Preliminary best practices in climate uncertainty quantification and communication

Preliminary best practices in climate uncertainty quantification and communication Deliverable 2.3 Authors: Charlotte Pascoe (STFC), Rutger Dankers (WR), Ángel G. Muñoz (BSC), Xavier Domingo (BSC) This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101056933. Document Information GRANT AGREEMENT 101056933 PROJECT TITLE Supporting and standardizing climate services in Europe and beyond PROJECT ACRONYM Climateurope2 PROJECT START DATE 01/09/2022 RELATED WORK PACKAGE W2 RELATED TASK(S) T2.4 LEAD ORGANIZATION UKRI AUTHORS Charlotte L. Pascoe (STFC), Rutger Dankers (WR), Ángel G. Muñoz (BSC), Xavier Domingo (BSC) SUBMISSION DATE 29 August 2025 DISSEMINATION LEVEL PU-Public History DATE SUBMITTED BY REVIEWED BY VISION (NOTES) 22 Feb 2024 Rutger Dankers Draft for internal review 29 Feb 2024 Charlotte Pascoe Judith Klostermann, Douglas Cripe Final version D2.3 Preliminary best practices in climate uncertainty quantification and communication | 1 29 Aug 2025 Charlotte Pascoe Update to address the requirements of the 2024 CE2 review. We added several sections: - Section 2.6: Uncertainty evaluation across timescales - Section 5.5: Uncertainty communication across timescales - Section 6: Climate service provision by the private Sector Please cite this report as: Pascoe, C. L., Dankers, R., Domingo, X., Muñoz, Á.G. (2025), Preliminary best practices in climate uncertainty quantification and communication. Update to D2.3 of the Climateurope2 project. Disclaimer: Funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Climate, Infrastructure and Environment Executive Agency (CINEA). Neither the European Union nor the granting authority can be held responsible for them. Table of Contents 1 Introduction..................................................................................................................................................................... 9 1.1 Uncertainty: concepts and definitions..................................................................................................................9 1.2 Sources of uncertainty in the climate-impact modelling chain................................................................ 10 1.2.1 Uncertainty and Time Horizons................................................................................................................ 10 1.2.2 Communication through the value chain.............................................................................................. 12 2 Current practices in uncertainty quantification................................................................................................13 2.1 Uncertainty quantification in climate modelling............................................................................................13 2.1.1 Multi-model ensemble.................................................................................................................................. 13 2.1.2 Perturbed-parameter ensemble............................................................................................................... 14 2.1.3 Initial condition ensemble............................................................................................................................14 2.2 Uncertainty quantification in climate data processing................................................................................15 2.3 Uncertainty quantification in climate impact assessment.........................................................................15 2.4 Passing uncertainty information on down the value chain........................................................................16 2.4.1 Uncertainty in decision making frameworks........................................................................................16 2.4.2 Example of passing uncertainty information on through the value chain................................18 2.5 Uncertainties in extremes.......................................................................................................................................20 2.5.1 Strategies for dealing with uncertainties in extremes......................................................................21 2.6 Uncertainty evaluation across timescales........................................................................................................22 3 Strategies for deep uncertainties........................................................................................................................... 24 3.1 Understanding deep uncertainty......................................................................................................................... 24 D2.3 Preliminary best practices in climate uncertainty quantification and communication | 2 3.2 Accounting for deep uncertainty..........................................................................................................................25 4 Best practices in communicating uncertainties.................................................................................................27 4.1 Formats for communicating uncertainty.......................................................................................................... 27 5 Examples of uncertainty assessment and communication............................................................................ 31 5.1 IPCC AR6....................................................................................................................................................................... 31 5.2 National climate scenarios: KNMI’23.................................................................................................................32 Establish the idea of uncertainty..................................................................................................................33 Present an even number of climate projections.....................................................................................33 Consider the most relevant climate risks..................................................................................................33 Include the lived experience of the audience..........................................................................................34 Using maps............................................................................................................................................................ 35 The language of uncertainty...........................................................................................................................35 5.3 Risk assessment: DNV..............................................................................................................................................36 Understanding Risk............................................................................................................................................36 Risk metrics and assumptions........................................................................................................................37 5.4 Climate interventions: Red Cross - Red Crescent Climate Centre.........................................................38 Humanitarian Sector.........................................................................................................................................38 Development Sector..........................................................................................................................................38 Anticipatory Action............................................................................................................................................39 Transparency and trust.....................................................................................................................................40 Climate risk narratives and storylines........................................................................................................40 Good decisions.....................................................................................................................................................42 5.5 Uncertainty communication across timescales..............................................................................................42 6 Climate service provision by the private sector................................................................................................ 44 6.1 Uncertainty quantification......................................................................................................................................45 6.1.1 Model ensembles.............................................................................................................................................45 6.1.2 Scenarios.............................................................................................................................................................45 6.1.3 Economics.......................................................................................................................................................... 46 6.2 Uncertainty communication.................................................................................................................................. 46 6.2.1 Format..................................................................................................................................................................46 6.2.2 Scenarios.............................................................................................................................................................46 6.2.3 Visual cues..........................................................................................................................................................46 6.3 Co-production and training.................................................................................................................................... 46 6.3.1 Clients’ background........................................................................................................................................47 6.3.2 Clients’ requests..............................................................................................................................................47 6.4 How can standardisation help? (or not).............................................................................................................47 6.4.1 Communication methods, phrasing, and vocabulary........................................................................47 6.4.2 Regulation and standardisation — advantages................................................................................... 48 6.4.3 Regulation and standardisation — disadvantages............................................................................. 48 6.4.4 Standardisation — considerations...........................................................................................................48 6.5 Trustability / openness / transparency...............................................................................................................49 D2.3 Preliminary best practices in climate uncertainty quantification and communication | 3 6.5.1 Publication in peer-reviewed journals....................................................................................................49 6.5.2 Proprietary knowledge.................................................................................................................................49 6.5.3 Competition and comparison between providers............................................................................. 49 6.5.4 Clients’ rights and responsibilities........................................................................................................... 49 7 Concluding Comments: Emerging Themes..........................................................................................................51 Start with the most relevant risks................................................................................................................51 Standard approach to uncertainty descriptions.................................................................................... 51 Use language the audience is familiar with (don’t say uncertainty)................................................51 There are multiple ways to evaluate and communicate uncertainty.............................................51 Use communication about uncertainty to build trust..........................................................................51 Precision of information should be relevant to the situation........................................................... 52 Understand existing narratives.....................................................................................................................52 Be aware of deep uncertainties.................................................................................................................... 52 8 References......................................................................................................................................................................53 9 Annex...............................................................................................................................................................................60 9.1 Interview questions...................................................................................................................................................60 List of tables Table 2.1: Overview of the main sources of uncertainty in climate data across timescales, plus strategies for evaluating the uncertainty. Based on established practices in the climate science literature (e.g., Hawkins & Sutton 2009; IPCC, 2021; Doblas-Reyes et al. 2013; Meehl et al., 2014).....22 Table 3.1: Quantification of the criticality of assumptions to risk assessment metrics, after Flage (2019).............................................................................................................................................................................................26 Table 4.1: Verbal descriptions of quantified uncertainty (or likelihood) in the guidance note for the IPCC Fifth Assessment Report (Mastrandrea et al., 2011). Note phrases ‘More likely than not’ (for probabilities above 50%), ‘Extremely likely’ (for probabilities above 95%) and ‘Extremely unlikely’ (for probabilities below 5%) have also been used ……………………………………………………………………………….…….28 Table 5.1: Examples of climate science terminology that have the potential to be misunderstood by the public and alternative options…………………………………………………………………………………………………….….…...35 Table 5.2: Overview of strategies for communicating climate data uncertainty across timescales…..…42 List of figures Figure 1.1: Sources of uncertainty in climate projections as a function of time horizon based on analysis of CMIP5 results, presented as a plume (a) and as a fraction of the total variance (b). (a) Projections of global mean decadal mean surface air temperature to 2100 together with a quantification of the uncertainty arising from internal variability (orange), model spread (blue) and RCP scenario spread (green). (b) Fraction of variance explained by each source of uncertainty. Note Figure (b) could be misinterpreted as showing that model spread is decreasing after the 2030s, while in fact it keeps growing throughout the century. From Chapter 11 of the IPCC WGI AR5 and Hawkins and Sutton (2009, 2011).........................................................................................................................................................11 D2.3 Preliminary best practices in climate uncertainty quantification and communication | 4 Figure 1.2: Three dimensions to categorise uncertainty: location, nature, and level (degree). Modified after Walker et al. (2010), Wilby and Dessai (2010)....................................................................................................12 Figure 2.1: The cascade of uncertainty illustrating the potential growth of the envelope of uncertainty and the scale of the uncertainty provenance task. In practice, a bottom-up approach that begins with climate impacts will reduce the communication challenge to a set of discrete pathways (grey triangle going upwards). Modified after Wilby and Dessai (2010).........................................................................................16 Figure 2.2: Illustration on how to define the risk and hazard probability density functions and related definitions of risk, using real maïze yield data for Guatemala (green time series curve). In this example, the key hazard is related to droughts, as measured by the Palmer Drought Standardised Index (PDSI, blue time series curve).............................................................................................................................................................19 Figure 2.3: Figure 2.3: Conceptual example illustrating the re-engineering of vulnerabilities and management of the related uncertainties………………………………………………………………………………….………..20 Figure 3.1: Continuum of uncertainty. The realm of probabilities and other methods to represent uncertainty when comparing the knowledge about outcomes with the knowledge about likelihoods. Modified after Dessai and Hulme (2003), Stirling (1998).........................................................................................24 Figure 3.2: Risk matrix combining a categorisation of the uncertainty associated with the use of a particular model with a categorisation of the consequences. Source: DNV (2021).......................................25 Figure 5.1: The IPCC AR6 approach for characterising understanding and uncertainty in assessment findings. The diagram illustrates the step-by-step process authors use to evaluate and communicate the state of knowledge in their assessment (Mastrandrea et al., 2010). Figure adapted from Mach et al. (2017).............................................................................................................................................................................................32 Figure 5.2: The four KNMI’23 scenarios for climate change in the Netherlands. The number of small blocks represents the extent of climate change around 2100 compared to 1991-2020. The four quadrant framework conveys the severity of climate risk factors associated with low and high CO2 emissions and wetter or drier climate. KNMI (2023)..................................................................................................34 Figure 5.3: Examples of making climate graphs easier to understand by using the same method to present both historical data and data for future projections, and providing context via a representation of the year-to-year variability of the historical period. a) Time-series of historical and projected summer temperature for the Netherlands (KNMI, 2023). b) The annual number of tropical days (observed and projected) for the Netherlands…………………………………………………………………………………….35 Figure 5.4: Examples of quantitative metrics for communicating risk…………………………………………………37 Figure 5.5: Quantitative metrics presented to summarise an assessment of risk are only part of the full picture. An evaluation of the assumptions made in the risk assessment process should also be communicated. Note that tacit assumptions have the potential to obscure a major aspect of the risk (see section 3 for more on the role of assumptions)....................................................................................................38 Figure 5.6: Early Action Plan (EAP) validation steps (Heinrich and Bailey, 2020)...........................................39 Figure 5.7: Understanding existing risk narratives and the co-creation of new climate risk narratives and adaptation pathways towards climate resilience for informal settlements in Lusaka, Zambia. Climate science information was presented during this process, such as flood maps from high resolution modelling, but it was available as print-outs on the walls and did not drive the narrative….40 D2.3 Preliminary best practices in climate uncertainty quantification and communication | 5 Figure 5.8: Climate Risk Narratives / Storylines for Lusaka, Zambia for three scenarios: 1. Hotter & drier, 2. Warmer & more erratic and extreme rainfall, 3. Warmer & more extreme rainfall……………..….41 D2.3 Preliminary best practices in climate uncertainty quantification and communication | 6 . About Climateurope2 Timely delivery and effective use of climate information is fundamental for a green recovery and a resilient, climate neutral Europe, in response to climate change and variability. Climate services address this through the provision of climate information for use in decision-making to manage risks and realise opportunities. The market and needs for climate information has seen impressive progress in recent years and is expected to grow in the foreseeable future. However, the communities involved in the development and provision of climate services are often unaware of each other and lack interdisciplinary and transdisciplinary knowledge. In addition, quality assurance, relevant standards, and other forms of assurance (such as guidelines, and good practices) for climate services are lagging behind. These are needed to ensure the saliency, credibility, legitimacy, and authoritativeness of climate services, and build two-way trust between supply and demand. Climateurope2 aims to develop future equitable and quality-assured climate services to all sectors of society by: ● Developing standardisation procedures for climate services ● Supporting an equitable European climate services community ● Enhancing the uptake of quality-assured climate services to support adaptation and mitigation to climate change and variability The project will identify the support and standardisation needs of climate services, including criteria for certification and labelling, as well as the user-driven criteria needed to support climate action. This information will be used to propose a taxonomy of climate services, suggest community-based good practices and guidelines, and propose standards where possible. A large variety of activities to support the communities involved in European climate services will also be organised. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 7 Executive Summary This deliverable collects examples of best practices about uncertainty quantification and communication in climate services, largely drawing from existing literature and reports. Examples and best practices were also collected from the Climateurope2 community during an online workshop on communicating climate uncertainty, held in November 2023. The document is organised in two main parts, discussing the current state-of-the-art and best practices in uncertainty quantification (Chapter 2), and uncertainty communication (Chapter 4). The deliverable also looks into emerging strategies to deal with deep uncertainties, i.e. those that cannot be quantified (Chapter 3), and it also discusses a number of recent real-world examples for assessing, quantifying and communicating uncertainty in climate information (Chapter 5). A set of eight main lessons learnt regarding preliminary best practices in climate uncertainty assessment and communication can be summarised as follows (see Chapter 6 for more details): 1. Always start with the most relevant risks for the target population 2. A standard approach to uncertainty assessment and communication is needed 3. Use language the audience is familiar with (don’t say uncertainty) 4. There are multiple ways to evaluate and communicate uncertainty 5. Communication about uncertainty builds trust 6. Precision of information should be relevant to the situation 7. Understand existing narratives 8. Be aware of deep uncertainties Update 2025/08 — This update introduces two key additions. First, sections 2.6 and 5.5 now address the influence of timescales on both the evaluation and communication of uncertainty. Second, section 6 incorporates perspectives from a subset of private-sector companies on uncertainty and standardisation. Keywords Climate services, Uncertainty, Communication, Risk, Assessment, Vulnerability D2.3 Preliminary best practices in climate uncertainty quantification and communication | 8 2.2 Uncertainty quantification in climate data processing The various processing steps typically applied to climate model data, including downscaling and bias correction, also have the potential to add uncertainty to the results. For example, an implicit assumption in many bias correction techniques is that biases found in the simulation of the past climate (typically by comparing the model simulations with observations or observation-based datasets) will be similar under future - and often very different - climate conditions. Although the limitations of these assumptions underpinning most bias correction techniques (and, by extension, statistical downscaling methods) have long been recognised (e.g., Ehret et al., 2012; Maraun et al., 2017), the effect of these on the outcomes of an analysis are often not assessed. The impact of the choice of different bias correction methods has been investigated in a number of studies, particularly in hydrological applications (e.g., Chen et al., 2011, 2013; Senatore et al., 2022), although it is still not common practice. Often the contribution of the bias correction method to the overall uncertainty in the results is found to be relatively smaller than the climate model or the scenario uncertainty, but in some cases the bias correction even changed the direction of the climate signal that was present in the original climate simulations (e.g., Huang et al., 2014). 2.3 Uncertainty quantification in climate impact assessment Similar to climate modelling, the uncertainty associated with the use of climate impact models can be evaluated through model intercomparisons, as is done in, e.g., the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP, https://www.isimip.org/). Comparatively fewer studies have looked at the effect of impact model parameter uncertainty, although techniques for quantifying both structural and parameter uncertainty, such as the Generalized Likelihood Uncertainty Estimation (GLUE) method (Beven & Binley, 1992) have been around for several decades. Most of these techniques rely on Monte Carlo simulations or similar approaches with the simulation results expressed as probability distributions of possible outcomes, as opposed to a single deterministic prediction. Fewer studies still have made a thorough end-to-end assessment of the uncertainties involved in the entire climate-impact modelling chain looking at all contributing factors,, as was done for the entire flood risk chain by Metin et al. (2018) or for hydrological impacts of climate change in Nepal by Aryal et al. (2019). While most studies adopt a “top-down” approach exploring the accumulation of uncertainty from emission scenarios to global climate response to regional or local impacts, some have also proposed a “bottom-up” approach starting from the impacted system and exploring how resilient it is to changes and variations in one or more climate variables (Van Bree & Van der Sluijs, 2014). Such a bottom-up approach is focused more on resilience of the system and how adaptation can make it less prone to uncertain and largely unpredictable changes in climate. An example of a more bottom-up approach is the use of impact-response surfaces where an impact model is used to evaluate the response of a system across a range of conditions. Not only does this allow for a more rigorous testing of the impact models (across many possible future conditions), but it also makes it possible to identify critical impact thresholds which might be missed if only a few climate scenarios are evaluated (Fronzek et al., 2022). Examples of this approach include the scenario-neutral approach proposed by Prudhomme et al. (2010) for fluvial flood impacts in the UK; and the application of response surface diagrams for evaluating climate change impacts on crop production by Van Minnen et al. (2000). Pirttioja et al. (2019) used a similar approach to evaluate adaptation options to crop yield shortfalls under climate change. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 15 2.4 Passing uncertainty information on down the value chain Especially when using a top-down approach, the different processing steps in a typical climate-impact modelling chain can give rise to what has been called a “cascade of uncertainty” (Wilby & Dessai, 2010; Figure 2.1). Figure 2.1: The cascade of uncertainty illustrating the potential growth of the envelope of uncertainty and the scale of the uncertainty provenance task. In practice, a bottom-up approach that begins with climate impacts will reduce the communication challenge to a set of discrete pathways (grey triangle going upwards). Modified after Wilby and Dessai (2010). Uncertainty information is generated upstream and passed on downstream. Each step of the cascade (or value chain) will have uncertainty associated with it that needs to be passed on and sufficiently understood to assess its relevance to the subsequent users of the data or information. While a climate service may need to keep a provenance record of the full depth and breadth of the cascade of uncertainty for its data holdings (figure 2.1), the specific uncertainty information to be communicated for any one study will only be that which is representative of the actual data used, and tailored to the intended audience. Each step in the cascade or value chain represents a different community of practice, so uncertainty information may need to be understood by actors that are far removed from the original work. Metadata ontologies such as the Simple Standard for Sharing Ontological Mappings (SSSOM) (Matentzoglu et al. 2022) provide a potential framework for maintaining a shared understanding of uncertainty information throughout the value chain. SSSOM is analogous to the I-ADOPT interoperability framework for observable properties developed by the Research Data Alliance (RDA) working group for InteroperAble Descriptions of Observable Property Terminology (Magagna et al., 2021) described in Climateurope2 Deliverable 2.1. However, SSSOM is also able to explicitly capture the imprecision, inaccuracy and incompleteness of mapped concepts. 2.4.1 Uncertainty in decision making frameworks One of the reasons why uncertainty information needs to be passed along the value chain is that it has the potential to make a difference to the decision that is being targeted by the climate service. At the same time, an uncertainty assessment has to be appropriate to the type of the decision being made, because of the time and effort being involved (Beven et al., 2018b) and the demands of clear communication. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 16 If the level of uncertainty can be described probabilistically, a classic risk approach (e.g., a cost-loss decision model) may be used. However, in a climate change context, this is rarely the case. The main drivers of climate change (which include economic development and population growth) are inherently uncertain, especially at the longer term, and can only be explored using a scenario approach. Moreover, at a detailed level our understanding of the Earth system is rather incomplete which may give rise to surprises, unforeseen effects and unanticipated impacts (sometimes referred to as ‘ignorance’). Therefore, decision-making frameworks are needed that can cope with scenario uncertainty and ignorance. Van Bree & Van der Sluijs (2014) describe three steps to account for uncertainties in a decision framework: 1. Identify and characterise sources of uncertainty; 2. Assess (weigh, appraise, and prioritise) sources of uncertainty; 3. Select and apply methods for dealing with uncertainties. The first step is likely to result in a long list of uncertainty sources and could be approached by analysing each step of the value chain or assessment process, or by considering where in the assessment the different types of uncertainty may occur. In the second step, the relative importance of each uncertainty source can be evaluated by its impact on the final decision or outcome. This may be done by performing a sensitivity analysis or, if quantification is not possible, could be based on expert judgement. In the third step, some of the key uncertainties and their impact on the final decision may be analysed and characterised in more detail. Van Bree & Van der Sluijs (2014) highlight that the uncertainties will need to be re-evaluated throughout the assessment process, as it may not be possible to identify, prioritise and characterise all sources of uncertainty right at the start. To be useful in a decision-making process, the uncertainties obviously need not only to be analysed and described, but also communicated to the decisionor policy-maker. It is therefore important to evaluate which uncertainties are most relevant for the decision at hand, and - if relevant - identify options that are robust given these uncertainties (Van Bree & Van der Sluijs, 2014). The way that information is presented can also have an impact on its interpretation and the decisions that are made on the basis of it. To avoid misinterpretation, decision makers should be provided with a fuller understanding of the context of the information that is given to them, yet not so much information that it would overload them. For instance, to limit the overload potential it can be useful to have a range of methods with which to communicate uncertainty and then to use those methods that are most relevant to the situation. The World Meteorological Organisation (2011) identifies a need for capacity development of both the users and providers of climate services to build the necessary understanding, confidence and skills to quantify and utilize probability and uncertainty better to support decisions and actions. Some practical examples of how uncertainty information is used to inform climate adaptation decisions can be found in Lourenço et al. (2014). D2.3 Preliminary best practices in climate uncertainty quantification and communication | 17 2.4.2 Example of passing uncertainty information on through the value chain An example of how quantitative uncertainty information can be used in decision-making is provided in this section. Technical products provided today by National Meteorological Services and related institutes mainly involve forecasts of precipitation and mean, maximum and minimum temperature. These products lead in general to different managerial actions for different sectors, and therefore stakeholders normally need to go one step further in order to use them for their specific interests, for a hazard only becomes a risk if there is a vulnerability to it. Therefore a hazard-oriented approach alone, which prioritises the sources of harm, is not enough because of its lack of information on the local sensitivity, exposure and adaptive capacity of the population. An adequate, comprehensive approach requires risk assessment, involving the difficult problem of satisfactorily quantifying both hazards and vulnerabilities for the specific sector (e.g. water availability, agriculture, health or energy managements), region of the world, and time scale (e.g. short-term, intraseasonal, seasonal, decadal, long-term) of interest (e.g. Muñoz et al. 2012). Unfortunately, due to the complicated character of the addressed problem, there is no unique methodology to quantify the risk, and different approaches are employed for different activities. These issues make it extremely difficult to compare risk indices on different regions of the world, or even among different sectors (e.g. agriculture and health) for the same geographical region. Moreover, in the cases where risk estimations are available, they tend to exclude information on the associated uncertainties. A way to circumvent these problems is to use a probabilistic risk management approach (e.g. Mora and Keipi, 2006; Mora, 2009; Muñoz et al., 2012), which permits to operationally define a probabilistic vulnerability distribution that is by construction consistent with both the hazard and risk probability density functions. At its core, this approach directly identifies risk with key indicators for decision makers, such as damage or cost, for which real data exists, and from which an empirical (or fitted) probability distribution can be obtained. To illustrate the case with a real-world example, consider the annual, total maïze yield production (in hectograms by hectare, Hg/Ha) for Guatemala, from 1961 to 2005 (Figure 2.2). The maïze yield can be translated into damage cost, for example in hundred of thousands US$, and a probability density function describing the probability distribution can be directly defined from the data record. For the sake of simplicity, assume that such a distribution follows a Gaussian distribution (as in the Figure 2.2) –nonetheless the distribution does not need to be a Gaussian one. In that case, a natural measure of uncertainty is provided by the dispersion parameter of the distribution, or the standard deviation in this case. A similar approach can be followed for the hazard, which in the example of Figure 2.2 is measured via the Palmer Drought Standardised Index (PDSI), which also conveys information on the related uncertainties via the corresponding probability distribution (see example in Figure 2.2). D2.3 Preliminary best practices in climate uncertainty quantification and communication | 18 Figure 2.2: Illustration on how to define the risk and hazard probability density functions and related definitions of risk, using real maïze yield data for Guatemala (green time series curve). In this example, the key hazard is related to droughts, as measured by the Palmer Drought Standardised Index (PDSI, blue time series curve). As suggested by Muñoz et al (2012), it is obvious that, given that the probability density functions for the risk and the hazard are known, a risk manager can mathematically define the associated vulnerability’s probability density function that is consistent with both the risk and the hazard ones. Furthermore, it is possible to then define which vulnerability distribution is required to obtain a risk distribution that the decision makers can cope with, i.e. a re-engineering of the vulnerability distribution given the knowledge of the present and future hazard distribution, and considering the desired or manageable risk distribution (Muñoz et al., 2012). For example, the risk manager might want to have a distribution that provides the most part of the probability distribution that corresponds to low values of risk. A possible choice for such a probability distribution is the exponential distribution (on the left panel of Figure 2.3). D2.3 Preliminary best practices in climate uncertainty quantification and communication | 19 Figure 2.3: Conceptual example illustrating the re-engineering of vulnerabilities and management of the related uncertainties. Overall, the approach enables decision makers not only to quantitatively assess risk and the related uncertainties, but also to support the choice of strategies and tasks that will help achieve a certain desired vulnerability distribution (that translates into concrete exposure, sensitivity and adaptive capacities for the population). 2.5 Uncertainties in extremes Key climate risks are often (though not always) associated with the occurrence of extreme events. Evaluating the nature and probability of these extremes including the possible effect of climate change is, however, subject to considerable uncertainty above and beyond the sources of uncertainty already discussed earlier in this report. Even the word “extreme” is in itself ambiguous, as there is no standard definition of when an event can be considered extreme. Some studies of climate extremes focus on the upper end of the distribution of a climate variable, for example the 90th percentile of daily maximum temperature, or the 95th or 99th percentile of daily rainfall (e.g., Donat et al., 2013; Contractor et al., 2021), which - depending on the precise definition - may be expected to be exceeded every year, if not multiple times per year (Myhre et al., 2019). Others focus on events that are much rarer: studies of changes in river flood frequency, for example, often use river flows with an annual probability of exceedance of 1% (average return period of 100 years, or a “once-in-100-years” flood), although return periods of 10 or 30 years have also been used (Kundzewicz et al., 2018b). The IPCC (2021) defines a climate extreme as “the occurrence of a value of a weather or climate variable above (or below) a threshold value near the upper (or lower) ends of the range of observed values of the variable. By definition, the characteristics of what is called extreme weather may vary from place to place in an absolute sense.” A key challenge in the estimation of especially high-impact extreme events to which society is not well adapted or prepared, such as severe flooding or intensive heatwaves, is that such events are typically very rare. This means that the sample size of these events - to study their nature and impacts, let alone perform a statistical analysis or detect any trends - can be very small or even non-existent. Take, for D2.3 Preliminary best practices in climate uncertainty quantification and communication | 20 example, the “once-in-a-100 years” flood event, or more precisely the river flow level with an annual probability of exceedance of 1%, which in many countries serves as a baseline for flood protection measures. The probability of exceeding this level at least once in a 100-year time series of data (observations or model-based) at the same location is only about 63%, meaning there is still a chance of 37% of not observing this river flow level at all. However, to detect a trend you would ideally need multiple events, and the probability of exceeding the 100-year flood level at least three times in a century is - statistically speaking - only about 8% (Cloke & Pappenberger, 2009). In reality, the statistical distribution is unlikely to remain stationary over such extended periods, not only due to climate variability and long-term climate change, but also because of land use change, changes in water management practices, engineering interventions, and other anthropogenic and environmental factors. In addition there may be inhomogeneities in the data, for example due to changes to the observation network, changes in the measurement method, or instrument drift. Measurements of extreme events can also be uncertain as instruments sometimes break down or wash away, or have been calibrated under different conditions. River flow, for example, is typically observed by measuring water level; the relationship between water height and flow volume (the rating curve) that is established during “normal” flow conditions may, however, no longer be valid during a flood (Di Baldassarre & Montanari, 2009; Hamilton & Moore, 2013). Because of the often limited sample size, the probability or frequency of extreme events is often estimated by fitting an extreme value distribution to a larger sample of the data, such as the annual maximum (or minimum) values or all peak values over a predefined threshold. Although these distributions provide a more robust estimate of extreme quantiles, these are nevertheless increasingly uncertain at more extreme levels, especially if the return period exceeds the length of the data record. Also the choice of the extreme value distribution (e.g., Generalised Extreme Value (GEV) vs. Gumbel) can bias return period estimates, as can the use of stationary models when the underlying distribution is changing due to climate or land use changes. Conversely, modelling non-stationarity introduces additional structural uncertainty in the estimates. Uncertainty in the evaluation of climate extremes may also arise from the communication and interpretation of some of the key concepts. A common source of misunderstanding is related to the concept of return period as reflected in phrases such as “a once-in-a-100-years” event or in short a “100-year event”. Statistically, such an event has an annual probability of exceedance of 1% which (in theory at least) applies every year, regardless of whether a similar event has just happened or not, The average recurrence interval between two events then works out as 100 years only when averaging over many events; statistically it is very well possible for multiple “100-year” events to happen within a relatively short timespan. However, many people interpret the concept of a 100-year event as an event that happens exactly once every 100 years, resulting in misunderstandings such as thinking that a new event is less likely when one has just occurred, or more likely when it has not happened for a long time (the “flood is due” effect; Grounds et al., 2018). Such misinterpretation of technical terms can result in a misperception of risk and ultimately negatively affect decision-making, for example around insurance or preparedness (Lee et al., 2021). But also other technical concepts related to the communication of uncertainty such as ranges, probabilities, and confidence levels can be misunderstood or underutilised by stakeholders. 2.5.1 Strategies for dealing with uncertainties in extremes Methods for evaluating and/or quantifying the uncertainty in climate extremes are to some extent similar to those discussed earlier in this chapter for climate data in general. For example, it is advisable to adopt a multi-model and multi-scenario approach to investigate the influence of scenario and model structural uncertainty on the outcomes, including the effects of impact model uncertainty D2.3 Preliminary best practices in climate uncertainty quantification and communication | 21 where appropriate. If statistical models are being used to estimate the likelihood of extremes, multiple extreme value distributions (e.g., GEV, Gumbel, Generalised Pareto) could be applied to test the robustness of the results. Likewise, covariates can be included into the extreme value distributions when this is justified (Coles, 2001; Katz et al., 2002). Methods also exist to calculate the parameter uncertainty in these distributions explicitly using bootstrapping or Bayesian inference (e.g., Gilleland, 2020; Cindrić & Pasarić, 2019). If the sample of extreme events is too small and/or too much affected by natural variability, a more robust signal of changes in extremes may sometimes be obtained by aggregating over larger regions. This has been found for climate variables such as precipitation (Fischer et al., 2013) but also may also apply to timeseries of flood events (Dankers & Kundzewicz, 2020). Large initial-condition climate model ensembles or millenial scale climate simulations provide an opportunity to estimate the role of natural variability in extreme-event probability giving more robust estimates (e.g., Huang et al., 2016; Maher et al., 2021). The he UNSEEN method (UNprecedented Simulated Extremes using ENsembles; Thompson et al., 2017) extends this approach to large ensembles of past seasonal or decadal forecast or hindcast data to identify plausible—but previously unseen—extreme weather events by sampling well beyond the observational record, and has been used by, e.g., Kay et al. (2024) to estimate the likelihood of unprecedented hydrological extremes that lie outside the range of historical observations. 2.6 Uncertainty evaluation across timescales Different sources of uncertainty dominate at different time horizons, and the approaches used to quantify or manage them vary accordingly. Table 2.1 provides a structured overview of the main sources of uncertainty—ranging from observational and modelling limitations to internal climate variability and scenario uncertainty—across the typical timescales of climate information (historical observations, seasonal prediction, decadal prediction, and long-term projections). The table also outlines common strategies to address these uncertainties, drawing on established practices in the climate science literature (e.g., Hawkins & Sutton 2009; IPCC, 2021; Doblas-Reyes et al. 2013; Meehl et al., 2014). Table 2.1: Overview of the main sources of uncertainty in climate data across timescales, plus strategies for evaluating the uncertainty. Based on established practices in the climate science literature (e.g., Hawkins & Sutton 2009; IPCC, 2021; Doblas-Reyes et al. 2013; Meehl et al., 2014). Time scale Main sources of uncertainty Strategies Historical climate (observations, reanalyses) ● Measurement errors (instrument calibration, homogenization issues) ● Changes in station locations/methods ● Incomplete spatial coverage ● Reanalysis model dependence ● Rigorous quality control and homogenization ● Use of multiple observational datasets (gauge, satellite, reanalysis) ● Metadata and bias corrections ● Data rescue & paleoclimate proxies for extension Seasonal prediction (months to ~1 year) ● Model structural uncertainty (initialisation, physics) ● Use multi-model ensembles (e.g., C3S, NMME) ● Ensemble initialisations to D2.3 Preliminary best practices in climate uncertainty quantification and communication | 22 ● Internal variability (ENSO, MJO, NAO etc.) ● Limited predictability horizon ● Initial and boundary condition uncertainty (SSTs, soil moisture, sea ice) sample internal variability ● Hindcast-based calibration and bias correction ● Probabilistic forecast products Decadal prediction (1–10 years) ● Internal climate variability (decadal oscillations, volcanic eruptions) ● Model drift and bias ● Initial condition data ● Moderate scenario uncertainty (esp. aerosols, near-term emissions) ● Initialised decadal prediction ensembles ● Multi-model comparisons ● Bias adjustment & drift correction ● Large ensembles to separate forced signal from variability Long-term climate projections (multi-decadal to century) ● Emission/scenario uncertainty (dominant beyond ~2050) ● Climate model structural uncertainty ● Internal variability (more important at regional scales) ● Use of multiple emission scenarios (SSPs/RCPs) ● Multi-model ensembles (CMIP, CORDEX) ● Large initial-condition ensembles for variability quantification ● Storyline approaches for communicating divergent outcomes D2.3 Preliminary best practices in climate uncertainty quantification and communication | 23 3 Strategies for deep uncertainties 3.1 Understanding deep uncertainty When a system is well understood and there are good measurements available, then uncertainty can be presented using statistical methods and probabilities as is commonly the case for weather forecasts. Uncertainties associated with limited knowledge, for example about future socio-economic and technological developments and about certain aspects of the climate system, are more appropriately conveyed with the use of scenarios that indicate “what if” situations. Figure 3.1 Continuum of uncertainty. The realm of probabilities and other methods to represent uncertainty when comparing the knowledge about outcomes with the knowledge about likelihoods. Modified after Dessai and Hulme (2003), Stirling (1998). Epistemic or deep uncertainty refers to a state of incomplete knowledge and understanding that goes beyond standard uncertainty. It involves fundamental ambiguity about the underlying system, its dynamics, and the relevant decision-making context. This concept is related to the notion of “vague uncertainties” of Budescu and Wallsten (1987) and similar concepts dating back to at least the 1920s. Deep uncertainty is often characterised by the inability to assign precise probabilities to future events or to fully comprehend the system's complexity. It often involves unknown unknowns, where potential future scenarios and their likelihoods are difficult to define. Deep uncertainty challenges traditional decision-making frameworks that assume a known and probabilistic future, requiring approaches that can navigate ambiguity, embrace scenario thinking, and incorporate adaptive strategies to account for the inherent unpredictability of certain situations. It also has implications for communication as overly precise numerical expressions of the likelihood of a particular event or outcome are potentially misleading. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 24 5 Examples of uncertainty assessment and communication The previous sections summarise what the literature has to offer on communicating uncertainties in climate information. We also wanted to gather some information about if and how these methods are used in practice. In this section we will present a number of examples of how the concept, methods and practices with regards to the evaluation and communication of uncertainty are implemented by the climate services community. With the exception of the first case study, the examples were collected from the Climateurope2 community during an online workshop on communicating climate uncertainty that was held in November 2023. In total, 83 people participated in the workshop. Three lectures were given by representatives from KNMI, DNV and the Red Cross / Red Crescent. During the workshop, participants discussed how users of climate information deal with uncertainties and how providers of climate services should communicate about uncertainties in ways that enable users to extract the information they need. Below, we introduce the uncertainty communication framework used by the IPCC and summarise the inputs from the three workshop lectures. 5.1 IPCC AR6 The IPCC framework for characterising knowledge and uncertainties (Mach et al. 2017) is a single framework (Figure 5.1) that could be applied consistently across working groups, spanning diverse disciplines and topics. This shared framework aimed to increase the comparability of assessment conclusions across all topics related to climate change, from the physical science basis to resulting impacts, risks, and options for response. The diagram in figure 5.1 illustrates the process IPCC AR6 authors used to evaluate and communicate the state of knowledge in their assessment. The process begins with evaluation of evidence and agreement (steps 1–3). Where possible, authors then evaluate confidence, synthesising evidence and agreement in one qualitative metric (steps 3–5). Where uncertainties can be quantified probabilistically, authors subsequently evaluate likelihood or a more precise measure of probability (steps 5–6). Note that the likelihood categories should be considered to have “fuzzy” boundaries (step 6 (CCSP, 2009)). Unless otherwise specified, assessment conclusions characterised probabilistically are underpinned by high or very high confidence. Authors present evidence/agreement, confidence, or likelihood terms with assessment conclusions, communicating their expert judgments accordingly. Example conclusions drawn from the IPCC AR6 are presented in the box at the bottom of the figure. (Mach et al. 2017) D2.3 Preliminary best practices in climate uncertainty quantification and communication | 31 Figure 5.1: The IPCC AR6 approach for characterising understanding and uncertainty in assessment findings. The diagram illustrates the step-by-step process authors use to evaluate and communicate the state of knowledge in their assessment (Mastrandrea et al., 2010). Figure adapted from Mach et al. (2017). 5.2 National climate scenarios: KNMI’23 The Royal Netherlands Meteorological Institute (Koninklijk Nederlands Meteorologisch Instituut - KNMI) is a national knowledge institute for weather, climate and seismology. The KNMI’23 climate scenarios (KNMI, 2023) are four new scenarios which outline what the future climate in The Netherlands could look like. With the new climate scenarios, KNMI offers guidelines for policy advisers and other professionals so they can make adequate decisions to ensure a safe, liveable and prosperous Netherlands in a changing climate. Janette Bessembinder from KNMI kindly shared with us the benefit of her experience of communicating climate information and uncertainty to wider society at the Communicating Climate Uncertainty workshop held by Climateurope21. 1 https://climateurope2.eu/news-events/events/events/communicating-climate-uncertainty D2.3 Preliminary best practices in climate uncertainty quantification and communication | 32 Establish the idea of uncertainty The first step is to establish the idea that there is not just one prediction for the future of the climate. KNMI’23 does this by presenting information in terms of two scenarios, one associated with high CO2 emissions and another with low CO2 emissions. The simple act of presenting two possible futures conveys the concept that the future is not known (is uncertain) and, more subtly, that in the long-term, the main source of that uncertainty is uncertainty associated with how humans will collectively behave. Present an even number of climate projections Climate communication practitioners should be aware of the impact of visual information. An audience will often interpret the middle or central projection of future climate as being the most likely when in fact no likelihood can be attributed. To avoid this bias, practitioners can present the audience with an even number of projections. Giving an even number of projections implicitly encourages the audience to think about which of the projections is the most relevant for their topic of interest. Consider the most relevant climate risks Climate communication practitioners should consider which are the most important climate risks for a specific audience. In The Netherlands there are distinct risks associated with dry years and with wet years, and sea-level rise is also a major concern. The most probable risks are often not the most relevant risks, so the simulated scenarios that inform an analysis should be those that do the best job at representing the risk situation that is of particular concern. KNMI’23 uses a four quadrant framework of high emissions vs low emissions (representing uncertainty about future socio-economic and technological developments) and drier climate vs wetter climate (representing uncertainty about the climate system). The framework is used to convey an indication of the severity of climate risk factors for temperature change, wetter winters, extreme summer showers, drought, and sea-level rise. (figure 5.2) This four quadrant format is able to convey whether year on year variability of a climate risk is to be expected. The severity of risk factors to do with temperature change and sea-level rise are only dependent on the emissions scenario. The severity of risk factors to do with wetter winters, extreme summer showers, and increased droughts are also dependent on whether the year is particularly wet or particularly dry. The four quadrant framework adopted by KNMI presents climate uncertainty in terms of a set of threats for which to be prepared. The assumption being that information about threats which are specific and relevant for an audience empower climate action. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 33 Figure 5.2: The four KNMI’23 scenarios for climate change in the Netherlands. The number of small blocks represents the extent of climate change around 2100 compared to 1991-2020. The four quadrant framework conveys the severity of climate risk factors associated with low and high CO2 emissions and wetter or drier climate. KNMI (2023). Include the lived experience of the audience For many people, the year-to-year variability of the climate system is more important and impactful than the slow shift of long-term climate means. Information about a projected future can be made more tangible and less abstract when it is framed in the context of the lived experience of the audience. Including year-to-year historical variability on graphs of historical and projected climate trends gives the audience an appreciation of what the variability they have experienced looks like in the context of the information that is presented about the future. For time series plots, this could be the use of 90% shading to show both the variability of the past climate as well as that of a projected future, with data for the year-to-year variability of the historical period overlaid (as in figure 5.3a). For box plots this could be overlaying the historical statistics with annual data points (as in figure 5.3b). Presenting historical data using the same method as data for the projected future, with context via a representation of year-to-year variability, makes the information on the graph easier to understand. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 34 (a) (b) Figure 5.3: Examples of making climate graphs easier to understand by using the same method to present both historical data and data for future projections, and providing context via a representation of the year-to-year variability of the historical period. a) Time-series of historical and projected summer temperature for the Netherlands (KNMI, 2023). b) The annual number of tropical days (observed and projected) for the Netherlands. Using maps An effective way to convey the geographical distribution of how the behaviour of a climatological index might be expected to change in the future is with a contour map. However, if the communication only provides one future instance it would convey an implicit bias towards the chosen scenario. The uncertainty of the future climate can be conveyed by showing at least two instances from different scenarios. People like to have access to climate information that is local to them. When presenting local information, it can be better to convey the values of climate indicators in terms of their absolute values rather than as percentage changes. There is an implicit assumption that data presented as a percentage change is true for the whole of an area rather than being specific to a locality. The language of uncertainty Much of the language that we use as climate scientists can have a very different meaning in public life, this is particularly pertinent for the topic of uncertainty communication. It can be better to use commonly understood descriptive language in place of scientific terminology. For instance, rather than using the word “uncertainty” which implies ignorance, a better choice could be the word “range”. But if the word uncertainty can not be avoided then it should be clearly explained. More examples of commonly misunderstood climate science terminology can be found in table 5.1 . Emphasising uncertainties too much may leave people “paralysed” and unsure of how to act. In climate change communication it is better to focus on what is known first. e.g. “all climate scenarios show that temperatures and extreme precipitation increase, even though there are some uncertainties”, rather than “there are many uncertainties about climate change, but we know that temperatures will change”. However, we should not obscure uncertainties. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 35 Table 5.1: Examples of climate science terminology that have the potential to be misunderstood by the public and alternative options. 5.3 Risk assessment: DNV DNV2 are experts in assurance (protection against events) and risk management for industry with a stated purpose to safeguard life, property and the environment. Andreas Hafver from DNV kindly gave us the benefit of his insights on risk management from an industry perspective and its relationship to uncertainty at the Communicating Climate Uncertainty workshop held by Climateurope2. Uncertainty is a part of the risk and should not be used to take risk less seriously. In fact more uncertainty is a reason to take a risk more seriously. However, people don’t like to hear about uncertainties, they want to have numbers and clear recommendations. Nevertheless, we should harness the power of uncertainty, because with an awareness of uncertainties comes an ability to assess risks more critically. More confidence can be placed in assessments of risk that are clear, in spite of the uncertainties. Understanding Risk There are many definitions for risk, a few examples are listed below: ● Cambridge Dictionary: “the possibility of something bad happening” ● IPCC AR6: “the potential for adverse consequences” ● ISO 31000: “the effect of uncertainty on objectives” ● ISO/IEC 61508: “combination of the probability of occurrence of harm and the severity of that harm” In general, the principle of risk is an assessment of the likelihood (degree of certainty) about a set of consequences that are either wanted or unwanted. 2 https://www.dnv.com/about/index.html# D2.3 Preliminary best practices in climate uncertainty quantification and communication | 36 A risk can be described in terms of a two axis framing of its consequences. Some consequences are wanted while others are unwanted, and some consequences are certain while others are uncertain. The severity of the consequences of a risk will be assessed differently by different stakeholders who will have different objectives, different knowledge and different ways of knowing. The values and objectives of a society will also change through time and things that were previously not considered as risks become important to consider and vice versa. Risk metrics and assumptions Risk assessments often communicate risk with plots showing quantitative risk metrics (Figure 5.4). For example, risk matrices of severity and likelihood, risk maps showing the geographical distribution of risk, risk radars which show how a range of scenarios score on different parameters, exceedance curves that show the frequency of events and their impact, and forecasts of some quantity of interest with some uncertainty around the predictions with maybe a critical limit to show uncertainty about when the limit will be reached. Figure 5.4: Examples of quantitative metrics for communicating risk. However, risk is more than what we can quantify with metrics. Any risk assessment involves choices and assumptions as part of the analysis. The metrics presented to describe a risk will only capture aspects of the full risk picture. Tacit assumptions made as part of the risk assessment process have the potential to obscure a major part of the risk picture. To convey a more complete picture of the risk, quantitative metrics could be accompanied by an evaluation which includes: the rationale for the metric; properties of the evidence used; degree of consensus; criticality of the assumptions; and the location, nature and degree of uncertainty (see sections 1.2 and 3). D2.3 Preliminary best practices in climate uncertainty quantification and communication | 37 Figure 5.5: Quantitative metrics presented to summarise an assessment of risk are only part of the full picture. An evaluation of the assumptions made in the risk assessment process should also be communicated. Note that tacit assumptions have the potential to obscure a major aspect of the risk (see section 3 for more on the role of assumptions). However, DNV stated that it is not always appropriate or indeed useful to communicate risk using quantitative methods (see also section 4 and 5.1). Even if a quantitative analysis is possible it may not be necessary, sometimes it is sufficient to provide a qualitative description of the risk to support a decision. On occasions where it is necessary to be more precise, the precision of the risk description should match what is supported by the knowledge. 5.4 Climate interventions: Red Cross - Red Crescent Climate Centre The mission of the Climate Centre is to support the Red Cross and Red Crescent Movement and its partners in reducing the impacts of climate change and extreme-weather events on vulnerable people. Christopher Jack is a climate science expert with the Climate Centre who is primarily engaged with forecast based finance and anticipatory action. Chris has become increasingly involved and passionate about integrating climate science into collaborative learning, co-production, and transdisciplinary action research processes. We are grateful to Chris for sharing his experience in this field with the Communicating Climate Uncertainty workshop held by Climateurope2. Humanitarian Sector Aid interventions in the humanitarian sector aim to limit the impact of rapidly emerging and evolving crises by anticipating emerging and evolving crises over weeks through months and possibly years. The humanitarian sector is characterised by high complexity with compounding and cascading risks and impact, high uncertainty with low quality or no data related to vulnerability as well as climate, and a high cost of failure. Failure to anticipate impacts can cost a lot more than acting in anticipation. The mis-allocation of funding and resources can cost lives. Within complex humanitarian contexts the relative contribution of weather and climate compared to other factors influencing a crisis is not always clear. Development Sector Aid interventions in the development sector aim to enable, support and encourage development in the face of climate variability and long-term change. The development sector is characterised by high complexity of climate risk governance (Who makes what decisions? Where does the money come from? There are multiple and contested agendas), and high uncertainty with lots of actors including D2.3 Preliminary best practices in climate uncertainty quantification and communication | 38 consultants and climate service providers. From a user perspective there is uncertainty about where to get information (who to listen to and which portal to use). Therefore, building trust and relationships is very important (who to trust and why to trust them). Anticipatory Action With anticipatory action, a warning of a crisis event allows action to be taken before the impacts occur. Forecast based Finance (FbF) is an example of an Early Action Plan (EAP) mechanism that enables disaster preparedness through humanitarian funding for early action based on in-depth forecast information and risk analysis. The release of funds to trigger early actions is agreed in advance and is based on forecasts exceeding specific thresholds (figure 5.6). Figure 5.6: Early Action Plan (EAP) validation steps (Heinrich and Bailey, 2020). 1. Risk assessment. Historical impact reports are used to identify risks, however, there remains a lot of uncertainty about how events impact communities. Impact reports require a broad participatory process involving national meteorological services and disaster management agencies collaborating with government and communities to identify beneficiaries (Who is most vulnerable? Who will benefit the most?). Broad involvement is used to achieve collectively supported decisions. 2. Identify forecasts. Forecasts from local institutions are prioritised as they will have a mandate and early warning systems in place. Trust in and a sense of ownership of local forecasts can be more important than forecast skill. 3. Define impact level. When will we trigger? The funding can’t support lots of false alarms so organizations aim to trigger only for high impact (1 in 5 years) events. Forecast skill may require accepting a certain number of false alarms or misses. It is important to collectively agree on an acceptable probability of false alarms and misses as these can degrade trust if the approach is not collectively owned. Collective agreement is difficult but is a core tenant of managing uncertainty in this context. 7. Multiple flexible triggers. Some uncertainty can be managed by using multiplied staged triggers. The first trigger has a higher false alarm rate but activates low cost preparedness actions. The second trigger has a lower false alarm rate and may even be a STOP that halts action. Flexible triggering moves away from “objective” triggers and allows for expert consensus on triggering. This allows for unanticipated factors not captured by the trigger model to be included (e.g. emerging conflict). D2.3 Preliminary best practices in climate uncertainty quantification and communication | 39 Transparency and trust Communication formats should transparently convey the nature of uncertainty being communicated and be readily comprehensible to ensure that decisions can be made based on an understanding of the uncertainties. Traditional, scientific formats for communicating uncertainty, such as technical graphs, can be difficult for non-scientists to comprehend. Although simplified information may be more easily understood, it may not provide sufficient depth of information to inform decisions. Finding the right balance may be best achieved using co-production (Coventry et al., 2019). Trust in climate information and those who communicate information should be measured and evaluated to assess how communication and engagement activities influence trust in information. Stakeholders often equate uncertainty with ‘not knowing’ and/or a lack of accuracy. This can reduce trust in using the information and in turn prevent action. Measuring trust can help identify communication approaches that foster shared understandings of uncertainty (Coventry et al., 2019). Climate risk narratives and storylines There is strong evidence that people use narratives to capture the essential meaning of complex evidence (see also section 3). When presented with complexity, people gravitate towards constructing a narrative. Even if someone is presented with a scientific graph, what they take away will be a narrative about what the graph means, and that meaning will be different for different people. Just as risk is different for different people, the way they engage with complexity is different too. People also tend to hold on to existing narratives and seek evidence that confirms these (confirmation bias). Figure 5.7: Understanding existing risk narratives and the co-creation of new climate risk narratives and adaptation pathways towards climate resilience for informal settlements in Lusaka, Zambia. Climate science information was presented during this process, such as flood maps from high resolution modelling, but it was available as print-outs on the walls and did not drive the narrative. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 40 From the provider’s side, feedback from clients is regularly collected and used to improve both the content and the way information is delivered. However, one recurring challenge is finding the person responsible for climate risk within a client’s organization. Few companies have a dedicated climate department, and risk managers often do not yet see the value of climate information—or are unsure how to integrate it into their processes. 6.3.1 Clients’ background Clients who are new to climate services are particularly vulnerable. They may not know what to ask for—because it is hard to define what you want if you don’t yet understand what is possible. This makes them more susceptible to confident, persuasive messages, even if those messages oversimplify the reality. Misunderstandings are common, and part of the learning process. For example, a colored map with grid cells every kilometer may look straightforward, but it could represent either interpolated observations or results from a complex physical model—two very different things. Skepticism also plays a role. Large climate models often have a poor reputation among non-experts. A common question is: “If weather forecasts can be wrong, how can we trust climate projections?” Addressing this requires careful explanation of the difference between predicting short-term weather and modeling long-term climate trends. For providers, the challenge lies in communicating clearly with people who may have little background knowledge. Successful communication happens when both sides start “speaking the same language.” Some sectors are already familiar with climate data, which makes this process faster. In other sectors, more time, training and adaptation are needed. 6.3.2 Clients’ requests Clients rarely approach providers with a clear idea of what format or product they need. They usually know they want climate information, but not the specifics. Most clients are interested in projections up to 20 years ahead. Beyond that, trust tends to decline, either because the data is seen as too uncertain or because clients feel they cannot realistically make decisions so far into the future. Expectations around uncertainty vary widely. Some clients prefer to avoid it altogether, wanting only the simplest possible outputs. Others take the opposite view: the more detail provided, the more valuable the information becomes, because they can build on it. There are also mixed cases: Some want uncertainty quantified at every grid cell. Others only want broad metrics describing potential change, distribution, or extent over time. Some know they want uncertainty included, but are unsure of the form it should take. In many cases, clients focus less on uncertainty itself and more on precision, accuracy, and skill. Some even test providers through pilot projects: they supply a dataset and expect the provider to demonstrate results that outperform their existing methods. 6.4 How can standardisation help? (or not) 6.4.1 Communication methods, phrasing, and vocabulary D2.3 Preliminary best practices in climate uncertainty quantification and communication | 47 Vague or overly emotional wording can be misleading. It also makes it harder for clients to compare different services. Clear phrasing guidelines for providers could help customers better understand what each service offers, manage expectations, and highlight the unique value proposition of each provider. Messages should be understandable by a broad audience—even children. At the same time, definitions should not be so rigid that they limit innovation in the very methods they aim to describe. Vocabulary also differs widely between industries. While certain climate-specific terms can be standardised, once communication moves into resilience and sector-specific risks, the language often becomes less precise and harder to unify. Any effort at standardization should therefore be developed in collaboration with private sector actors and market players, to ensure it reflects real needs and practices. 6.4.2 Regulation and standardisation — advantages From the client’s perspective, regulation and standardization can bring a number of benefits. They raise awareness and help define the minimum information that should be delivered, ensuring that at least a baseline of quality is met. Once services are standardised, their price often decreases, making them more accessible. At the same time, standardization can increase trust, for example by introducing certification systems that distinguish between different levels of quality according to the effort and methodology used. This kind of certification not only justifies different price levels but also allows companies to position themselves in different parts of the market. Smaller firms and startups, in particular, could gain credibility from such systems, as official standards would give them a stronger basis to demonstrate their value in comparison with larger corporations. 6.4.3 Regulation and standardisation — disadvantages There are also clear downsides. If climate services become too similar under a standardised framework, the market risks sliding into commoditization, where every provider delivers the same minimum product. This drives prices and profits down, which in turn can force smaller providers out of the market, reduce competition, and leave space mainly for large corporations. In such a scenario, innovation suffers, because there is little incentive to go beyond the minimum requirements. Some fear that regulation would shift the market toward compliance-driven “regtech” products, designed mainly to satisfy rules rather than to improve climate services themselves. Certification processes, if they are too slow or rigid, could also become an obstacle, especially for small and fast-moving companies. Moreover, there is an argument that startups are already able to compete effectively without such certifications, and that additional layers of bureaucracy would not add value. 6.4.4 Standardisation — considerations Even if regulation and standardization are pursued, important questions remain. How can standardisation be made downstream when upstream there are still efforts to be made? For example, the provision of information through climate model ensembles is not yet fully standardised, which raises doubts about how far downstream services can be expected to align. It is also important to remember that many clients approach smaller companies because they want more than simple compliance; they are looking for tailored services and additional insights. This leads to a deeper question: what exactly should be the objective of standardization? If the goal is to guarantee the same level of quality everywhere, for any user at any time, is this something climate science can realistically provide? And even if it can, how do we define “quality” in a way that works across all sectors? Existing resources already provide partial answers. Organizations such as ECMWF and C3S have developed robust libraries that many providers treat as guidelines or informal standards. One option D2.3 Preliminary best practices in climate uncertainty quantification and communication | 48 could be to require companies that choose not to use these resources to make that clear to clients. In terms of climate model ensembles, the CMIP framework already functions like an international standard, similar to an ISO, by enforcing specific formats and protocols. In addition, the IPCC already offers some guidelines on how to communicate uncertainty and what wording to use. These examples show that parts of the system are already standardised, which raises further questions about where additional regulation is truly necessary. 6.5 Trustability / openness / transparency An important question for climate service providers is whether they would be willing to link their own success to that of their clients—for example, by earning money only when their service leads to profits. Many companies would not accept such terms, which highlights the difference between offering information and taking on business risk. Some voices in the field argue that providers should not be afraid of communicating openly about the limits of their work, including acknowledging that uncertainty is both very large and extremely difficult to quantify. There is also increasing demand for companies that promote “magical” algorithms or deliver purely qualitative outputs to demonstrate how they validate their results. Transparency about validation is essential to build trust. 6.5.1 Publication in peer-reviewed journals Publishing in peer-reviewed journals can help strengthen credibility and foster trust, but the picture is not entirely straightforward. Even when providers are willing to publish, not all types of data or sources can be openly shared. The degree of openness in scientific journals also varies: publishing does not always mean being fully transparent about every aspect of a service. Moreover, methodologies evolve over time, which means that what was described in a paper a few years ago may no longer reflect the current practice. This creates a gap between published material and the living, evolving methods that providers use in their services. 6.5.2 Proprietary knowledge Some companies choose to protect their intellectual property through strict industrial secrecy, keeping everything developed in-house under tight control. Confidentiality agreements allow them to disclose more detail to clients when necessary, but here too, openness exists in shades of grey. In practice, keeping knowledge proprietary rarely prevents companies from attracting new clients, since high-level information is usually sufficient. In fact, as mentioned earlier in the context of client requests, some customers test providers directly by running pilot projects, offering their own data and asking the provider to demonstrate that their results outperform the client’s current state of the art. 6.5.3 Competition and comparison between providers Competition between providers often touches on the question of transparency. Some companies complain that others are not open enough, and suggest that if all methodologies were clearly laid out, it would be easier to see which approach is stronger. Cultural differences also shape how providers present themselves. In some countries, humility in communication is mistaken for weakness or uncertainty, while in Europe clients increasingly see humility as a mark of seriousness and credibility. Another issue arises in the context of regionalisation and downscaling, where providers debate whether complex new techniques should be openly disclosed. The discussion becomes even sharper when artificial intelligence is involved, since many clients view AI methods as a “black box.” 6.5.4 Clients’ rights and responsibilities Clients ultimately have the right to make informed decisions based on the services they receive. This makes clear documentation not just useful, but in many cases essential, and some argue it should even be compulsory. To protect both sides, some providers include disclaimers that customers must sign, D2.3 Preliminary best practices in climate uncertainty quantification and communication | 49 confirming that they understand the type of data they are receiving and the risks associated with its use. Such practices underline the shared responsibility between provider and client in managing climate information and the decisions built upon it. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 50 7 Concluding Comments: Emerging Themes This section includes concluding remarks, in the form of key “emerging themes” or lessons learnt. A lot of work has been done in the scientific community to describe sources of uncertainty and quantify uncertainty where this is possible. Also, sophisticated ways have been developed to visualise uncertainties. However, communicating about these uncertainties to users outside of the scientific community remains a huge challenge. Some efforts have been made, for example, by the IPCC but this issue is far from being solved. From this exercise we gather the following ways forward: Start with the most relevant risks An exploration of vulnerabilities can be used to discover which climate risks are of most concern. From there a climate communication practitioner (ideally in collaboration with the clients they are supporting) can discern which aspects of climate information are most relevant to those risks and hence the relevant uncertainty space to describe. To support the work of climate communication practitioners a climate service provider will need to provide sufficient information about its data holdings to enable the relevant uncertainty information to be extracted. Standard approach to uncertainty descriptions The scope of climate science is broad. The implications of its findings need to be understood by many different research communities and also be communicated to wider society. A standard approach to describing and quantifying uncertainty will facilitate the passing of information between different communities of practice. Such an approach should consider not only the climate science component, but also the complexities regarding socio-economic vulnerability, and hence social sciences should be involved. Use language the audience is familiar with (don’t say uncertainty) The vocabulary of science can have very different meanings in public life. Therefore, when dealing with climate services and decision makers, it is usually better to use commonly understood descriptive language in place of scientific terminology, along with multiple contextualised examples illustrating the concepts and approaches. Particular care should be taken with uncertainty communication, for instance, rather than using the word “uncertainty” which implies ignorance, a better choice could be the word “range”, and rather than presenting a likelihood as a percentage instead refer to it using the framing of odds. There are multiple ways to evaluate and communicate uncertainty Decision makers require the best, reliable information to optimally accomplish their job. Although a measure of uncertainty must be always conveyed, there are multiple ways to measure and communicate it. The most used way is to communicate uncertainty via probabilities of occurrence (or not occurrence) of a certain event, but very often decision makers consider the use of odds -and relative oddsmore understandable and actionable than actual probabilities. It is also possible to communicate uncertainty by a combination of the expected value (e.g. the ensemble median) and uncertainty bars (e.g. the ensemble interquartile range). In other contexts, it may be more appropriate not to communicate probabilistic information at all and instead explore the use of plausible storylines to help communicate climate risk and highlight specific vulnerabilities. Use communication about uncertainty to build trust The existence of high uncertainty should be seen as a reason to engage with information, particularly if it relates to a vulnerability. Climate services should be transparent about the uncertainty evaluation of their information. With an awareness of uncertainties comes an ability to assess risks more D2.3 Preliminary best practices in climate uncertainty quantification and communication | 51 critically. More confidence can be placed in assessments of risk that are clear, in spite of the uncertainties. Precision of information should be relevant to the situation General statements of uncertainty can be sufficient. Even if a quantitative analysis is possible it may not be necessary. The precision of a risk description should match what is needed. The presenting of high resolution data will lead audiences to assume a high degree of certainty that may not be justified. Risk descriptions should match what the knowledge supports. Understand existing narratives People use narratives to capture the essential meaning of complex evidence. Until the existing narratives and how people are thinking about risk and decision making are well understood, it is very hard to introduce actionable climate information and related services. Once there is a clear idea of existing narratives, it is key to build trust through mutual learning, embrace the diversity and contradictions of people with competing views and understanding, and work with humility and transparency to co-create new climate risk narratives or storylines of plausible future events. Be aware of deep uncertainties The conventional approach to representing uncertainty in climate services is probabilistic, typically based on ensembles of climate model simulations. In the face of deep uncertainties, the limitations of this approach are becoming increasingly apparent. It is therefore important to extend existing methodologies to include strategies for accounting for and communicating deep uncertainty, including the recording of assumptions made in the analysis process and evaluating the impact of these on the final outcome or decision, and the co-creation of narratives or storylines to improve risk awareness and strengthen decision-making. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 52 8 References Aryal, A., Shrestha, S. & Babel, M.S. (2019): Quantifying the sources of uncertainty in an ensemble of hydrological climate-impact projections. Theor Appl Climatol 135, 193–209, https://doi.org/10.1007/s00704-017-2359-3 Barclay, J., Robertson, R., and Armijos, M. T. (2023): Scientists as storytellers: the explanatory power of stories told about environmental crises, Nat. Hazards Earth Syst. Sci., 23, 3603–3615, https://doi.org/10.5194/nhess-23-3603-2023. Beven KJ, Almeida S, Aspinall WP et al (2018) Epistemic uncertainties and natural hazard risk assessment—part 1: a review of different natural hazard areas. Nat Hazards Earth Syst Sci 18:2741–2768. https://doi.org/10.5194/nhess-18-2741-2018 Budescu, D.V. & T.S. Wallsten (1987): Subjective estimation of precise and vague uncertainties. In G. Wright & P. Ayton (Eds.), Judgmental forecasting, Chapter 4, p. 63–82.John Wiley & Sons Ltd, Chichester, UK. Budescu, D.V., H.-H. Por, S.B. Broomell & M. Smithson (2014): The interpretation of IPCC probabilistic statements around the world. Nature Climate Change, 4(6):508–512, http://dx.doi.org/10.1038/nclimate2194. CCSP 2009: Best Practice Approaches for Characterizing, Communicating, and Incorporating Scientific Uncertainty in Climate Decision Making, United States Climate Change Science Program, National Oceanic and Atmospheric Administration, Washington, DC (2009) https://books.google.co.uk/books?id=aN9YWryjrrgC&ots=n9Lbg6Tje8&lr&pg=PA21#v=onepage&q& f=false Chen, J., F.P. Brissette & R. Leconte (2011): Uncertainty of downscaling method in quantifying the impact of climate change on hydrology. Journal of Hydrology 401(3–4), 190-202, https://doi.org/10.1016/j.jhydrol.2011.02.020. Chen, J., F.P. Brissette, D. Chaumont & M. Braun (2013): Performance and uncertainty evaluation of empirical downscaling methods in quantifying the climate change impacts on hydrology over two North American river basins. Journal of Hydrology 479, 200-214, https://doi.org/10.1016/j.jhydrol.2012.11.062. Cindrić, K, & Z. Pasarić (2019): Modelling Dry Spells by Extreme Value Distribution with Bayesian Inference. In: Vilibić, I., Horvath, K., Palau, J. (eds) Meteorology and Climatology of the Mediterranean and Black Seas. Pageoph Topical Volumes. Birkhäuser, Cham. https://doi.org/10.1007/978-3-030-11958-4_11. Cloke, H.L. & F. Pappenberger (2009): Ensemble flood forecasting: A review. Journal of Hydrology 375, 613-626, https://doi.org/10.1016/j.jhydrol.2009.06.005. Coles, S. (2001): An introduction to statistical modeling of extreme values. Springer-Verlag, London, https://doi.org/10.1007/978-1-4471-3675-0. Contractor, S., M.G. Donat & L. V. Alexander (2021): Changes in Observed Daily Precipitation over Global Land Areas since 1950. J. Climate, 34, 3–19, https://doi.org/10.1175/JCLI-D-19-0965.1. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 53 Corner, A., Lewandowsky, S., Phillips, M. and Roberts, O. (2015) The Uncertainty Handbook. Bristol: University of Bristol. Available from: https://climateoutreach.org/reports/uncertainty-handbook/ Corner, A., Shaw, C., and Clarke, J. (2018): Principles for effective communication and public engagement on climate change: A Handbook for IPCC authors, Climate Outreach, Oxford, UK. Coventry, K.R., J. Harold, I. Lorenzoni, J. Kavonic, I. Soukeyna Diop, E. Visman (2019): Improving methods of communicating climatic uncertainties to aid decision-making: Project report and guidelines prepared for Future Climate for Africa, Future Resilience for African Cities and Lands https://www.fractal.org.za/wp-content/uploads/2019/02/FCFA-Report-Communicating-Climate-Cha nge-to-Decision-Makers.pdf Dankers, R. & Z.W. Kundzewicz (2020): Grappling with uncertainties in physical climate impact projections of water resources. Climatic Change 163, 1379–1397, https://doi.org/10.1007/s10584-020-02858-4 Daron, J.D., S. Lorenz, P. Wolski, R.C. Blamey & Christopher Jack (2015): Interpreting climate data visualisations to inform adaptation decisions. Climate Risk Management 10, 17-26, https://doi.org/10.1016/j.crm.2015.06.007. Dessai, S., & M. Hulme (2003): Does climate adaptation policy need probabilities?, Climate Policy, 4, DOI:10.1080/14693062.2004.9685515 Det Norske Veritas (DNV) (2021): Assurance of simulation models. DNV Recommended Practice DNV-RP-0513, available from https://www.dnv.com/, accessed February 2024. Di Baldassarre, G & A. Montanari (2009): Uncertainty in river discharge observations: a quantitative analysis, Hydrol. Earth Syst. Sci., 13, 913–921, https://doi.org/10.5194/hess-13-913-2009. Doblas-Reyes, F.J., J. García-Serrano, F. Lienert, A.P. Biescas & L.R.L. Rodrigues (2013): Seasonal climate predictability and forecasting: status and prospects. WIREs Clim Change, 4: 245-268. https://doi.org/10.1002/wcc.217. Donat, M. G., et al. (2013): Updated analyses of temperature and precipitation extreme indices since the beginning of the twentieth century: The HadEX2 dataset, J. Geophys. Res. Atmos., 118, 2098–2118, https://doi.org/10.1002/jgrd.50150 Ehmele, F., Kautz, L.-A., Feldmann, H., and Pinto, J. G. (2020): Long-term variance of heavy precipitation across central Europe using a large ensemble of regional climate model simulations, Earth Syst. Dynam., 11, 469–490, https://doi.org/10.5194/esd-11-469-2020. Fischer, E., U. Beyerle & R, Knutti (2013): Robust spatially aggregated projections of climate extremes. Nature Clim Change 3, 1033–1038, https://doi.org/10.1038/nclimate2051. Flage, R., 2019: Implementing an uncertainty-based risk conceptualisation in the context of environmental risk assessment, with emphasis on the bias of uncertain assumptions, Civil Engineering and Environmental Systems, 36:2-4, 149-171, DOI: 10.1080/10286608.2019.1702029 D2.3 Preliminary best practices in climate uncertainty quantification and communication | 54 Fronzek, S., Y. Honda, A. Ito, J.P. Nunes, N. Pirttioja, J. Räisänen, K. Takahashi, E. Terämä, M. Yoshikawa &, T.R. Carter (2022): Estimating impact likelihoods from probabilistic projections of climate and socio-economic change using impact response surfaces. Climate Risk Management 38, 100466, https://doi.org/10.1016/j.crm.2022.100466. Gilleland, E. (2020): Bootstrap Methods for Statistical Inference. Part II: Extreme-Value Analysis. J. Atmos. Oceanic Technol., 37, 2135–2144, https://doi.org/10.1175/JTECH-D-20-0070.1. Grounds, M.A., J.E. LeClerc & S. Joslyn (2018): Expressing Flood Likelihood: Return Period versus Probability. Wea. Climate Soc., 10, 5–17, https://doi.org/10.1175/WCAS-D-16-0107.1. Hamilton, A.S. & R.D. Moore, (2012): Quantifying Uncertainty in Streamflow Records. Canadian Water Resources Journal / Revue Canadienne Des Ressources Hydriques, 37(1), 3–21, https://doi.org/10.4296/cwrj3701865. Hawkins, E., and R. Sutton, (2009): The Potential to Narrow Uncertainty in Regional Climate Predictions. Bull. Amer. Meteor. Soc., 90 (8), 1095–1108, https://doi.org/10.1175/2009BAMS2607.1 Hawkins, E., and R. Sutton, (2011): The potential to narrow uncertainty in projections of regional precipitation change. Clim Dyn 37:407–418. https://doi.org/10.1007/s00382-010-0810-6 Heinrich, D., and M. Bailey, (2020): Forecast-based Financing and early action for droughts: guidance notes for the Red Cross Red Crescent, British Red Cross French Red Cross Red Cross / Red Crescent Climate Centre. Available from https://www.preventionweb.net/publication/forecast-based-financing-and-early-action-droughts-gui dance-notes Höppner, C. , M. Buchecker & M. Bründl (2010): Risk communication and natural hazards. Technical Report D5.1 Version 2.1, CapHaz project, Birmensdorf, Switzerland, available from http://caphaz-net.org/outcomes-results/CapHaz-Net_WP5_Risk-Communication.pdf. Huang, S., Krysanova, V. and Hattermann, F.F. (2014), Does bias correction increase reliability of flood projections under climate change? A case study of large rivers in Germany. Int. J. Climatol., 34: 3780-3800. https://doi.org/10.1002/joc.3945 IPCC, 2021: Annex VII: Glossary [Matthews, J.B.R., V. Möller, R. van Diemen, J.S. Fuglestvedt, V. Masson-Delmotte, C. Méndez, S. Semenov, A. Reisinger (eds.)]. In Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 2215–2256, doi:10.1017/9781009157896.022. Katz, R.W. & M. Ehrendorfer (2006): Bayesian Approach to Decision Making Using Ensemble Weather Forecasts. Wea.Forecasting, 21(2):220–231, http://dx.doi.org/10.1175/waf913.1. Katz, R.W., M.B Parlange & P. Naveau (2002): Statistics of extremes in hydrology. Advances in Water Resources 25, 1287-1304, https://doi.org/10.1016/S0309-1708(02)00056-8. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 55 Kaye, N.R., A. Hartley & D. Hemming (2012): Mapping the climate: guidance on appropriate techniques to map climate variables and their uncertainty. Geoscientific Model Development, 5(1):245–256, http://dx.doi.org/10.5194/gmd-5-245-2012. Kay, A.L., N. Dunstone, G. Kay, V. A. Bell & J. Hannaford (2024).: Demonstrating the use of UNSEEN climate data for hydrological applications: case studies for extreme floods and droughts in England, Nat. Hazards Earth Syst. Sci., 24, 2953–2970, https://doi.org/10.5194/nhess-24-2953-2024. Kelder, T., Müller, M., Slater, L.J. et al. (2020): Using UNSEEN trends to detect decadal changes in 100-year precipitation extremes. npj Clim Atmos Sci 3(47), https://doi.org/10.1038/s41612-020-00149-4. KNMI, 2023: KNMI’23 climate scenarios for the Netherlands, KNMI, De Bilt, KNMI Publication 23-03a. https://cdn.knmi.nl/system/ckeditor/attachment_files/data/000/000/308/original/KNMI23_climate_s cenarios_user_report.pdf Knutti R., D. Masson & A. Gettelman (2013): Climate model genealogy: generation CMIP5 and how we got there. Geophys Res Lett 40:1194–1199. https://doi.org/10.1002/grl.50256. Kundzewicz Z.W., Krysanova V., Benestad R.E., et al (2018a) Uncertainty in climate change impacts on water resources. Environ Sci Pol 79:1–8. https://doi.org/10.1016/J.ENVSCI.2017.10.008 Kundzewicz, Z.W., V. Krysanova, R. Dankers, Y. Hirabayashi, S. Kanae, F.F. Hattermann, S. Huang, P.C.D. Milly, M. Stoffel, P.P.J. Driessen, P. Matczak, P. Quevauviller &H.-J. Schellnhuber (2016): Differences in flood hazard projections in Europe – their causes and consequences for decision making. Hydrological Sciences Journal, 62(1), 1–14, https://doi.org/10.1080/02626667.2016.1241398. K.D. Lee, G.L. Torell & S. Newman (2021): A once-in-one-hundred-year event? A survey assessing deviation between perceived and actual understanding of flood risk terminology. Journal of Environmental Management 277, 111400, https://doi.org/10.1016/j.jenvman.2020.111400. Lourenço, T. C., A. Rovisco, A. Groot, C. Nilsson, H.-M. Füssel, L. van Bree &, R.B. Street (Eds.) (2014): Adapting to an Uncertain Climate: Lessons From Practice. Springer, Cham, https://doi.org/10.1007/978-3-319-04876-5. Mach K. J., M. D. Mastrandrea, P. T. Freeman, C. B. Field, 2017: Unleashing expert judgment in assessment, Global Environmental Change, 44, 1-14 Magagna, B., G. Moncoiffe, M. Stoica, A. Devaraju, J. A. Pamment, S. Schindler and R. Huber, (2021). The I-ADOPT Interoperability Framework: a proposal for FAIRer observable property descriptions. EGU General Assembly 2021, EGU21-13155, https://doi.org/10.5194/egusphere-egu21-13155 Maher, N., Milinski, S., Suarez-Gutierrez, L., Botzet, M., Dobrynin, M., Kornblueh, L., et al. (2019). The Max Planck Institute Grand Ensemble: Enabling the exploration of climate system variability. Journal of Advances in Modeling Earth Systems, 11, 2050–2069. https://doi.org/10.1029/2019MS001639 Maher, N., S. Milinski & R. Ludwig (2021): Large ensemble climate model simulations: introduction, overview, and future prospects for utilising multiple types of large ensemble, Earth Syst. Dynam., 12, 401–418, https://doi.org/10.5194/esd-12-401-2021. D2.3 Preliminary best practices in climate uncertainty quantification and communication | 56