Estimating TCR using an integrated model-observation framework that accounts for spatio-temporal variability and pre-industrial radiative imbalances.

Apart from a few exceptions (e.g. Aldrin et al. 2012, Skeie et al. 2013) TCR estimates with EBMs are based on global data. Since these estimates don't represent the true spatial-temporal behaviour for observed temperature as well as external forcing (Marvel et al. 2015), we have developed a two-box EBM framework that accounts for these effects. In addition, external forcing from anthropogenic aerosol and GHGs tends to have different response times in comparison to volcanic stratospheric aerosols. Using PMIP3 and an extended ensemble of HadCM3 simulations (Euro500; Schurer et al. 2014) GCM simulations for the pre-industrial period, we obtain the fast and slow response time constants required in the EBM. With the most recent anthropogenic and natural forcing estimates, we test a range of TCR values against observations. The TCR/ECS ratio necessary to achieve that goal is taken from CMIP5 as observationally OHC-based estimates are notoriously unreliable. Given that observed and modelled OHC estimates are in agreement (Cheng et al. 2016), we argue that this should be the standard procedure the make inferences about ECS. Alternatively, it should be distinguished between equilibrium and effective climate sensitivity. The preliminary best estimate for TCR is 1.6K (1.1-2.2K) with an associated ECS value of 2.9K (2.0-4.0K). This is in good agreement with other D&A techniques that do use spatio-temporal patterns as well (e.g. Jones et al. 2016, Gillet et al. 2013). Correcting for natural ENSO variability and tas/tos-related inaccuracies (Richardson et al. 2016) further increases the robustness of the estimated sensitivity range. Our results also indicate that the small radiative imbalance caused by the period of very strong volcanic eruptions just before the CMIP5 historical period starts (1809-1840) has noteworthy implications for the response to later volcanic eruptions and the temperature evolution after 1850. Simply put, CMIP5-type simulations are slightly more sensitive to volcanic eruptions than PMIP3-type simulations. This has been pointed out in the literature before (e.g. Gleckler et al. 2006, Stenchikov et al. 2009, Gregory et al. 2010). We therefore argue that more PMIP3-type of experiments are needed to factor in the planetary energy imbalance caused by earlier volcanic eruptions.