 Original article
 Open Access
Applying Bayesian model averaging for uncertainty estimation of input data in energy modelling
 Monika Culka^{1}Email author
https://doi.org/10.1186/s1370501400219
© Culka; licensee Springer. 2014
Received: 2 May 2014
Accepted: 30 September 2014
Published: 24 October 2014
Abstract
Background
Energy scenarios that are used for policy advice have ecological and social impact on society. Policy measures that are based on modelling exercises may lead to far reaching financial and ecological consequences. The purpose of this study is to raise awareness that energy modelling results are accompanied with uncertainties that should be addressed explicitly.
Methods
With view to existing approaches of uncertainty assessment in energy economics and climate science, relevant requirements for an uncertainty assessment are defined. An uncertainty assessment should be explicit, independent of the assessor’s expertise, applicable to different models, including subjective quantitative and statistical quantitative aspects, intuitively understandable and be reproducible. Bayesian model averaging for input variables of energy models is discussed as method that satisfies these requirements. A definition of uncertainty based on posterior model probabilities of input variables to energy models is presented.
Results
The main findings are that (1) expert elicitation as predominant assessment method does not satisfy all requirements, (2) Bayesian model averaging for input variable modelling meets the requirements and allows evaluating a vast amount of potentially relevant influences on input variables and (3) posterior model probabilities of input variable models can be translated in uncertainty associated with the input variable.
Conclusions
An uncertainty assessment of energy scenarios is relevant if policy measures are (partially) based on modelling exercises. Potential implications of these findings include that energy scenarios could be associated with uncertainty that is presently neither assessed explicitly nor communicated adequately.
Keywords
 Uncertainty
 Energy modelling
 Assessment methods
 Bayesian model averaging
Background
Energy scenarios are quantitative or qualitative output from mathematic models^{a} of the energy system, or, systematic, consistent thinking in qualitative terms about the energy system. Quantitative energy models can be classified as topdown models, typically macroeconomic models with focus on energy economics and bottomup models, typically technology oriented processbased models. Different mathematical descriptions of the target system are possible, such as general equilibrium models (e.g. E3ME [1]), linear programs (e.g. TIMES [2]), stochastic models (e.g. [3], especially [4]) or mixed complementary problems (e.g. [5]). If an energy model has an objective function to be minimised or maximised, it is an optimisation model. Opposed to these, simulation models simulate consequences over time of key assumptions. Frequently used terms in that context are business as usual scenario or reference scenario. An energy scenario can describe both, key assumptions about relevant input variables in energy economics, and the result of a model run, the model output. In this text, the term energy scenario refers to the results of quantitative models. Amongst the most important input variables, also called key assumptions or assumption framework, are for example shares of specific electricity generation capacities, population growth assumptions, fuel price assumptions or gross domestic product assumptions. These key assumptions can be varied to produce alternative scenarios that can be assessed quantitatively by means of energy models or qualitatively in different storylines. The choice which key assumptions are considered in a study strongly depends on the aim of the research and is often coupled to specific questions regarding energy futures or political considerations cf. [6],[7]. One of the main aims of such energy scenarios are statements about the future, be it possible, probable, normative or deterministic statements. These statements may serve for political advice, or are the basis for (energy) political decision making. Examples to illustrate this function are the Energiekonzept der Bundesregierung [8] in Germany that is  at least inspired  by a modelling exercise, or, on a European level, the Energy Roadmap 2050 [9] that refers to a modelling exercise, detailed in part 2/2 of their publication.
The scenario report for the Energiekonzept introduces its findings with a clarification on what energy scenarios, as described in the report, are meant to be: ‘Scenarios describe possible futures. They do not claim to represent the most likely development from today’s perspective.[…] Depending on the definition of relevant parameters (‘Eckpunkte’), next to the derived scenarios, many other pathways in the future of the German energy supply are possible that are not under scrutiny in this work’ [10].
So, what discerns such a possibilistic statement (an energy scenario) from any other possibilistic energy scenario that, admittedly, may serve the same purpose of fulfilling the German energy supply? And if there is no difference, of what profit is the modelling exercise altogether? One answer can be found in the document, the developed target scenarios comprise consistent pathways of longterm energy economic developments [10].
The question arises, what consistent in this context means. Consistent with expectations about the future, consistent with past evidence, consistent with the mathematical model framework or consistent in a rather abstract sense that there are no contradictions in the energy scenarios. However interpreted, the question remains, what additional value, with respect to any other possible pathway, does a modelbased pathway into the German energy future provide? Consistency, in the broadest sense could be understood in and by itself as possibility; for, if a statement is selfcontradictory, it is not possible. Hence, consistency is no unique characteristic of possible modelbased statements. However, it is possible that some energy scenarios are not even consistent in a noncontradictory sense. WeimerJehle has developed a systematic approach to ensure that assumption preselection and associated effects on society, economy and environment are consistent and evaluated in a transparent way [11]. If such a transparent approach in the phase of key assumption selection or scenario construction is not provided, it remains at least questionable, if energy scenarios are not selfcontradictory.
An added value could be derived, if modelbased energy scenarios were contrasted to other possible scenarios, particularly, if an uncertainty assessment was carried out for the energy scenario. In the case of a quantitative uncertainty assessment, the possibility of comparing different scenarios with respect to their adherent uncertainty could indicate to what extent and in what respect the scenario can be discerned from other scenarios. An energy scenario that is possible begs the question how possible it is, given key assumptions about the future. The question is thus, how uncertain is an energy scenario?
The main objective of this text is a discussion which requirements an uncertainty assessment should fulfil. Based on these requirements, a method, Bayesian model averaging (BMA) is debated as possible candidate to satisfy them. The main arguments are that an energy modelling exercise that is enriched with an uncertainty assessment which satisfies the requirements is (1) comprehensible in terms of its associated uncertainty and (2) should contribute to a complete understanding of energy scenarios, especially if they are used in decision support.
Uncertainties in energy scenarios can have different sources and can be of different kinds. Walker et al. have presented a concise summary of existent uncertainty of modelbased energy scenarios, based on the location, level and nature of the uncertainty [12]. According to them, generic locations can be context, model uncertainty, inputs, parameter and outcome. The uncertainty estimation method discussed in this text, BMA for input variables to energy models, can assess uncertainty in the location input and parameter uncertainty, and, to an extent, context uncertainty that concerns the modelled boundaries of the system. The presented method does not aim to evaluate error propagation within a specific energy model. Walker describes the level of uncertainty in terms of determinism, statistical uncertainty, scenario uncertainty, recognised ignorance and indeterminacy, i.e. total ignorance. BMA for input variables to energy models represents uncertainty based on probabilistic assessment and hence can also range from certainty to total ignorance. However, the use of statistical data renders the assessment itself prone to statistical uncertainty. Finally, Walker et al. describe the nature of uncertainty as epistemic uncertainty or variability where epistemic uncertainty is due to the imperfection of our knowledge and variability refers to inherent variability, especially present in human and natural systems. The BMA approach can evaluate both natures of uncertainty in statistical terms. If input variables to energy models are exposed to variability, data fit of model results with respect to statistical data will indicate that exposure.
Uncertainty due to variability can also be addressed by stochastic modelling. A stochastic modelling approach aims to represent (natural) variability within the model, e.g. [13],[14]. Such uncertainty analyses are mainly applicable in physical systems. Energy models that also represent economic, political, environmental and social aspects of an energy system regard the system from a broader perspective. However, stochastic uncertainty assessments could be beneficial for (parts of) energy models that are exposed to variability such as electricity generation modelling based on wind or solar cf. [15],[16]. Epistemic uncertainty is less naturally defined in a probabilistic framework, and hence, one of the objectives of uncertainty quantification is the reformulation of epistemic uncertainty as variability [17]. If epistemic uncertainty is contained in the energy model, the model results are likely to be more uncertain than the input variables. The BMA approach accounts for that by establishing a lower bound of uncertainty.
Recent evaluations have investigated the nature of energy scenarios and their limitations in terms of legitimate inference from model output [6]. In contrast to a vivid discussion of model quality and legitimate inference, as can be observed in climate modelling [18][21], energy models have not invoked a similar discussion. Whilst climate modelling developed a framework for the treatment and communication of uncertainties [22][24], energy models and resulting scenarios lack such a systematic approach for uncertainty qualification (or quantification). However, investigating uncertainties in models is necessary for quality assessment of model results and reliability of results, especially if such results figure in policy advice.
The next chapter will investigate existing uncertainty assessments in energy modelling and focus on the strengths and weaknesses of those. From these considerations, general requirements that an uncertainty assessment for energy modelling should satisfy are retrieved. In the following section, presently applied methods for uncertainty evaluation are discussed, including expert elicitation, robustness analysis, model fit, variety of evidence and standard statistical analysis. As research regarding uncertainty evaluation for energy models is not yet as advanced as for example in climate science, a substantial part of discussion is based on examples of other disciplines, especially climate science. The next two sections firstly address critique on Bayesian approaches and present the uncertainty assessment based on BMA for input variables of energy models. The last section summarises results and discusses the method critically.
Existing uncertainty assessments in energy modelling
Data of already existing publications are compared with the results of this work, such that an evaluation of existing approaches to uncertainty assessment in the context of energy economics is examined on the basis of two examples.
Walker et al. [12] have investigated energy modelrelated uncertainties with respect to their nature and occurrence. Their definition of uncertainty being ‘any departure from the unachievable ideal of complete determinism’, allows for a conceptual investigation of all relevant uncertainties, by defined categories, reaching from determinism to total ignorance. The provided tool, an uncertainty matrix, should be used to identify model outcome uncertainty according to their level and nature. It is not clear in what terms the matrix should be evaluated, yes/no, much/little, or, in a numeric scale that is not provided. This approach allows for an illustrative representation of uncertainties involved in modelling. However, the method seems to end with a delicate categorization rather than with a valuable assessment. Indeed, awareness of the location, level and nature of uncertainty is important information; nonetheless, the method does not provide insight into how the uncertainty should be assessed or into what way uncertainty of, for example, recognised ignorance in the location model structure bears on the uncertainty of model outcomes.
The second example of an assessment is the numeral unit spread assessment pedigree (NUSAP) method to assess qualitative and quantitative uncertainties in the targets image energy model regional (TIMER) energy model, part of RIVMs IMAGE model [25]. Firstly, by means of a comprehensive checklist for model quality assurance, key loci and sorts of uncertainties in the TIMER modelling process are identified. Model structure uncertainties were analysed by a metaanalysis of similarities and differences of six energy models. A sensitivity analysis for model parameters in terms of magnitude of influence has been carried out. A NUSAP expert elicitation workshop has systematically assessed those parameters in the following dimensions: proxy, empirical basis, theoretical understanding, methodological rigour and validation. And finally, a diagnostic diagram is produced [25].
This evaluation provides interesting insight in the TIMER model, the uncertainties associated with the modelling process and the model results. However, there is some critique that relates to the applicability and required expert knowledge for such an assessment. Firstly, the method is rather modelspecific. Secondly, the method relies mainly on expert elicitation for both experts in modelling and energy economics. Experts in modelling are likely to be experts for especially the model they work with. Large and complex models tend to require long periods of vocational adjustment before a model is fully understood. And due to intellectual property considerations, some models cannot be assessed by ‘foreign’ modellers. Another critical remark considers the output of the assessment. The diagnostic diagram, as presented in section 6.8 of document [25], is difficult to understand. The diagnostic diagram is based on the results of expert elicitation.
These three evaluations all represent the uncertainty associated with the learning rates of nuclear power production. There are several interpretations possible: either, the dissent in the expert groups indicates that the experts do not dispose of deep understanding of the question^{c}.
Or, one expert group is correct and the others are wrong. Or, the limited number of experts does not allow for the results to converge to an unambiguous assessment result. Or, the strong dissent indicates that uncertainty is high. This last reading is not without further ado more justified than any of the other readings. Yet, it is the only one that actually assesses the uncertainty of the parameter in question. This example illustrates the necessity of an intersubjective requirement in an assessment method. An intersubjective assessment method would render the result of an uncertainty assessment less dependent on the specific individuals that carry out the analysis. If an assessment is based on expert elicitation and possible interpretations of dissent (and consent) are not explicit, the method itself contributes to uncertainty. For then, the uncertainty of the assessment method itself and the uncertainty of the model assessed are present, and it can be difficult to cleave them apart. It is necessary to stress that expert elicitation is an important tool. However, due to practical limitations^{d}, issues of the method in and by itself, such as convergence in findings and trustworthiness of findings, need to be addressed and evaluated.
 1)
give a clear indication how reliable the findings are (uncertainty assessment)
 2)
be applicable independent of assessor’s expertise (intersubjectivity)
 3)
be applicable to different models (comparability of results)
 4)
incorporate qualitative and quantitative aspects (complete representation)
 5)
be intuitively understandable and straightforward to communicate (scale requirements)
 6)
be reproducible and unambiguous.

The assessment method does not solely rely on expert elicitation, although valuable subjective expert knowledge can be included.

BMA could provide a versatile tool for the assessment of complex interrelated statistical data.

Requirements that should be satisfied by an uncertainty measurement method are met.
An explicit uncertainty assessment of energy scenarios that satisfies these requirements would increase transparency of assumption uncertainty and thus model results. The aims of the text are to present a methodology, BMA for input variables of energy models, that satisfies these requirements and to infer quantitative uncertainty estimations from input parameters to energy models. Recipients of energy scenarios could gain a better understanding regarding the uncertainty of model results (i.e. energy scenarios) what might impact their function as decision support or basis for decision, especially if energy scenarios are used for policy advice, leading to far reaching ecological, financial and societal consequences.
Methods
Quantitative methods versus qualitative methods for uncertainty assessments
In the quest of an appropriate uncertainty assessment for energy scenarios, climate science may provide a suitable starting point as the uncertainty assessment discussion in climate sciences is more advanced than in energy economics.

Confidence in the validity of a finding, based on the type, amount, quality and consistency of evidence (e.g. mechanistic understanding, theory, data, models, expert judgement) and the degree of agreement. Confidence is expressed qualitatively.

Quantified measures of uncertainty in a finding expressed probabilistically (based on statistical analysis of observations or model results, or expert judgement) [23].
By means of a confidence matrix, a likelihood scale (expressed as probabilities) and probability distribution functions, the three working groups of the AR5 are to evaluate uncertainties associated with their findings. The likelihood scale for (subjective) quantitative assessment of uncertainty is recommended to be applied only in cases with high or very high confidence [29]^{f}.
The IPCC uncertainty assessment thus relies on both, qualitative and quantitative ways to describe reliability of findings. Apparently, if quantitative assessments are applicable, they should be used preferable to qualitative assessment. Qualitative uncertainty assessment is applied in cases of deep uncertainty^{g}, where uncertainty cannot be quantified. Qualitative uncertainty assessment faces several challenges. The problem of linguistic ambiguity seems to be the predominant problem when uncertainty is qualitatively assessed. In the guidance, note the level of confidence is defined using five qualifiers: very low, low, medium, high and very high [23]. It synthesises the author teams’ judgments about the validity of findings as determined through evaluation of evidence and agreement. It is arguable if there is a common understanding of such categories amongst individuals and hence, the question arises, whether the evaluation of agreement actually depicts the uncertainty of the finding in question or rather the ambiguity in understanding of the term used. Also, there is no clear indication how much agreement is necessary for the affiliation to a certain category^{h}. And, finally, it is unclear in which way agreement can be associated with high confidence and in turn with uncertainty (judgments about the validity of findings). One interpretation could be that high confidence (inter alia based on high agreement) means low uncertainty; however, this could not hold true in cases where agreement is high that the level of uncertainty is high for a finding (e.g. due to the stochastic nature of a process). Moreover, this reading also faces the criticism that it is thinkable that even with high agreement, the finding is not at all certain, and all assessors could collectively be wrong in their valuation. The other reading, that high confidence means high uncertainty, next to being counterintuitive, does not reflect that agreement sometimes does give an indication for the truth of a finding.
Qualitative assessment methods, even if normalised to summary terms (IPCC) seem to intrinsically depend on not only a subjective comprehension of summary terms but also subjective opinion of the assessor. This can be advantageous or disadvantageous, depending on the expertise of the assessor and the communication of relevant information that influenced the assessment. In any case, such assessment methods lack the important property of generating reproducible assessments. If a different group of experts assessed the results, the uncertainty assessment of a specific finding might turn out to be significantly different, even if a sound reasoning underpins the assessment. As Krueger et al. point out, expert opinion in modelling will benefit from formal, systematic and transparent procedures [30]. Intersubjective reproducibility is a necessity if a finding is called robust. A qualitative assessment is likely to be not as efficient in evaluating robustness as a quantitative, standardised assessment, given the problem of linguistic ambiguity and subjectivity of the assessment method. A quantitative approach that uses a method that can be standardised and applied independent of the expertise of the assessor would presumably yield higher agreement.
However, qualitative uncertainty assessments have the important benefit of putting findings into perspective of the state of art of modelling and the present knowledge about processes and/or assumptions. If a finding is based on limited knowledge, it cannot represent a certain statement and has to be supplemented with information regarding the validity of findings.
Quantitative assessment methods often face the critique of being perceived with more precision than justified [31],[32], especially [33] when he discusses Nowotny’s perspective. This could be the case where probability density functions (pdfs) can be produced but are themselves based on uncertain input. In such cases, communication (qualitatively or quantitatively) of the uncertainty related to the pdfs is necessary. An advantage of quantitative methods is an unambiguous representation. The intuitive understanding, even in the simplest form, for example, a scale from one to ten, ten representing high uncertainty, might allow the recipient of such an assessment a clear understanding. This is, indeed, not unproblematic. For one, there is an intrinsic assumption that must be clarified if not true, which is that the scale units are uniform in size^{i} or a logarithmic scale. Even more intuitively understandable appears to be a probabilistic statement. However, regarding the perception of probabilistic uncertainty assessments, Patt et al. report that changes of equal magnitude in assessed probabilities can have different effects in decisionmaking experiments. For example, a change of 10 percentage points from 90% to 100% impacts choices of test persons differently than a change from 50% to 60% [34]. Nonetheless, a probabilistic statement is in itself less susceptible to interpretational errors or misunderstandings than a qualitative statement that uses  again  words for interpretation that might be ambiguous.
Another relevant advantage of quantified assessments is the simplicity and comparability of results. The benefits of retrieving a result that can be compared to results, say, some years ago are obvious: applying the same quantitative method can be used to evaluate scientific progress in modelling and scientific understanding (if uncertainty decreases) or illustrate that the nature of a process is more complex than assumed years ago (if uncertainty increases). However, a critique formulated by Kandlikar et al. [35] is that biases can result if simple schemes that attempt to represent uncertainty in a uniform manner across many different contexts are depending on how much detail is presented in the information. This effect is analysed by ignorance aversion. Indeed, quantification, or for that matter qualification to summary terms, can result in a loss of detail and reasoning. It is not totally clear how bias can be created through such a process. However, it is the very task of an uncertainty assessment to transform information of various kinds (quantitative, qualitative, narrative, implicit assumptions, etc.) into a form that can be understood without the profound expertise that is necessary to accomplish the uncertainty assessment itself.
The question whether a quantitative assessment method is preferable to a qualitative assessment method cannot be answered purely by evaluating the respective (dis) advantages. There are practical limitations that may render a quantitative assessment impossible. However, for the development of an uncertainty assessment in the context of energy economics, relevant differences to climate science prevail that might justify the preferable use of quantitative methods.
Climate science and energy economics
The discussion concerning confirmation of climate models may serve as orientation, and energy model evaluation could profit from these considerations. Lloyd [20] concludes that climate models should not be judged primarily or solely on the basis of what they are weak at. This is an important aspect to remember when evaluating energy models as well. Generally, her approach to confirmation ‘takes it as a matter of degree; models can accrue credit and trustworthiness upon being supported by empirical evidence as well as by theoretical derivation’. Lloyd illustrates the strengths in terms of confirmation as model fit, variety of evidence, independent support for aspects of the models and robustness for climate models. These concepts, bearing on the reliability of models, can be applied to energy models as well.
Model fit

unanticipated strong political decisions such as closing of mines in the UK, feedin tariffs in Germany and world climate change concerns;

unexpected energy requirements, like the transport behaviour and the rush for gas;

definition and availability of statistical data [36].
The main difference when comparing climate models with energy models is that energy models represent and simulate a wellunderstood system with mainly economic drivers. In contrast, climate modelling has its challenges in representing chaotic systems with at least partially little understood causal relationships and magnitudes of impact of system components. It was, at least in principle, possible to know today with sufficient accuracy how the energy system will look like in a given point of time in the future. The problem is that many interests must be met and decisions not tend to be of durable nature as political, environmental or economic circumstances change. This is one reason why energy roadmaps, energy strategies and energy programs on a political level are important. These commitments to a specific system state in future allow for energy modellers to accordingly define constraints in models and consequently investigate  using models  different paths to meet the desiderata. Results of such model simulations may be costeffective, environmentalfriendly, socially accepted or other (possibly optimised) system development paths. The reason why model fit of energy models yields a poor record in the past hence is not (primarily) due to little understanding of the system but must rather be contributed to influences on the system of radical nature that cannot be anticipated. Moreover, such radical impacts (e.g. political reorientation) do not lead to any improvement of energy models or target system understanding for their nature is vested in societal decisions that can and should not be anticipated, hence allowing evolvement of society.
Variety of evidence
Lloyd refers in her analysis to the fact that climate models can accurately predict other variables than, for example, global mean temperature. This type of confirmation translated to energy models could be interpreted as correctly predicted installed capacity for electricity generation, fuel mix and the like. Variety of evidence in energy models has close relation to the constraints and assumptions built into the model. As Knutti et al. [38] state, it is due to the physical principles known to be true, such as conservation of mass, energy and momentum, that can be applied and transferred across hierarchies of models that confidence in climate models is justified. This physical nature of climate modelling can only partially be applied in energy models. Energy systems have physical limitations, e.g. land use, maximum solar radiation or exhaustible resources; however, the system state is mainly dependent on economic drivers, law requirements and incentive policy. These are not physical, lawobeying mechanisms, although, a kind of causeandeffect relation can be observed. Due to this different nature of the modelled system, variety of evidence cannot be applied for confirmation in the same sense as climate models are verified by true evidence. The same is true for independent support for aspects.
Robustness
Lloyd applies a robustness analysis developed by Weisberg [39] that puts forward a robust theorem of the general form: ceteris paribus, if [common causal structure] obtains, then [robust property] will obtain. The causal structure captured in the respective models seems to be the key difference between climate models and energy models. Causal structures in energy models, depending on the model in question, are for example, inverse supply functions [2], whereas in climate modelling, for example, thermodynamic laws are applied^{j}[40]. It seems that climate science partially due to ignorance of (components of) the target system face epistemic uncertainty. Stevens and Bony [41] analyse that for example, tropical precipitation over land and consequently vegetation dynamics are poorly understood. As a result the understanding of the carbon cycle is limited.
It is necessary to clarify the applied interpretation of implication used to analyse Weisberg’s theorem. Let A denote the antecedent ‘common causal structure’ and B denote the consequent ‘common property’ of the theorem. The ‘if…then’ clause can be interpreted in different ways. A strict material implication in its truth functional sense means that A is false or B is true [42]. Another interpretation would be a logical implication to state that B is already logically implicit in A. This interpretation means that it is a logical consequence that a common causal structure implies a common property to obtain (ceteris paribus). Another interpretation of implication (A implies B) is that B is deducible from A by logical reasoning. To prove that it is logically deducible that a common property obtains if a common causal structure obtains would surpass the scope of this text. But, it may well be possible to do so. Weisberg departs in this question from Levins, Orzack and Sober and clarifies that robustness analysis is effective at identifying robust theorems, and, whilst it is not itself a confirmation procedure, robust theorems are likely to be true [39]. It is important that a theorem as put forward by Weisberg does not presuppose the truth of A. In other words, the theorem does not claim to guarantee that if a common causal structure obtains, this implies that a robust property will obtain (ceteris paribus). In this sense, the theorem is much weaker than one would wish for an uncertainty analysis. If robust theorems according to Weisberg are likely to be true, the only case that is unlikely is the one where A is true and B is false, for this renders the theorem to be false. Hence, the unlikely case is that if common causal structure obtains then robust property will not obtain, ceteris paribus. But as indicated by the example of Stevens and Bony, the antecedent ‘common causal structure’ (A) can well be false. The use of Weisberg’s theorem does not indicate if B (robust property will obtain) is true or false if A is false.
Hence, if common causal structure changes in climate models due to new understanding, robustness, defined as such, does not allow inference to the truth of the associated robust property or uncertainty.
In the case of energy models, causal structures face less uncertainty of epistemic nature, but rather uncertainty due to social or political underdeterminism of future developments. In this case, robustness could indicate some degree of certainty. However, it is not straightforward to conclude from robust results to uncertainty, even in a qualitative manner.
Another challenging issue in this respect is the ceteris paribus clause. A common approach for robustness analysis is scenario technique. The choice of parameters that are defined stable (ceteris paribus) and parameters or constraints that are varied significantly influences the results of energy models. It is therefore a choice, what results appear robust, for any result could be in principle produced by choice of parameters (e.g. by technology prices in cost optimization models). Hence, robustness as an indicator for uncertainty in energy scenarios has limited potential for uncertainty assessment of energy model results.
A discussion of Bayesian approaches
Probabilistic interpretation of uncertainty assessments is considered valuable, as the IPCC guidance note for treatment of uncertainties specifies [23]. Uncertainty and risk are to be assessed to the extent possible, and if appropriate probabilistic information is available, special attention to highconsequence outcomes should be given.
Probabilistic uncertainty assessments satisfy requirements 1 (clear indication how reliable the findings are), 5 (intuitively understandable and straightforward to communicate) and 6 (reproducible and unambiguous), if the methodology of assessment is a standardised process. As well in the IPCC guideline notes, as in the approach by Walker,as in the NUSAP method statistical knowledge is considered as knowledge with little inherent uncertainty. It seems thus appropriate to consider an assessment method that is based on statistical data and produces probabilistic uncertainty assessment results.
Bayesian statistics could provide such a method. As Bernardo [43] points out, the comprehension of probability in Bayesian statistics corresponds precisely to the sense in which this word is used in everyday language. This quality corresponds to satisfying requirement 5 (intuitively understandable and straightforward to communicate): the understanding of probability as a conditional measure of uncertainty associated with the occurrence of a particular even, given the available information and accepted assumptions. Bernardo stresses that a conditional probability measure is dependent on two arguments, the event E with the uncertainty to be measured and the conditions C of the measurement, ‘absolute’ probabilities do not exist [43].
In typical applications, one is interested in the probability of some event E given the available data D, the set of assumptions A which one is prepared to make about the mechanism which has generated the data, and the relevant contextual knowledge K which might be available. Thus, Pr (ED, A, K) is to be interpreted as a measure of (presumably rational) belief in the occurrence of the event E, given data D, assumptions A and any other available knowledge K, as a measure of how “likely” is the occurrence of E in these conditions [43].
With p(Dω) being a formal probability model for some (unknown) value of ω, the probabilistic mechanism which has generated the observed data D; p(ωK) being the prior probability distribution over the sample space Ω, describing the available (expert) knowledge K about the value of ω prior to the data being observed and p(ωD,A,K) being the posterior probability density.
representing the integrated likelihood of model M_{ K }. θ_{ k } is the vector of parameters of model M_{ K }, pr (θ_{ k }M_{ k }) is the prior density of θ_{ k } for model M_{ k }, pr(Dθ_{ k }, M_{ k }) is the likelihood and pr(M_{ k }) is the prior probability that M_{ k } is the true model. For a regression model θ = β, σ^{2}, all probabilities are implicitly conditional on the set of all models being considered.
Critique that has been offered for Bayesians includes but is not restricted to scepticism versus prior probabilities [46] and interpretational aspects [47],[48] and in response [49],[50]. Some arguments are also briefly presented by Gelman [51]. It outreaches the possibilities within this text to discuss all of them; therefore, the focus will lie on critique related to Bayesian methods in the context of climate models and energy models. One of the main objections to the use of Bayesian methods is the arbitrariness of the prior distribution. In the context of climate science, Betz [46] argues that the dependence on (1) the specific prior probability distribution over the initially considered hypotheses and (2) the climate model used for probability estimates of climate sensitivity obtained by Bayesian learning is problematic. According to Betz, the choice of prior distribution is an arbitrary assumption and  in the context of climate modelling, with limited sample sizes  entail that the final posterior probability is a function of the initial prior (which is arbitrary). This critique of prior distribution influence on posterior probabilities is a wellknown and not a new objection to Bayesian analysis cf. [52],[53].
Thus, the arbitrariness of the prior distribution is considered to be problematic. Put in Bayesian terms, an expert elicitation result is nothing but a collection of prior probabilities and though, this method is used for uncertainty assessment in climate modelling as well as energy modelling. If one accepts that a Bayesian statistician is an expert, the claim can be formulated even stronger, namely, that a Bayesian approach exactly satisfies requirement 4 (incorporate qualitative and quantitative aspects). This is to say that by means of prior distributions not only historical data (the likelihood) is used to assess uncertainty, but also a qualitative, subjective expert judgement can be incorporated. This not only renders the prior distribution choice a relevant tool for a complete representation but also responds to another critique that is often brought up against statistical methods in general, namely, that past evidence cannot provide for future developments. By means of a prior distribution, the likelihood of past events is relativized and both are possible, the recognition of the world as it is (was) and the representation of how this evidence is to be evaluated with respect to the future. It could be considered thus as a distinct virtue that prior probabilities depend on expert judgement rather than being problematic. The argument that subjective criteria can enrich a (statistical) model rather than disempower its findings due to lack of objectivism is also put forward by Isaac. His ‘integrated subjectivism’ also characterises a Bayesian model as the simplest form of integrating subjective knowledge and objective likelihoods with the aim of ‘transforming a scientific model into a decisiontheoretic one in which objective parameters (about the world) and subjective parameters (about the agent) peacefully coexist’ [54].
Requirement 4 (incorporate qualitative and quantitative aspects, i.e. complete representation) is satisfied more explicitly with a BMA evaluation that uses informative priors. However, it has been argued that even improper priors (aka noninformative) or weak priors (i.e. flat priors) contain information about the subjective certainty of the modeller, e.g. [55].

precision: the doctrine that uncertainty may be represented by a single probability or an unambiguously specified distribution;

prior knowledge of sample space: the assumption that all possible outcomes (the sample space) and alternatives are known beforehand.
Indeed, the problem of deceptive preciseness of probability distributions needs to be addressed when an uncertainty assessment is based on probabilities. One mean to that end could be a transparent documentation of data used and assumptions made for the uncertainty assessment. Again, comparing with the predominant assessment method, expert elicitation, such critique could hold here as well, however, ambiguity in expert elicitation results seems to be perceived as less problematic. Another mean to that end could be a systematic sensitivity analysis in Bayesian terms. In this effort, a variety of prior probabilities and its effect on posterior probabilities could yield important insight, possibly even in cooperation with expert elicitation to define priors that are suitable^{k}. BMA for input variables of energy models addresses this critique by evaluating a lower bound of uncertainty. Another possible way that is not investigated in the text could be the computation of interval probabilities that specify an interval of uncertainty for an input variable. However, due to considerations of ignorance, a lower bound seems more appropriate that respects the fact that unknown or intentionally ignored influences might increase uncertainty by a not specified amount.
Prior knowledge about the sample space Ω seems to pose more a problem in climate science than in energy economics. Possible outcomes and alternatives in energy economics are likely to be more predictable than in climate science. For example, in climate science, it might be true that a possible outcome is unknown due to interdependencies that are not well understood or orders of magnitude of effects that outrange expectations and the sample space does not account for that possibility. For example, if consequences of unprecedented gaseous concentrations (as in the past low O_{3} in the stratosphere [56] or more recently high CO_{2} concentrations in the atmosphere) are modelled, Ω might not be complete. In energy economics, some nonexplicit assumptions such as that the target system will exist in a comparable way within the time horizon and geographic scope of the model will simplify the treatment and assessment of the sample space. This is not due to insufficient modelling techniques, but rather to science being an evolving matter that naturally develops with new insight, new measurement techniques and scientific understanding. However, the critique is certainly valid in the context of energy scenarios if key assumptions are considered such as gross domestic product (GDP) growth, future energy prices or population growth. Even if sound forecast data from statistical sources are available^{l}, these assumptions could be associated with deep uncertainty and possibly, the sample space Ω is not complete. This fact might belong to the realm of recognised ignorance, as Walker et al. term it. Especially for such key assumptions, an uncertainty assessment that evaluates as many potential influences on the key assumption as possible is adequate.
One possibility of limiting such deep uncertainty in the context of energy economic models is a deliberate choice of system boundaries. In addition to typically topological, economic or sectorial system boundaries and subsystem units, social systems can and should be detailed in energy models, see [57]. In energy models, as in climate models, one can intentionally define system boundaries to represent parts of the integrated (energy) system with simplified connections across the system boundaries. However, for climate models that are concerned with questions of global impact and consistent regional interpretation, meaningful results can only be obtained within a global system boundary. IPCC [58] specifies that only general circulation models (GCMs) have the potential of consistent estimates of regional climate change which are required in impact analysis. Energy models can be designed to depict a certain part of the global energy economic system, hereby possibly increasing uncertainty due to ignorance of effects on a larger scale, and possibly reducing uncertainty within the system boundaries as Ω becomes more complete. It thus seems to be a tradeoff between chosen ignorance (due to system boundaries) and recognised ignorance (that one is aware of but cannot address). The BMA uncertainty assessment for input variables to energy models respects these uncertainties by formulating a lower bound of uncertainty.
It is worth discussing whether such uncertainties are better assessed with qualitative methods than in quantitative methods in probabilistic terms. The choice of key assumptions and their related uncertainty clearly limits the inferences that can be drawn from model results. However, the assessment of such deep uncertainties could be endeavoured in Bayesian terms.
The Bayesian endeavour
The key idea is to assess the uncertainty of the input variables on the left side of the graph in Bayesian terms and thusly define a lower bound of uncertainties associated with model results (model output). If one accepts the premise that model output cannot be less uncertain than model input, this lower bound could be defined by the uncertainty of the input variables. It is important to stress at this point that the BMA method for input variables does not replace an energy model, e.g. LP, MCP or a CGE model, to name just a few that are a common practice in energy economics. The aim is rather to assess uncertainties of input variables that are specific for a given model by means of BMA. This process should render transparent that independent of the predictive power of an energy model the sheer use of variables that are inherently uncertain leads to model outcomes that must reflect that uncertainty. It can and should not be the aim of an energy model to present results as more or less certain than they are due to the nature of a nondeterministic world which the target system is based in. The structure, nature, scope, aim and mathematical formulation of energy models are highly diversified. For a given energy economic question, many different potential energy models can be designed to provide an answer. However, any model that could be designed will have input variables that are more or less uncertain. The aim of the proposed method is providing an estimation of these uncertainties independent of the specific (dis) advantages a given model holds with respect to other energy models that could answer the question.
The predominant assessment method, expert elicitation, of uncertainty is used as reference. An expert elicitation process makes use of expert knowledge to assess how uncertain an assumption or a finding is. But what exactly is expert knowledge? The supposition is that expert knowledge is based on understanding of causal relationships, (long) record of observation or research, inclusion and exclusion of relevant factors and an intuitive ‘feel’ for the field of expertise. At least these virtues should be met by a Bayesian approach as well, together with the requirements previously defined.
The understanding of causal relationships  in the context of energy economics  refers to the ability of understanding market mechanisms, micro and macroeconomic processes, social processes, etc. Consider the example of energy prices in Figure 4. If an assumption regarding the future energy price of, for example, natural gas is to be defined, it would be necessary to think of influences that impact the natural gas price, for example, resources, (global) demand, infrastructure, efficiency of devices and the like. These influences need not be assessed in qualitative terms or subjective opinion of an expert, for there are statistical data available. If such statistical data are not readily available, it might be necessary to look for a suitable statistical representation of the influence, e.g. for consumer acceptance [59], or methods described by [60] with respect to the food industry. A sound record of research and a long record of observation can be translated in statistical terms in sufficient large sample sizes of the statistical data. This might pose a problem if time series are short or the influence record is short.
The causal relationships, or how an influence bears on the input variable in question, in the example, the natural gas price, could be represented in a mathematical relation, e.g. a linear regression model. A regression model representing the dependent variable, natural gas price, and the explanatory variables, the influences, could capture causal relationships and the magnitude of impact of an influence on the input variable. Note that, nonlinear models could be applied also, but for the analysis of the impact of an influence on the dependent variable (that is, the input variable in an energy model), it suffices to evaluate whether the influence increases or decreases the dependent variable and with what order of magnitude (that is, the coefficient estimate). This is straightforward standard statistical work. But this would not respect that the representation with a linear model itself increases uncertainty, for one might choose the wrong explanatory variables (influences) or not enough. This problem can be addressed by BMA.
On the abscissa, the coefficient value for the variable in the linear regression model is quantified. The ordinate represents the probability density for the coefficient value (i.e. the rate of change of the conditional mean of the natural gas price conditional on the change of GDP). The double conditional standard deviation (2× cond. SD) is indicated in the red dotted line. An equivalent chart can be produced for every explanatory variable of the competing models. The PIP of this variable is 96.1% what reflects that if the variable was contained in a model, competing models were less successful in explaining the data. In other words, the PIP is the sum of PMPs for all models wherein a covariate was included. The shape of the probability density and the low range of double standard deviation (approx. 0.4 to 1.4) indicate that variation from the conditional expected value (cond. EV) is rather low.
In practice the approach can be detailed in several steps. In step one, relevant input variables, or all input variables  depending on the size of the energy model under scrutiny  are identified, e.g. GDP^{o} within the energy models’ system boundary. In the next step, statistical data of economic, ecological, social or from other disciplines is gathered that is suspected to influence the input variable (e.g. statistical data concerning, industrial production, import and export, taxes and subventions, birth and death rates, education, etc.), including statistical data of the input variable. This input variable (GDP) in the uncertainty assessment becomes the dependent variable on these influences. Note that, in contrast to other methods, there are hardly practical limitations to the amount of influences that can be considered, for BMA by means of a MCMC sampler investigates the model space and ranks explanatory variables (influences) according to their PIP. The next step is the definition of the form of mathematical representation, e.g. a multivariate linear regression^{p}. As many potential explanatory variables are defined, the question is what variables should be included in the model. BMA estimates models for all possible combinations of explanatory variables and constructs a weighted average over all of them. Then, the choice of a suitable prior distribution is defined, e.g. Zellner’s gprior [61],[62]. If the integrated likelihood is constant over all models, the PMP is proportional to the marginal likelihood of a specific candidate model, i.e. the probability of the data given that model times a prior probability. The prior probability reflects how probable the expert thinks the model is before looking at the data [63]. The thence generated models with highest PMPs can be evaluated, and a model that best represents the dependent variable (e.g. GDP) can be chosen. Finally, the uncertainty estimation for the input variable is derived from the PMP of the model chosen.
An additional feature that is not the focus of this text is the possibility of generating predictive distribution functions from the chosen model that consistently with past evidence and expert judgement represent the dependent variable for given assumptions of explanatory variables. This could foster consistency in the choice of key assumptions.
The interpretation of BMA results as uncertainty can be straight forward if uncertainty is suitable defined. To that end, a definition that is based on probability is introduced.
Definition: Uncertainty equals the probability that statement S might not be true.
Given, by means of BMA, a PMP is calculated for an uncertainty model (e.g. a PMP of 13%^{q} for a model that represents the natural gas price), uncertainty  by definition  would be at least 87% for the dependent variable. This would mean that the input variable ‘natural gas price’ to an energy model holds an uncertainty of at least 87%, even if all relevant explanatory variables are considered. Hence, the results of a model including an assumption about the natural gas price cannot be less uncertain than 87%.
In other words, the PMP reflects the probability that the input variable thusly described matches data. For a model with a PMP of 13%, the associated uncertainty would be at least 87%. A clarifying statement of the following form could accompany model results.
“In consideration of expert judgement, statistical data of influence X_{1}, influence X_{2}, influence X_{3},…, of the last 25 years, the uncertainty that the input variable can be described as such is at least 87%.”
For every influence X_{1}, X_{2},…, the PIP indicates the explanatory contribution of the influence and 1PIP indicates the uncertainty that the influence contributes to explaining the dependent variable of a given model (typically the one with the highest PMP). In the example, the uncertainty that GDP explains the natural gas price (together with the other explanatory variables) of the chosen model is 3.99% (1  0.9601 = 0.0399). Such an assessment clearly satisfies requirement 1 as uncertainty expressed as probability density is a clear indication how reliable the findings are.
The third virtue of expert knowledge, inclusion and exclusion of relevant factors, could be achieved by this standardised method, hereby satisfying requirement 2 (applicable independent of assessor’s expertise).
The approach would limit many intuitive over or underestimations of impact of influences on variables that figure as input variables in energy models. It is thinkable that different experts evaluate individual influences as more/less relevant for the assumption of an input variable (e.g. a natural gas price assumption) thereby generating ambiguousness and dissent. A standardised method, relying on statistical data, i.e. knowledge with little associated uncertainty in and by itself, could yield significant improvement in uncertainty assessment for energy models. However, as expert knowledge is an important part of assessment methods, it is possible to take this by prior probability specification into account.
A key quality of the BMA method for input variables is that model uncertainty of the linear regression model itself, and thus, the assessment method’s uncertainty is quantified in probabilistic terms. This is a distinct advantage of the method as opposed to purely statistical or qualitative methods. Other methods that are applied in uncertainty analysis, for example, standard statistical analysis or purely qualitative methods ignore that source of uncertainty. A standard regression analysis is conditional on the assumed statistical model, and the analyst may be uncertain whether it is the best representation. If an expert Delphi [64],[65] is carried out opinions are rarely scrutinised for their correctness or compliance with statistical evidence. However, if an expert is asked, how probable she thinks her evaluation is, a prior distribution could be constructed.
Another requirement previously defined is the applicability to different energy models (comparability of results), requirement 3. As indicated by Figure 4, the assessment method is concerned with input data to energy models and is hence independent of the mathematical model that consequently processes the input. The uncertainty assessment method would be applicable for different kinds of models common in energy economics, LP’s, MCP’s, CGE’s, stochastic models or even qualitative models that use input variables.
Requirement 4 (inclusion of qualitative and quantitative aspects) to assure a complete representation can be achieved through prior probabilities and statistical data. The resulting posterior probabilities and the probabilistic interpretation of uncertainty are straightforward to communicate, as demanded in requirement 5 (intuitively understandable and straightforward to communicate).
Finally, requirement 6 demands for reproducibility and unambiguousness. Given assessors use the same set of data, the results of BMA are reproducible. However, a source of ambiguousness could be prior probability choice. This lies, as previously discussed, in the very nature of expert judgement. A sensitivity analysis to evaluate such ambiguousness could both, increase understanding of the BMA method within this context, and indicate to what extent expert elicitation has to be put in perspective to statistical data.
Results and discussion
Results of applying BMA to energy model input variables are PMPs of competing models for input variables of an energy model. The PMPs of the input variables can then be used to define quantitatively the associated uncertainty of the specific input variable. The method respects previously defined requirements. The result is an uncertainty assessment of the form: applied input variable X has an associated uncertainty of at least Y%. Results of the model are thus associated with an uncertainty of at least Y%. Note, that such a result demands acceptance of the premise that model results cannot be less uncertain than model input.
Existing uncertainty assessments for energy models provide evaluations of energy models or energy scenarios. However, the approaches discussed in this text lack some qualities in the context of energy modelling that BMA for input variable uncertainty estimation could provide.
The method described by Walker et al. is rather a classification of uncertainty than an assessment that explicitly states uncertainty of results (requirement 1). On the other hand, methods that are applied in classical uncertainty quantification^{r} such as statistical analysis, stochastic modelling or error propagation computation, although being explicit, treat uncertainties in a mechanistic way that does not respect the various social and political aspects (requirement 4). Methods that mainly rely on expert elicitation, such as the NUSAP method might lack reproducibility of results and objectiveness (requirements 2 and 6). BMA could potentially combine the desired qualities. Intuitively understandable (requirement 5) uncertainty assessments that can be produced for different energy models (requirement 3) only dependent on the respective input variables the model demands could provide relevant insight in uncertainties that are associated with model input. As potential consequence of applying BMA for input variable uncertainty, transparency regarding model results with respect to the reliability of such findings could be evaluated and communicated. Moreover, input variables could be classified according to their adherent uncertainty if the method is applied. And finally, but left for further research, the possibility of generating predictive densities by means of BMA could lead to consistent input variable values that respect influences across system boundaries of a specific model.
All uncertainty assessment methods have advantages and disadvantages. In spite of the successful fulfilment of previously defined requirements, the BMA approach for input variables has deficits that need to be discussed.
A rather practical issue stems from the fact that the approach is parametric. This means that in practice, many different input variables need to be assessed if large and complex models are analysed and a significant amount of data collection and preparation seems necessary. One way, which proved successful in the NUSAP method, for reduction of assessment variables is a classification of input variables and a consequent sensitivity analysis to discern highly relevant input variables [25]. Such a procedure could be suitable for the BMA approach as well.
Another issue might arise if input variables yield individual uncertainties of different orders of magnitude. The question then arises whether the least certain defines the uncertainty or if model dependent interpretation of individual uncertainties (of individual input variables) would be meaningful. It is not straightforward to see where in the mathematical core of energy models input variables are processed, and hence, tracing back results to individual inputs could be difficult. A form of metaanalysis, as proposed by [66],[67] could possibly give relevant insight regarding the uncertainty significance of individual input variables across studies of different model applications, as done for studies in medicine (psychotherapy) [68].
And finally, an issue could arise if an energy model incorporates aspects or effects that are relatively ‘new’?, e.g. unconventional gas in Europe. Due to data scarcity and lacking maturity of available processes, a Bayesian approach to assess such input data would be difficult. The same problem of data scarcity can occur if scenario assumptions are not explicit, e.g. social, or psychological assumptions. If data are available for such assumptions, their bearing on an input variable to an energy model can be incorporated by the BMA method and hence could increase transparency in that aspect. If data are not available, it must be communicated that the aspect is not part of the uncertainty assessment.
In the light of increased transparency, intersubjective independency, quantitative explicit results, comparability of results and methodological advantages (reproducibility and inclusion of subjective expert judgements), the expected value of assessing input variable uncertainty of energy models amounts to a better understanding of the associated uncertainty of an energy scenario. This is valuable information for the evaluation of results of different models. For example, topdown models are distinct from bottomup models with regard to the assumptions they apply. The proposed uncertainty assessment could potentially add value for the comparison of energy scenarios that stem from different models. In addition, but left to further research, is the potential of scrutinising model ensembles^{s}, as demonstrated by [69] in the context of weather forecasts.
Conclusions
BMA for uncertainty quantification of input variables could potentially satisfy the requirements that a versatile applicable and standardised method for uncertainty assessment in energy economic modelling demands. Given the described advantages and disadvantages, it is at least worth discussion whether such an approach could improve the assessment itself and consequently could put the inferences and policy recommendations based on model results in perspective. This in turn should enable stakeholders and decision makers to include reported uncertainties in their decision making processes and increase trust in scientific findings. Trust in scientific findings is not solely generated by unerring model results but also by acknowledgement and transparency of uncertainties respecting that reality is not strictly deterministic.
Further research should be undertaken concerning the critical remarks and potential solutions for the application of the method. To this end, firstly an application of the approach to different models should yield insight allowing for further improvement of the approach.
Endnotes
^{a}For further information on energy models as referred to in this text see [70][72], or [73] in Germany.
^{b}The green kite is spanned up by the minimum scores in each group for each pedigree criterion; the orange kite is spanned up by the maximum scores. The orange band between the green kite and the red area represents expert disagreement on the pedigree scores for that variable. In some cases, the picture was strongly influenced by a single deviating low score given by one of the six experts. In those cases, the light green kite shows what the green kite would look like if that outlier would have been omitted.
^{c}Given that the same experts evaluate many fields of modelassociated uncertainties, it is thinkable that the expertise in some areas is not as sound as one would expect.
^{d}e.g. Limited number of experts, limited knowledge of experts.
^{e}However, Smithson [74] has made a strong case that ‘in all tasks, preciseconflictive sources were viewed as less credible than ambiguousconsensual ones even when subjects expressed preference for the preciseconflictive alternative.’ what suggests that the requirement of intersubjectivity has more relevance in terms of acceptance of the assessment.
^{f}This might trace back to a criticism of [35] on the interdependence of likelihood and confidence: ‘When an event is said to be extremely likely (or extremely unlikely) it is implicit that we have high confidence’.
^{g}As defined by [35], uncertainty that results from myriad factors both scientific and social and consequently is difficult to accurately define and quantify.
^{h}Qualification of the degree of agreement: summary terms: low, medium or high [29].
^{i}By that is meant that the uncertainty captured between 2 and 3 is not more/less than the uncertainty captured between 4 and 5, or for that matter 5 and 6.
^{j}For more information see for example [75][77].
^{k}e.g. Experts could be questioned what probability to a qualitative assessment like ‘surely no more/less than’ can be attributed. In that way, a subjective degree of belief can lead to a subjective prior distribution. See also [78],[79].
^{m}One could also use a Bayesian Markov chain Monte Carlo (BMCMC), as for example, Kim et al. have done to determine optimum tender prices [82].
^{n}This result stems from work not published yet, available from the author. Abbreviations: GDP gross domestic product, PIP posterior inclusion probability, Cond. EV Conditional expected value, SD standard deviation. Note that, the shape of the probability distribution offers a further indication of the reliability of the conditional expected value.
^{o}Note that, input variables may vary considerably between models. For example, a bottom up model as the TIMES linear program [2] does not enter economic performance directly in form of a GDP input variable. Instead, such macroeconomic assumptions must be translated in sectorial demands, e.g. megakilogrammes of crude steel demand. This transformation is often done based solely on expert judgement for a given sector of an energy model. A BMA uncertainty assessment could improve transparency with regard to that process. An example of a demand forecast is industrial production is provided by [83].
^{p}Of course other models can be applied as well, for example, generalised linear models, proportional hazard models or logistic regressions.
^{q}A posterior model probability (PMP) of ca. 11% is a rather poor model representation of observed data. However, similar approaches in other contexts show that low PMPs are not unusual, e.g. infrastructure PMP 0.39 [84], econometric context PMP 0.3 [85], medical context PMP 0.17 in a dataset on primary biliary cirrhosis [44] or [86]. For an explicit application to forecasts, see [69]. For a BMA example in the context of hydrology, where the BMA method was coupled with a maximum likelihood estimation proposed by Taplin, see [87].
^{r}Mainly applied in engineering.
^{s}Ensembles as used by Raftery are model results in which a model is run several times with different initial conditions or model physics. This might be applicable for energy models as well where different key assumptions are applied or key assumptions are varied.
Declarations
Acknowledgements
The author Monika Culka would like to thank the reviewers for their comments that helped improving the manuscript.
Authors’ Affiliations
References
 Pollitt H: E3ME Technical Manual, Version 6.0 Cambridge Econometrics. 2014.Google Scholar
 Loulou R, Remme U, Kanudia A, Lehtila A, Goldstein G: Documentation for the TIMES Model. PART I. IEA ETSAP. 2005.Google Scholar
 Meerschaert MM: Stochastic models. In Mathematical Modeling: Elsevier. S. Academic, Boston; 2013:251–299.Google Scholar
 Wallace SW, Fleten S: Stochastic programming models in energy. In Stochastic Programming, vol 10. Handbooks in Operations Research and Management Science. Elsevier,; 2003:637–677.Google Scholar
 Gabriel SA: Complementarity Modeling in Energy Markets. International Series in Operations Research & Management Science. Springer, New York, London; 2010.Google Scholar
 Dieckhoff C: Energieszenarien. Konstruktion, Bewertung und Wirkung  “Anbieter” und “Nachfrager” im Dialog. Energieszenarien. KIT Scientific Publishing, Karlsruhe; 2011.Google Scholar
 Grunwald A: Energy futures: diversity and the need for assessment. Futures 2011, 43(8):820–830. doi:10.1016/j.futures.2011.05.024View ArticleGoogle Scholar
 Bundesregierung Deutschland (2010) Energiekonzept der Bundesregierung vom. September 2010. Online: http://www.bundesregierung.deGoogle Scholar
 Commission E: Energy Roadmap 2050. Energy. Publications Office of the European Union, Luxembourg; 2012.Google Scholar
 EWI GWS PROGNOS (2010) Energieszenarien für ein Energiekonzept der Bundesregierung. Projekt Nr. 12/10 des Bundesministeriums für Wirtschaft und Technologie, Basel/Köln/OsnabrückGoogle Scholar
 WeimerJehle W: Crossimpact balances: a systemtheoretical approach to crossimpact analysis. Technol Forecast Soc Change 2006, 73(4):334–361. doi:10.1016/j.techfore.2005.06.005 doi:10.1016/j.techfore.2005.06.005 10.1016/j.techfore.2005.06.005View ArticleGoogle Scholar
 Walker WE, Harremoes P, Rotmans J, van der Sluijs JP, van Asselt MBA, Janssen P, Krayer von Krauss MP (2005) Defining uncertainty: a conceptual basis for uncertainty management in modelbased decision support. Integr Assessment 4(1). doi:10.1076/iaij.4.1.5.16466Google Scholar
 Michaelides PG, Fassois SD: Experimental identification of structural uncertainty  an assessment of conventional and nonconventional stochastic identification techniques. Eng Struct 2013, 53: 112–121. doi:10.1016/j.engstruct.2013.03.033View ArticleGoogle Scholar
 Kovacevic RM, Paraschiv F: Mediumterm planning for thermal electricity production. OR Spectrum 2014, 36(3):723–759. doi:10.1007/s00291–013–0340–9 doi:10.1007/s0029101303409 10.1007/s0029101303409View ArticleMathSciNetGoogle Scholar
 Sura P: Stochastic analysis of southern and pacific ocean Sea surface winds. J Atmos Sci 2003, 60(4):654–666. doi:10.1175/1520–0469(2003)060<0654:SAOSAP>2.0.CO;2View ArticleGoogle Scholar
 ZavalaGaray J, Moore AM, Perez CL, Kleeman R: The response of a coupled model of ENSO to observed estimates of stochastic forcing. J Climate 2003, 16(17):2827–2842. doi:10.1175/1520–0442(2003)016<2827:TROACM>2.0.CO;2View ArticleGoogle Scholar
 Smith RC: Uncertainty Quantification. Theory, Implementation, and Applications. Computational Science & Engineering Series. SIAMSociety for Industrial and Applied Mathematics, PA USA; 2013.Google Scholar
 Oreskes N, ShraderFrechette K, Belitz K: Verification, validation, and confirmation of numerical models in the earth sciences. Science 1994, 263(5147):641–646. 10.1126/science.263.5147.641View ArticleGoogle Scholar
 Betz G: Der Umgang Mit Zukunftswissen in der Klimapolitikberatung. Eine Fallstudie Zum stern review. Philos Nat 2008, 45(1):95–129. 10.3196/003180208787332369View ArticleMathSciNetGoogle Scholar
 Lloyd EA: Confirmation and robustness of climate models. Philos Sci 2010, 77(5):971–984. 10.1086/657427View ArticleGoogle Scholar
 Parker WS: IIconfirmation and adequacyforpurpose in climate modelling. Aristotelian Soc Suppl 2009, 83(1):233–249. doi:10.1111/j.1467–8349.2009.00180.xView ArticleGoogle Scholar
 Mastrandrea MD, Mach KJ: Treatment of uncertainties in IPCC assessment reports: past approaches and considerations for the fifth assessment report. Clim Change 2011, 108(4):659–673. doi:10.1007/s10584–011–0177–7 doi:10.1007/s1058401101777 10.1007/s1058401101777View ArticleGoogle Scholar
 Mastrandrea MD, Field CB, Stocker TF, Edenhofer O, Ebi KL, Frame DJ, Held H, Kriegler E, Mach KJ, Matschoss PR, Plattner GK, Yohe G: The IPCC AR5 guidance note on consistent treatment of uncertainties: a common approach across the working groups. Clim Change 2011, 108(4):675–691. doi:10.1007/s10584–011–0178–6View ArticleGoogle Scholar
 Barrett S, Dannenberg A: Climate negotiations under scientific uncertainty. Proc Natl Acad Sci USA 2012, 109(43):17372–17376. doi:10.1073/pnas.1208417109 doi:10.1073/pnas.1208417109 10.1073/pnas.1208417109View ArticleGoogle Scholar
 van der Sluijs JP, Potting J, Risbey J, van Vuuren D, de Vries B, Beusen A, Heuberger P, Corral Quintana S, Funtowicz S, Kloprogge P, Nujiten D, Petersen A, Ravetz J: Uncertainty assessment of the IMAGE/TIMER B1 CO2 emissions scenario, using the NUSAP method, Report No: 410 200 104. 2002.Google Scholar
 Rai V: Expert elicitation methods for studying technological change under uncertainty. Environ Res Lett 2013, 8(4):041003. doi:10.1088/1748–9326/8/4/041003View ArticleGoogle Scholar
 Le Maître OP, Knio OM: Spectral Methods for Uncertainty Quantification with Applications to Computational Fluid Dynamics. Scientific computation, Springer, Dordrecht, New York; 2010.View ArticleGoogle Scholar
 Council IA: Climate Change Assessments. Review of the Processes and Procedures of the IPCC. Inter Academy Council, Amsterdam, The Netherlands; 2010.Google Scholar
 Mastrandrea MD, Field CB, Stocker TF, Edenhofer O, Ebi KL, Frame DJ, Held H, Kriegler E, Mach KJ, Matschoss PR, Plattner GK, Yohe GW, Zwiers FW: Guidance Note for Lead Authors of the IPCC Fifth Assessment Report on Consistent Treatment of Uncertainties. Intergovernmental Panel on Climate Change (IPCC), CA, USA; 2010.Google Scholar
 Krueger T, Page T, Hubacek K, Smith L, Hiscock K: The role of expert opinion in environmental modelling. Environ Model Software 2012, 36: 4–18. doi:10.1016/j.envsoft.2012.01.011View ArticleGoogle Scholar
 Der Sluijs V, Jeroen P, Craye M, Funtowicz S, Kloprogge P, Ravetz J, Risbey J: Combining quantitative and qualitative measures of uncertainty in modelbased environmental assessment: the NUSAP system. Risk Anal 2005, 25(2):481–492. doi:10.1111/j.1539–6924.2005.00604.xView ArticleGoogle Scholar
 Aven T, Pörn K: Expressing and interpreting the results of quantitative risk analyses. Rev Discuss Reliability Eng Syst Saf 1998, 61(12):3–10. doi:10.1016/S0951–8320(97)00060–4View ArticleGoogle Scholar
 Jaeger C: Risk, Uncertainty, and Rational Action. Risk, Society, and Policy Series. Earthscan, London; 2001.Google Scholar
 Patt A, Dessai S: Communicating uncertainty: lessons learned and suggestions for climate change assessment. Comptes Rendus Geoscience 2005, 337(4):425–441. doi:10.1016/j.crte.2004.10.004View ArticleGoogle Scholar
 Kandlikar M, Risbey J, Dessai S: Representing and communicating deep uncertainty in climatechange assessments. Comptes Rendus Geoscience 2005, 337(4):443–455. doi:10.1016/j.crte.2004.10.010View ArticleGoogle Scholar
 Pilavachi PA, Dalamaga T, Rossetti di Valdalbero D, Guilmot JF: Expost evaluation of European energy models. Energy Policy 2008, 36(5):1726–1735. doi:10.1016/j.enpol.2008.01.028View ArticleGoogle Scholar
 Bezdek RH, Wendling RM: A half century of longrange energy forecasts: errors made, lessons learned, and implications for forecasting. J Fusion Energ 2002, 21(34):155–172. doi:10.1023/A:1026208113925View ArticleGoogle Scholar
 Knutti R, Furrer R, Tebaldi C, Cermak J, Meehl GA: Challenges in combining projections from multiple climate models. J Climate 2010, 23(10):2739–2758. doi:10.1175/2009JCLI3361.1View ArticleGoogle Scholar
 Weisberg M: Robustness analysis. Philos Sci 2006, 73(5):730–742. 10.1086/518628View ArticleMathSciNetGoogle Scholar
 Gent PR, Danabasoglu G, Donner LJ, Holland MM, Hunke EC, Jayne SR, Lawrence DM, Neale RB, Rasch PJ, Vertenstein M, Worley PH, Yang ZL, Zhang M: The community climate system model version 4. J Climate 2011, 24(19):4973–4991. doi:10.1175/2011JCLI4083.1View ArticleGoogle Scholar
 Stevens B, Bony S: What are climate models missing? Science 2013, 340(6136):1053–1054. doi:10.1126/science.1237554View ArticleGoogle Scholar
 Hughes RIG: A Philosophical Companion to FirstOrder Logic. Hackett Pub. Co., Indianapolis; 1993.Google Scholar
 Bernardo JM: Probability and Statistics. In Encyclopedia of Life Support Systems (EOLSS). Edited by: Bernardo JM. Bayesian Statistics. Eolss Publishers, Paris; 2003.Google Scholar
 Hoeting JA, Madigan D, Raftery AE, Volinsky CT (1999) Bayesian model averaging: a tutorial. Stat Sci 382401Google Scholar
 Raftery AE, Madigan D, Hoeting JA: Bayesian model averaging for linear regression models. J Am Stat Assoc 1997, 92(437):179–191. 10.1080/01621459.1997.10473615View ArticleMathSciNetGoogle Scholar
 Betz G: Probabilities in climate policy advice: a critical comment. Clim Change 2007, 85(12):1–9. 10.1007/s1058400793139View ArticleGoogle Scholar
 Thompson B: A critique of Bayesian inference. In The Nature of Statistical Evidence. Springer, New York; 2007:84–96. 10.1007/9780387400549_9Google Scholar
 Gelman A, Shalizi CR: Philosophy and the practice of Bayesian statistics. Br J Math Stat Psychol 2013, 66(1):8–38. doi:10.1111/j.2044–8317.2011.02037.xView ArticleMathSciNetGoogle Scholar
 Mayo DG: The errorstatistical philosophy and the practice of Bayesian statistics: comments on Gelman and Shalizi: ‘Philosophy and the practice of Bayesian statistics’. Br J Math Stat Psychol 2013, 66(1):57–64. doi:10.1111/j.2044–8317.2012.02064.xView ArticleMathSciNetGoogle Scholar
 Burstyn I, Kromhout H: A critique of Bayesian methods for retrospective exposure assessment. Ann Occup Hyg 2002, 46(4):429–431. doi:10.1093/annhyg/mef058View ArticleGoogle Scholar
 Gelman A: Objections to Bayesian statistics. Bayesian Anal 2008, 3: 445–450. doi:10.1214/08BA318View ArticleMathSciNetGoogle Scholar
 Winkler RL: The assessment of prior distributions in Bayesian analysis. J Am Stat Assoc 1967, 62(319):776. doi:10.2307/2283671View ArticleGoogle Scholar
 Pierce DA, Folks JL: Sensitivity of Bayes procedures to the prior distribution. Oper Res 1969, 17(2):344–350. doi:10.1287/opre.17.2.344View ArticleGoogle Scholar
 Isaac AM: Model uncertainty and policy choice: a plea for integrated subjectivism. Stud Hist Philos Sci Part A 2014, 47: 42–50. doi:10.1016/j.shpsa.2014.05.004View ArticleMathSciNetGoogle Scholar
 van Dongen S: Prior specification in Bayesian statistics: three cautionary tales. J Theor Biol 2006, 242(1):90–100. doi:10.1016/j.jtbi.2006.02.002View ArticleMathSciNetGoogle Scholar
 Stolarski R, Bojkov R, Bishop L, Zerefos C, Staehelin J, Zawodny J: Measured trends in stratospheric ozone. Science 1992, 256(5055):342–349. 10.1126/science.256.5055.342View ArticleGoogle Scholar
 Luhmann N: Social Systems. Stanford University Press, Stanford, Calif, Writing science; 1995.Google Scholar
 Intergovernmental Panel on Climate Change (2013) What is a GCM? Guidance on the Use of Data. Online: www.ipccdata.orgGoogle Scholar
 Strickert DP: Estimating consumer acceptance limits. Commun Stat Theory Methods 1990, 19(7):2365–2472. doi:10.1080/03610929008830327View ArticleGoogle Scholar
 N’s T, Brockhoff PB, Tomić O: Statistics for sensory and consumer science. Wiley, Chichester, West Sussex, Hoboken, NJ; 2010.Google Scholar
 De Finetti B, Goel Prem K, Zellner A: Bayesian Inference and Decision Techniques. Essays in Honor of Bruno de Finetti. NorthHolland, Sole distributors for the USA and Canada, Elsevier Science Pub. Co, Amsterdam, New York, New York, NY, USA; 1986.Google Scholar
 Zellner A, Hong C: Forecasting international growth rates using Bayesian shrinkage and other procedures. J Econom 1989, 40(1):183–202. doi:10.1016/0304–4076(89)90036–5View ArticleGoogle Scholar
 Zeugner S: Bayesian Model Averaging with BMS for BMS version 0.3.0. 2011.Google Scholar
 Adler M, Ziglio E: Gazing into the Oracle. The Delphi Method and its Application to Social Policy and Public Health. Jessica Kingsley Publishers, London; 1996.Google Scholar
 Ayyub BM: Elicitation of expert opinions for uncertainty and risks. CRC Press, Boca Raton, Fla; 2001.View ArticleGoogle Scholar
 Glass GV: 9: integrating findings: the metaanalysis of research. Rev Res Educ 1977, 5(1):351–379. doi:10.3102/0091732X005001351View ArticleMathSciNetGoogle Scholar
 Hedges LV: Statistical Methodology in MetaAnalysis. ERIC Clearinghouse on Tests, Measurement, and Evaluation, Princeton, NJ; 1982.Google Scholar
 Smith ML, Glass GV: Metaanalysis of psychotherapy outcome studies. Am Psychol 1977, 32(9):752–760. doi:10.1037/0003–066X.32.9.752View ArticleGoogle Scholar
 Raftery AE, Gneiting T, Balabdaoui F, Polakowski M: Using Bayesian model averaging to calibrate forecast ensembles. Mon Weather Rev 2005, 133(5):1155–1174. doi:10.1175/MWR2906.1View ArticleGoogle Scholar
 Capros P, Paroussos L, Fragkos P, Tsani S, Boitier B, Wagner F, Busch S, Resch G, Blesl M, Bollen J: Description of models and scenarios used to assess European decarbonisation pathways. Energ Strategy Rev 2014, 2(34):220–230. doi:10.1016/j.esr.2013.12.008View ArticleGoogle Scholar
 Jebaraj S, Iniyan S: A review of energy models. Renew Sustain Energ Rev 2006, 10(4):281–311. doi:10.1016/j.rser.2004.09.004View ArticleGoogle Scholar
 Bhattacharyya SC, Timilsina GR: A review of energy system models. Int J Energ Sector Manage 2010, 4(4):494–518. doi:10.1108/17506221011092742View ArticleGoogle Scholar
 Fahl U: Energiemodelle zum Klimaschutz in liberalisierten Energiemärkten: die Rolle erneuerbarer Energieträger. Umwelt und Ressourcenökonomik, LIT; 2004.Google Scholar
 Smithson M: Conflict aversion: preference for ambiguity vs conflict in sources and evidence. Organ Behav Hum Decis Process 1999, 79(3):179–198. doi:10.1006/obhd.1999.2844View ArticleGoogle Scholar
 Community Earth System Model CESM (2014) Models. CO, USAGoogle Scholar
 Program for Climate Model Diagnosis and Intercomparison (2013) About the WCRP CMIP3 MultiModel Dataset Archive at PCMDI. Online: http://wwwpcmdi.llnl.gov/Google Scholar
 Meehl GA, Covey C, Taylor KE, Delworth T, Stouffer RJ, Latif M, McAvaney B, Mitchell JFB: THE WCRP CMIP3 Multimodel dataset: a new era in climate change research. Bull Am Meteorol Soc 2007, 88(9):1383–1394. doi:10.1175/BAMS88–91383View ArticleGoogle Scholar
 Eicher TS, Papageorgiou C, Raftery AE: Default priors and predictive performance in Bayesian model averaging, with application to growth determinants. J Appl Econ 2011, 26(1):30–55. 10.1002/jae.1112View ArticleMathSciNetGoogle Scholar
 Ley E, Steel MF: Mixtures of priors for Bayesian model averaging with economic applications. J Econom 2012, 171(2):251–266. doi:10.1016/j.jeconom.2012.06.009View ArticleMathSciNetGoogle Scholar
 Department of Economic and Social Affairs, Population Division (2013) World Population Prospects: The 2012 Revision, Highlights and Advance Tables. New York. Working Paper No. ESA/P/WP. 228Google Scholar
 OECD (2012) Main Economic Indicators. OECD Publishing, Main Economic Indicators  complete database,Google Scholar
 Kim S, Kim G, Lee D: Bayesian Markov chain Monte Carlo model for determining optimum tender price in multifamily housing projects. J Comput Civ Eng 2014, 28(3):06014001. doi:10.1061/(ASCE)CP.1943–5487.0000297View ArticleGoogle Scholar
 Feldkircher M: Forecast combination and Bayesian model averaging: a prior sensitivity analysis. J Forecast 2012, 31(4):361–376. doi:10.1002/for.1228View ArticleMathSciNetGoogle Scholar
 Wesonga R: Bayesian model averaging: an application to the determinants of airport departure delay in Uganda. AJTAS 2014, 3(1):1. doi: 10.11648/j.ajtas.20140301.11View ArticleGoogle Scholar
 Fernandez C, Ley E, Steel MFJ: Model uncertainty in crosscountry growth regressions. J Appl Econ 2001, 16(5):563–576. doi:10.1002/jae.608View ArticleGoogle Scholar
 Chua CL, Suardi S, Tsiaplias S: Predicting shortterm interest rates using Bayesian model averaging: evidence from weekly and high frequency data. Int J Forecast 2013, 29(3):442–455. 10.1016/j.ijforecast.2012.10.003View ArticleGoogle Scholar
 Neuman SP: Maximum likelihood Bayesian averaging of uncertain model predictions. Stoch Environ Res Risk Assess 2003, 17(5):291–305. doi:10.1007/s00477–003–0151–7View ArticleGoogle Scholar
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.