D2.7 Understanding Drivers for Behavioural Change in the Living Labs

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101037293 Deliverable D2.7 Understanding Drivers for Behavioural Change in the Living Labs June 2025 This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101037293 Innovating Climate services through Integrating Scientific and local Knowledge Deliverable Title: Understanding Drivers for Behavioural Change in the Living Labs Author(s): Györgyi Bela, Lotte C.F.E. Muller, Marije Schaafsma Contributing Authors(s): Béla Mihalik, Vera Fabók Date 29 May 2025 Suggested citation: Bela, G., et al. (2025) Understanding Drivers for Behavioural Change in the Living Labs Availability: ☒ PU: This report is public [Please select] ☐ CO: Confidential, only for members of the consortium (including the Commission Services) Document Revisions: Author Revision Date Györgyi Bela, Lotte C.F.E. Muller, Marije Schaafsma, Béla Mihalik, Vera Fabók, Lili Varga First Draft February 2025 Györgyi Bela, Béla Mihalik, Lili Varga Second Draft April 2025 Györgyi Bela (editor) Review by LL leaders April 2025 Györgyi Bela & Marije Schaafsma Final Editing June 2025 Györgyi Bela (editor) Final Review October 2025 D2.7 – Understanding Drivers for Behavioural Change in the Living Labs i Executive Summary The main objective of the task that is reported on in this deliverable (Task 2.4) is to identify and analyse the drivers of behavioural change in the Living Labs of the I-CISK project, focusing on how climate services (CS) encourage informed decision-making for local climate adaptation. This deliverable (D2.7) examines how the different prototype CS developed in the seven I-CISK Living Labs (LLs) instigate behavioural change. Each of these LL, established in the Netherlands, Spain, Italy, Greece, Georgia, Lesotho, and Hungary, has a unique socio-ecological context, creating barriers for use of CS to foster adaptive behaviour. The research integrates multiple methods, namely document analysis, workshops, user-feedback sessions, a choice experiment survey, and a serious game, to assess both the design and the effectiveness of locally tailored CS. The CS prototypes across the I-CISK LLs show most promise when they are co-designed with end-users, blend global forecasts with local knowledge and observations, and present information through intuitive, low-barrier interfaces. The survey among recreationists in the LL Rijnland revealed that users do consult climate information but change behaviour only when forecasts address concrete pain points (e.g., extra costs or detours). Players of the serious game in the LL Budapest encountered barriers in governance fragmentation, mirroring real-life obstacles in the uptake of CS and ability to change actions. Tailored, actionable data— whether irrigation advice, urban-heat maps or low-flow alerts—therefore remain among the main stimuli for adaptation. However, in all LLs, major barriers inhibit CS use and adaptation behavioural change. Limited budgets, siloed governance cultures and a lingering preference for emergency-response routines slow down the uptake of proactive, forecast-based planning. Trust gaps persist where uncertainty is high or observational networks are thin. This suggests, in turn, that transparent communication of model limits, continuous stakeholder engagement and proof-of-concept pilots are essential to build credibility. When supportive governance frameworks and stakeholder platforms (MAPs) are in place, these frictions ease and CS can more readily be used in decision-making. Overall, the results emphasise the importance of user-centred designs in developing CS. This fosters adaptive behaviour and highlights the necessity of understanding and addressing the diverse factors, including adaptation triggers, barriers, enablers, and other drivers, that interact to influence user behaviour. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs ii Table of Contents 1 Introduction ................................................................................................................ ................................. 1 2 Methodology ............................................................................................................................................... 2 3 Summary of the literature review ............................................................................................................... 3 4 Discrete Choice Experiments in the Rijnland Living Lab .............................................................................. 5 4.1 Introduction ......................................................................................................................................... 5 4.2 Setup of the discrete choice experiments ........................................................................................... 5 4.3 Model ................................................................................................................................................... 7 4.4 Sample Descriptors ........................................................................................................ ...................... 9 4.5 Results ............................................................................................................................................... 11 4.6 Conclusions ........................................................................................................................................ 13 5 Reflection on Living Lab Climate Services ................................................................................................. 15 5.1 Rijnland Living Lab, the Netherlands ................................................................................................. 15 5.1.1 Basic information about the Living Lab and its climate services ............................................... 15 5.1.2 Adaptation drivers ..................................................................................................................... 15 5.1.3 Socio-economic/political/governance barriers and enablers ................................................... 16 5.1.4 Influence of other drivers .......................................................................................................... 17 5.1.5 Key findings ................................................................................................................................ 17 5.2 Living Lab in Los Pedroches, Andalucia, Spain ................................................................................... 18 5.2.1 Basic information about the Living Lab and its climate services ............................................... 18 5.2.2 Adaptation drivers ..................................................................................................................... 18 5.2.3 Socio-economic/political/governance barriers and enablers ................................................... 19 5.2.4 Influence of other drivers .......................................................................................................... 20 5.2.5 Key findings ................................................................................................................................ 20 5.3 Living Lab in Emilia-Romana, Italy ..................................................................................................... 22 5.3.1 Basic information about the Living Lab and its climate services ............................................... 22 5.3.2 Adaptation drivers ..................................................................................................................... 22 5.3.3 Socio-economic/political/governance barriers and enablers ................................................... 23 5.3.4 Influence of other drivers .......................................................................................................... 23 5.3.5 Key findings ................................................................................................................................ 23 5.4 Living Lab in Crete, Greece ................................................................................................................ 25 5.4.1 Basic information about the Living Lab and its climate services ............................................... 25 5.4.2 Adaptation drivers ..................................................................................................................... 25 5.4.3 Socio-economic/political/governance barriers and enablers ................................................... 26 5.4.4 Influence of other drivers .......................................................................................................... 26 D2.7 – Understanding Drivers for Behavioural Change in the Living Labs iii 5.4.5 Key findings ................................................................................................................................ 27 5.5 Living Lab in Budapest, Hungary ........................................................................................................ 28 5.5.1 Basic information about the Living Lab and its climate services ............................................... 28 5.5.2 Adaptation drivers ..................................................................................................................... 28 5.5.3 Socio-economic/political/governance barriers and enablers ................................................... 29 5.5.4 Influence of other drivers .......................................................................................................... 30 5.5.5 Key findings ................................................................................................................................ 30 5.6 Living Lab in the Alazani-Iori Basin , Georgia ..................................................................................... 31 5.6.1 Basic information about the Living Lab and its climate services ............................................... 31 5.6.2 Adaptation drivers ..................................................................................................................... 31 5.6.3 Socio-economic/political/governance barriers and enablers ................................................... 32 5.6.4 Influence of other drivers .......................................................................................................... 32 5.6.5 Key findings ................................................................................................................................ 33 5.7 Living Lab in Southern Lesotho and Senqu Valley, Lesotho .............................................................. 34 5.7.1 Basic information about the Living Lab and its climate services ............................................... 34 5.7.2 Adaptation drivers ..................................................................................................................... 34 5.7.3 Socio-economic/political/governance barriers and enablers ................................................... 35 5.7.4 Influence of other drivers .......................................................................................................... 35 5.7.5 Key findings ................................................................................................................................ 35 6 Serious Game in the Budapest LL: “Thermaform your city” ..................................................................... 37 6.1 Synergy with the Climate Service in the Budapest Living Lab ........................................................... 37 6.2 Addressing Adaptation Drivers .......................................................................................................... 37 6.3 Key findings ........................................................................................................................................ 39 7 Cross living labs synthesis .......................................................................................................................... 40 7.1 Basic information on Living Lab prototypes ...................................................................................... 41 7.2 Adaptation drivers: design, accessibility, content, and trustworthiness .......................................... 41 7.3 Socio-economic, political, and governance barriers and enablers .................................................... 41 7.4 Influence of other drivers: socio-economic, ecological, and governance factors ............................. 42 7.5 Using choice experiments and serious games to understand adaptation behaviour ....................... 42 7.6 Overall evaluation and key findings .................................................................................................. 44 8 Summary .................................................................................................................................................... 45 9 References ................................................................................................................................................. 46 Appendices D2.7 – Understanding Drivers for Behavioural Change in the Living Labs iv List of Figures Figure 1 Example choice card of the choice experiment in LL Rijnland. ............................................................. 6 Figure 2 Case study area of the choice experiment in LL Rijnland (Waterrecreatie Nederland, 2024). ............. 8 Figure 3 Years in which respondents of the LL Rijnland survey have experienced summer droughts ............. 10 Figure 4 Stated experience with drought related problems among respondents of the LL Rijnland survey. ... 10 Figure 5 Action cards for ‘Theraform Your City” board game ........................................................................... 38 Figure 6 Elements of the cross LL synthesis ...................................................................................................... 40 D2.7 – Understanding Drivers for Behavioural Change in the Living Labs v List of Tables Table 1 Attribute levels and their respective units of the choice experiment in LL Rijnland. ............................. 7 Table 2 Number of types and sources of CS used by recreationists prior to boating trip(s) in the survey of LL Rijnland. ............................................................................................................................................................... 9 Table 3 Expected changes in the frequency and intensity of droughts in the coming 5 years by respondents in the LL Rijnland survey. ....................................................................................................................................... 11 Table 4 Mixed logit model results of the choice experiment in LL Rijnland. ..................................................... 12 Table 5 Willingness to pay for the attributes in the choice experiment of LL Rijnland. ................................... 12 D2.7 – Understanding Drivers for Behavioural Change in the Living Labs vi Glossary Acronym Definition AIC Akaike Information Criterion API Application Programming Interface C3S Copernicus Climate Change Service CDS Climate Data Store CEMS Copernicus Emergency Management Services CMIP World Climate Research Programme’s Coupled Model Intercomparison Project CORDEX Coordinated Regional Climate Downscaling Experiment CS Climate Services CSIS Climate Services Information Systems D… (pl. D2.7) Deliverable (e.g., D2.7 = Deliverable 2.7) DRR Disaster Risk Reduction GIS Geographic Information System GEO Group on Earth Observations GEOSS Global Earth Observation System of Systems GUI Graphical User Interface I-CISK Innovating Climate Services through Integrating Scientific and local Knowledge IoT Internet of Things IPCC Intergovernmental Panel on Climate Change IWMI International Water Management Institute KNMI Royal Netherlands Meteorological Institute (Koninklijk Nederlands Meteorologisch Instituut) LL Climate Services Living Labs NHMS National Hydro-meteorological Service MOOC Massive Open Online Course OGC Open Geospatial Consortium R R (Statistical Computing Environment) RWS Rijkswaterstaat (Netherlands) S2S Sub-seasonal to Seasonal SEASS SEAS5 (Seasonal Ensemble System 5) is the European Centre for Medium-Range Weather Forecasts (ECMWF) global seasonal-forecast model, providing up-to-seven-month ensemble predictions of key climate variables. SG Serious Game SMHI Swedish Meteorological and Hydrological Institute TRL Technology Readiness Level UNCCD United Nations Convention to Combat Desertification UNDRR United Nations Office for Disaster Risk Reduction UNFCCC United Nations Framework Convention on Climate Change WCRP World Climate Research Programme WFD Water Framework Directive WMO World Meteorological Organization WTP Willingness to Pay WP Work Package D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 1 1 Introduction This deliverable, which has been developed as a part of Task 2.4, aims to explore and analyse the drivers of behavioural change in the Living Labs that have been established in the project, with respect to the adoption and use of climate services (CS). This task is part of Work Package (WP) 2 of the I-CISK project, and builds on the outputs of Tasks 2.1–T2.3. Task 2.4 assessed end-user responses to the potential/prototype CS applications and tools developed in the I-CISK project (primarily in WP3, WP5 and WP6, each within the context of the Living Labs (LLs) in WP1). Specifically, Task 2.4 evaluated the potential effect of the CS ex-ante by (a) assessing how these services instigated adaptive behaviour – such as changes in skill levels, practices, knowledge, and resource allocation in livelihood and economic activities – compared to the baseline established in Tasks 2.1– 2.3, and (b) determining whether the CS comprehensively addressed the existing and emerging needs of all end-users to help build climate resilience. A key part of this project involved investigating how the characteristics of CS (e.g. design, accessibility, content, accuracy, frequency, language, media) affect decision-making across different end-user groups. The study revealed socio-economic, political and governance barriers that hinders the use of new CS and the adaptive behaviours they aim to support. Furthermore, the study investigated the impact of additional socio-economic and ecological factors in order to clarify the intricate set of elements that influence decision-making and behavioural change. To achieve this, Task 2.4 employed a range of research methods to investigate the drivers and barriers to the uptake and use of CS. The outcomes of Task 2.4 are linked to WP4, which examines the broader consequences of behavioural change for other stakeholders and sectors, and WP5, which aims to improve the design of climate service information systems (CSIS) in the Living Labs. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 8 choice experiment in software R (version 4.4.2; R Core Team 2021) using package ‘Apollo’, (version 0.3.4; Hess & Palma, 2019). To estimate the mixed logit model, we used 5000 Sobol draws. We additionally wanted to further our understanding of the current information landscape that the recreationists experience and embedded the DCE in a larger survey. We first asked about the recreationists’ boat characteristics, recreation experience(s), current CS use, perceptions of droughts and their experienced drought-related recreation impediments. Next, we presented the recreationists with an example choice card, followed by 6 choice cards. We ended the survey with questions about whether they took all the choice card information into account during their decision-making, and demographic information. Note: the different coloured lines in the map on the right side indicate alternative routes. Figure 2 Case study area of the choice experiment in LL Rijnland (Waterrecreatie Nederland, 2024). We made use of various methods to disseminate the survey and reach as many recreationists as possible. We conducted face-to-face dissemination at ship locks while boaters were waiting to pass through the lock, as well as at an in-water boat show (HISWA te Water). We collaborated with the Rijnland Water Authority, Watersportverbond Noord Holland Zuid, and Waterrecreatie Nederland to distribute the survey. Additionally, we reached out to individual harbours along the standing-mast route (see Figure 2), leveraging their local communication channels to maximise dissemination. This means we also include recreationists from outside of the LL Rijnland specifically, but we include them as they are members of harbours along the standing-mast route, which is affected during summer droughts. An advantage of our wide sample is that we gain insights into the varying impacts of drought; drought impacts throughout the region vary widely, and are often dependent on local infrastructure (Philip et al. 2020; Watersportverbond, 2019). Respondents could fill in the survey on their device of choice as the survey supported a computer, mobile phone or tablet format. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 9 We collected data over the period between July 14th until December 31st, 2024. According to the Dutch national weather service, the summer in this period was characterised as matching the record-breaking warm and wet summer of 2023 (Koninklijk Nederlands Meteorologisch Instituut, 2025). 4.4 Sample Descriptors A total of 308 recreationists completed the survey, and after excluding non-boat owners and those who do not currently have their boat at a harbour, we analysed the choices made by 258 respondents. Of the respondents in our sample, 33% owned a sailboat and 67% a motorboat. The majority were male (81%), with females comprising 17%, and the median age fell between 60 and 69 years. Notably, our sample gender and age distribution differs from findings by the Netherlands Bureau voor Toerisme & Congressen and Waterrecreatie Nederland (2022), which reported an even gender split (51% male) and a predominant age range of 35 to 54 across various boating groups (cabin motorized, cabin sail, sloop, open, and open sailboats). However, these figures are not directly comparable, as their data was collected nationally in 2021 at an individual level, whereas our 2024 data focuses on households within the Rijnland LL region. The older demographic represented in our sample reflects Rijnland LL MAP workshop discussions, in which harbour management expressed concerns about an aging recreational boating population. We also collected information about respondents’ current recreation, and their CS use. In our sample, yearly harbour costs range from €0 to €10,000, averaging €1458 (std. dev. €1,221), reflecting variations in boat size and harbour pricing. The distance from residence to harbour varies from 0 to 220 km, with a mean of 25 km (std. dev. 37 km). Respondents used information prior to their boating trip(s): wind (89%), rainfall (76%), temperature (69%), tides (30%), sun strength UV (14%), wave (14%), cloud cover (11%), day length (7%) information. They also used ship lock (42%), bridge (40%), route (27%), swimming water quality (6%), and other (5%) information. Only 2% of the recreationists stated that they use no information. Recreationists use on average four of the climate indicators we listed in our survey before their boating trips (std. dev. = 2) (see Table 2). Table 2 Number of types and sources of CS used by recreationists prior to boating trip(s) in the survey of LL Rijnland. Number of CS Percentage of respondents types or Sources CS Types CS Sources 0 2% 1% 1 3% 22% 2 13% 38% 3 21% 26% 4 18% 9% 5 19% 4% 6 13% - 7 9% - 8 1% - 9 2% - 0 2% 1% 1 3% 22% Note: table is based on the total sample size of 258 respondents. Mostly recreationists access this information via mobile applications (79%), and websites (71%). Own experiences or instincts (40%) are also cited frequently by recreationists, indicating the use and prevalence of local knowledge. Beyond this, respondents make use of (marine VHF) radio (10%), printed media (10%), TV (8%), social media (6%), social network (5%), other (3%) sources. When selecting other information sources, respondents mostly state using specialised onboard radar or navigation systems. Only 1.16% of respondents D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 10 stated that they do not source any information. Recreationists consult on average two (St. Dev = 1) of the listed sources for their climate information (see Table 2). Despite their stated use of climate information, only 29% of respondents identified the 2022 drought, and 7% the 2018 drought, shown in Figure 3. More respondents identified having experienced summer droughtrelated obstacles, with problematic pondweed (50%) (see Figure 4). Note: figure is based on the total sample size of 258 respondents. Figure 3 Years in which respondents of the LL Rijnland survey have experienced summer droughts Note: figure is based on the total sample size of 258 respondents. Figure 4 Stated experience with drought related problems among respondents of the LL Rijnland survey. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 11 We asked individuals who had experienced drought-related recreation problems to rate the severity of these issues using a scale from 0 (no hindrance) to 4 (very much hindrance). The presence of pondweed (mean = 2.77, std. dev. = 1.04, n = 128) and blue algae (mean = 2.13, std. dev. = 1.04, n = 116), as well as the lack of ship lock opening times (mean = 2.14, std. dev. = 1.19, n = 64), were identified as the most severe hindrances. In contrast, issues related to inaccessible routes—due to low water levels (mean = 1.64, std. dev. = 0.94, n = 44) or high water levels (mean = 1.67, std. dev. = 0.78, n = 12)—and limited (mean = 1.74, std. dev. = 1.00, n = 92) or changed (mean = 1.60, std. dev. = 0.90, n = 40) ship lock opening times were rated as less hindering. Despite being experienced by many recreationists, the expansion of bridges due to high temperatures (mean = 1.49, std. dev. = 0.91, n = 113) was perceived as the least disruptive to their recreation. In addition to their experiences with the frequency and intensity of drought impacts, we also asked respondents about their expectations for droughts over the next 5 years. Most respondents anticipate an increase in both the frequency (mean = 0.45, std. dev. = 0.69) and intensity (mean = 0.52, std. dev. = 0.64) of droughts in the coming 5 years (Table 3). However, a significant number of respondents expect no change in either the frequency or severity of droughts in the coming 5 years Table 3 Expected changes in the frequency and intensity of droughts in the coming 5 years by respondents in the LL Rijnland survey. Expected change of droughts (scale) Frequency Intensity Strongly decrease (=-2) 1% 0% Decrease (=-1) 7% 4% No change (=0) 40% 42% Increase (=1) 51% 51% Strongly increase (=2) 1% 3% Note: Table is based on the total sample size of 258 respondents. 4.5 Results We find that climate information is used alongside cost and distance information when deciding to remain at their current harbour or move to an alternative harbour. As expected, recreationists prefer harbours with lower costs, that are closer to their homes and are less vulnerable to summer droughts. They favour shorter waiting times at ship locks, fewer route detours, and minimal pondweed on the boating route. Table 4 shows the results of the mixed logit model. All the mean estimates of the attribute coefficients, except the ASC, are negative and statistically significant. The positive and significant mean estimate of ASC reflects that recreationists have a preference to remain at their current harbour. We also find that recreationists do not use the information presented on the choice cards in a homogeneous way for decision-making. In Table 4, the standard deviation estimates for the attributes are also all significant, indicating that, for example, not all recreationists respond to changes in drought level in the same way. To compare which information recreationists are most responsive to, we evaluate the willingness to pay (WTP) estimates, shown in Table 5. Recreationists experience, on average, a 399 euros welfare gain to remain at their current harbour, as is reflected by the WTP for the ASC component in Table 5. The 95% confidence interval of the WTP estimate also excludes zero, which means that most respondents would gain welfare by remaining at their current harbour. This could reflect the costs of relocating their boat to another harbour, the social connections they have established at their current harbour, or a preference for maintaining the status quo. The remaining attributes have a negative WTP, suggesting that recreationists would experience an inconvenience, or loss of welfare. To move to a harbour further away from their place of residence, a recreationist would experience a welfare loss of 6 euros per kilometre. Looking at the drought chance D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 12 coefficient, for each extra summer drought expected in the coming 5 years, recreationists would experience a welfare loss of 65 euros for this inconvenience. For each additional hour of waiting time at ship locks during droughts, recreationists would experience a welfare loss of 51 euros. For each additional kilometre of detours along the route, a welfare loss of 66 euros would be felt. Lastly, for each percentage point of pondweed along the boating route, recreationists would experience 6.50 euros of welfare loss. All confidence intervals of the estimates do not include zero, implying that most respondents experience inconvenience due to summer droughts and their impacts. Table 4 Mixed logit model results of the choice experiment in LL Rijnland. Attribute Mean estimate Std. dev. estimate Alternative specific constant (AS C ) 2.134*** (0.240) 2.411*** (0.249) Yearly harbour fees (Cost) -1.425*** (0.131) 1.249*** (0.160) Distance from harbour to place of residence (Distance) -0.321*** (0.091) 0.217** (0.072) Chance of summer drought (Drought) -0.351*** (0.064) 0.356** (0.111) During drought: extra ship lock wait time (Wait) -0.274*** (0.038) 0.194* (0.079) During drought: length of detour (Detour) -0.304*** (0.057) 0.397*** (0.085) During drought: Pondweed presence (Plant) -0.344*** (0.056) 0.490*** (0.084) Attribute Mean estimate Std. dev. estimate No. individuals 258 No. observations 1548 AIC 2133.7 Log-likelihood -1052.85 Notes: Significance is indicated with asterisks, where *** = p<0.001, **= p<0.01, *=p<0.05. Robust standard errors are shown in brackets. Attribute Cost is lognormally distributed, while all other attributes are normally distributed. Table 5 Willingness to pay for the attributes in the choice experiment of LL Rijnland. WTP for Median (95% Confidence interval) Alternative specific constant (ASC) 399.35 (255.88, 583.50) Distance from harbour to place of residence (Distance) -5.97 (-10.09, -2.62) Chance of summer drought (Drought) -65.29 (-101.43, -38.63) During drought: extra shiplock wait time (Wait) -51.37 (-77.25, -32.73) During drought: length of detour (Detour) -5.64 (-8.81, -3.32) During drought: Pondweed presence (Plant) -6.45 (-10.22, -3.77) Notes: Willingness to pay estimates are calculated based on the mixed logit model estimates, and following Krinsky & Robb (1986, 1990). Using these WTP estimates, we can determine at which point recreationists would choose to relocate to a different harbour. Assuming the average distance of 25 km from their residence to the (alternative) harbour and considering representative conditions at a harbour in LL Rijnland over the past five years (one summer drought, a four-hour ship lock wait time, a 15 km detour, and 30% pondweed coverage along the recreation route), we find that removing ship lock wait times and detours alone would not be enough to motivate recreationists to switch harbours. In this scenario, for recreationists to consider moving, the alternative harbour would have to have less than 14% pondweed coverage along the route. This suggests that while D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 13 pondweed presence is a key factor in decision-making, recreationists also have a strong preference for staying at their current harbour. If the alternative harbour were 40% farther from their home (35 km in this case) but had the same drought risk, recreationists would only be willing to switch if there were no detours, no shiplock wait times, and less than 3% pondweed coverage along the route. If the alternative harbour were 80% farther away (45 km), even with no wait times, no detours, and no pondweed presence, recreationists would still prefer to remain at their current harbour. This highlights the reluctance of recreationists to travel farther to their harbour, even under improved drought conditions. 4.6 Conclusions We find that recreationists currently make use of various types and sources of CS. Many recreationists are aware of or have experienced summer droughts, and in the coming five years expect the frequency and severity of them to increase. By collaborating closely with MAP members in designing our DCE, we were able to explore both their use of summer drought information and their behavioural responses to it. Our results show that while recreationists are responsive to tailored drought impact information, they have a strong preference to remain at their current harbours. Although they acknowledge and utilize drought information, behavioural changes—such as switching harbours—are only likely if conditions deteriorate significantly. This finding is reassuring for harbour management in Rijnland, as it suggests that climate change alone is unlikely to drive large-scale shifts in harbour membership. However, it also highlights a key challenge in measuring the impact of CS through observed behaviour. Since CS use may not immediately translate into visible behavioural changes, methods like DCEs provide valuable insights into hypothetical but realistic decision-making scenarios. As a limitation of this study, we have a relatively small sample size; however, we accounted for this by using a mixed logit model instead of more data-intensive models to evaluate preference heterogeneity. The very wet summer of 2024 could also have influenced the perceived realism of our DCE, and hence our study results. However, in working closely with LL Rijnland MAP members, as well as pretesting the DCE and survey with recreationists outside of I-CISK, we focussed on designing a realistic DCE. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 14 I-CISK methodology influencing further research The I-CISK design and use of DCEs to explore behavioural responses to CS has inspired further study already. In 2024, I-CISK partner IVM conducted a study among cocoa farmers’ preferences for solar powered irrigation pumps in Ghana in collaboration with the International Water Management Institute (IWMI) and Cornell University. In Ghana, CS could provide the information needed to encourage climate change adaptation strategies, as their adoption is generally low (Gbodji et al., 2023). In this setting, we evaluated whether CS information about droughts (rainfall and consecutive dry days) could support the adoption of solar irrigation. This is ongoing research, but a summary of this study to date is below: Cocoa farmers in Ghana have not widely adopted solar-powered irrigation, despite the recognized potential to mitigate the impacts of climate change on cocoa production. The study assesses farmers' preferences for climate and loan financing information for solar irrigation adoption. We conducted a discrete choice experiment with 631 cocoa farmers across 8 regions of Ghana. In four treatment groups, we varied the availability of consecutive dry days and previous season reference information. In all treatment groups, rainfall information is provided alongside forecast variation information and the financial conditions for solar irrigation adoption. Using a mixed logit model, we find that farmers responded strongly to loan payback period, and the fraction of their farm that would be irrigated. This aligns with previous findings that few cocoa farmers in Ghana have access to credit and hence struggle to afford climate change adaptation. Farmers showed limited sensitivity to climate and forecast variation information. We find that they would not be willing to pay for CS providing rainfall or rainfall variance information. We do not find a significant difference in cocoa farmer preferences for solar irrigation when comparing the treatment groups with and without additional dry days information. Although when consecutive dry days information was provided, we find evidence of it being used during decision making. Using a latent class model, we further evaluate heterogeneity amongst cocoa farmer preferences, where we find that perceptions of climate change, education and stated attribute non-attendance have an effect. Farmers with above-average education who recognize the reality of climate change and believe the weather can be predicted are more likely to consider irrigation loan costs in conjunction with factors such as the proportion of land to be irrigated and the loan repayment period. These farmers are also less inclined to favour the status quo situation without adopting a solar irrigation system. This study underscores the need to integrate financial support with CS to enhance climate adaptation in Ghana’s cocoa sector. Additionally, while uncertainty is a major concern in climate forecasting and prediction for CS, in the context of adopting solar power, this does not seem to be relevant information to cocoa farmers in Ghana. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 15 5 Reflection on Living Lab Climate Services In this section we report on the findings of the document analysis and further information provided by the living lab teams. For each of the seven living labs we first provide basic information about the living lab and the climate services that have been established within the project. We explore the drivers of adaption, and the socio-economic, political, governance barriers and enablers for these drivers. A reflection is provided for other drivers, followed by key findings. These findings are derived through a detailed document analysis, and verified with the respective living lab teams. Where appropriate, quotes from these documents are provided in italics, with clear indication of the deliverables and documents these are sourced from. 5.1 Rijnland Living Lab, the Netherlands 5.1.1 Basic information about the Living Lab and its climate services The LL Rijnland CS is specifically designed to encourage proactive adaptive behaviour among water managers, agricultural stakeholders, and other users by providing timely, actionable drought alerts. As described, "The ICISK CS ... is a drought alert service" (D3.3, Section 4.4.1) that incorporates key components, including "Subseasonal to seasonal (S2S) lead time drought alerts," "S2S forecasts of cumulative precipitation deficit," and "Climate change projections and expected impact on drought hazards" (D3.3, Section 4.4.1). This combination of comprehensive forecasting and alert mechanisms offers users critical lead time, enabling them to "prepare and implement operational mitigation measures in case of an eminent drought" (D3.3, Section 4.4.3). The underlying technical architecture of the CS is built on a cloud-based infrastructure using Docker containers, Python, Flask API, and React, and ensures reliable data integration from sources such as SMHI and Rijkswaterstaat (D5.2, Sections 4.3.1 and 4.3.2). This robust system guarantees that forecast data is updated monthly ( D3.4, Sections 4.1–4.2). Importantly, the CS design supports user engagement by featuring intuitive visualizations such as graphs showing "historic gauge levels and predicted future water levels at the gauge station Lobith" (D5.2, Section 4.3.2). Regular one-to-one sessions and mock-up discussions with stakeholders have informed the interface design and functionality, ensuring that the service reflects user needs and preferences (D3.3, Section 4.4.3; D5.2, Section 4.3.3). By increasing awareness of both shortand long-term drought risks and fostering active discussion on adaptation measures, the LL Rijnland CS aims to empower stakeholders to make informed decisions. Its ability to provide operationally relevant forecasts and easily interpretable visualizations plays a pivotal role in shaping adaptive behaviour in the face of climate-induced water challenges. 5.1.2 Adaptation drivers The LL Rijnland CS is designed to facilitate adaptive behaviour by providing clear, actionable drought alerts and water management information to stakeholders. Its design emphasizes simplicity and accessibility; for example, "The time series visualisation of the gauge station at Lobith received the highest priority and was developed first" (D5.2, Section 4.3.2), ensuring that the most critical data is immediately visible. The interface also offers "additional front-end designs... following a colour-based indicator scheme (green to red) per week" specifically created for the general public who may not seek detailed technical data (D5.2, Section 4.3.2). Furthermore, the platform’s user-centric approach is highlighted by features that require "no programming skills" and allow users to easily switch between views of historical and forecasted data, thereby empowering all users to interact with complex climate information through simple actions like clicking and hovering (D5.2, Section 4.3.2). D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 16 In terms of content, the service delivers highly relevant and timely information. The gauge at Lobith is described as providing "very important information of the water management in Rijnland and beyond," with "discharge values at this strategic location [being] critical for national water management" (D5.2, Section 4.3.2; to be reported in D3.4, Section 4.1). Alerts are tied to specific indicators, such as "drought thresholds established by Rijkswaterstaat," which guide operational decisions during critical low-flow periods, ensuring that users have sufficient lead time—"[the alert service aims] to provide users with more time to decide on, prepare, and implement operational mitigation measures" (D3.3, Section 4.4.3). This timeliness is reinforced by forecast lead times ranging from one week to seven months. Data accuracy is maintained through rigorous processing methods; the service integrates local data in its "downscaling and bias correction performed by SMHI," ensuring improved forecast reliability (D5.2, Section 4.3.1). Although model evaluations indicate challenges—such as underprediction at shorter lead times and overprediction at longer ones, evidenced by findings like "the Mean Absolute Error (MAE)... shows an overall increase with lead time" ( D3.4, Section 4.1)—the use of comprehensive skill assessments, including metrics like Probability of Detection and Probability of False Detection, demonstrates the CS ability to reliably identify drought events ( D3.4, Section 4.2). By combining an accessible, user-friendly design with clear, operationally relevant content and robust data processing, the LL Rijnland CS effectively supports adaptive decision-making. Its integrated approach— leveraging simplified visualisations, dynamic data interaction, and precise alerts—ensures that stakeholders are well-equipped to implement timely mitigation measures and engage in strategic long-term planning, ultimately fostering a more resilient water management system. 5.1.3 Socio-economic/political/governance barriers and enablers The LL Rijnland CS is well-positioned to influence adaptive behaviour by leveraging strong stakeholder engagement and robust data integration. Its development has been enriched by "extensive Mock-Up discussion with the LL hosts and their stakeholders" and "one-to-one sessions with Multi actor Platform (MAP) members [collected] feedback" (D5.2, Section 4.3.2; D3.3, Section 4.4.3), which ensure that the service is closely tailored to real user needs. The use of open data from Rijkswaterstaat (the national water management agency) and the Royal Netherlands Meteorological Institute (KNMI) and the inclusion of local data in downscaling and bias correction further enhance the accuracy and relevance of the forecasts (D5.2, Section 4.3.1), while integrating local knowledge into dynamic drought alert thresholds and alert messages helps translate scientific data into actionable insights. Socio-economically, the service addresses the needs of key sectors, as "the main stakeholders considered in this LL are the water tourism and agricultural sectors" (D3.3, Section 4.4). Features like simplified visualizations designed for the general public and traffic light-based symbols for waterways ensure that the information is accessible and engaging for users with diverse interests, from tourism impacts like "closing of locks at certain locations" to agricultural concerns such as irrigation water salinity (D5.2, Section 4.3.2; D5.2, Section 4.3.3). From a governance perspective, the involvement of institutional actors such as the Rijnland Water Authority underscores the CS integration into the water management framework. The CS provides operational drought forecasts at S2S lead times and customized alerts tailored to sector-specific needs, addressing a notable gap in existing public monitoring services (D3.3, Section 4.4.4). By combining detailed user feedback, advanced data processing, and clear governance integration, the LL Rijnland CS is set to empower stakeholders with timely and actionable information, ultimately driving proactive adaptation measures in water management. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 17 5.1.4 Influence of other drivers The ecological and hydrological context of the LL Rijnland plays a crucial role in shaping adaptive behaviour. The service capitalizes on the fact that "the LL region is mostly flat and below sea level" (D3.3, Section 4.4), which, combined with its reliance on "the Rhine [as] the major riverine water source in The Netherlands" (D5.2, Section 4.3.2), underscores the region's sensitivity to water scarcity and quality (salinity) issues. Faced with "increasing challenges posed by climate change, such as extended drought periods and rising temperatures" (Reported in D3.4, Section 4.1), the service monitors key indicators like the "cumulative precipitation deficit starting from 1 April, up to 31 October" (D3.3, Section 4.4.1) to define the drought season. Moreover, observing that "too low a discharge at Lobith is an indicator for possible salt intrusion from the North Sea" (D3.3, Section 4.4.1) alerts users to potential mitigation measures Rijnland water authority will take, which in turn adversely affects their own intended use of the water. By integrating these targeted hydrometeorological and water use insights into its forecasting and alert systems, the CS empowers stakeholders to make informed, timely adaptation decisions that enhance regional resilience against droughts in a changing climate. 5.1.5 Key findings The Rijnland CS prototype is positioned to potentially influence adaptive behaviour by providing users with extended lead times that allow for both immediate operational responses and long-term strategic planning. As stated, "The alert service aims to provide users with more time to decide on, prepare, and implement operational mitigation measures..." (D3.3, Section 4.4.3), while also aiming "[to] foster discussion on, developing, and deciding on strategic long-term adaptation measures" (D3.3, Section 4.4.3). This dual focus ensures that both immediate actions and future adaptation strategies are supported. The CS benefits include addressing specific stakeholder needs with longer forecast lead times (over 14 days), integrating diverse data such as alerts, forecasts, and climate change information, and utilizing a user-friendly and accessible interface that incorporates local knowledge. Notably, features like the visualization of Lobith discharge data—an essential national indicator—demonstrate the system's operational relevance (D3.4, D5.2). However, challenges remain, such as limitations in forecast accuracy and the need for further enhancements in bias correction techniques for inputs like ECMWF’s SEAS5 seasonal ensemble, as well as ensuring effective mobile phone styling and simplified views for broader accessibility. The LL Rijnland CS, through agile development and ongoing stakeholder feedback, is well-equipped to empower water managers, farmers, and other decision-makers with the reliable, timely information needed to build more resilient adaptation strategies. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 24 Climate Service platform provides a comprehensive and intuitive way to monitor and forecast river flow, setting the ground for a climate service that supports effective water resource management in the Emilia Romagna region" (D5.2, section 4.2.2). This user-friendly design allows stakeholders to "visualise real-time data, forecasts, and statistical analyses," which aids in navigating the complexities of water resource allocation. However, real-world challenges remain. One key lesson learned is that despite existing policy approaches, the broader application of forecasts on resource availability is still limited by a dependence on real-time monitored values and historical patterns, a practice recently challenged by unprecedented drought conditions in 2022 (D2.6, section 6.2). For many irrigation consortia, the forecasts are already proving instrumental in both "the short to mid-term management of water distribution and the long-term strategic planning," highlighting their potential to enhance resilience and improve adaptive strategies (D2.6, section 6.3). Looking forward, there are clear suggestions for further development to improve the CS impact. Future enhancements are planned to refine the interface based on ongoing one-to-one stakeholder interviews, ensuring that the CS better meets regional needs and effectively triggers decision chains (D5.2, section 4.2.1). Additionally, more accurate short to medium-term discharge forecasts could enable a more nuanced approach to water withdrawal regulations, bridging the gap between adaptation and coping strategies (D2.6, section 6.3). This iterative, stakeholder-driven improvement process will be crucial for transforming forecast information into sustained, proactive adaptation measures. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 25 5.4 Living Lab in Crete, Greece 5.4.1 Basic information about the Living Lab and its climate services The Living Lab in Crete has developed a comprehensive CS prototype that supports a multi-sectoral approach towards the tourism product, by addressing the needs of multiple users and sectors (tourism; water allocation and reservoir management; transportation infrastructure) at the scales which serve both operational (seasonal forecasting) as well as long-term planning needs (decadal projections). The CS integrates a broad range of climate indicators, including "Essential Climatic Variables, Hydrological Variables and indicators, Climatic Tourism Indicators, and Infrastructure Indicators" (D5.2, Section 4.4), operationally providing 14 key variables and indexes, ensuring updated and relevant data. Designed to support decision-making for the tourism sector as well as in the spectrum of the cross-cutting interactions of various sectors around the tourism sector, the CS delivers seasonal forecasts and end-ofcentury predictions based on climate essential variables (e.g., air temperature, precipitation, humidity) sourced from ECMWF through the Copernicus Climate Change Service and hydrological forecasting from SMHI. It also incorporates local datasets, such as meteorological historical data, geological maps and digital elevation models, to tailor the information to the specific needs of the region. The user interface, organized into multiple tabs, makes complex information accessible even to those with limited technical skills. Extensive stakeholder engagement, through interviews, e-meetings, and full MAP workshops, has ensured that the CS addresses local climate information gaps and adaptation needs. Overall, by combining state-ofthe-art forecasts, tailored local data, and a user-friendly design, the LL Crete CS effectively empowers stakeholders in the tourism sector as well as cross-cutting sectors (hospitality, water management, transportation) to make proactive, informed adaptation decisions in the face of climatic challenges. 5.4.2 Adaptation drivers The climate services that have been developed in the Crete Living Lab can significantly influence user adaptation if it addresses the core needs of stakeholders while overcoming barriers of trust, uncertainty, and variable capacity. The presence of “MAP members generally characterised by high educational background… and good ability to understand the goals of the I-CISK project” (D2.6, Section 8) suggests that user-friendliness and clarity could resonate well with this audience, especially given their “increased awareness on available tools (early warning systems, case-specific and user-friendly) and uncertainty related information” (D2.6, Section 8, Table 3). Accessibility challenges persist, as “Initial stakeholder engagement showed that different members… had various levels of awareness about existing CS… readiness to specifically express their CS needs” (D2.6, Section 8). Moreover, barriers such as “lack of trust on seasonal forecasting and climate information” and “financial limitations / in house technical skills” (D2.6, Section 8, Table 3) can undermine uptake. Yet the content itself has clear potential to foster adaptive decisions if delivered reliably and at the right times. From “a forecast in September about the coming wet period” (D2.6, Section 8, Table 1) to “the need for an operational seasonal forecast of sea-surges… to support better planning” (D2.6, Section 8, Additional measures…), stakeholders repeatedly emphasize the value of timely, relevant projections. With “all seasonal indicators… updated on a monthly basis with a forecasting horizon of seven months” (D5.2, Section 4.4), the service targets this demand for frequent updates, but uncertainties—“…large uncertainties” (D2.6, Section 8, Table 3)—can still deter full reliance. Ultimately, building trust and clarifying forecast limitations will be vital for the CS to truly shape user behaviour, ensuring that false predictions do not lead to “added costs and… re-planning” (D2.6, Section 8, Additional measures...) and instead encourage proactive adaptation. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 26 5.4.3 Socio-economic/political/governance barriers and enablers The prototype CS has the potential to positively shape user adaptation behaviour, despite significant socioeconomic, political, and governance barriers. Key socio-economic challenges include "financial limitations" and "lack of in-house technical skills" (D2.6, Section 8), both of which could limit the ability of stakeholders to fully benefit from available climate information. Additionally, the presence of competing priorities was highlighted, noting that "if costs are too high and the need for extra water is not that high, then maybe other projects and works should be prioritized" (D2.6, Section 8). Politically, there is potential conflict arising from municipalities competing for scarce water resources, putting additional political pressure on water management bodies. This contributes to a broader "lack of willingness to act" (D2.6, Section 8), hindering proactive adaptation efforts. Governance challenges are evident, particularly the "low experience of Civil Protection on flood related risks" and an overall "lack of maturity on climate adaptation" within institutions (D2.6, Section 8). Despite these barriers, several enablers create opportunities for effective use of the CS. Notably, high-capacity actors such as the water management organization (OAK) and tourist resorts have been identified, accompanied by increased stakeholder awareness of available climate tools and the pressing nature of climatic threats. Additionally, policy support, such as the "Development of Flood Risk Management Plans," has been highlighted as an enabling factor, suggesting that when stakeholders possess sufficient resources, capacity, and awareness, the CS can effectively catalyse proactive and informed adaptation behaviour. 5.4.4 Influence of other drivers The LL Crete CS has significant potential to positively influence adaptive behaviour, particularly by aligning closely with socio-economic, ecological, and governance drivers. Given that the LL Crete explicitly "focuses on the tourism sector," recognizing that "climate is a critical factor for tourism" and directly impacts water availability and economic viability, the CS delivers targeted climate information essential for effective planning in tourism-dependent economies (D2.6, Section 8; D5.2, Section 4.4). Ecological vulnerabilities amplify the importance of the CS, as Crete faces increasing threats from droughts, floods, heatwaves, and other extreme events. By integrating indicators like the “outflow from subcatchment” and “Wet period total discharge”, the CS provides stakeholders with actionable information to support ecological risk assessments to inform resilient planning. Indexes like the "Instability Index," which assesses landslide risks based on various environmental factors, and “wind frequency” for specific thresholds support infrastructure (road network and ports) protection (D5.2, Section 4.4). The CS therefore plays a crucial role in enhancing resilience by delivering targeted forecasts and warnings, supporting more robust operational planning for transportation networks and outdoor tourism (Section 8.3). Governance and institutional contexts further enhance CS impact. High-capacity organizations such as OAK, which already "incorporates adaptation actions into development planning," and structured initiatives like the "Development of Flood Risk Management Plans," provide supportive policy frameworks that facilitate proactive adaptation (D2.6, Section 8). Additionally, by leveraging authoritative data sources from Copernicus and SMHI, the CS builds credibility and trust among stakeholders. Overall, by effectively addressing and integrating socio-economic needs, ecological threats, and supportive governance frameworks, the LL Crete CS is strategically positioned to encourage meaningful and sustained adaptive actions among users. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 27 5.4.5 Key findings The Living Lab Crete has shown that tailored climate indicators, considering sector interlinkages, supporting a multi-sectoral approach as well as timely operational delivery significantly enhance the relevance and usability of the CS for local stakeholders. However, varying levels of stakeholder awareness, understanding, and readiness pose critical challenges, indicating that targeted, user-specific engagement approaches are essential. A fundamental barrier identified is the widespread lack of trust in seasonal forecasting and climate information, emphasizing the need for increased transparency and clearer communication around uncertainties. Furthermore, combining multiple data sources, such as global forecasts, hydrological models, and local datasets, has proven essential in creating actionable, context-specific CS. Operational indicators like the Tourism Climatic Index, temperature related indexes, Wet Period Flow Volume, Instability Index and Wind frequency for specific thresholds and directions have demonstrated strong potential for supporting decisionmaking in water management, tourism planning, and infrastructure maintenance. Overall, building stakeholder trust, addressing administrative and technical capacity barriers, and continuously improving the timeliness and reliability of provided forecasts are crucial steps to ensure the CS effectively influences adaptive behaviour in the region. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 28 5.5 Living Lab in Budapest, Hungary 5.5.1 Basic information about the Living Lab and its climate services The Living Lab in Budapest has developed a comprehensive and integrated urban heat visualization service. This innovative CS effectively integrates static drone-based thermal maps, a web-based interactive GIS platform, and real-time sensor dashboards. It provides high-resolution, georeferenced urban heat exposure maps alongside spatial data layers, such as Land Surface Temperature (LST) and OpenStreetMap (OSM). The platform is specifically designed to support data-driven decision-making, enhance public awareness, and enable adaptive planning through comparative analysis of before-and-after scenarios. One component of this LL is the "Convolutional Neural Network (CNN) component, designed to accurately predict surface emission temperatures during intense heatwaves" (Deliverable 5.2 section 4.5). The platform visually integrates these predictive insights with detailed "orthophotos, thermal imagery (observed and predicted), and an OpenStreetMap (OSM) integration" (Deliverable 5.2 section 4.5). Additionally, the LL employs "drone-based heat mapping, citizen-generated ground measurements, and advanced energy balance modelling techniques" (Deliverable 3.3 section 4.3). Complementing the core CS, the LL features the CityZcan monitoring dashboard and Internet-of-Things sensor box. This dashboard integrates real-time data collection on air temperature, humidity, particulate matter (PM) 2.5, and thermal images, engaging citizens directly in the data collection process. It delivers live microclimate updates. Furthermore, the LL includes an urban heat data analytics toolkit. This toolkit facilitates detailed analysis of urban heat patterns and identifies vulnerable neighbourhoods through script-based processing of drone thermal imagery. It aids planning, evaluation, and visual communication through statistical graphs, segmentation masks. Lastly, the urban heat awareness and gamification toolkit, offers a diverse range of interactive and educational tools. It includes educational videos, public awareness materials such as banners, a documentary film, childfriendly visuals, and a card-based serious game designed to foster understanding and adaptive behaviour toward extreme heat events. Targeted at diverse groups such as citizens, students, and community organizations, this toolkit enhances climate literacy and encourages participatory engagement. 5.5.2 Adaptation drivers One of the critical adaptation triggers identified in this LL is the intentional focus on user-friendly design and broad accessibility of the CS. The platform's straightforward, intuitive interface, available via a "publicly accessible web platform" (Deliverable 2.6 section 5), features interactive maps where users can view multiple data layers (orthophotos, observed and predicted thermal images, OSM), utilize "zooming in and out and allows easy switching between different map layers" (Deliverable 5.2 section 4.5), and will soon be able to "measure distances, areas, and temperature variations directly" (Deliverable 5.2 section 4.5). This ease of use lowers the technical barrier for stakeholders. Accessibility is further bolstered by the planned "open source" publication of the CNN model and raw data, which not only promotes transparency but also "allow[s] users to generate heat maps for any urban environment... fostering greater community engagement and collaboration" (Deliverable 5.2 section 4.5), particularly among civic tech communities and academic partners. The content provided by the CS acts as a powerful lever for adaptation by directly addressing information gaps and decision-making needs. Historically, decision-making in Budapest "favoured empirical wisdom from direct experience over more sophisticated climate modelling or data-driven insights" (Deliverable 2.6 section 5), and existing tools like the "National Heat Warning system... does not consider local specificities and heat island D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 29 anomalies" (Deliverable 2.6 section 5). The LL Budapest CS triggers action by offering highly relevant, localized, and actionable information. It "facilitates effective decision-making... at the individual/building blocks and district level by identifying hotspots at the street and block level, thus adaptation measures can be implemented where they are most needed" (Deliverable 2.6 section 5). This targeted information supports diverse adaptation strategies, providing a "basis for technological interventions, such as the need to use reflective materials" (Deliverable 2.6 section 5), encouraging "initiatives such as shading, greening" (Deliverable 2.6 section 5), and offering an "efficient way to monitor and evaluate UHI mitigation measures, allowing for continuous refinement of strategies" (Deliverable 2.6 section 5). Furthermore, data accuracy and trustworthiness are key levers, especially given the past reliance on experiential knowledge. While barriers like "limited financial resources, space constraints, and the absence of robust amateur networks pose significant challenges" (Deliverable 2.6 section 5), the CSaims to build user confidence through tangible, high-resolution data and transparent methods. The use of detailed drone measurements provides a strong observational foundation (Deliverable 3.3 section 4.3), and the "CNN model used for predicting thermal images is under continuous development for improved accuracy" (Deliverable 5.2 section 4.5). By providing reliable hotspot identification and demonstrable insights, the CS directly counters the limitations of previous approaches and triggers more informed, localized adaptive measures, helping to prioritize interventions effectively despite resource constraints. 5.5.3 Socio-economic/political/governance barriers and enablers Several interconnected socio-economic and governance-related barriers hinder broader climate adaptation in Budapest’s dense urban environment. Primarily, "scarce financial resources" and the fact that "long-term adaptation solutions are present, but they are generally expensive" pose significant hurdles (Deliverable 2.6 section 5). Compounding this, "limited urban space restrict[s] the implementation of extensive greening and resilience projects" (Deliverable 2.6 section 5), especially as "the districts are overcrowded" (Deliverable 2.6 section 5) and face competing demands, such as for parking. From a governance perspective, the current adaptation strategy is described as "comprehensive yet fragmented", with "resilience efforts... traditionally driven by individuals rather than a unified community effort" (Deliverable 2.6 section 5). This indicates a "limited use of shared resources to collectively address climate risks" (Deliverable 2.6 section 5). Consequently, "systematic adaptation at a larger scale, such as comprehensive urban planning or the integration of building-specific resilience measures, remains uncommon" (Deliverable 2.6 section 5). Decision-making processes further complicate this, having "typically favoured empirical wisdom from direct experience over more sophisticated climate modelling or data-driven insights" (Deliverable 2.6 section 5). This "reliance on historical patterns rather than predictive, climate-based information" (Deliverable 2.6 section 5) can limit effective long-term planning. Additionally, while a national heat alert system exists, it lacks local specificity (Deliverable 2.6 section 5), and crucial tools like municipal "heat health action plans" are absent in the LL districts (Deliverable 2.6 section ). Furthermore, the prevalence of "protected-heritage buildings mostly from the late 19th and early 20th centuries" (Deliverable 3.3 section 4.3) may add regulatory complexity to implementing certain physical adaptations. Conversely, notable enablers within this LL include "the proactive engagement of local municipalities, the National Public Health Institute (OKI), Eötvös Loránd University, and civil society organizations" (Deliverable 2.6 section 5). Existing institutional frameworks, such as the Hungarian heat alert system, provide foundational support for adaptation activities, although they currently lack detailed urban specificity. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 30 5.5.4 Influence of other drivers Other drivers significantly influencing adaptation include socio-economic, ecological, and governance factors. Societal drivers such as "heightened awareness generated through citizen science, volunteer activities, and social media campaigns effectively mobilize public interest and participation" (Deliverable 2.6 section 5). Financial constraints remain another significant hurdle, potentially mitigated by leveraging cost-effective technological solutions like drones and AI-driven analytics. Ecologically, the intensity of urban heat islands demands urgent interventions, emphasizing strategies like "green infrastructure (e.g., trees, green walls, and water bodies) to enhance urban cooling and biodiversity" (Deliverable 2.6 section 5). 5.5.5 Key findings The LL Budapest CS significantly promotes adaptive behaviour among users by clearly "pinpointing heat hotspots and offering precise, actionable insights". The CS notably "raises public awareness and actively encourages localized adaptation initiatives such as greening, reflective construction materials, and urban agriculture" (Deliverable 2.6 section 5). Its strength lies in continuous monitoring capabilities and iterative improvement through ongoing data refinement and stakeholder feedback. Key findings indicate that urban heat islands pose critical challenges in densely built districts, requiring focused interventions. The current adaptation strategies are fragmented, driven largely by individual efforts, and heavily reliant on traditional experience rather than data. Existing heat alerts lack specificity at the local scale, significantly limiting their effectiveness. Additionally, financial and spatial constraints notably restrict extensive greening initiatives. Nevertheless, volunteer-based data collection has proven feasible, enhancing community engagement and data availability. The application of AI-driven predictive analytics, particularly the CNN model, demonstrates strong potential for accurately forecasting urban heat patterns and thus improving adaptation strategies. In conclusion, the climate-service prototype already provides strong behavioural levers – high-resolution heat maps, AI-driven forecasts, and citizen-generated data that encourage residents and planners to take action – but to fully unlock its adaptive impact, supportive governance, integrated urban-climate policies, and continued technical refinement will be necessary. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 31 5.6 Living Lab in the Alazani-Iori Basin , Georgia 5.6.1 Basic information about the Living Lab and its climate services The LL Alazani-Iori River Basin CS is designed to positively influence adaptive behaviour by enhancing water resource management through accurate, actionable information. The CS prototype, described as "a Water Resource Management Service" (D3.3, Section 4.2.1), integrates a sophisticated streamflow prediction system that serves as the foundation for hydrological drought monitoring and forecasting, which is critical for managing water allocation between agriculture and hydropower (D2.6, Section 4.3). Its robust architecture is built on a cloud-based infrastructure using Docker containers, with a front-end powered by node-js, ReactJS, OpenLayers, and Highcharts, and a back-end based on pygeoapi (D5.2, Section 4.6.1). This technical framework not only ensures data accuracy through bias adjustment, downscaling, ensemble statistics, and skill assessment, but also delivers a user-friendly visualization interface, making complex data readily comprehensible. Moreover, the CS is tailored to address the regional need for timely climate information. It enriches agrometeorological bulletins with key climate variables, such as precipitation, temperature, and drought indicators, which guide water management and on-farm operations (D2.6, Section 4.3). The qualitative insights obtained via focus group discussions and semi-structured interviews with MAP members further ensure that the platform meets the practical needs of its users. As a result, by combining robust data processing with intuitive, targeted visualizations, the CS empowers local stakeholders to make informed decisions that enhance resilience in the face of drought and competing water demands. 5.6.2 Adaptation drivers The LL Alazani-Iori River Basin CS platform is designed to drive adaptive behaviour by making complex climate data accessible and actionable. Its interface is built to be intuitive and user-friendly: "The front-end of the Georgia Climate Service platform is designed to be intuitive and user-friendly..." (D5.2, Section 4.6.2). The platform features a well-organized layout, with a header, left sidebar, main map display, and right panel for data visualizations, which facilitates seamless navigation and decision-making. Visual tools such as the "Ensemble Time Series," "Box Plot," and "Interactive Map" not only enhance data comprehension, but also enable users to drill down into specific areas of interest, supporting targeted, adaptive responses. Accessibility is a key design goal: "the Georgia Climate Service platform is designed with a strong emphasis on accessibility, ensuring that users with limited programming skills can easily interact with the system" (D5.2, Section 4.6.2). This commitment to ease-of-use is reinforced by features like a graphical interface that avoids command-line complexity, though challenges remain, for instance, data availability is limited due to gaps in the hydrological network (D3.3, Section 4.2.2). Despite this, the system provides timely information, with a Data Ingestor that updates data monthly and forecasts available up to "[Forecasts up to] 6 months ahead" (D5.2, Section 4.6.2). The content of the CS is highly relevant and action-oriented. It addresses the critical need for reliable climate information pertaining to water availability and drought, which guides water allocation and agricultural planning processes (D2.6, Section 4.1; D2.6, Section 4.3). This is particularly evident where the service is used to enrich agrometeorological bulletins that support decisions on crop choice, irrigation, and timing (D2.6, Section 4.3). The careful design of the interaction flow ensures that stakeholders can swiftly interpret the data, thus enabling effective decision-making. Data accuracy and trustworthiness are also core to the potential impact of the CS. The CS leverages highquality, bias-corrected data from SMHI: "The provided data [from SMHI] are already bias-corrected to enhance accuracy." (D5.2, Section 4.6.1), and applies rigorous processing methods, including bias-adjustment, D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 32 downscaling, and skill assessments, to refine the forecasted river flow rates using the drained area hydro method (D5.2, Section 4.6.1). However, a challenge remains: while the system is engineered for precision, many farmers still rely on their past experiences and peer learning, indicating that acceptance of scientific forecasts requires further trust-building (D2.6, Section 4.2). By combining an intuitive design, strong accessibility features, timely and relevant content, and robust data processing techniques, the LL Alazani-Iori River Basin CS has the potential to significantly influence users’ adaptive behaviour. Its effective translation of sophisticated climate data into actionable insights can empower stakeholders to manage water resources more efficiently, despite existing challenges in data availability and trust in forecast outputs. 5.6.3 Socio-economic/political/governance barriers and enablers The LL Alazani-Iori River Basin CS has strong potential to drive adaptive behaviour by addressing key socioeconomic and governance challenges while leveraging robust stakeholder collaboration. On the socioeconomic side, it is clear that adaptation is critical given that many "farmers with sufficient financial resources also transitioned to greenhouse farming," highlighting that resource constraints remain a significant barrier for others (D2.6, Section 4.2). Additionally, the CS must contend with the "competing need to supply water for hydropower generation," which illustrates the complex trade-offs between agricultural and energy priorities in the region (D2.6, Section 4.3). Governance obstacles also pose challenges, such as "acute water losses due to improper infrastructure" and the fact that "the hydrological network is still being reconstructed following its virtual collapse after the soviet period" (D3.3, Section 4.2.2). These issues not only affect data quality and availability, but also reflect lingering institutional and resource constraints that can hamper effective climate adaptation. On the enabling side, however, the presence of a well-established MAP that "includes a diverse group of stakeholders," ranging from CS providers to individual farmers, supports co-production and shared decision-making (D2.6, Section 4.1). Moreover, non-climate related socio-economic factors, such as reliance on agriculture as the main livelihood source, fluctuations in local market prices, and the need for livelihood diversification, underscore the broader context in which the CS operates (D2.6, Section 4.1; D2.6, Section 4.2). These factors, combined with policy influences like "integrated management to the EU Water Framework Directive" and the role of the National Environmental Agency in operating critical networks, create both challenges and opportunities. While "competing demands for water in the region" (D2.6, Section 4.3) highlight the need for coordinated action, they also stress the importance of tailored, data-driven strategies to optimize resource use. By addressing these multi-dimensional challenges and leveraging strong institutional partnerships and stakeholder engagement, the LL Alazani-Iori River Basin CS is well-positioned to encourage and support adaptive behaviour. Its ability to provide accurate, context-specific climate information, despite legacy infrastructural and resource limitations, will be critical for informing decision-making and enhancing resilience across the region. 5.6.4 Influence of other drivers The LL Alazani-Iori River Basin CS is poised to drive adaptive behaviour by delivering crucial, context-specific ecological insights. For example, as stated, "Climate change models predict a steady increase in temperature... precipitation is expected to decline... most pronounced between May – October. This coincides not only with periods of low flows in the Alazani River, but also with periods of peak irrigation water demand" (D2.6, Section 4.2). This observation underscores the urgency for adaptation, as it clearly links changing climatic conditions with increased water stress during critical agricultural periods. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 33 Moreover, the persistent challenges in the region are well captured by the acknowledgement that "farming communities in the region also suffer from droughts, hail, and sudden onset high intensity precipitation events" (D2.6, Section 4.2), highlighting the frequent disruption of agricultural activities. In response, practical adaptations, such as "installing windbreaks to reduce effects of wind erosion during droughts" (D2.6, Section 4.2), offer tangible measures to mitigate these impacts. Finally, the region has a diverse and challenging topography: "Due to the complex mountainous topography and highly diverse climate settings, Georgia is subject to climate-related hazards such as floods, flash floods, landslides, debris flow/mudflow, snow avalanches, hailstorms, windstorms and droughts" (D3.3, Section 4.2). Therefore, the ability of the CS to provide detailed, actionable forecasts is essential. By integrating these drivers into its design, the LL Alazani-Iori River Basin CS can empower local stakeholders to implement targeted adaptation strategies. 5.6.5 Key findings The LL Alazani-Iori River Basin CS is designed to actively support adaptive behaviour by providing timely and precise information for decision-making. It aims to deliver "[CS provides] better information... to guide on-farm operations... as well as to support water management and planning..." (D2.6, Section 4.3), thereby informing critical decisions such as crop selection, irrigation timing, and resource allocation. This service enriches existing agrometeorological bulletins, offering farmers actionable insights that translate complex climate data into practical strategies ("[CS used to] enrich currently provided agrometeorological bulletins to farmers, thereby informing strategies such as crop choice, need for irrigation and timing etc." D2.6, Section 4.3). The key benefits include tailored streamflow forecasts that address competing water demands, advanced processing techniques ensuring data accuracy, and a highly accessible, user-friendly interface supported by a robust cloud-based platform and integration of multiple data sources. However, challenges remain, such as limited real-time observational data, competing water needs, reliance on external data providers, and the fact that many farmers continue to rely on traditional knowledge over scientific forecasts. Overall, the LL Alazani-Iori River Basin has successfully identified local adaptation needs and established effective data pipelines and skill assessment methodologies. Yet, to further enhance the influence of the LL Alazani-Iori River Basin CS on adaptive behaviour, efforts should continue to improve visualization, adjust external projections to local conditions, strengthen the real-time monitoring network, and better bridge scientific forecasts with local decision-making practices. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 40 7 Cross living labs synthesis Efforts in I-CISK LL focused on developing and evaluating climate services (CS) aimed at encouraging behavioral change and promoting climate adaptation. Task WP2 2.4 used various methodological approaches, including discrete choice experiments (DCEs), serious games, participatory workshops, document analysis, and stakeholder feedback to understand how CS can effectively encourage adaptive behavior. The context-appropriate application of methods, based on the guidelines of the I-CISK co-creation framework (D2.8), enabled robust synthesis across LLs while respecting local realities, stakeholder capacity, and CS maturity levels. A fundamental literature review (Muller et al., 2024) revealed the widespread use of forecasts, particularly among agricultural producers, as well as the critical impact of local contextual factors and the persistent challenge of translating increased climate awareness into concrete adaptation measures. These findings directly influenced the development of the methods used in Task 2.4. For example, the finding that improved perception does not always lead to immediate behavioral change motivated the use of serious games and DCEs to measure behavioral intentions and decision thresholds, rather than just measuring awareness. Within the I-CISK LLs, we examined the ability of each CS prototype to promote adaptive behavior by overcoming barriers and providing facilitating factors. Below, we summarize the most important findings from the cross-laboratory experiences, including a structured, cross-laboratory comparison of methodological approaches, empirical depth, behavioral drivers, and barriers, as well as explicit links to WP 4 and 5. Figure 6 Elements of the cross LL synthesis D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 41 7.1 Basic information on Living Lab prototypes All the Living Labs in I-CISK created web-based climate service prototypes tailored to specific regional contexts. The prototypes varied significantly in purpose: ● Rijnland (Netherlands): A drought alert service for water management, agriculture, and tourism. ● Andalusia (Spain): An agricultural and forestry planning tool. ● Crete (Greece): A service focusing on climate-sensitive tourism. ● Emilia Romagna (Italy): Climate forecasts supporting multi-sector water management. ● Alazani-Iori (Georgia): A water resource management platform for agriculture and hydropower. ● Lesotho: A system addressing drought and cold-wave risks, integrating early warning and anticipatory actions. ● Budapest (Hungary): An urban heat visualisation service, supported by predictive modelling and sensor data. Despite their differences, all prototypes employed cloud-based infrastructure, combined global forecast products (ECMWF, Copernicus, SMHI) with local observational data, and engaged MAPs through stakeholder interviews, workshops, and iterative feedback. 7.2 Adaptation drivers: design, accessibility, content, and trustworthiness In line with Task 2.4’s goal to identify which features of CS encourage practical uptake and behavioural change, each LL placed strong emphasis on user-friendly design and accessibility. Diverse approaches, such as clear maps, colour coding, and local-language interfaces, were adopted to lower technical barriers and expand the pool of potential users. Many LLs also introduced context-specific indicators (ranging from drought severity thresholds to tourism climate indices) tailored to the real-world challenges that various stakeholder groups face. Issues of data accuracy and reducing the associated lack of trust are fundamental to influencing behaviour. Despite adopting advanced forecasting techniques (e.g. downscaling, bias correction), uncertainties still persisted, particularly at smaller spatial scales or for longer-term forecasts. These uncertainties underscore why Task 2.4 highlights the importance of gaining user trust, not only by integrating local and citizen-collected data but also through transparent discussions of forecast reliability. Sustained stakeholder engagement thus emerged as a powerful tool for aligning scientific outputs with everyday decisions, demonstrating in practice how CS can prompt shifts in habits, planning cycles, and investments. 7.3 Socio-economic, political, and governance barriers and enablers The results also call for understanding how broader societal and institutional contexts shape whether and how CS lead to actual changes in practices. All LLs encountered socio-economic constraints, such as limited funding or insufficient technical capacity, which directly impacted whether new forecasts and tools were readily adopted. Meanwhile, governance structures, often geared towards historical protocols, were sometimes slow to integrate forecast-based measures into standard procedures, reinforcing reactive rather than proactive approaches. Nonetheless, clear examples of enablers appeared where local administrations and stakeholders were both mandated and motivated to cooperate on climate issues (e.g., through resilience planning or formal MAPs). Supportive governance frameworks—such as municipal resilience plans, updated water-allocation rules and D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 42 sector-specific adaptation guidelines— often simplified the institutional hurdles, allowing new CS to become embedded in official processes. Such experiences reflect conditions under which CS effectively resonate with diverse end-users and lead to measurable shifts in behaviour, whether through resource allocation decisions, policy reforms, or collective adaptation planning. 7.4 Influence of other drivers: socio-economic, ecological, and governance factors Reflecting that multiple drivers can interact to shape behavioural change, each LL confirmed that a wider set of factors intersect in climate adaptation. The specific design, demand for, and potential utility of the CS in each LL were shaped by broader environmental, societal, and institutional drivers. Ecological vulnerabilities were paramount: Rijnland's low-lying geography and drought/salinity risks, Andalusia's agricultural drought sensitivity, Crete's tourism sector exposure to climate variability, Emilia Romagna's complex water competition under drought and flood risks, and Alazani-Iori's water stress exacerbated by challenging topography and multiple hazards. The CS were explicitly designed to address these ecological pressures. Socioeconomic drivers, such as the economic importance of climate-sensitive sectors (agriculture, tourism, water management) and competition for resources, directly influenced CS priorities and the type of information provided. Governance drivers, including the role of national or regional agencies, reliance on specific data providers, and the need to align with existing planning frameworks, also played a significant role in shaping CS development and integration pathways. 7.5 Using choice experiments and serious games to understand adaptation behaviour Within Task 2.4, two methodologies, namely choice experiments that were applied in the Rijnland Living Lab, and the “Thermaform your city” serious game in the Budapest Living Lab emerged as complementary means of understanding drivers of behaviour. A brief comparison of these is provided in the following table. Dimension Discrete Choice Experiments “Thermaform your city” game Focus & Approach Individual, survey-based choice tasks assessing willingness to pay or trade off attributes Cooperative, scenario-based gameplay requiring real-time group decisions Data Type Primarily quantitative (mixed logit models, willingness to pay estimates) Mixed (observing negotiation, resource sharing, and group dynamics) Insights Provided Quantifies which factors (cost, distance, drought severity) most influence individual choices Reveals collective behaviour, negotiation challenges, trust dynamics, and governance bottlenecks Key Strengths  Yields numerical insights  Captures heterogeneous preferences  Highly scalable  Immersive and participatory  Real-time adaptation feedback  Highlights governance fragmentation Potential Limitations  Less insight into social interaction or institutional constraints  Hypothetical scenarios  Harder to capture systematic quantitative data  Timeand facilitation-intensive Usefulness for Task 2.4  Pinpoints high-impact CS attributes  Identifies thresholds for behavioural change  Shows how multiple stakeholders respond to CS data  Illustrates how governance and trust shape adaptation Synergy When Combined DCEs clarify individual-level drivers; the serious game highlights group-level negotiation, governance, and real-time learning D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 43 7.6 Implications for WP4 (system-level implications) and WP5 (CSIS design) While D2.7 identifies behavioural drivers and barriers at the individual and institutional levels, these findings directly support the WP4 analysis of two-way feedback loops between adaptation measures and CS use (D4.5) and the WP5 iterative refinement of the I-CISK climate service platform (D5.2–D5.4). This section clearly articulates these relationships. 7.6.1 WP4 findings: behavioural incentives, barriers, and the risk of maladaptation The behavioural incentives and barriers identified in D2.7 directly feed into WP4's analysis of two-way feedback loops between adaptation measures and CS use (D4.5). Specifically:  Communication of mistrust and uncertainty: The findings in D2.7 highlight that stakeholders turn away when the uncertainty of CS is poorly communicated (e.g., Crete, Georgia). This is consistent with the power asymmetries identified in D4.5 and the risk that reliance on short-term CS may exacerbate poor adaptation outcomes if users do not trust the accuracy of forecasts or if certain sectors (e.g., tourism on the island of Crete) dominate access to CS compared to other groups (e.g., agriculture).  Governance fragmentation as a barrier: The Budapest serious game revealed that strong governance structures prevent coordinated adaptation, mirroring D4.5's 'Fixes that Fail' and 'Band-Aid Solutions' archetypes, where short-term CS-driven responses (e.g., emergency irrigation in Andalusia, lock closures in Rijnland, watering hot asphalt in Budapest) delay transformative adaptation and risk locking in unsustainable practices. As documented in D4.5, "Maladaptation is inherently difficult to study because it unfolds over long timescales and often emerges from well-intentioned actions… Models like the one presented here fill this gap by offering foresight into how today's decisions affect tomorrow's vulnerabilities."  Behavioural thresholds and feedback loops: The Rijnland DCE quantified exact thresholds at which recreationists would switch harbours (e.g., 14% pondweed coverage, €65 WTP for avoiding extra drought), informing WP4 scenario modelling by providing empirical thresholds for when CS information triggers measurable behavioural shifts vs. status-quo persistence. These thresholds could be critical inputs for D4.5's dynamic models exploring how individual adaptation decisions aggregate into systemic consequences (e.g., increased water demand, resource competition, cascading vulnerabilities).  Power asymmetries and sectoral priorities: D2.7 identified competing stakeholder demands (e.g., tourism vs. agriculture in Crete) that directly inform the D4.5 archetype “Success to the successful.” 7.6.2 Informing WP5: Refining the CS design. The insights provided by D2.7 regarding design, accessibility, content and trustworthiness informed the iterative refinements made to the I-CISK CSIS platform (see D5.2, D5.3 and D5.4) directly. Specifically:  User-defined thresholds and local knowledge integration: Feedback from LLs emphasised the need for transparent communication of forecast limits and the integration of local observational data (e.g. historical drought experiences and farmer-defined impact thresholds). As documented in D5.4, "A central motivation behind the development of the Climate Services is to lower the entry barriers for non-specialist users... By making climate data readily available, the Climate Service empowers users to configure tailored applications that reflect their specific contexts and decision-making needs".  Iterative feedback and usability improvements: The Rijnland and Budapest LLs conducted multiple rounds of user testing (documented in D5.3 and D5.4), leading to refinements such as improved navigation, enhanced data visualisation, and training materials tailored to stakeholders' technical capacity. These refinements address the findings in D2.7 that CS usability failures (e.g. coarse spatial D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 44 scales, jargon, and unclear purpose) are primary barriers to adoption (see Section 3 for the literature review and Sections 4–6 for the LL case studies).  Multi-sectoral dashboard design addressing power asymmetries: The Crete LL's 'Umbrella Service' approach (D5.4, Section 4.4) integrates multiple bespoke indices for tourism, water and agriculture. This directly responds to the identification in D2.7 of competing stakeholder needs and the risk that prioritising one sector over another (e.g. tourism over agriculture) could exacerbate vulnerabilities. This approach aligns with the 'Success to the Successful' archetype in D4.5.  CS delivery in low-capacity contexts: The findings in D2.7 on the critical role of intermediaries (e.g. the Red Cross in Lesotho and extension officers in Georgia) informed the development of WP5's simplified interfaces, localised language support (e.g. a Georgian translation) and intermediary training modules (see D5.4). This ensures that CS is accessible in contexts where direct end-user engagement is limited due to literacy, connectivity, or institutional barriers. 7.7 Overall evaluation and key findings The CS prototypes hold substantial promise in supporting both immediate operational choices (e.g. responding to imminent drought or heatwaves) and longer-term adaptation strategies (like urban greening or integrated water-resource planning). Yet these experiences also spotlight three persistent barriers to catalysing widespread behavioural change: 1. Forecast accuracy: Without reliable, location-specific data, stakeholders may disregard climate information, reverting instead to established habits. 2. Communicating clearly: Complex models require clear explanations, emphasising the likely range of outcomes rather than presenting misleading certainties. 3. Bridging the science–practice gap: Many sectors remain more comfortable with traditional, experiential knowledge, underscoring the need for ongoing trust-building and demonstration of practical benefits. Ultimately, Task 2.4 and the I-CISK LL experiences converge on a shared insight: CS can accelerate adaptation only when user-centric design, co-production, and transparent, evidence-based communication are in place. By translating climate data into action-oriented guidance and embedding it into routine decision-making, these CS can indeed drive the behavioural shifts that underpin resilient, climate-ready communities. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 45 8 Summary Deliverable D2.7 examines how climate services (CS) can instigate behavioural change across diverse Living Lab contexts in the Netherlands, Spain, Italy, Greece, Georgia, Lesotho, and Hungary. Each CS prototype demonstrates that co-design—integrating local data, addressing user-specific needs, and providing intuitive interfaces—enhances user acceptance and potential impact. Systemic governance, social, and economic obstacles such as siloed budgeting, uncertain forecasts, and traditional reactive culture, continue to act as barriers for CS to instigate adaptation behaviour. Bridging these gaps requires transparent communication of limitations, iterative stakeholder engagement to refine service usability, and alignment with policy frameworks that support proactive, forecast-based planning. Ultimately, climate intelligence, i.e. the combination of high-quality data, local knowledge, trust-building processes, and effective governance, will only translate into sustained adaptation if embedded within end user decision-making. By refining forecast accuracy, ensuring accessible design, and consistently involving stakeholders through co-creation processes, climate services stand a far greater chance of shaping real-world behaviour in an era of mounting climate risks. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 46 9 References The reference of literature review of chapter 3. ChoiceMetrics (2021). Ngene 1.3 user manual and reference guide. ChoiceMetrics. Retrieved March 3, 2025 from https://choice-metrics.com/NgeneManual130.pdf Gbodji, K. K., Quarmine, W., & Minh, T. T. (2023). Effective demand for climate-smart adaptation: A case of solar technologies for cocoa irrigation in Ghana. Sustainable Environment, 9(1), 2258472. Glenk, K., Meyerhoff, J., Akaichi, F., & Martin-Ortega, J. (2019). Revisiting cost vector effects in discrete choice experiments. Resource and Energy Economics, 57, 135-155. Hess, S., & Daly, A. (Eds.). (2024). Handbook of choice modelling. Edward Elgar Publishing. Hess, S., & Palma, D. (2019). Apollo: A flexible, powerful and customisable freeware package for choice model estimation and application. Journal of choice modelling, 32, 100170. Koninklijk Nederlands Meteorologisch Instituut. (2025). Jaar 2024: Extreem warm en zeer nat met vrijwel de normale hoeveelheid zon. KNMI. Retrieved March 3, 2025, from https://www.knmi.nl/nederlandnu/klimatologie/maand-en-seizoensoverzichten/2024/jaar Krinsky, I., & Robb, A. L. (1986). On approximating the statistical properties of elasticities. The review of economics and statistics, 715-719. Krinsky, I., & Robb, A. L. (1990). On approximating the statistical properties of elasticities: A correction. Review of Economics & Statistics, 72(1), 189-190. Mediafonds Provincie Utrecht. (2025, February 14). Fonteinkruid vloek of zegen. Retrieved March 12, 2025, from https://mediafondsprovincieutrecht.nl/producties/fonteinkruid-vloek-of-zegen/ Muller, L. C. F. E., Schaafsma, M., Mazzoleni, M., & Van Loon, A. F. (2024). Responding to climate services in the context of drought: A systematic review. Climate Services, 35, 100493. https://doi.org/10.1016/j.cliser.2024.100493 Netherlands bureau voor Toerisme & Congressen, & Waterrecreatie Nederland. (2022, March). Watersportonderzoek 2021. Watersportonderzoek. Retrieved March 12, 2025, from https://waterrecreatienederland.nl/wp-content/uploads/2022/04/Overkoepelende-rapportageWatersportonderzoek-2021-waterrecreatie-nederland.pdf Philip, S. Y., Kew, S. F., Van Der Wiel, K., Wanders, N., & Van Oldenborgh, G. J. (2020). Regional differentiation in climate change induced drought trends in the Netherlands. Environmental Research Letters, 15(9), 094081. R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/. SurveyEngine (2025). Recruitment & Software for Preference Research. Retrieved March 12, 2025, from https://surveyengine.com/ Train, K. E. (2009). Discrete choice methods with simulation. Cambridge university press. Varen doe je Samen. (2018, August 2). Grote Sluis Spaarndam sluit voor recreatievaart. Retrieved March 12, 2025, from https://www.varendoejesamen.nl/nieuws/grote-sluis-spaarndam-sluit-voor-recreatievaart D2.7 – Understanding Drivers for Behavioural Change in the Living Labs 47 Waterrecreatie Nederland. (2024, April 25). Staande Mast Route (2024). ArcGis Map Viewer. Retrieved March 12, 2025, from https://www.arcgis.com/apps/mapviewer/index.html? webmap=eff1932899354fd0995b913afd776bc2 Watersportverbond. (2019, June 18). Nieuw Sluisbeleid Sluis Spaarndam. Retrieved March 12, 2025, from https://www.watersportverbond.nl/nieuws/nieuw-sluisbeleid-sluis-spaarndam/ Haarlemse Zeil Vereniging (2022, July 12). Beperkt schutten sluis Spaarndam. Retrieved March 12, 2025, from https://www.haarlemsezeilvereniging.nl/2022/07/12/beperkt-schutten-sluis-spaarndam/ I-CISK Deliverables Cited D2.1: Egan, K., Emerton, R., Prudhomme, C., Baugh, C., Masih, I., Hernández-Mora, N., Mazzoli, P., Ziogas, A., Chitishvili, V., Castellana, D., & van Andel, S. J. (2024, June). Preliminary report: Information on climate-service needs and gaps (I-CISK Deliverable 2.1, Iteration 3) [Project deliverable]. I-CISK Consortium. D2.6 De Stefano, L., Ropero Szymañska, N., Hernández-Mora, N., Bagli, S., Bela, G., Mazzoli, P., Rastogi, S., Renzi, F., van Andel, S. J., van den Homberg, M., & Ziogas, A. (2024, June). User-centred validation of the integration of climate action information (Deliverable 2.6). I-CISK Consortium D3.3 Pesquer, L., Pechlivanidis, I., Castellana, D., Chitishvili, V., Egan, K., Mazzoli, P., Ziogas, A., van Andel, S. J., Batlle, A., Prat, E., & Werner, M. (2024, October). Benchmarking tailored climate services for local applications using local knowledge and data (Deliverable D3.3). I-CISK Consortium. D3.4: Baugh, C., Egan, K., Klein Holkenborg, E., Castellana, D., Dotta Correa, D., van Andel, S. J., Emerton, R., & Prudhomme, C. (2025, June). Assessment of existing and tailored climate services using a range of userdriven evaluation metrics [Project deliverable in progress]. I-CISK Consortium. D5.2 Bagli, S., Mazzoli, P., Luzzi, V., Renzi, M., Renzi, F., Sartini, A., Gräler, B., Schnell, J., Bela, G., Ziogas, A., Castellana, D., & Werner, M. (2024, July). Climate data and back-end components (Task 5.2) and front-end components (Task 5.3) [Project deliverable]. I-CISK Consortium D2.7 – Understanding Drivers for Behavioural Change in the Living Labs I Appendix 1 Capturing living-lab climate service insights: evidence review and structured reflection template Building the evidence base for a structured framework to gather and analyse Living Lab insights This focused review, which surveys literature published between 2019 and 2024, was prepared to underpin the structured framework presented in the next section of this document. The chapter builds on the broader synthesis provided in Chapter 3: while that earlier overview mapped the thematic landscape of climateservices research, the present summary focuses on methodology – specifically, Living-Lab design and development considerations that practitioners draw on to encourage behaviour change among climateservice users. This document also serves as a concise summary of the literature-scanning work that has been conducted since the project's outset. To date, 40 peer-reviewed articles have been curated, all of which have been published after 2019 and clearly demonstrate European or Australian relevance. Climate service design and accessibility factors Without user-centred design climate services often face a usability gap: information or visual presentation fails end users and thus cannot effectively support adaptation decisions. (Raaphorst et al., 2020) developed a 12-item usability-gap typology highlighting typical design flaws. A common problem is climate information presented at overly large spatial/temporal scales in highly technical language, making it vague or hard to interpret for local decision-makers. Gaps can also concern validity, readability and interactivity—e.g. whether recommended actions fall within a user’s competence; whether language and symbols match the audience’s mental frames; and whether users possess the visual literacy required for interpretation. The purpose and message of a climate service must be explicit (it should not invite action before users understand the data), and the information must be trustworthy, relevant and—if needed—modifiable (e.g. to test scenarios). Interactivity matters: map interfaces should allow zoom, colour and layer customisation so the service can address specific questions. Researchers emphasise iterative user-centred design from early development stages, ideally via Living Lab testing. (Jacob et al., 2025) show that co-development improves the three “U’s”: usefulness (content fits the decision context), usability (easy access and handling) and use (wide transferability and scalability). Another crucial factor is customisation: services should address diverse user needs (e.g. urban planners require different data than farmers) and, where possible, let users select relevant information or scenarios. Literature further stresses easy accessibility—not only technologically (online platforms, open data) but also linguistically and visually—tailored to users’ capabilities and interpretive frames. Key design and accessibility principles therefore include context-specific content, clear communication, interactive features and user involvement—all of which raise stakeholder commitment and the practical uptake of climate services in adaptation decisions. Climate service design and accessibility factors It is clear that uptake of climate services is being hampered by a number of factors. These include limited funds, staff and guidance, inaccessible scientific language, missing governance frameworks and unreliable, user-specific data (Lourenço et al., 2019; Vincent et al., 2020). Many potential users are unaware of existing or credible services, and unequal power relations can exclude vulnerable groups (Jacob et al., 2025). Early, continuous and inclusive stakeholder engagement supported by transparent methods, open data and trust building is vital. This ensures services answer real needs (Street et al., 2021). Capacity-building and political backing are essential for embedding climate information in routine decisions. D2.7 – Understanding Drivers for Behavioural Change in the Living Labs II It is clear that usability remains a common failure point. A 12-item typology highlights flaws such as coarse scales, jargon, unclear purpose and poor interactivity (Raaphorst et al., 2020). Services must supply clear, context-specific, trustworthy data, enable scenario testing and map customisation, and let varied users tailor outputs to their tasks (Soares & Dessai, 2019). Iterative, user-centred design improves usefulness, usability and real-world uptake (Jacob et al., 2025). Broader Socio-Economic, Ecological and Governance Drivers of Adoption The adoption of climate services is dependent on the local mix of resources, politics and governance. Access to funding (e.g. EU grants) expands cities' and farms' capacity to co-develop services, while strong political commitment—expressed through mandatory climate plans, data repositories or certification schemes— creates clear incentives to use them (Doblas-Reyes et al., 2024). Standards that require certified climate information in adaptation planning are essential for building confidence and accelerating uptake. It is clear that ecological pressures and social dynamics are also factors in this situation. It is clear that direct experience of floods, droughts or heat waves sharply raises demand for climate data, and that degraded soils or dwindling water reserves keep that demand high (Biella et al., 2024). Climate-aware communities, farmer cooperatives and trans-local networks (e.g. the Covenant of Mayors) spread best practices. Conclusions Literature highlights that the key to successful climate-service development and application is the systematic collection and integration of participants’ experiential knowledge. The Living Lab approach has proven effective by facilitating joint learning and innovation under real-world conditions with multiple actors. Five overarching lessons emerge:  Knowledge-gathering methods: frameworks such as the Living Lab learning framework support structured synthesis of participant insights.  Design and accessibility factors: user-centredness, contextualised information, visual communication and interactivity drive acceptance.  Barriers and enablers: resource and knowledge gaps and weak coordination hinder uptake, while trust-building, inclusive cooperation, tailoring and capacity building foster success.  Broader drivers: adoption depends on political support, economic conditions and ecological pressures.  Evaluation approaches: measuring real adaptation impact requires structured frameworks that combine continuous learning with formal indicators References Boon, E., Wright, S. J., Biesbroek, R., Goosen, H., & Ludwig, F. (2022). Successful climate services for adaptation: What we know, don’t know and need to know. Climate Services, 27, 100314. https://doi.org/10.1016/j.cliser.2022.100314 Boon, E., Body, N. S., & Biesbroek, R. (2025). Developing and testing an evaluation framework for climate services for adaptation. Climate Services, 38, 100549. https://doi.org/10.1016/j.cliser.2025.100549 Bhatta, A., Vreugdenhil, H., & Slinger, J. (2025). Harvesting Living-Lab outcomes through learning pathways. Current Research in Environmental Sustainability, 9, 100277. https://doi.org/10.1016/j.crsust.2024.100277