by Judith Curry
On the uncertainty monster thread, a scenario was defined as a plausible but unverifiable description of how the system and/or its driving forces may develop in the future. Scenarios may be regarded as a range of discrete possibilities, often with no a priori allocation of likelihood. An example is the future greenhouse gas emission scenarios used to force global climate models.
I hadn’t paid much attention to issues surrounding emissions scenarios, assuming that whoever was putting these together was doing a reasonable job, with new emissions scenarios being prepared for IPCC’s 5th Assessment report. Recently, I encountered a paper by Betz, entitled “Underdetermination, model-ensembles and surprises: on the epistemology of scenario-analysis in climatology” and the first few sentences of the abstract really caught my attention:
“As climate policy decisions are decisions under uncertainty, being based on a range of future climate change scenarios, it becomes a crucial question how to set up this scenario range. Failing to comply with the precautionary principle, the current scenario methodology of the IPCC seems to violate international environmental law, in particular a provision of the United Nations Framework Convention on Climate Change.”
Now I’m no fan of the precautionary principle, but I found this too irresistible not to investigate.
The IPCC’s scenarios of future climate change have two elements: (1) creation of a suite of emissions scenarios and (2) climate model simulations forced by the emission scenarios. This post focuses on (1); (2) will be the topic of next week’s post.
Background on emissions scenarios
Basic references that describe emissions scenarios are:
- Moss et al. The next generation of scenarios for climate change research and assessment
- IPCC’s Special Report on Emissions Scenarios (SRES)
- CCSP SAP 2.1 Scenarios of Greenhouse Gas Emissions and Atmospheric Concentrations
- O’Niell & Nakicenovic Learning from global emissions scenarios
Moss et al. describe emissions scenarios as:
“Emissions scenarios are descriptions of potential future discharges to the atmosphere of substances that affect the Earth’s radiation balance, such as greenhouse gases and aerosols. Along with information on other related conditions such as land use and land cover, emissions scenarios provide inputs to climate models. They are produced with integrated assessment models based on assumptions about driving forces, such as patterns of economic and population growth, technology development, and other factors. Over time, the information provided by integrated assessment models and used in climate models has become increasingly comprehensive, including time-dependent emissions of radiatively significant gases and particles, precursor pollutant compounds, and land cover and use. . . Emissions scenarios for climate change research are not forecasts or predictions, but reflect expert judgments regarding plausible future emissions based on research into socioeconomic, environmental, and technological trends represented in integrated assessment models.”
Scenarios used in the IPCC 3rd and 4th Assessment Report (the SRES scenarios) were produced to provide a common basis for climate change projections, mitigation analyses, and impact assessments. The SRES scenarios were developed using storylines (or narratives of the future) derived from a wide range of driving forces.
For the AR5 (Moss et al.) there is a redesigned scenario process that starts with scenarios of radiative forcing for 2100. The four different forcing scenarios used correspond to the following peak levels of atmospheric CO2: 490 ppm, 650 ppm, 850 ppm, 1370 ppm. These radiative forcing trajectories are not associated with unique socioeconomic or emissions scenarios, and the role of models in the scenario builiding is to develop a time line of emissions and an internally consistent scenario that includes the other long lived greenhouse gases plus aerosols. Also new to the AR5 are storylines associated with possible mitigation policies.
The Kaya identity
A simple way of thinking about emissions scenarios is enabled by the Kaya identity, which frames the determination of IPCC emissions scenarios using four different inputs: population growth projections ; Gross Domestic Product (GDP) per capita trends; energy use per unit of GDP; and emissions per unit energy consumed. The identity is expressed in the form (pulled from the Wikipedia article):
F = P * (G / P) * (E / G) * (F / E) = P * g * e * f
where F is global CO2 emissions from human sources, P is global population, G is world GDP and g = (G/P) is global per-capita GDP, E is global primary energy consumption and e=(E/G) is the energy intensity of world GDP, and f=(F/E) is the carbon intensity of energy. Play with the online Kaya calculator here.
Criticisms of the IPCC SRES scenarios
As per the Wikipedia, the IPCC SRES scenarios have been criticized for their choice of exchange rates, estimates of energy intensity, estimates of resource availability, and emission rates after 2000 that are more rapid than projected.
O’Neill and Nakicenovic argue that the range of scenarios is too narrow: the scenarios should be expanded to address specific questions and needs; the scenarios do not adequately address the full range of uncertainties in emissions over the next few decades; and there has been inadequate investigation of what might constitute the “barely feasible” scenario and under what conditions such a scenario might be created.
The new scenarios for the AR5 seem to address many of the criticisms, although it is not clear that a CO2 concentration of 1370 ppm in 2100 is high enough to qualify as a “barely feasible” scenario.
Modal inductivism and falsification
At the heart of Betz’s argument regarding IPCC scenarios violating the precautionary principle is the concept of modal logic.
Modal logic extends propositional logic to include the classification of propositions according to whether they are contingently true or false, possible, impossible, or necessary. Betz argues for the logical necessity of considering future climate scenarios as modal statements of possibilities.
Betz contrasts two general methodological principles for constructing the IPCC climate scenario range: modal inductivism and modal falsificationism. Modal inductivism states that a certain statement about the future is possibly true if and only if it is positively inferred from our relevant background knowledge. By this token, the story lines and models used to generate the IPCC SRES emissions scenarios, along with climate model simulations forced by these scenarios, are scenarios of the future determined from modal inductivism.
Modal falsificationism further permits creatively constructed scenarios to be accepted as long as the scenarios cannot be falsified by being incompatible with background knowledge. Developing a suite of scenarios by modal falsification is a two step process: the first step is coming up with as many potential future scenarios as possible (using inductive as well as other more creative methods) and then submitting these future scenarios to tests in order to see which ones can be discarded as impossible. The AR5 scenarios seem more consistent with a modal falsification approach, but it is unclear how the top emission value (associated with 1370 ppm) was determined; the criteria seems to have been plausibility rather than barely feasible. The O’Neill and Nakicenovic concept of the “barely feasible” scenario is consistent with modal falsification.
So how does scenario creation by modal inductivism violate the precautionary principle?
Definitions of the precautionary principle have two key elements (Wikipedia):
- an expression of a need by decision-makers to anticipate harm before it occurs. Within this element lies an implicit reversal of the onus of proof: under the precautionary principle it is the responsibility of an activity proponent to establish that the proposed activity will not (or is very unlikely to) result in significant harm.
- the establishment of an obligation, if the level of harm may be high, for action to prevent or minimise such harm even when the absence of scientific certainty makes it difficult to predict the likelihood of harm occurring, or the level of harm should it occur. The need for control measures increases with both the level of possible harm and the degree of uncertainty.
Betz’s argument A18 basically states that if the precautionary principle is the normative premise, scenarios developed by modal inductivism exclude plausible catastrophic scenarios (anticipating harm) and so violate the precautionary principle.
Although not mentioned by Betz, the concept of possibility distribution seems very useful for characterizing a collection of modal statements and therefore to characterizing a group of scenarios.
Possibility theory is an imprecise probability theory driven by the principle of minimal specificity that states that any hypothesis not known to be impossible cannot be ruled out. In contrast to probability, possibility theory describes how likely an event is to occur using the dual concepts of the possibility and necessity of the event, which makes it easier to capture partial ignorance. A possibility distribution distinguishes what is plausible versus the normal course of things versus surprising versus impossible. Possibility theory can be interpreted as a non-numerical version of probability theory or as a simple approach to reasoning with imprecise probabilities.
An interesting application of possibility theory to the risk of terrorism attacks is here, see especially Fig 3.2 on p 3-7 that illustrates the possibility distribution graphically.
JC’s modest suggestion
The strategy for developing emissions scenario for the AR5 seems to be much improved over the previous strategy used in the SRES. Nevertheless, Betz’s argument about about the the IPCC emissions scenarios violating the precautionary principle seems valid. Even if the precautionary principle is not used, other strategies for decision making under deep uncertainty require information on the “barely feasible” worst case scenario.
So how might the IPCC proceed in this regard? First, the complicated models that develop emissions scenarios don’t seem to be necessary for forcing the climate models; simply specifying a value of CO2 concentration (with the other greenhouse gases and anthropogenic aerosol) at 2100 along with a simple time trajectory is sufficient to force the climate model. The value of the emission models would be in establishing the “barely feasible” worst case scenario and the conditions under which this scenario might be created, and in rejecting more extreme scenarios.
The individual scenarios in the IPCC scenario suites (both SRES and AR5) are implicitly regarded as equally plausible. Armed with Kaya’s identity (or the more sophisticated emission models), modal falsification, and the possibility distribution, it seems that there is a feasible and credible method for establishing the relative likelihood of the different radiative forcing scenarios. Inverse modeling using Kaya’s identity could identify the number of different pathways among the various combinations of possible input variables that could result in a specific radiative forcing scenario (say +/- 10%) . The number of different combinations of variables that would produce a particular forcing scenario would provide some sense of the likelihood of that scenario (with the barely feasible scenario having only one combination of variables, and so being the least likely). A further embellishment could be provided by ranking the input values for at least some of the input variables in terms of their likelihood (from necessary to barely feasible). This would provide a rationale for the size of the bar (on the possibility to necessity scale) related to that particular scenario.
It would also be useful to assess the likelihood for each radiative forcing scenario in terms of the likelihood of exceeding the value of the scenario. For example, exceeding the 400 ppm scenario by 2100 would probably be classified as necessary, whereas exceeding the barely feasible strategy would be ranked as impossible.
This seems to me to be a plausible way to tame the scenario uncertainty monster, but since I have only been thinking about this for a short time and am no expert on this topic, these ideas could be off the mark in some way. In any event, Betz makes a compelling argument for including the “barely feasible” scenario, without which the IPCC’s scenarios and their use by the UNFCCC is arguably fails to comply with the precautionary principle.
I look forward to your assessment of this and suggestions on fleshing out these ideas.