The bias varies by scenario category, here for 1.5°C scenarios with "no or low overshoot" as highlighted in #SR15. Should REMIND & POLES get more weight then GCAM or MERGE?
2/
2/
I often take medians of scenarios to simplify communication, trying to reveal the "key characteristics" of 1.5°C (trying to find robust messages). The scenario community does not like this approach
3/
3/
One approach is to select individual scenarios (figure shows 4 #SR15 illustrative scenarios), which is fine in principle, but suffers the same problem. This helps shows how scenarios vary, but the choice is inherently value based (not statistical).
5/
5/
An alternative is to plot all scenarios & label them, something I have done with #IPCC #AR5 scenarios (as have others). This makes it quite clear that models can give very different outcomes... Figure shows CCS in <2°C scenarios. cicero.oslo.no/en/posts/klima…
6/
6/
More robust, perhaps, is that model groups flag one scenario as their base case to be used in model ensembles (select one 1.5°C scenario from each model framework). This was done in some figures in #AR5. Other scenarios can be used for sensitivity analyses. 7/
Important to not that emission scenarios cannot be treated statistically as climate scenarios. Many of the same problems exist, but it is much harder to get a representative statistical sample of the scenario space (easier to sample climate space for a given scenario). 8/8
