Presentation on theme: "How fossil fuel CO 2 uncertainty impacts estimates of carbon exchange and variability Kevin Gurney, Yang Song, Jianhua Huang, Kevin Coltin, Alex Garden."— Presentation transcript:
How fossil fuel CO 2 uncertainty impacts estimates of carbon exchange and variability Kevin Gurney, Yang Song, Jianhua Huang, Kevin Coltin, Alex Garden School of Life Sciences/School of Sustainability, Global Institute of Sustainability, Arizona State University, Tempe, AZ, 85281, USA
Introduction A focus on power plant uncertainty: More available data, particularly in US Large proportion of total (~40%) Examine US and global cases Examine National inventories - total spread Descriptive statistical characteristics of spans Impacts of national span on high resolution fossil fuel CO 2 data product (FFDAS) and atmospheric CO 2 Inversions
Vulcan Gurney et al., Env. Sci & Tech, 2009
Hestia See poster 217, Today
FFDAS+ Uncertainty quantification in all of these data products remains challenging
US Power Plants (almost 40% of US FFCO 2 ) Comparing EPA derived data (CEMs primarily) to DOE/EIA derived data (fuel consumption-based calculation)
US Power Plants continued Examine hourly data where one can isolate the “methods” used. CEMs is primary, but there are 6 alternative methods Require that every hour of a month be 100% single method Aggregate to month for comparison to DOE/EIA Mean: -1.8% aka “signed bias” CEMs
US Power Plants continued Method/type Average IRD (%) lo/hi quartile IRD (%)Data (%)CO2 (%) CEMs-1.8%-5 / %64.4% alternative 19.4%-21 / %0.01% alternative 2-5.7%-10 / %0.2% alternative %-31 / %0.06% alternative 49.3%-3 / % alternative 5-4.3%-10 / %35.2% alternative 6NS Why? Are the CEMs measurements biased lo/hi? Are the Fuel consumption estimates biased hi/lo? ……Stay tuned…. See poster 221, Today
Global Power Plants Currently the only available global dataset is CARMA. That includes more than 60,000 power plants worldwide. CARMA provides plant location and estimated CO 2 emission for each power plant. Statistical model based on WEPP & US data We analyze national publicly disclosed data relative to model prediction to assess accuracy.
Global Power Plants continued Country location bias (km) Total emit signed bias (%) Mean emit signed bias (%) Mean emit unsigned bias (%) USA ~05.9%-8.3%17% Canada %-25%95% India %2.7%19% S Africa %0.3%7.0% Australia50.0NA United Kingdom26.110%2.4%42% Germany6.1-26%-2.9%51% Italy10.037%-7.8%73% Netherlands4.0-27%-7.8%52% Spain %13%57% To the extent we want to utilize remote platforms to calibrate or monitor large point sources, these distance biases are unacceptable Use of power plants as “standalone” data sources in global FFCO2 estimation requires better accuracy
Global Fossil Fuel CO 2 Emission Inventories Carbon Dioxide Information Analysis Center (CDIAC) UN energy data from annual energy surveys International Energy Agency (IEA) Annual IEA energy surveys and UN energy data for non-member nations Sectoral approach (bottom up) vs. Reference approach (top down) Energy Information Administration (EIA) EIA’s review of National reports British Petroleum (BP) BP’s review of National reports
Macknick’s Harmonized Database Macknick, J. (2011) Energy and CO 2 emission data uncertainties. Carbon Manage. 2, 189−205. Energy Sources Survey (IEA, CDIAC) vs National Reports (EIA, BP) Categories International bunkers, biomass waste, cement production, gas flare… Conversions Calorific value Emission factors Units Cubic meters vs feet Short ton vs tonnes Etc.
World emission unadjusted World emission harmonized (Cement production from CDIAC; Gas flaring from EIA; no wastes, renewables and land-use emissions) ---harmonized data from Macknick, 2011 Time Mean Percentage span = 7.7% Time Mean Percentage span = 7.6% Teragrams of carbon Year
Implications – atmospheric CO 2 inversion Run “hi” and “lo” through TransCom 3 Level 2 inversions Scales the global fossil total No change in spatial pattern RESULTS: Total Land Uptake for the decade of the 1990’s “hi”………………………… -1.3 GtC/year “lo”………………………… GtC/year Total Land Uptake for the decade of the 2000’s “hi”……………………… GtC/year “lo”……………………… GtC/year High resolution inversion underway – implications will be more significant for some global regions.
= “hi”= “lo” = error range
-200%200%0% Impact on FFDAS Span as prior uncertainty (diagonal cov matrix only)
Conclusions Exploration of fossil fuel uncertainty remains challenging The single largest contributor and sector previously considered the best known – power plant emissions: US: potentially large error at facility level. Source of bias being investigated. I think we can solve. CARMA database: requires revision. Underway - VENTUS Discrepancies among national FFCO2 inventories varies through time and by country, even after “harmonization” The median percentage span averaged over countries and time is around 12.5% (mean = 15%) Impact of the span on fossil fuel CO 2 on the inversion results is large: difference exceeds the current “within” inversion error. Use of this as national uncertainty has significant impact on the FFDAS result ATMOSPHERIC MEASUREMENTS WILL HELP (OF COURSE), BUT IT ISN’T ENOUGH – THE DATA PRODUCTS MUST IMPROVE