asia-Pacific Integrated Model) developed by NIES, Kyoto University, and other research institute.
Future plans for REASItem Description
Years
1980-2007 → 2010 → ?
2015, 2030 → 2050?Reference/Alternative policyfor all countries
Scenario AIMSpatial Reso. 0.25 degree × 0.25 degreeTemporal Reso. Monthly → Weekly/Diurnal ?
Adjoint inverse modeling of NOx emissions over eastern China
using satellite observations of NO2 vertical column densities
Jun-ichi Kurokawa1,*, Toshimasa Ohara1
Keiya Yumimoto2, Itsushi Uno2
1. National Institute for Environmental Studies, Tsukuba, Japan2. Research Institute for Applied Mechanics, Kyushu University, Japan
IntroductionIt is essential to prepare accurate emission inventories for
atmospheric chemistry modeling.
Bottom-up Approach Inverse Modeling
Combinations of activity statistics and source- or region-specific EFs.There are uncertainties for statistics, parameters, and temporal and grid allocation profiles.The publication of basic statistics is generally a couple of years behind.
We developed a 4DVAR data assimilation system for the optimization of NOx emissions based on a regional CTM and satellite observation data
Emissions are optimized to reduce the differences between simulated and observed data.It is a powerful method that helps solving the problems of bottom-up approach.
Model description – 4DVAR
0 T
First Guess
Optimal Solution
X(t)
Observation
X 0*
X 0b
In our inverse modeling system, surface NOx emissions are assigned as the parameters to be adjusted and then, the results will be the optimized emissions.
4 -Dimensional VARiational ( 4 DVAR) data assimilation
Simulation
The cost function is defined as a function of some input parameters in a numerical model.
In 4DVAR data assimilation system, the parameters are adjusted to minimize the value of the cost function.
Model description – RC4-NOxA Priori
Emissions CMAQ
Forward Model
Adjoint Model
Optimization
Updated Emissions
CMAQ
A PosterioriEmissions
P(NOx), L(NOx)NO2/NOx
Vd(NO), Vd(NO2)
P(NOx), L(NOx)NO2/NOx
Vd(NO), Vd(NO2)
RAMS/CTM-4DVAR-NOx
ObservationGOME
NO2 Column
Met. FieldsRAMS
(Pielke et al.,1992)
i = i + 1
The Convergence Criterion
i = 1
Initial Condition
No
Yes
( in offline manner)
The Convergence Criterion
Norm of the gradient of the cost function is reduced by 1% with respect to the initial norm.
Base Model:Yumimoto and Uno, 2006 RAMS/CTM-4DVAR-CO
A parameterized NOx Chemistry scheme was introduced into RAMS.
Key parameters such as chemical production and loss terms of NOx were calculated in advance using full chemistry model CMAQ
① Forward Model → Cost Function
② Adjoint Model → Gradient of Cost Function
③ Optimization → Update of Emissions
④ Iteration (① ~ ③) ※ For each iteration step, parameters for NOx chemistry are also updated by running CMAQ.
⑤ Emissions passing the convergence criterion are the a posteriori emissions.
Inversion experiments’ set-up
RAMS, CMAQ modeling domainwith NOx emissions in July 2002Horizontal Resolution : 80kmVertical Layer : RAMS/CMAQ 23/14 with stretching grid layers (150m at the surface)EC : eastern China, BR : the Beijing regionNCP : the North China Plain, YD : the Yangtze Delta
【 Target Region for Emission Optimization 】 The emission optimization and comparison between observed and model-simulated results were performed only over EC region in order to save the computational costs.
【 A Priori Emissions 】 ・ REAS 1.1 NOx emissions for fuel combustion and soil sources. ・ TRACE-P for biomass burning emissions. ・ Seasonality is only considered for soil
emissions. ・ The uncertainty of 500% is assigned for the background errors.【 Observation Data 】 ・ Monthly averaged GOME NO2 vertical column densities (VCDs) ・ The observation errors are assumed to be the MAX (the absolute error 1.0×1015 molec./cm2, 20% of the observed NO2 VCDs)【 Simulation period 】 July 1996-2002
Comparison of observed and model-simulated NO2 VCDs
GOME Without Assimilation
1996
1999
2002
GOME-observed NO2 VCDs
NO2 VCDs increased clearly from both 1996 to 1999 and 1999 to 2002, especially around BR, NCP, and YD.Overall, NO2 VCDs increased nearly monotonically from 1996 to 2002.
NO2 VCDs without assimilation
NO2 VCDs over EC increased monotonously from 1996 to 2002. However, the rate of increase was obviously lower than that of the GOME observation.The temporal variations over BR, NCP, and YD were much different from each other and from those of GOME observation.
Assimilated
Assimilated NO2 VCDs
The spatial distribution and time evolution of GOME-observed data were generally reproduced.
A posteriori NOx emissions
-A priori emissions over BR were reduced by optimization during the whole simulation period. → The reductions at extremely high emission grid cells might be excessive because GOME has a coarse horizontal resolution and the NO2 VCDs for each observation grid were smoothed values. → In northern China, more energy is required for heating in winter, thus emissions in summer might be lower than annual average. - In NCP, emissions increased both in 1999 and 2002, but the area of increase was different. In 2002, emissions largerly increased in YD, in contrast to the changes in 1996 and 1999. - A priori emissions in 1996 were reduced in a wide area where soil NOx emissions were relatively high. → Soil NOx emissions in 1996 were prepared by linear interpolation of data for 1990 and 2001. Interannual differences are also important for emissions from soil sources.
Differences between a priori and a posteriori NOx emissions (a posteriori minus a priori)
Emission differencesbetween 1999 and 1996
(1999 minus 1996)
A p
rior
i em
issi
ons
Emission differencesbetween 2002 and 1999
(2002 minus 1999)
・ In general, changes in the spatial distributions of a posteriori emissions are larger than those of a priori emissions, especially in the eastern part of EC.
・ The area of increasing emissions from 1996 to 1999 and from 1999 to 2002 are around BR, YD, and the center of NCP both in a priori and a posteriori emissions.
・ The amounts of increase of a posteriori emissions are larger than those of a priori emissions. In particular, the differences are extremely large around YD.
Trends of NOx emissions before and after optimization
A p
oste
rior
i em
issi
ons
Trends of NOx emissions before and after optimization
The North China Plain
Eastern China The Beijing Region
The Yangtze Delta It is suggested that REAS 1.1 underestimates the rate of increase of NOx emissions from 1996 to 2002.
All trends are normalized to 1999
The rates of increase of a posteriori emission (▲) were larger than those of a priori emissions (■) and their trends became similar to those of GOME (●).
Running RC4-NOx through a whole year is needed to evaluate annual emission amounts and their trends.
The increase in emissionsrates from 1996 to 2002became much larger overeach region. EC : 19%→49% BR : 14%→63% YD : 23%→101% NCP : 20%→54%
Conclusion
REAS reveals that anthropogenic emissions in Asia show marked increases in these two decades.
It is important not only to extend the target years but also consider the introducion of new technologies for Asian emissoin inventories.
RAMS/CTM-4DVAR-NOx which optimizes NOx emissions using GOME observed NO2 VCDs was developed and applied to NOx emissions over eastern China in July from 1996 to 2002.
- The increase in emissions rates from 1996 to 2002 became larger over eastern China from a priori to a posteriori emissions.
- It is suggested that REAS underestimates the rate of increase of NOx emissions for China from 1996 to 2002.
Thank You for your Attention !