AI and climate resilience governance

Summary While artificial intelligence (AI) offers promising solutions to address climate change impacts, it also raises many application limitations and challenges. A risk governance perspective is used to analyze the role of AI in supporting decision-making for climate adaptation, spanning risk assessment, policy analysis, and implementation. This comprehensive review combines expert insights and systematic literature review. The study’s findings indicate a large emphasis on applying AI to climate “risk assessments,” particularly regarding hazard and exposure assessment, but a lack of innovative approaches and tools to evaluate resilience and vulnerability as well as prioritization and implementation process, all of which involve subjective, qualitative, and context-specific elements. Additionally, the study points out challenges such as difficulty of simulating complex long-term changes, and evolving policies and human behavior, reliance on data quality and computational resources, and the need for improved interpretability of results as areas requiring further development.

), building models that mimic reality (1), studying emergent phenomena (1) b.Computer vision: simulating natural processes (1), processing satellite images (1), identifying specific buildings that are vulnerable to flooding such as informal settlements (1), determining areas of risk or compare impacts before and after an event (1) c.Data mining: Automating tasks e.g.data collection on information about disasters (2), Combining and harmonizing existing datasets on adaptation and vulnerability (1) d.Humans and AI: targeted and tailored warnings before hazard events (1) e. Knowledge representation and reasoning: modeling of different dimensions and scales (1), prediction in complex systems (1) f.Machine learning: prediction of disasters, early warning (1), vulnerability analysis (1), data processing and analysis (1), Better deal with uncertainties in probabilistic forecasts (1) g.Natural language processing: Sentiment analysis on the acceptance of adaptation measures (1) h.Robotics: Future carbon sequestration (1) i. Fuzzy inference: offers a way to interpret changes in terms of risk j.Neural networks: help to model the relationship between variables for classification and prediction General a. Understanding better decision-making, behavioral processes (2) b. simulating the diffusion processes of certain technologies (1) c. understanding the barriers of uptake of disaster risk reduction measures (1) d. data analysis and inference (1) e.To gain understanding of how to implement early-warning systems/indicators or do contingency planning.(1) 14.In which of the following use-cases or application areas, based on your experience and knowledge, can AI/ML techniques be applied at a greater scale or frequency?(please select all that apply) General Elaboration a.They are all areas where we have a good understanding of the connection between the desired latent variable and the observable proxy b.These fields are required to use new technology to bring sustainability c.ML offers more tools to study the structural elements of these fields of application, for understanding their dynamics, and improve the prediction capacity.d. items I selected are those for which the broadest coverage and time series are expected to exist.e. Availability of data, need for preparation for uncertain and unpredictable events in the future.
f. Scenario building and understanding the consequences of systemic changes to governance structures could be useful applications in all the areas.g. scientific understanding of these systems is very poorly understood, and in that regard, AI/ML will drive a deeper understanding of how natural (and human) systems work, which will indirectly improve overall h.they can support ongoing processes, identify hidden complexities and help explain possible outcomes i.There is a "Goldilocks" issue for our work where global data may not be an actionable granularity, whilst asset level engineering studies are prohibitively expensive.I would envision AI/ML to be a viable middle road to improve granularity in a cost-effective way.j.This is a systems component that I believe AI can help us see and address.We are good at working on the parts but less sighted on the interdependencies.
Technique-Specific k.Operational DRR: simulate complex processes, such as damage processes or evacuation; identifying damages and impacts (3), early warning systems and forecasting (2) l.Risk finance: (Ripple) Effects in complex systems (1), support damage model applications for insurance (3) m.Agriculture: for planning and monitoring crops (1) n.Transport and critical infrastructure: data-driven models can be made of infrastructure black-outs in the event of climate/weather extremes (1) o.Water governance: ML models can help assess the management of resources (2) p. Supply chain management: Efficiency optimization (1), exploring disruption propagation in network (1), ethical/ emissions flagging (1) q.Climate adaptation: To identify, understand and predict Adaptation Responses (2) r.Urban heat: identifying heat islands (1), models can be built relating health outcomes to environmental variables; predictive models can be made to assess the effects of proposed urban adaptation measures (1) s.Adaptation behavior and psychology: use of agent-based models to model human behavior (2)

Sources of data collection
The survey questions have been distributed across a variety of networks and mailing lists to ensure a broad and diverse set of respondents.These channels include:  SIMSOC mailing list: A mailing list for announcements, news and discussion related to the use of computer simulation in the social sciences. Risk-Kan mailing list: Knowledge Action Network on Emergent Risks and Extreme Events  Network of researchers and practitioners from Zurich Flood Resilience Alliance: https://floodresilience.net/ Please indicate your areas of expertise, responsibility, or research focus (e.g., operational disaster risk management, risk finance, transportation, climate adaptation, AI & Machine learning, supply chain management, etc.).On a scale of 1 to 5, with 1 being 'not familiar at all' and 5 being 'very familiar', how would you describe your familiarity with the topic of AI/ML?What, in your view, are the challenges or limitations that hamper the use of AI/ML methods for climate change adaptation and disaster risk management?Please elaborate where possible.(n=40) a. Lack of expertise/capacity/knowledge (e.g.few researchers trained in both AI/ML and DRR/environmental sciences) (8) b.Lack of training/historical data for AI (e.g. of extreme events with low probability); need to operate in extremes outside training data (6) c.Poor data availability and quality, especially socioeconomic data (e.g.
j. Disaster response: NLP on mentions of disaster topics; movement tracking, classification and segmentation for search and rescue (2) 9. [For those who have not used AI in Q6] Do you think the application of AI/ML methods could be useful to your work/study related to climate change adaptation and disaster risk management?If yes, please explain how and where it could be useful?(n=16) a. Yes (14) i.Early predictions/ warning (4) ii.Simpler and faster hazard modelling (1) iii.Satellite imagery for hazard/vulnerabilities/exposure mapping (2) iv.Post-disaster recovery (2) v. Scenario-planning and sensemaking among professionals (a.Agent-based modelling: Planning and forecasting future, Studying and prediction of actors' behavior (3), assessing policy interventions or to project how populations may adapt to new climate scenarios (