Presentation Information
[AAS08-P38]Estimation of Anthropogenic CO2 Emissions in the Kanto Metropolitan Region Using LETKF with Wind-Dependent Localization
*Yutaka Arai1, Ryoichi Imasu2, Qiao Wang2, Hiroaki Kondo3 (1.Graduate School of Frontier Sciences, The University of Tokyo, Department of Natural Environmental Studies, 2.Atmosphere and Ocean Research Institute, The University of Tokyo, 3.National Institute of Advanced Industrial Science and Technology)
Keywords:
carbon dioxide,emission estimation,regional model,satellite data,data assimilation,localization radius
Carbon dioxide (CO2) in the Earth's atmosphere is one of the major greenhouse gases (GHGs) that strongly influence global warming. Accurate warming assessments require understanding the global carbon cycle, particularly the behavior of atmospheric CO2. Atmospheric CO2 concentrations have increased rapidly since the Industrial Revolution, primarily due to anthropogenic emissions associated with large-scale fossil fuel consumption. Although urban areas cover only a small fraction of the Earth's surface, they account for a large share of emissions and are therefore a key focus of carbon cycle research.
Although detailed CO2 emission datasets for urban regions have been developed in recent years, inconsistencies in construction methods, diverse emission sources, and heterogeneous statistics introduce significant uncertainties. Consequently, further improvement in the accuracy of urban CO2 emission estimates is required.
With recent advances in numerical modeling and observational technologies, data assimilation has become widely used for estimating CO2 emissions. Data assimilation is a Bayesian statistical technique that derives optimal estimates by combining observations, model simulations, and their uncertainties.
In this study, we aim to reduce these uncertainties and improve the characterization of CO2 emissions in urban regions by estimating emission sources through data assimilation using the regional model AIST-MM (National Institute of Advanced Industrial Science and Technology-Mesoscale Model) and satellite observations from OCO-3 (Orbiting Carbon Observatory-3).
AIST-MM is a one-way nesting model, in which the outer domain covering the Kanto metropolitan region is simulated at a horizontal resolution of approximately 10 km, and the results are used to drive an inner domain over the Kanto metropolitan region at a resolution of approximately 2 km. MSM-GPV (MesoScale Model Grid Point Value) data were used for nudging the meteorological fields in the outer domain, and NOAA OISST v2.1 (Daily Optimum Interpolation Sea Surface Temperature) was used for sea surface temperature. Anthropogenic CO2 emissions were based on EAGrid2000-Japan and adjusted using annual emission values from the Greenhouse Gas Inventory Office, "Japan's GHG Emissions Data (FY1990-2023, Final Figures)." Biospheric CO2 fluxes from plant photosynthesis, plant respiration, and soil respiration were taken from BEAMSdiurnal-V2f.
In densely populated regions such as the Kanto metropolitan region, CO2 emissions from human respiration are non-negligible. To estimate these emissions, we constructed a population distribution dataset using the "500 m Mesh Future Population Projection Data (FY2018 National Institute of Population and Social Security Research, shapefile version)" from the National Land Numerical Information, Ministry of Land, Infrastructure, Transport and Tourism, Japan (https://nlftp.mlit.go.jp/ksj/index.html). The average daily respiration of a human is assumed to be 1 kg CO2 person-1 day-1. Using this value as the baseline, we developed an hourly emission factor that reflects a typical daily human activity cycle. Finally, the CO2 emissions from human respiration in each grid cell were calculated by multiplying this hourly factor by the population data.
For data assimilation, we employed the Local Ensemble Transform Kalman Filter (LETKF), extended to include CO2 emission variables. While conventional assimilation typically uses circular localization based on distance, this study introduces a localization approach that accounts for wind fields associated with atmospheric advection, aiming to enhance the stability and accuracy of the assimilation.
Currently, we are implementing a method that dynamically adjusts the localization radius according to wind direction and wind speed, for future operational implementation.
Although detailed CO2 emission datasets for urban regions have been developed in recent years, inconsistencies in construction methods, diverse emission sources, and heterogeneous statistics introduce significant uncertainties. Consequently, further improvement in the accuracy of urban CO2 emission estimates is required.
With recent advances in numerical modeling and observational technologies, data assimilation has become widely used for estimating CO2 emissions. Data assimilation is a Bayesian statistical technique that derives optimal estimates by combining observations, model simulations, and their uncertainties.
In this study, we aim to reduce these uncertainties and improve the characterization of CO2 emissions in urban regions by estimating emission sources through data assimilation using the regional model AIST-MM (National Institute of Advanced Industrial Science and Technology-Mesoscale Model) and satellite observations from OCO-3 (Orbiting Carbon Observatory-3).
AIST-MM is a one-way nesting model, in which the outer domain covering the Kanto metropolitan region is simulated at a horizontal resolution of approximately 10 km, and the results are used to drive an inner domain over the Kanto metropolitan region at a resolution of approximately 2 km. MSM-GPV (MesoScale Model Grid Point Value) data were used for nudging the meteorological fields in the outer domain, and NOAA OISST v2.1 (Daily Optimum Interpolation Sea Surface Temperature) was used for sea surface temperature. Anthropogenic CO2 emissions were based on EAGrid2000-Japan and adjusted using annual emission values from the Greenhouse Gas Inventory Office, "Japan's GHG Emissions Data (FY1990-2023, Final Figures)." Biospheric CO2 fluxes from plant photosynthesis, plant respiration, and soil respiration were taken from BEAMSdiurnal-V2f.
In densely populated regions such as the Kanto metropolitan region, CO2 emissions from human respiration are non-negligible. To estimate these emissions, we constructed a population distribution dataset using the "500 m Mesh Future Population Projection Data (FY2018 National Institute of Population and Social Security Research, shapefile version)" from the National Land Numerical Information, Ministry of Land, Infrastructure, Transport and Tourism, Japan (https://nlftp.mlit.go.jp/ksj/index.html). The average daily respiration of a human is assumed to be 1 kg CO2 person-1 day-1. Using this value as the baseline, we developed an hourly emission factor that reflects a typical daily human activity cycle. Finally, the CO2 emissions from human respiration in each grid cell were calculated by multiplying this hourly factor by the population data.
For data assimilation, we employed the Local Ensemble Transform Kalman Filter (LETKF), extended to include CO2 emission variables. While conventional assimilation typically uses circular localization based on distance, this study introduces a localization approach that accounts for wind fields associated with atmospheric advection, aiming to enhance the stability and accuracy of the assimilation.
Currently, we are implementing a method that dynamically adjusts the localization radius according to wind direction and wind speed, for future operational implementation.