Search

Abstract Artificial Intelligence applications are rapidly expanding across weather, climate, and natural hazards. AI can be used to assist with forecasting weather and climate risks, including forecasting both the chance that a hazard will occur and the negative impacts from it, which means AI can help protect lives, property, and livelihoods on a global scale in our changing climate. To ensure that we are achieving this goal, the AI must be developed to be trustworthy, which is a complex and multifaceted undertaking. We present our work from the NSF AI Institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography (AI2ES), where we are taking a convergence research approach. Our work deeply integrates across AI, environmental, and risk communication sciences. This involves collaboration with professional end-users to investigate how they assess the trustworthiness and usefulness of AI methods for forecasting natural hazards. In turn, we use this knowledge to develop AI that is more trustworthy. We discuss how and why end-users may trust or distrust AI methods for multiple natural hazards, including winter weather, tropical cyclones, severe storms, and coastal oceanography.

Abstract Artificial intelligence (AI) and machine learning (ML) pose a challenge for achieving science that is both reproducible and replicable. The challenge is compounded in supervised models that depend on manually labeled training data, as they introduce additional decision‐making and processes that require thorough documentation and reporting. We address these limitations by providing an approach to hand labeling training data for supervised ML that integrates quantitative content analysis (QCA)—a method from social science research. The QCA approach provides a rigorous and well‐documented hand labeling procedure to improve the replicability and reproducibility of supervised ML applications in Earth systems science (ESS), as well as the ability to evaluate them. Specifically, the approach requires (a) the articulation and documentation of the exact decision‐making process used for assigning hand labels in a “codebook” and (b) an empirical evaluation of the reliability” of the hand labelers. In this paper, we outline the contributions of QCA to the field, along with an overview of the general approach. We then provide a case study to further demonstrate how this framework has and can be applied when developing supervised ML models for applications in ESS. With this approach, we provide an actionable path forward for addressing ethical considerations and goals outlined by recent AGU work on ML ethics in ESS.

IntroductionWe conducted mental model interviews in Aotearoa NZ to understand perspectives of uncertainty associated with natural hazards science. Such science contains many layers of interacting uncertainties, and varied understandings about what these are and where they come from creates communication challenges, impacting the trust in, and use of, science. To improve effective communication, it is thus crucial to understand the many diverse perspectives of scientific uncertainty.MethodsParticipants included hazard scientists (n = 11, e.g., geophysical, social, and other sciences), professionals with some scientific training (n = 10, e.g., planners, policy analysts, emergency managers), and lay public participants with no advanced training in science (n = 10, e.g., journalism, history, administration, art, or other domains). We present a comparative analysis of the mental model maps produced by participants, considering individuals’ levels of training and expertise in, and experience of, science.ResultsA qualitative comparison identified increasing map organization with science literacy, suggesting greater science training in, experience with, or expertise in, science results in a more organized and structured mental model of uncertainty. There were also language differences, with lay public participants focused more on perceptions of control and safety, while scientists focused on formal models of risk and likelihood.DiscussionThese findings are presented to enhance hazard, risk, and science communication. It is important to also identify ways to understand the tacit knowledge individuals already hold which may influence their interpretation of a message. The interview methodology we present here could also be adapted to understand different perspectives in participatory and co-development research.

Abstract Introduction As wildfire smoke events increase in intensity and frequency in the Pacific Northwest, there is a growing need for effective communication on the health risks of smoke exposure. Delivery through a trusted source or intermediary has been shown to improve reception of risk communication messages. This is especially salient in rural and tribal communities who may be hesitant to trust information from state and federal agency sources. This study aims to identify and characterize trusted sources for smoke risk information in the Okanogan River Airshed Emphasis Area (ORAEA), a rural region of North Central Washington state that is heavily impacted by smoke from wildfires and prescribed fire. Methods The research team conducted a qualitative study using data collected through key informant interviews and focus groups to assess the role of various sources and intermediaries in disseminating smoke risk information. We used a consensual coding approach in NVivo Qualitative Analysis Software to sort data into preliminary categories, which were grouped into themes using a thematic analysis approach. We used member checking and iterative feedback processes with local project partners throughout the project to ensure credibility of results. Results Through the analysis, we identified three themes characterizing trusted sources for smoke risk communication in the ORAEA. These themes were: (1) local and tribal sources of information are perceived as more trustworthy than state and federal government sources, (2) trustworthiness is determined by an evaluation of multiple factors, in particular, perceived credibility, quality of information, and relationship with the source, and (3) conservative political ideology and perceived parallels with COVID-19 communication influence perception of trust. Within each theme, we identified several sub-themes, which contributed additional nuance to our analysis. Conclusion This study provides insights into which sources of information are trusted by rural and tribal community members in the ORAEA and why. Results from our study emphasize the importance of relationships and collaboration with local and tribal partners in smoke risk communication. In this paper, we discuss implications for state and federal agency practitioners and present recommendations for how to work with local and tribal partners on smoke risk communication.

Abstract Lahars pose a significant risk to communities, particularly those living near snow-capped volcanoes. Flows of mud and debris, typically but not necessarily triggered by volcanic activity, can have huge impacts, such as those seen at Nevado Del Ruiz, Colombia, in 1985 which led to the loss of over 23,000 lives and destroyed an entire town. We surveyed communities around Mount Rainier, Washington, United States, where over 150,000 people are at risk from lahar impacts. We explored how factors including demographics, social effects such as perceptions of community preparedness, evacuation drills, and cognitive factors such as risk perception and self-efficacy relate to preparedness when living within or nearby a volcanic hazard zone. Key findings include: women have stronger intentions to prepare but see themselves as less prepared than men; those who neither live nor work in a lahar hazard zone were more likely to have an emergency kit and to see themselves as more prepared; those who will need help to evacuate see the risk as lower but feel less prepared; those who think their community and officials are more prepared feel more prepared themselves; and benefits of evacuation drills and testing evacuation routes including stronger intentions to evacuate using an encouraged method and higher self-efficacy. We make a number of recommendations based on these findings including the critical practice of regular evacuation drills and the importance of ongoing messaging that focuses on appropriate ways to evacuate as well as the careful recommendation for residents to identify alternative unofficial evacuation routes.

Warning systems are essential for providing people with information so they can take protective action in response to perils. Systems need to be human-centered, which requires an understanding of the context within which humans operate. Therefore, our research sought to understand the human context for Earthquake Early Warning (EEW) in Aotearoa New Zealand, a location where no comprehensive EEW system existed in 2019 when we did this study. We undertook a survey of people's previous experiences of earthquakes, their perceptions of the usefulness of a hypothetical EEW system, and their intended responses to a potential warning (for example, Drop, Cover, Hold (DCH), staying still, performing safety actions). Results showed little difference in perceived usefulness of an EEW system between those with and without earthquake experience, except for a weak relationship between perceived usefulness and if a respondent's family or friends had previously experienced injury, damage or loss from an earthquake. Previous earthquake experience was, however, associated with various intended responses to a warning. The more direct, or personally relevant a person's experiences were, the more likely they were to intend to take a useful action on receipt of an EEW. Again, the type of experience which showed the largest difference was having had a family member or friend experience injury, damage or loss. Experience of participation in training, exercises or drills did not seem to prompt the correct intended actions for earthquake warnings; however, given the hypothetical nature of the study, it is possible people did not associate their participation in drills, for example, with a potential action that could be taken on receipt of an EEW. Our analysis of regional differences highlighted that intentions to mentally prepare on receipt of a warning were significantly higher for Canterbury region participants, most likely related to strong shaking and subsequent impacts experienced during the 2010–11 Canterbury Earthquake Sequence. Our research reinforces that previous experience can influence earthquake-related perceptions and behaviors, but in different ways depending on the context. Public communication and interventions for EEW could take into consideration different levels and types of experiences of the audience for greater success in response.

Given the growing use of Artificial intelligence (AI) and machine learning (ML) methods across all aspects of environmental sciences, it is imperative that we initiate a discussion about the ethical and responsible use of AI. In fact, much can be learned from other domains where AI was introduced, often with the best of intentions, yet often led to unintended societal consequences, such as hard coding racial bias in the criminal justice system or increasing economic inequality through the financial system. A common misconception is that the environmental sciences are immune to such unintended consequences when AI is being used, as most data come from observations, and AI algorithms are based on mathematical formulas, which are often seen as objective. In this article, we argue the opposite can be the case. Using specific examples, we demonstrate many ways in which the use of AI can introduce similar consequences in the environmental sciences. This article will stimulate discussion and research efforts in this direction. As a community, we should avoid repeating any foreseeable mistakes made in other domains through the introduction of AI. In fact, with proper precautions, AI can be a great tool to help reduce climate and environmental injustice. We primarily focus on weather and climate examples but the conclusions apply broadly across the environmental sciences.

Natural hazards and disaster reconnaissance investigations have provided many lessons for the research and practice communities and have greatly improved our scientific understanding of extreme events. Yet, many challenges remain for these communities, including improving our ability to model hazards, make decisions in the face of uncertainty, enhance community resilience, and mitigate risk. State-of-the-art instrumentation and mobile data collection applications have significantly advanced the ability of field investigation teams to capture quickly perishable data in post-disaster settings. The NHERI RAPID Facility convened a community workshop of experts in the professional, government, and academic sectors to determine reconnaissance data needs and opportunities, and to identify the broader challenges facing the reconnaissance community that hinder data collection and use. Participants highlighted that field teams face many practical and operational challenges before and during reconnaissance investigations, including logistics concerns, safety issues, emotional trauma, and after-returning, issues with data processing and analysis. Field teams have executed many effective missions. Among the factors contributing to successful reconnaissance are having local contacts, effective teamwork, and pre-event training. Continued progress in natural hazard reconnaissance requires adaptation of new, strategic approaches that acquire and integrate data over a range of temporal, spatial, and social scales across disciplines.

In 2016, the National Science Foundation (NSF) funded a multi-institution interdisciplinary team to develop and operate the Natural Hazards Reconnaissance Facility (known as the “RAPID”) as part of the Natural Hazards Engineering Research Infrastructure (NHERI) program. During the following 2 years, the RAPID facility developed its instrumentation portfolio and operational plan with input from the natural hazards community, the facility’s leadership team, and an external steering committee. In September 2018, the RAPID began field operations, which continue today and include instrumentation, software, training, and support services to conduct reconnaissance research before, during, and after natural hazard and disaster events. Over the past 2 years, the RAPID has supported the data collection efforts for over 60 projects worldwide. Projects have spanned a wide range of disciplines and hazards and have also included data collection at large-scale experimental facilities in the United States and abroad. These projects have produced an unprecedented amount of high-quality field data archived on the DesignSafe cyberinfrastructure platform. This paper describes the RAPID facility’s development, instrumentation portfolio (including the mobile application RApp), services and capabilities, and training activities. Additionally, overviews of three recent RAPID-supported projects are presented, including descriptions of field data collection workflows, details of the resulting data sets, and the impact of these project deployments on the natural hazard fields.

Prior research suggests that the pandemic coronavirus pushes all the “hot spots” for risk perceptions, yet both governments and populations have varied in their responses. As the economic impacts of the pandemic have become salient, governments have begun to slash their budgets for mitigating other global risks, including climate change, likely imposing increased future costs from those risks. Risk analysts have long argued that global environmental and health risks are inseparable at some level, and must ultimately be managed systemically, to effectively increase safety and welfare. In contrast, it has been suggested that we have worry budgets, in which one risk crowds out another. “In the wild,” our problem-solving strategies are often lexicographic; we seek and assess potential solutions one at a time, even one attribute at a time, rather than conducting integrated risk assessments. In a U.S. national survey experiment in which participants were randomly assigned to coronavirus or climate change surveys (N = 3203) we assess risk perceptions, and whether risk perception “hot spots” are driving policy preferences, within and across these global risks. Striking parallels emerge between the two. Both risks are perceived as highly threatening, inequitably distributed, and not particularly controllable. People see themselves as somewhat informed about both risks and have moral concerns about both. In contrast, climate change is seen as better understood by science than is pandemic coronavirus. Further, individuals think they can contribute more to slowing or stopping pandemic coronavirus than climate change, and have a greater moral responsibility to do so. Survey assignment influences policy preferences, with higher support for policies to control pandemic coronavirus in pandemic coronavirus surveys, and higher support for policies to control climate change risks in climate change surveys. Across all surveys, age groups, and policies to control either climate change or pandemic coronavirus risks, support is highest for funding research on vaccines against pandemic diseases, which is the only policy that achieves majority support in both surveys. Findings bolster both the finite worry budget hypothesis and the hypothesis that supporters of policies to confront one threat are disproportionately likely also to support policies to confront the other threat.

With the anticipated boom in the ‘blue economy’ and associated increases in industrialization across the world’s oceans, new and complex risks are being introduced to ocean ecosystems. As a result, conservation and resource management increasingly look to factor in potential interactions among the social, ecological and economic components of these systems. Investigation of these interactions requires interdisciplinary frameworks that incorporate methods and insights from across the social and biophysical sciences. Risk assessment methods, which have been developed across numerous disciplines and applied to various real-world settings and problems, provide a unique connection point for cross-disciplinary engagement. However, research on risk is often conducted in distinct spheres by experts whose focus is on narrow sources or outcomes of risk. Movement toward a more integrated treatment of risk to ensure a balanced approach to developing and managing ocean resources requires cross-disciplinary engagement and understanding. Here, we provide a primer on risk assessment intended to encourage the development and implementation of integrated risk assessment processes in the emerging blue economy. First, we summarize the dominant framework for risk in the ecological/biophysical sciences. Then, we discuss six key insights from the long history of risk research in the social sciences that can inform integrated assessments of risk: (1) consider the subjective nature of risk, (2) understand individual social and cultural influences on risk perceptions, (3) include diverse expertise, (4) consider the social scales of analysis, (5) incorporate quantitative and qualitative approaches, and (6) understand interactions and feedbacks within systems. Finally, we show how these insights can be incorporated into risk assessment and management, and apply them to a case study of whale entanglements in fishing gear off the United States west coast.

Objectives. To examine commonalities and gaps in the content of local US heat action plans (HAPs) designed to decrease the adverse health effects of extreme heat. Methods. We used content analysis to identify common strategies and gaps in extreme heat preparedness among written HAPs in the United States from jurisdictions that serve municipalities with more than 200 000 residents. We reviewed, coded, and analyzed plans to assess the prevalence of key components and strategies. Results. All 21 plans evaluated incorporated data on activation triggers, heat health messaging and risk communication, cooling centers, surveillance activities, and agency coordination, and 95% incorporated information on outreach to at-risk populations. Gaps existed in the specific applications of these broad strategies. Conclusions. Practice-based recommendations as well as future areas of research should focus on increasing targeted strategies for at-risk individuals and expanding the use of surveillance data outside of situational awareness. ( Am J Public Health . 2023;113(5):559-567. https://doi.org/10.2105/AJPH.2022.307217).