CarbonTracker CT-NRT.v2017

CarbonTracker CT-NRT.v2017 is an extension of the standard CarbonTracker CO2 analysis system, designed to provide more timely model results. It was created under funding from NASA, for use in the OCO-2 program. CT-NRT uses "real-time" ERA-interim transport in TM5.

CT-NRT.v2017 differs from our standard CarbonTracker projects in several ways:

  • Assimilation of provisional CO2 observations. In collaboration with measurement scientists from NOAA Global Monitoring Division and Environment Canada's Greenhouse Gas Measurement Program. (i.e., fewer obs and less quality control). CT-NRT.v2017 uses observational data from two ObsPacks: Prior to 2016-Jan-01, data are extracted from the GLOBALVIEWplus_v2.1_2016-09-02 ObsPack. After 2016-Jan-01, data come from the NRT_v3.3_2017-04-19 ObsPack.
  • Use of a different prior flux model. Our standard land biosphere flux prior is not available in near-real time, so we had to develop an alternative first-guess flux estimate. For this, we take the climatology of optimized land, wildfire, and ocean fluxes from the latest standard CarbonTracker release (for this release, that is CT2016). Since the majority of flux variability comes from the land biosphere, a land flux anomaly model also has been developed. This is a simple regression of CT2016 land flux anomalies as a function of anomalies of precipitation, shortwave radiation, and temperature. The statistical flux anomaly model was developed for each CarbonTracker ecoregion, and provides daily estimates of NEE anomaly. The radiation and temperature data for this model come from the ERA-interim meteorology used by TM5; the precipitation anomaly comes from the Global Precipitation Climatology Project. Although this flux prior does not reliably reproduce interannual variability, it is arguably a more statistically optimal prior than that for our standard release since it already represents the long-term mean CO2 sinks that we know exist. Since the priors for our standard release do not represent these sinks, the standard CarbonTracker requires that CO2 observations correct these biased priors.
  • Changes to represent the 2015-2016 El Niño This release represents a significant change to the CarbonTracker methodology, implemented to simulate the 2015-2016 El Niño. This is the first major El Niño to have occurred since the beginning of the CarbonTracker project, and some modifications were needed to the assimilation scheme. Those two changes were:
    1. Increase the assimilation window length from 5 weeks to 12 weeks. This means that the last week for which the fluxes were optimized using a full 12 weeks of data starts on 2017-Feb-04. Results are presented for the following 3 months, but each subsequent week has one week less of optimization. That means fluxes in the week starting 2017-Feb-11 have seen only 11 weeks of optimization; those in 2017-Feb-18 have seen 10 weeks, etc.
    2. Inflate prior covariance for land regions by a factor of 4 (in variance space; factor of 2 in std. dev. space).
    These changes are experimental, and we'd appreciate any feedback on the performance of CT-NRT.v2017. If there are no problems detected with this assimilation scheme, we may implement these modifications for CT2017.

Versions CT-NRT.v2017 extends from 2015-12-12 (near the end of CT2016) through 2017-04-29. Results can be downloaded from CT-NRT.v2017 starts from a CT2016 initial condition of 2015-12-12 using the eic-b2oifmi suite member.

Previous releases of CT-NRT are available from