by Judith Curry
Health risks arise from the interaction of uncertain future climatic changes with complex ecological, physical, and socio-economic systems, which are simultaneously affected by numerous other changes, e.g. globalisation, demographic changes, and changes in land use, nutrition, health care quality. Policymaking on adaptation to health risks of climate change thus faces substantial uncertainty. – Wardekker et al.
Health risks of climate change: An assessment of uncertainties and its implications for adaptation policy
JA Wardekker, A deJong, L vanBree, WC Turkenburg, J van der Sluijs
Abstract
Background: Projections of health risks of climate change are surrounded with uncertainties in knowledge. Understanding of these uncertainties will help the selection of appropriate adaptation policies.
Methods: We made an inventory of conceivable health impacts of climate change, explored the type and level of uncertainty for each impact, and discussed its implications for adaptation policy. A questionnaire-based expert elicitation was performed using an ordinal scoring scale.
Experts were asked to indicate the level of precision with which health risks can be estimated, given the present state of knowledge. We assessed the individual scores, the expertise weighted descriptive statistics, and the argumentation given for each score. Suggestions were made for how dealing with uncertainties could be taken into account in climate change adaptation policy strategies.
Results: The results showed that the direction of change could be indicated for most anticipated health effects. For several potential effects, too little knowledge exists to indicate whether any impact will occur, or whether the impact will be positive or negative. For several effects, rough “order-of-magnitude‟ estimates were considered possible. Factors limiting health impact quantification include: lack of data, multi-causality, unknown impacts considering a high-quality health system, complex cause-effect relations leading to multi-directional impacts, possible changes of present-day response-relations, and difficulties in predicting local climate impacts. Participants considered heat-related mortality and non-endemic vectorborne diseases particularly relevant for climate change adaptation.
Conclusions: For possible climate related health impacts characterised by ignorance, adaptation policies that focus on enhancing the health system‟s and society‟s capability of dealing with possible future changes, uncertainties and surprises (e.g. through resilience, flexibility, and adaptive capacity) are most appropriate. For climate related health effects for which rough risk estimates are available, „robust decision-making‟ is recommended. For health effects with limited societal and policy relevance, we recommend focusing on no-regret measures. For highly relevant health effects, precautionary measures can be considered. This study indicated that analysing and characterising uncertainty by means of a typology can be a very useful approach for selection and prioritization of preferred adaptation policies to reduce future climate related health risks.
[Link] to online paper.
Excerpts:
Health risks arise from the interaction of uncertain future climatic changes with complex ecological, physical, and socio-economic systems, which are simultaneously affected by numerous other changes, e.g. globalisation, demographic changes, and changes in land use, nutrition, health care quality. Policymaking on adaptation to health risks of climate change thus faces substantial uncertainty.
For example, quantitative (health) risk approaches handle statistical uncertainties quite well, but fail to tackle other types of uncertainty. Resilience-oriented approaches, on the other hand, can cope well with ignorance and surprises, but are less appropriate when statistical uncertainty prevails. Thus, the level and nature of uncertainty have important implications for selecting appropriate adaptation approaches and for policy choices regarding their implementation.
Level of Precision Scale
1 Effective ignorance: Knowledge of the factors that govern this effect is so weak that we are effectively ignorant.
2 Ambiguous sign or trend: Some effect is expected, but its sign or trend is not clear. There are plausible arguments either direction (effect could be positive, could be negative; could increase or decrease).
3 Expected sign or trend: It is clear what the sign and trend of the effect will be. However, there is no plausible or reliable information on how strong it will be.
4 Order of magnitude: It is possible to give a rough indication of the magnitude of the effect, a qualitative scoring (e.g. 1–10 scale), or a rough comparison with other effects.
5 Bounds: It is possible to estimate the bounds for the distribution of the effect, e.g. its 5/95 percentiles (effect is only 5 % likely to be more than … and only 5 % likely to be less than …). However, the shape of the distribution, or best-guess estimates, cannot be provided.
6 Full probability density function: It is possible to provide a full probability density function; the bounds as well as the shape of the distribution.
Notes on the expert elicitation methodology:
A list of 33 potential health impacts of climate change was identified and grouped into eight health themes. Level of precision scores were elicited from 21 participating experts (see Methods section).
As respondents were asked to what extent they were able to estimate the risk, it is relevant to explore whether the score resulted from the state of knowledge or from the respondent‟s personal level of knowledge, skills and familiarity with risk assessment techniques such as modelling, statistical techniques, and expert elicitation. Personal lack of knowledge or skill was explicitly checked in the argumentation as potential bias. It appeared to play a minor role, with lack of knowledge appearing only on a few occasions for scores of 2 or 1. Another measure to the same end was to track scores by generalists and by subject-matter experts separately. These scores corresponded fairly well. However, most of the argumentation focused on the availability of basic data and models, the degree to which the system dynamics are understood, and the knowledge gaps and complexities that exist. As such, the scores should be interpreted as whether it is appropriate to quantify the health risks for specific effects given the state of knowledge, rather than whether it is possible to produce a number in one way or another. In a few instances, for low scoring effects, respondents made arguments that the impacts could be low or high considering e.g. constraints posed by the high quality healthcare system or considering the current incidence. Consequently, it may be possible to further scope some low scoring risks, at least to some extent, using for instance imprecise, ordinal or qualitative/comparative approaches. Further investigation would be required to assess the scope to which this is possible and appropriate.
Scores and arguments for the relevance of effects varied between experts, although the general ordering and, for the high-scoring effects, the general line of reasoning is relatively clear. Results should be seen as indicative, as they may vary over time, group of respondents, and country. An interesting issue, for example, is the potential influence of recent (extreme) events. Such events may influence public perception and therefore the societal salience of effects. Current public perception played a role (although not a major role) in the arguments for heat-related effects, referring to the 2003 European heat wave. It also played a role for vector-borne diseases, although the arguments related to the potential role it could play due to e.g. the “fright factors‟ associated with the effect, rather than current public perception due to recent events. Recent events might also influence expert scorings when they reveal vulnerabilities that had been unknown or not sufficiently perceived before. Again, this seems to play a role for heat-related effects in reference to the 2003 heat wave. This certainly is a valid reason to consider the effect relevant, and one that may remain relevant over time.
Being based on expert elicitation, results should be treated with some care. The sample of participants is always a limited subset of the total expert-population and situational factors influence the composition of the panel (e.g., who is well-known in the field, who has time to participate). Therefore, results are not necessarily representative. Rather, they give an approximation, and the lines of reasoning behind the scores provide valuable insights into the issue studied. Given the broad coverage of relevant subfields, relative consistency in scores and arguments for most health effects, and consistency with the literature, we consider the findings robust enough to support the general conclusions.
Conclusions
Knowledge regarding health risks of climate change is characterised by large gaps and deep uncertainties. Planned adaptation to these risks requires profound understanding of the level of uncertainty of available knowledge of anticipated health effects. This study presents a systematic exploration and appraisal of uncertainties regarding climate change-related health risks. Using a six point scale, experts were asked to indicate the level of precision with which health risk estimates can be made, given the present state of knowledge. The study focussed on The Netherlands.
The experts assessed that, for most of the 33 (potential) health effects identified, it is possible to indicate its sign of change, but not its magnitude. Individual scores varied, generally between being unable to indicate the direction of change and being able to calculate the rough “order-of-magnitude‟ of the impacts. Factors that were often indicated to limit quantification include: limited data (in general and country-specific), the multi-factorial nature of the health issues (many important non-climatic drivers of change), and unknown impacts considering a high-quality health system.
For some effects, rough estimates of the order-of-magnitude were deemed possible: heat- and cold-related mortality, the oak processionary caterpillar, microbial contamination of swimming/recreation water, flood-related mortality and air quality-related effects. For these effects, data and impact assessment models are available. However, the availability of locally-specific data is relatively limited, there are many confounding factors, present-day response-relationships may change, and changes in local extreme weather events, such as heat waves, are still difficult to project for the future.
For allergic eczema, flood-related exposure to dangerous substances, wasps, UV-related weakening of the immune system, and epidemics of non-endemic vector-borne diseases it may not be possible to even indicate the direction of change. The latter, however, differs per specific disease: for some, effects are unlikely, for others, unknown. In addition to the difficulties noted above, the cause-effect relations of these effects are often highly complex and impacts are likely multi-directional.
These results suggest that, among various alternative approaches to climate change adaptation under uncertainty, approaches that focus on enhancing the health system‟s and society‟s capability of dealing with changes, uncertainties and surprises (for example by increasing resilience, flexibility, and adaptive capacity) are most suitable for adapting to the health impacts of climate change. Furthermore, we advise assessing the availability of “no-regret‟ options, which make economic or societal sense due to co-benefits or health benefits in the current climate, and the „climate and health‟ co-benefits of adaptation policy on other policyissues.
For more quantifiable effects, we recommend exploring the robustness of various policy strategies under a range of plausible outcomes, at least in a qualitative/semiquantitative way. Such analyses can contribute to setting preferred levels of ambition for adaptation efforts. For highly relevant effects, precautionary measures and other highly specific, costly or rigorous adaptations are also a relevant option, although it is advisable to enhance the flexibility of such options and to assess the associated risks (e.g. of these options becoming an overinvestment or resulting in detrimental side-effects).
Because nature, extent and rate of climate change and its health impacts are uncertain, understanding the relative level of relevance and uncertainty is crucial to making rational choices in adaptation policies and for possible adjustments if climate change effects occur slower, faster, or just different than earlier expected. Similar to e.g. Ebi [34] we argue that, to reduce climate change-related health risks, flexible, adaptive, multilevel and dynamic adaptation strategies should be developed. This study indicated that analysing and characterising uncertainty by means of a typology can be a very useful approach for selection and prioritization of preferred adaptation policies to reduce future climate related health risks.
JC comment: This paper comes from the Dutch postnormal science group. Postnormal science is a particular view of the interface between science and policy for complex problems with deep uncertainties that are associated with value commitments and involvement of an extended peer community (see this previous post).
This paper is an excellent example of application of postnormal science ideas to policy making. Relative to the UNFCCC/IPCC model, this method places uncertainty front and center, uses a carefully crafted expert elicitation (rather than a group consensus building process targeted at a specific policy, and matches the policy options to the level of uncertainty/ignorance.
Based upon the comments on the previous postnormal post, many of you do not ‘get’ postnormal science and think it condones pseudo science as a basis for policy. This is absolutely not the case, and I hope this paper will clarify the methods of postnormal science.