bligh8 cited the first reference in the "Hansen et al paper: 3+ meters SLR by 2100" thread; however, I cite it in this thread as I believe that it provides a great example of how, in my opinion, AR6 should present the significance of the 'deep uncertainty' of the potential impacts of the many dynamical Earth Systems that AR6 is almost certain to be overconfident about. The Bakker et al (2017) reference presents a worked example of how to calibrate a scenario (see the first image for part of the calibration process using both paleo ,and observed, data) for the potential collapse of the WAIS this century. Bakker et al (2017) then effectively summarize the findings from their calibrated scenario, w.r.t. to its impact on SLR, in an easily understood plot (see the second attached image) of Sea level with time and scenarios with varying degrees of 'deep uncertainty. Also, I note that Bakker et al (2017) indicate that: "Around 2040-2050, a large and uncertain contribution of the GIS becomes important …"
Alexander M. R. Bakker, Tony E. Wong, Kelsey L. Ruckert & Klaus Keller (2017), "Sea-level projections representing the deeply uncertain contribution of the West Antarctic ice sheet", Scientific Reports 7, Article number: 3880; doi:10.1038/s41598-017-43134-5
http://www.nature.com/articles/s41598-017-04134-5Abstract: "There is a growing awareness that uncertainties surrounding future sea-level projections may be much larger than typically perceived. Recently published projections appear widely divergent and highly sensitive to non-trivial model choices. Moreover, the West Antarctic ice sheet (WAIS) may be much less stable than previous believed, enabling a rapid disintegration. Here, we present a set of probabilistic sea-level projections that approximates the deeply uncertain WAIS contributions. The projections aim to inform robust decisions by clarifying the sensitivity to non-trivial or controversial assumptions. We show that the deeply uncertain WAIS contribution can dominate other uncertainties within decades. These deep uncertainties call for the development of robust adaptive strategies. These decision-making needs, in turn, require mission-oriented basic science, for example about potential signposts and the maximum rate of WAIS-induced sea-level changes."
Extract: "Our sea-level projections are constructed to support robust decision frameworks by i) being explicit about the relevant uncertainties, both shallow and deep; ii) communicating plausible ranges of sea-level rise, including the deep uncertainties surrounding future climate forcings and potential WAIS collapse; and iii) tending to err on the side of underconfident versus overconfident when possible.
Model design. We design the projections to be probabilistic where reasonable and explicit about deep uncertainties (e.g. resulting from non-trivial model choices) when needed. Robust decision frameworks often apply plausible rather than probabilistic ranges to represent and communicate uncertainties. In the case of sea-level projections, the bounding of the plausible range usually involves both a probabilistic interpretation of the surrounding uncertainties and estimates of which probabilities are still relevant. For example, a full disintegration of the major ice sheets is often not taken into account because the probabilities of this occurring are considered too
small to be relevant. What probability is relevant is highly dependent on the decision context and therefore it makes sense to be explicit about the probabilities. Moreover, probabilities are the easiest and most unambiguous way to communicate uncertainties.
Our projections are designed to highlight the relatively large deep uncertainties, notably those resulting from future climate forcings and those surrounding potential WAIS collapse (even though representations of deep uncertainty often implicitly encompass probabilistic interpretations). The future climate forcing is, to a large extent, controlled by future human decisions.
The probability of a WAIS collapse is potentially much larger than previously thought due to the combined effects of Marine Ice Sheet Instability (MISI), ice cliff failure and hydrofracturing. The discovery of this new mechanism puts earlier expert elicitations in a different light as it is unclear if those were based on this combined effect. One approach when faced with deeply uncertain model structures and priors is to present a potential WAIS collapse as deeply uncertain by means of a plausible range. We stress that this range is not meant to represent an implicit probabilistic projection of the WAIS contribution to sea-level rise.
We merge some small deep uncertainties into the probabilistic part of the projections. According to Herman et al. “… a larger risk lies in sampling too narrow a range (thus ignoring potentially important vulnerabilities) rather than too wide a range which, at worst, will sample extreme states of the world in which all alternatives fail”. Thus, in the context of informing robust decision making, it can be preferable to be slightly under- than slightly overconfident. To minimize the risk of producing overconfident projections we only use observational data with relatively uncontroversial and well-defined error structure.
Model setup. We use a relatively simple (39 free physical and statistical parameters), but a mechanistically motivated model framework to link transient sea-level rise to radiative concentration pathways applying sub-models for the global climate, thermal expansion (TE), and contributions of the Antarctic ice sheet (AIS), Greenland ice sheet (GIS) and glaciers and small ice caps (GSIC) (see Methods). This approach extends on the semi-empirical model setup recently reported by Mengel et al..
We use a Bayesian calibration method, wherein paleoclimatic data is assimilated with the AIS model separately from the calibration for the rest of the model, which assimilates only modern observations. Modern model simulations are then run at parameters drawn from the two resulting calibrated parameter sets (AIS and rest-of-model) and compared to global mean sea-level (GMSL) data (see Methods). Only model realizations which agree with each GMSL data point to within 4σ are admitted into the final ensemble for analysis. 4σ was chosen so the spread in the model ensemble characterizes well the uncertainty in the GMSL data (Fig. 1f). We choose, at this time, not to use paleo-reconstructions nor reanalyses, beyond incorporating a windowing
approach into our calibration method for the Antarctic ice-sheet parameters. This choice is motivated by the highly complex and uncertain error structure of these data sets. Failure to account for such complex error structure can result in considerable overconfidence, especially for low-probability events."
Furthermore, Bakker et al (2017) cite the second linked reference which provides a worked example of how the potential bias of a current model can be quantified by comparing its projections against the projections of a dynamical model with 'deep uncertainty', in this cases one that includes the dynamical mechanism of cliff failures and hydrofracturing w.r.t. to SLR contributions from the Antarctic ice sheet. To the best of my understanding none of the Earth System Models in CMIP6 include the dynamical cliff failures, and hydrofracturing, mechanisms, apparently due to 'deep uncertainty'. Nevertheless, even if CMIP6/AR6 do not present projections including the impacts of the dynamical cliff failures, and hydrofractuing, mechanisms, they could still numerically present the potential bias of their projections by following the methodology presented by Ruckert et al (2017), & in this regards see the last two attached images.
Kelsey L. Ruckert, Gary Shaffer, David Pollard, Yawen Guan, Tony E. Wong, Chris E. Forest &Klaus Keller (2017), "Assessing the impact of retreat mechanisms in a simple Antarctic ice sheet model using Bayesian Calibration", PLoS ONE, 12, 1-15,
https://doi.org/10.1371/journal.pone.0170052 http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0170052Abstract: "The response of the Antarctic ice sheet (AIS) to changing climate forcings is an important driver of sea-level changes. Anthropogenic climate change may drive a sizeable AIS tipping point response with subsequent increases in coastal flooding risks. Many studies analyzing flood risks use simple models to project the future responses of AIS and its sea-level contributions. These analyses have provided important new insights, but they are often silent on the effects of potentially important processes such as Marine Ice Sheet Instability (MISI) or Marine Ice Cliff Instability (MICI). These approximations can be well justified and result in more parsimonious and transparent model structures. This raises the question of how this approximation impacts hindcasts and projections. Here, we calibrate a previously published and relatively simple AIS model, which neglects the effects of MICI and regional characteristics, using a combination of observational constraints and a Bayesian inversion method. Specifically, we approximate the effects of missing MICI by comparing our results to those from expert assessments with more realistic models and quantify the bias during the last interglacial when MICI may have been triggered. Our results suggest that the model can approximate the process of MISI and reproduce the projected median melt from some previous expert assessments in the year 2100. Yet, our mean hindcast is roughly 3/4 of the observed data during the last interglacial period and our mean projection is roughly 1/6 and 1/10 of the mean from a model accounting for MICI in the year 2100. These results suggest that missing MICI and/or regional characteristics can lead to a low-bias during warming period AIS melting and hence a potential low-bias in projected sea levels and flood risks."
Extract: " We calibrate a simple AIS model (that does not include a cliff instability mechanism nor is able to capture regional characteristics) with observational constraints over the past 240,000 years using a Bayesian inversion considering the heteroskedastic nature of the data. Using the hindcasts and projections, we compare our results to those from a pre-calibration method and expert assessments with potentially more realistic models. We approximate how neglecting fast processes (i.e., the MICI mechanism) in an AIS model can lead to biases in the AIS hindcasts and projections during warming periods. For the specific example considered, we show how missing MICI produces a lower mean hindcast (roughly 26% or 1 m smaller) during the LIG, a period when the marine ice sheet is suggested to have deglaciated. Additionally, the model is unable to account for roughly 96 and 100% of future AIS contributions predicted by a physically more realistic model accounting for MISI, MICI, and hydro-fracturing yet reproduces the projected median melt in other expert assessments in the year 2100. Overall, accounting for retreat mechanisms can potentially increase warming period AIS melt and reduce model discrepancy."