+ The ability of satellite-based CO2 measurements to constrain carbon cycle science: from GOSAT to OCO-2 Chris O’Dell 1 & Hannakaisa Lindqvist 1 1 Colorado State University, Fort Collins, CO, USA
+ Acknowledgments 2 ACOS Team (JPL, CSU) Christian Frankenberg, David Crisp, Annmarie Eldering, Mike Smyth, James McDuffie, Michael Gunson, Lukas Mandrake, Albert Chang, Brendan Fisher, Vijay Natraj, Igor Polonsky, Thomas Taylor, Robert Nelson CarbonTracker Model Output (NOAA) Andy Jacobson et al. MACC Model Data (LSCE) Frederic Chevallier Univ. of Edinburgh Model Data (UoE) Liang Feng, Paul Palmer
+ 3 XCO2 precisions of 1 – 2 ppm are needed on regional scales to improve our knowledge of carbon cycle phenomena
+ 2009: Greenhouse Gases 4 Observing SATellite (GOSAT)
+ 1. Unbiased GOSAT retrievals should 5 help constrain CO2 sources & sinks Theoretical work shows that bias-free GOSAT observations reduce surface carbon flux uncertainties . Chevallier et al. (2011) found uncertainty reductions of 20-60% over land using OSSEs, including the effects of transport model uncertainty. Maksyutov et al. (2013) found Percent Uncertainty reduction in surface fluxes brought by GOSAT relative to uncertainty reductions of 15-50% over surface observations (GLOVALVIEW) many land areas relative to alone. From Maksyutov et al. (2013). GLOBALVIEW , for real GOSAT observations.
+ 6 2. Biases in GOSAT data can lead to large errors on inverted fluxes. Basu et al. (2013) found that a 0.8 ppm bias between land and ocean in GOSAT retrievals was enough to turn the global lands from a sink to a source. Chevallier et al. (2014) looked at inversions of ACOS and UoL GOSAT data, using mutiple inversions systems, found that both satellite biases and transport errors can lead to unrealistic inferred surface fluxes. As a result, very few consistent flux inversion results have resulted from GOSAT XCO 2 observations so far.
+ 7 SO… 1. How large are errors in raw GOSAT retrievals? 2. How large are the errors after bias correction?
+ RAW GOSAT XCO2 Errors 8 Raw GOSAT errors can be many ppm, and are often correlated with geophysical parameters such as surface albedo. X CO2 Error [ppm] 2 μm Surface Albedo 2 μm Surface Albedo
+ ACOS Bias Correction Approach 9 Error vs. Models (Land gain H) Bias-correction parameters MUST agree between TCCON & Error vs. TCCON (Land gain H) MODELS Variables identified via semi- automated procedure. Corrections are typically 0-2 ppm.
+ 2. How large are the remaining biases? 10 Method 1: Different regressions Scheme 1: Albedo_3, Fs, CO2 Vertical Gradient Scheme 2: Sig3/Sig1, Fs, CO2 Vertical Gradient June, Land Gain H Most areas have differences ≤ 1 ppm
+ 11 How large are the remaining biases? Comparing different algorithms Most areas have differences ≤ 2 ppm Before Bias Correction After Bias Correction July 2009 Inter-algorithm Standard Deviations for 5 GOSAT algorithms: (RemoTeC, NIES, PPDF-S, UoL, ACOS) From Takagi et al. (2014)
+ XCO2 comparisons to models 12 Compare retrieved XCO2 to models directly Only use modelled XCO2 values from fluxes optimized against surface data Large (> 1-2 ppm) systematic differences are probably NOT from data biases! These diffferences are what inversions will use to change fluxes. Model Biosphere/ Transport Inversion Fires Type CarbonTracker CASA/GFED TM5/ECMWF EnKF 2013ei MACC v12.2 ORCHIDEE LMDZ/ECMWF Variational Univ. Edinburgh CASA/GFED GEOS- EnKF 3 CHEM/GEOS5
All sounding statistics: 13 Tells us little On average: models give lower values compared to ACOS* • (ACOS overall level set via TCCON comparisons) Don’t learn much otherwise •
+ Monthly Averages 14 January 2010 ACOS – CT2011oi (ppm) ACOS – MACC2011 (ppm) CT2011_oi not enough positive flux • in Equatorial Africa Problematic MACC fluxes over • India, appear linked to seasonal cycle of uptake & respiration. ACOS - UoL (ppm) MACC has too strong S.H. sinks? • (seen via ocean data)
+ Monthly Averages 15 January 2010 ACOS – CT2013ei (ppm) ACOS – MACC2012 (ppm) CT2011_oi not enough positive flux • in Equatorial Africa Problematic MACC fluxes over • India, appear linked to seasonal cycle of uptake & respiration. ACOS - UoL (ppm) MACC has too strong S.H. sinks? • (seen via ocean data)
16 India Clear amplitude problem • Differences as large as 3.1 with CASA seasonal cycle ppm in monthly averages! vs. obs. MACC seasonal cycle better • amplitude, but phasing problem.
17 African Sahel Differences as large as 3.2 Large differences, missing • ppm in monthly averages! respiration signal or biomass burning in Dec-Feb. MACC shows generally • better agreement. No obs. April-October! •
+ OCO-2 vs. GOSAT data density 18 32 day repeat Cycle September 2010 GOSAT Observations 4x4 degree boxes OCO-2 Simulations
+ Summary 19 Direct inversions with GOSAT XCO2 are hampered by both model issues and observation biases. Direct comparison of XCO2 between Models and Observations is potentially useful to diagnose both model issues and observation biases. Retrieval biases tend to be ~ 1 ppm . Significantly larger model/observation differences point to model deficiencies. Several potential model weaknesses seen : Poor model seasonal cycle characterization in India Poor model representation of African Sahel (esp CT+UoL) See Poster P-26 (Lindqvist/Schuh) for detailed model/ACOS comparisons.
+ Open Questions 20 How can we best use some of these robust model- observation differences? Push simultaneous assimilation of GROUND and SPACE- BASED observations (e.g., CarbonTracker!) Work to improve the biosphere priors directly? Observational data gaps leave us blind in many regions and times of year – how much will OCO-2 mitigate this?
+ Backup 21
+ On Transcom Regions: 22 Getting better… OCEANS LANDS UoL MACC v12.2 CT2013 • Larger regional differences between GOSAT & Models • Substantial differences between the three Models in certain regions. • Largest Land differences over South America, Boreal regions • Smaller differences over ocean
+ ACOS Truth Proxies: 23 TCCON & Models TCCON : Models : SRON/KIT/Basu Colocation Use soundings where all models • • Described in Guerlet et al., 2013 agree to within ~1 ppm. • Yields larger number of accurate Model mean is best guess. • • colocations Models: MACC, CT2011_oi, U. • Data from 2009-2012, 15+ stations Edinburgh (x2), NIES (x2), D. Baker • TM5 Accepted Rejected Mar/Apr/May
+ Temperate North America 25
+ Monthly Averages 26 January 2010 ACOS – CT2011oi (ppm) ACOS – MACC2011 (ppm) CT2011_oi not enough positive flux • in Equatorial Africa Problematic MACC fluxes over • India, appear linked to seasonal cycle of uptake & respiration. ACOS - UoL (ppm) MACC has too strong S.H. sinks? • (seen via ocean data)
+ Monthly Averages 27 January 2010 ACOS – CT2011oi (ppm) ACOS – MACC2011 (ppm) CT2011_oi not enough positive flux • in Equatorial Africa Problematic MACC fluxes over • India, appear linked to seasonal cycle of uptake & respiration. MACC Fluxes kgC/m 2 /yr MACC has too strong S.H. sinks? • (seen via ocean data)
28 Differences as large as 3.1 Clear amplitude problem • ppm in monthly averages! with CASA seasonal cycle vs. obs. MACC seasonal cycle better • amplitude, but phasing problem.
29 Sahara For comparison: the Saharan region
Australia + 30 Nov 2009 Dec 2009 Larger emissions seen in GOSAT data Jan 2009 Forest fires prevalent in • Australia in December- January
31 Amazon Large bias between models & obs! • GOSAT retrievals or model issue? • Potential causes? • Data gaps leave us blind ½ the year! •
+ Regional differences generally don’t align 32 with Transcom-3 regions!
Amortized Complexity of Zero- Knowledge Proofs Revisited: Achieving Linear Soundness Slack Ronald Cramer (CWI) Ivan Damgrd (AU) Chaoping Xing (NTU) ChenYuan (NTU) Eprint 2016/681 Integer One-Way Function (iOWF) maps integers to finite
582 views • 32 slides
A Computational Understanding of Classical (Co)Recursion P a ul Downen a nd Zen a M. Ariol a PPDP 2020, September 8 10 Topic Topic Both programs and proofs with loops Topic Both programs and proofs with loops (Co)Recursion and
1.74k views • 145 slides
Divide and conquer Philip II of Macedon Divide and conquer 1) Divide your problem into subproblems 2) Solve the subproblems recursively, that is, run the same algorithm on the subproblems (when the subproblems are very small, solve them from
1.39k views • 116 slides
Scalable Gaussian Processes Zhenwen Dai Amazon September 4, 2018 @GPSS2018 Zhenwen Dai (Amazon) Scalable Gaussian Processes September 4, 2018 @GPSS2018 1 / 55 Gaussian process Input and Output Data: X = ( x 1 , . . . , x N ) y = ( y 1 ,
620 views • 57 slides
Ground-Based Nuclear Explosion Monitoring R&D Validation of Regional Seismic Travel Time (RSTT) Predictions and Use in Event Location Stephen C. Myers 1 Michael L. Begnaud 2 Sanford Ballard 3 Abelardo L. Ramirez 1 , Michael E. Pasyanos 1 , W.
829 views • 28 slides
1 CSCI 350 Ch. 7 Scheduling Mark Redekopp Michael Shindler & Ramesh Govindan 2 Overview Which thread should be selected to run on the processor(s) to yield good performance? Does it even matter? Does the common case of
976 views • 54 slides
Large-Scale Data Fusion for Improved Model Simulation and Predictability Ahmed Attia Mathematics and Computer Science Division (MCS), Argonne National Laboratory (ANL) Thanks to Adrian Sandu: VTech Emil M. Constantinescu: ANL/UChicago Alen
1.16k views • 102 slides
Seismic Modeling, Migration and Velocity Inversion Full Waveform Inversion Bee Bednar Panorama Technologies, Inc. 14811 St Marys Lane, Suite 150 Houston TX 77079 May 30, 2014 Bee Bednar (Panorama Technologies) Seismic Modeling, Migration
850 views • 53 slides
CS688: Web-Scale Image Retrieval Inverted Index Sung-Eui Yoon ( ) Course URL: http://sgvr.kaist.ac.kr/~sungeui/IR Class Objectives Discuss re-ranking for achieving higher accuracy Spatial verification Query expansion
539 views • 29 slides
Formulation of the . . . Formulation of the . . . Mathematical and Main Idea of the Paper Pareto Optimality Computational Aspects of a Genetic Algorithm: . . . Selecting a Single Model Joint Inversion Paper by M. Moorkamp, A. G. Jones,
412 views • 7 slides
Formal Verification of a WCET Estimation Tool Sandrine Blazy 1 , Andr Maroneze 1 , David Pichardie 2 , Isabelle Puaut 1 1 University of Rennes 1 France 2 ENS Rennes, France 08/07/2014 1/30 Motivation Formal methods in industry Formal
751 views • 47 slides
lecture 1 - two's complement - floating point numbers - hexadecimal Mon. January 11, 2016 Car odometer (fixed number of digits) If you know what "modular arithmetic" is (MATH 240), then you recognize this: addition of integers
453 views • 33 slides
Fall 2017 :: CSE 306 Paging in Virtual Memory Nima Honarmand (Based on slides by Prof. Andrea Arpaci-Dusseau) Fall 2017 :: CSE 306 Problem: Fragmentation Definition: Free memory that cant Segment A be usefully allocated Segment B
509 views • 39 slides
Aspect-Oriented Programming with Dependency Injection Mark Seemann @ploeh Cross-Cutting Concerns Security Security Auditing Auditing Logging Logging Caching Caching Metering Metering Stability patterns Stability patterns Transactions
622 views • 26 slides
Estimating long-run coefficients and bootstrapping in large panels with cross-sectional dependence using xtdcce2 2019 London Stata User Group Meeting Jan Ditzen Heriot-Watt University, Edinburgh, UK Center for Energy Economics Research and
634 views • 49 slides
Angel Investors Critical Initiators of Startups and Job Creation SEC Advisory Council on Small and Emerging Companies September 17, 2013 David Verrill ACA Chairman and Founder/Managing Director, Hub Angels Marianne Hudson ACA
1.01k views • 47 slides
#FORUMCON19 Opportunity Zones, Impact Investing and Loan Guarantees: What is the Role for PSOs? Melanie Audette , Senior Vice President and Partner Engagement, Mission Investors Exchange, @melaudette Lyn Hunter , Director, Regional
426 views • 26 slides
Should Cyber-Insurance Providers Invest in Software Security? Aron Laszka 1 and Jens Grossklags 2 1 University of California, Berkeley 2 Pennsylvania State University Software Vulnerabilities Most software products suffer from vulnerabilities
448 views • 22 slides
Assessing the causal attribution of impact investments www.enterprise-development.org 1 o Overview of key findings from DCED Todays study on fund managers and assessing attributable impact session o Audience views on the role of different
511 views • 18 slides
The natural mathematics arising in information theory and investment Thomas Cover Stanford University Page 1 of 40 Felicity of mathematics We wish to maximize the growth rate of wealth. There is a satisfactory theory. The strategy achieving
659 views • 40 slides
Recommendation: Invest in Increasing Supply of High Quality ECE Programs to Address Shortage Minnesota: Partnering to boost the supply of high-quality early care and education providers Build Your Own Success to date Customized Approaches to
391 views • 9 slides
Game Theory -- Lecture 2 Patrick Loiseau EURECOM Fall 2016 1 Lecture 1 recap Defined games in normal form Defined dominance notion Iterative deletion Does not always give a solution Defined best response and Nash
546 views • 41 slides
4Q16 Supplem ental Slides John C. R. Hele Chief Financial Officer Table of Contents Page Effect of Derivative Losses... 3 Net Derivative Gains (Losses). 4 2016
566 views • 10 slides
Lecture: Continuous Time Models with Investment Applications Simon Gilchrist Boston Univerity and NBER EC 745 Fall, 2013 Brownian Motion Brownian motion (Wiener process): Continous time stochastic process with three properties: Markov
1.01k views • 59 slides