Download presentation
Presentation is loading. Please wait.
1
CarboEurope Open Science Conference
Regional scale CO2 budget constraints from concentration and flux measurements Christoph Gerbig, Ravan Ahmadov, Stefan Körner Max-Planck-Institute for Biogeochemistry Acknowledgements: MetAir (Bruno Neininger, Joel Giger, Han Bär) many CEIP Fluxtower PIs Danke für die Einladung. CarboEurope Open Science Conference Crete , November 2006
2
Overview Motivation Modeling system: WRF– VPRM – STILT
Application: CERES Uncertainties, their characterization and possibilities for propagation Closing remarks
3
Motivation Data-Model-Fusion: Bridging the gap:
utilizing information from mixing ratio measurements & eddy flux measurements in a consistent model Bridging the gap: between tall tower and global model Regional scale budget requires: interpreting regional signals in PBL measurements dominant scales of variability: diurnal-synoptic (time) vegetation patterns (~ km) Quantitative fusion: appropriate uncertainties for each datastream, especially model uncertainties in biosphere and transport
4
WRF-VPRM-STILT modeling system
Forward modeled CO2 WRF-VPRM-STILT modeling system WRF Weather Research and Forecasting Model ECMWF meteorology VPRM Vegetation Photosynthesis and Respiration Model Eddy flux data
5
VPRM Vegetation Photosynthesis and Respiration Model
[Pathmathevan et al., submitted to GBC], based on Xiao et al. [2004] Optimization of parameters α, β, λ, and PAR0 vegetation classes (5) SYNMAP land cover [Jung et al., 2006] NEE = GEE + R ECMWF, NCEP, WRF or site measurements MODIS surface reflectances 8 day, 500 m Eddy Cov. data [many CE site PI’s]
6
2005 CEIP-EC data vs. VPRM (driven by site meteorology)
Spatial gradients: deciduous forests Spatial gradients: evergreen forests 2005 CEIP-EC data vs. VPRM (driven by site meteorology) Captures: hourly variations (radiation), site-site variations () diurnal fluxes (June-July) diurnal
7
2005 CEIP-EC data vs. VPRM (driven by site meteorology)
Phenology captured at most sites 8-day aggregated fluxes 8-day
8
2005 CEIP-EC data vs. VPRM (driven by site meteorology)
Residuals for 1-day aggregated fluxes PDF 1/[µmoles/m2/s] NEE [µmoles/m2/s] ORCHIDEE (Chevallier et al., 2006) 2005 CEIP-EC data vs. VPRM (driven by site meteorology) Lorentz (Cauchy) Gauss
9
2005 CEIP-EC data vs. VPRM (driven by site meteorology)
Closer to sources/sinks is challenging: strong spatial and temporal variations in surface fluxes in the near field of measurement locations, combined with strong variations in transport and mixing (mixed layer height variations, frontal systems) residuals 1-day aggregated fluxes
10
WRF-VPRM-STILT modeling system
Forward modeled CO2 WRF-VPRM-STILT modeling system measured CO2 CO2 global boundary conditions LMDZ, TM3 WRF Weather Research and Forecasting Model ECMWF meteorology VPRM Vegetation Photosynthesis and Respiration Model Eddy flux data
11
WRF-VPRM vs. Aircraft data (CERES campaign)
Measurement WRF-VPRM Closer to sources/sinks is challenging: strong spatial and temporal variations in surface fluxes in the near field of measurement locations, combined with strong variations in transport and mixing (mixed layer height variations, frontal systems) MetAir Eco-Dimona
12
Respired CO2 signal 10 ppm surface
See Poster by Ravan Ahmadov et al., RT23 Closer to sources/sinks is challenging: strong spatial and temporal variations in surface fluxes in the near field of measurement locations, combined with strong variations in transport and mixing (mixed layer height variations, frontal systems)
13
WRF-VPRM-STILT modeling system
Forward Inverse modeled CO2 WRF-VPRM-STILT modeling system measured CO2 WRF STILT Weather Research and Forecasting Model Stochastic Time Inverted Lagrangian Transport Model ECMWF meteorology VPRM VPRM parameter optimization scalars for R, GEE Vegetation Photosynthesis and Respiration Model Can’t afford 2 km resolution mesoscale for long term regional inversion. Use ECMWF driven STILT transport. regional scale CO2 budget Eddy flux data
14
STILT-ECMWF footprints hourly, 10 km (dynamic)
Closer to sources/sinks is challenging: strong spatial and temporal variations in surface fluxes in the near field of measurement locations, combined with strong variations in transport and mixing (mixed layer height variations, frontal systems)
15
Uncertainties involved in inversions (continental stations)
Order of magnitude smaller over the ocean! => challenge over the continent CAUSE of representation error: land vs continent; continental variability: could be reproduced by running particle model and degrading upstream sources/sinks IMPLICATION for using ocean data vs land
16
Uncertainties involved in inversions (continental stations)
Order of magnitude smaller over the ocean! => challenge over the continent CAUSE of representation error: land vs continent; continental variability: could be reproduced by running particle model and degrading upstream sources/sinks IMPLICATION for using ocean data vs land
17
Uncertainties involved in inversions (continental stations)
Source of uncertainty Type Magnitude Reference Transport Model Advection PBL mixing Convection Transport Model + Flux Model Grid resolution Flux Model Aggregation Measurement Precision, accuracy winds uncertain + spatial flux variability = mixing ratios uncertain Order of magnitude smaller over the ocean! => challenge over the continent CAUSE of representation error: land vs continent; continental variability: could be reproduced by running particle model and degrading upstream sources/sinks IMPLICATION for using ocean data vs land
18
Uncertainties involved in inversions (continental stations)
Source of uncertainty Type Magnitude Reference Transport Model Advection ~ 5 ppm (summertime) Lin and Gerbig, 2005 PBL mixing Convection Transport Model + Flux Model Grid resolution Flux Model Aggregation Measurement Precision, accuracy Order of magnitude smaller over the ocean! => challenge over the continent CAUSE of representation error: land vs continent; continental variability: could be reproduced by running particle model and degrading upstream sources/sinks IMPLICATION for using ocean data vs land
19
Temporal correlation: ~ 12 hours
daytime bias and random error in zi Variogram of differences zi(ECMWF)- zi(RS) Day: 99 km length scale Night: 229 km length scale Temporal correlation: ~ 12 hours Mixing heights zi derived from T, RH and winds using Bulk Richardson method
20
Uncertainties involved in inversions
Source of uncertainty Type Magnitude Reference Transport Model Advection ~ 5 ppm (summertime) Lin and Gerbig, 2005 PBL mixing Convection Transport Model + Flux Model Grid resolution Flux Model Aggregation Measurement Precision, accuracy Order of magnitude smaller over the ocean! => challenge over the continent CAUSE of representation error: land vs continent; continental variability: could be reproduced by running particle model and degrading upstream sources/sinks IMPLICATION for using ocean data vs land
21
Uncertainties involved in inversions (continental stations)
Source of uncertainty Type Magnitude Reference Transport Model Advection ~ 5 ppm (summertime) Lin and Gerbig, 2005 PBL mixing Convection ? Transport Model + Flux Model Grid resolution Flux Model Aggregation Measurement Precision, accuracy Order of magnitude smaller over the ocean! => challenge over the continent CAUSE of representation error: land vs continent; continental variability: could be reproduced by running particle model and degrading upstream sources/sinks IMPLICATION for using ocean data vs land
22
Uncertainties involved in inversions (continental stations)
Source of uncertainty Type Magnitude Reference Transport Model Advection ~ 5 ppm (summertime) Lin and Gerbig, 2005 PBL mixing Convection ? Transport Model + Flux Model Grid resolution ~ km (summertime) Gerbig et al., 2003 Flux Model Aggregation Measurement Precision, accuracy Spatial statistics of multiple profile measurements (COBRA experiments) Order of magnitude smaller over the ocean! => challenge over the continent CAUSE of representation error: land vs continent; continental variability: could be reproduced by running particle model and degrading upstream sources/sinks IMPLICATION for using ocean data vs land
23
Uncertainties involved in inversions (continental stations)
Source of uncertainty Type Magnitude Reference Transport Model Advection ~ 5 ppm (summertime) Lin and Gerbig, 2005 PBL mixing Convection ? Transport Model + Flux Model Grid resolution ~ 1 200km (summertime) Gerbig et al., 2003 Flux Model Aggregation depending on Aggregation and Model Gerbig et al., 2006 Measurement Precision, accuracy pseudo data experiment, varying a-priori covariance length scale Order of magnitude smaller over the ocean! => challenge over the continent CAUSE of representation error: land vs continent; continental variability: could be reproduced by running particle model and degrading upstream sources/sinks IMPLICATION for using ocean data vs land
24
Uncertainties involved in inversions (continental stations)
Source of uncertainty Type Magnitude Reference Transport Model Advection ~ 5 ppm (summertime) Lin and Gerbig, 2005 PBL mixing this work Convection ? Transport Model + Flux Model Grid resolution ~ 1 200km (summertime) Gerbig et al., 2003 Flux Model Aggregation depending on Aggregation and Model Gerbig et al., 2006 Measurement Precision, accuracy 0.1 ppm (targeted) WMO Order of magnitude smaller over the ocean! => challenge over the continent CAUSE of representation error: land vs continent; continental variability: could be reproduced by running particle model and degrading upstream sources/sinks IMPLICATION for using ocean data vs land
25
Closing remark VPRM as a diagnostic biosphere model captures NEE on relevant spatial and temporal scales WRF-VPRM is an invaluable tool to bridge the gap between global model and tall tower (see poster R. Ahmadov) The VPRM-STILT allows spatially resolved retrieval of NEE from mixing ratio observations To utilize long term & large scale information from mixing ratio observations, we first need to model (or parameterize) the short term & small scale with minimal bias Uncertainties need characterization (covariances) and propagation Order of magnitude smaller over the ocean! => challenge over the continent CAUSE of representation error: land vs continent; continental variability: could be reproduced by running particle model and degrading upstream sources/sinks IMPLICATION for using ocean data vs land
26
Thank you.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.