The “uncertainty monster” strikes again
We’ve been highly critical for some time of the paper in summer 2015 by Karl et al. that claimed “the pause” or hiatus went away once “properly adjusted” ocean surface temperature data was applied to the global surface temperature dataset. Virtually everyone in the climate skeptic community considers Karl et al. little more than a sleight of hand.
No matter, this paper published today in Nature Climate Change by Hedemann et al. not only confirms the existence of “the pause” in global temperature, but suggests a cause, saying “…the hiatus could also have been caused by internal variability in the top-of-atmosphere energy imbalance“.
That’s an important sentence, because it demonstrates that despite many claims to the contrary, CO2 induced forcing of the planetary temperature is not the control knob, and natural variability remains in force.
Also of note, see the offset as designated by the two colored X’s in Figure 1:
Models and observations don’t even begin to match.
The subtle origins of surface-warming hiatuses
Christopher Hedemann, Thorsten Mauritsen, Johann Jungclaus & Jochem Marotzke
Nature Climate Change (2017) doi:10.1038/nclimate3274
Received 12 July 2016 Accepted 17 March 2017 Published online 17 April 2017
During the first decade of the twenty-first century, the Earth’s surface warmed more slowly than climate models simulated1. This surface-warming hiatus is attributed by some studies to model errors in external forcing2, 3, 4, while others point to heat rearrangements in the ocean5, 6, 7, 8, 9, 10caused by internal variability, the timing of which cannot be predicted by the models1. However, observational analyses disagree about which ocean region is responsible11, 12, 13, 14, 15, 16. Here we show that the hiatus could also have been caused by internal variability in the top-of-atmosphere energy imbalance. Energy budgeting for the ocean surface layer over a 100-member historical ensemble reveals that hiatuses are caused by energy-flux deviations as small as 0.08 W m−2, which can originate at the top of the atmosphere, in the ocean, or both. Budgeting with existing observations cannot constrain the origin of the recent hiatus, because the uncertainty in observations dwarfs the small flux deviations that could cause a hiatus. The sensitivity of these flux deviations to the observational dataset and to energy budget choices helps explain why previous studies conflict, and suggests that the origin of the recent hiatus may never be identified.
From the Introduction:
The surface temperature of the Earth warmed more slowly over the period 1998–2012 than could be expected by examining either most model projections or the long-term warming trend1. Even though some studies now attribute the deviation from the long-term trend to observational biases17, 18, the gap between observations and models persists. The observed trend deviated by as much as −0.17 °C per decade from the CMIP5 (Coupled Model Intercomparison Project Phase 5; ref. 19) ensemble-mean projection1—a gap two to four times the observed trend. The hiatus therefore continues to challenge climate science.
The coupled climate model MPI-ESM1.1 is forced with CMIP5-prescribed historical forcing from 1850 until 2005, and extended until 2015 with the RCP4.5 scenario (see Methods). When the red line lies above the grey line, at least one ensemble member is experiencing a hiatus, defined as a deviation of more than 0.17 °C per decade below the ensemble mean. This deviation is the same as the gap between the CMIP5 ensemble mean (black cross) and the observed (yellow cross) GMST trends for the period 1998–2012. Contours represent the number of ensemble members in bins of 0.05 °C per decade.
From our analysis of observational estimates, we are unable to exclude the TOA anomaly as a possible cause of the recent hiatus. Referencing the observations to an alternative energy budget (rather than that of the large ensemble) could shift the absolute position of the green and yellow crosses in Fig. 3c. However, their relative distance from one another and the size of their error bars would not change.
The role of the TOA and the ocean in each hiatus can be determined by comparing their relative contributions to the flux-divergence anomaly. For hiatuses in the large historical ensemble, the negative (cooling) anomaly is caused entirely by the TOA in 12% of cases and by the ocean in 24% of cases. In the remainder (64%), the negative anomaly is caused by the TOA and ocean acting together (bottom left quadrant of Fig. 3c). TOA variability is therefore involved in 76% of all hiatuses.
We conclude that the TOA may have been a source of significant internal variability during the hiatus. Our conclusions are not an artefact of model-generated TOA variability29—the large ensemble produces TOA variability that is similar to that in the observational record (Supplementary Fig. 7). Rather, our conclusions are based on a simple yet robust principle, namely that the Earth’s surface layer has a small heat capacity. The surface temperature can therefore be influenced by small variations in the large yet mutually compensating fluxes that make up this layer’s energy budget. Comparing the small variability in the TOA imbalance with the total TOA imbalance under global warming27, 30 obscures the significance of these small variations for the hiatus.
This is the true dilemma at the heart of the hiatus debate: the variability in ocean heat content alone has no power to explain the hiatus, and the measure that can—the surface-layer flux divergence—is dwarfed by observational uncertainty. While there are attempts to fill the gaps in observations with ocean reanalyses such as ORAS4 (refs 9,23), the resulting data are of questionable integrity during the hiatus14, 21 and, as we show, disagree with the budget based on CERES21 and WOA22. Even if these disagreements could be reconciled, the process of anchoring satellite observations with ocean heat uptake makes the contributions from TOA and ocean difficult to disentangle, because their absolute difference is unknown. Therefore, unless the uncertainty of observational estimates can be considerably reduced, the true origin of the recent hiatus may never be determined.
The MPI-ESM1.1 model version was used to generate the large ensemble and is available at http://www.mpimet.mpg.de/en/science/models/mpi-esm.html. Computer code used in post-processing of raw data has been deposited with the Max Planck Society: http://pubman.mpdl.mpg.de/pubman/faces/viewItemFullPage.jsp?itemId=escidoc:2353695.
Raw data from the large ensemble were generated at the Swiss National Computing Centre (CSCS) and Deutsches Klimarechenzentrum (DKRZ) facilities. Derived data have been deposited with the Max Planck Society (http://pubman.mpdl.mpg.de/pubman/faces/viewItemFullPage.jsp?itemId=escidoc:2353695). Supplementary Fig. 7 uses TOA flux reconstructions provided by R Allan40 (http://www.met.reading.ac.uk/~sgs01cll/flux) and satellite observations provided by the NASA CERES project31 (http://ceres.larc.nasa.gov). For observational estimates in Fig. 3c, we make use of data provided by the NOAA World Ocean Atlas22(https://www.nodc.noaa.gov/OC5/3M_HEAT_CONTENT) and by the ECMWF Ocean Reanalysis System 4 (ref. 9; http://icdc.zmaw.de/projekte/easy-init/easy-init-ocean.html).
Supplementary Information (3,616 KB)
Note: about 30 mins after publication, some grammatical and spelling errors were corrected, and a subtitle added.