is sensitive to model assumptions. Together, the models do not cover all the plausible futures.
Kram said that, in order to use the scenarios to study IAV questions, it would be necessary to map levels of vulnerability against the climate signal given by the scenarios. It might be very useful to analyze two or three levels of vulnerability against several strengths of climate signal. A different approach is needed for mitigation, Kram said, and the mitigation community may want to examine whether particular RCPs are consistent with certain assumptions about technology or about potential international agreements.
In summary, Kram said that these RCPs, especially the middle ones, are consistent with a very broad range of futures in terms of key socioeconomic variables. This suggests that looking at a small set of socioeconomic scenarios might be very useful in IAV research.
In the discussion, the following issues were raised:
The most recent scenarios tend not to cover the full socioeconomic space. For example, most of the estimates of population are near the UN median projection.
The conclusions about low correlations are based only on a first pass and reflect a selection bias in the scenarios and the lack of examination of interactions in which some combinations of economic assumptions are inconsistent with some pathways.
There is debate in the community about whether any 2.6 scenario is reasonable. That scenario was created as an “only-if” scenario—the idea was that 2.6 could be reached only if all the assumptions in the scenario are met.
The potential to change net emissions with agricultural technology and increased uptake of carbon in soils may not have been fully examined in all the models.
Interaction between the IAM and IAV communities will be needed to consider interaction effects and to develop skeletally described scenarios for the RCP forcing levels for the IAV community. Such analyses would need conceptual design, for example, to decide on how to focus on particular vulnerable groups.
An econometric approach might be used to capture interactions.
Many scenarios cannot reach 2.6. It might be useful to develop a 3.7 scenario—a value that is in a lot of other analyses. Second-best solutions are important for reaching 3.7.
Is it sensible to put probabilities on the scenarios? This has been tried with simpler models, for example, using multimodel analysis of key assumptions underlying the RCPs Monte Carlo analysis.