Bridging the Water

Abstract

Artificial intelligence is often presented as a means of overcoming remoteness. For islands, the claim is attractive. AI-assisted diagnosis may increase the reach of scarce specialists; educational tools may give learners access to subjects that cannot be staffed locally; virtual assistants may make government information available without a journey to the capital; and predictive systems may support transport, environmental monitoring and emergency response. Yet distance is not a single obstacle that technology can remove. A service may become easier to contact while the people who design it, decide eligibility, interpret evidence and answer for errors become more remote.

This working paper examines AI in island health, education, administration and resilience. It distinguishes operational AI from proposed systems and from the digital infrastructure on which AI depends. The cases include outer-island clinical communication in Tuvalu, AI-assisted magnetic-resonance imaging across Vancouver Island, preparedness for AI-enabled public-health warning systems in Trinidad and Tobago, generative-AI provision in Hawaiʻi’s public schools, Singapore’s Pair suite for public officers, and AI-assisted environmental observation through Digital Earth Pacific. The evidence does not support a general claim that AI either centralises or decentralises island services. Outcomes depend on institutional design.

The paper identifies three possible relationships between AI and island services. AI can extend local capability, substitute for local provision, or extract data and decision-making authority into distant systems. It proposes an Island AI Services Test based on place fit, capability, human presence, authority, data stewardship, infrastructure continuity, archipelagic distribution and workforce effects. The central policy principle is that digital access should not become a rationale for physical withdrawal. AI should reduce avoidable journeys and professional isolation while preserving local people, institutions and routes of redress.

Keywords: artificial intelligence; islands; remoteness; public services; digital health; education; public administration; resilience; centralisation

Evidence note

The references are limited to sources whose existence and bibliographical details were checked against publisher, government, intergovernmental or institutional records. Complete texts or the relevant pages were inspected wherever accessible. DOI, repository and stable institutional links are supplied so that readers can examine the evidence. The paper distinguishes deployed AI, AI pilots, readiness studies and non-AI digital infrastructure. A digital service is included only when it demonstrates a condition on which later AI use depends; it is never presented as an AI deployment. Institutional reports of benefits are attributed to the reporting institution and are not treated as independent impact evaluations.

1. Can AI bridge the water

Island residents routinely pay a distance premium for essential services. A hospital referral may require a flight or boat journey. A school may be unable to recruit a specialist teacher. A government office may sit on another island. Maintenance engineers, social workers and emergency coordinators may cover scattered communities through systems designed around the principal island. These are not simply problems of kilometres. They arise from schedules, weather, cost, limited professional pools, administrative centralisation and the practical difficulty of maintaining specialist services for small populations.

AI appears to offer a compelling response. A model can assist a clinician without travelling, help a teacher prepare material, search government records, translate a query or identify patterns in satellite and public-health data. The marginal cost of another digital interaction may be low compared with the cost of moving a specialist. These capabilities deserve serious attention. Small Island Developing States face fixed costs and shortages across the principal components of health systems, and the burden of maintaining specialist functions can be proportionately greater than it is for larger states (Meessen et al., 2024, pp. i4–i8).

The difficulty lies in the phrase “overcome distance”. Technology can reduce the time needed to obtain information, but it can also move professional work, discretion and institutional authority away from the island where a service is experienced. A resident may receive an immediate automated answer while finding it harder to reach a named official. A local clinic may gain diagnostic support while the system becomes dependent on a distant vendor, cloud service or connection. A school may acquire new learning material while teaching is increasingly standardised around content produced elsewhere.

The first three AI-slands working papers developed an island-centred approach to sovereignty, culture and policy (Jennings, 2026a; 2026b; 2026c). This paper applies that approach to services. Its question is not whether AI can operate across water. It plainly can. The question is whether it leaves island communities with stronger services, greater practical capability and clearer authority than they had before.

2. Distance has several forms

Geographic distance matters because services involve bodies, equipment and relationships. An image can travel while a patient still needs a scanner. A remote consultation can advise a nurse but cannot supply medicines, repair a generator or evacuate a patient. Digital systems therefore alter some components of distance more readily than others.

Four forms should be distinguished.

Access distance is the time, cost and difficulty of reaching a service. AI may reduce it by answering routine questions, supporting remote interpretation or directing a person to the right service.

Capability distance is the separation between a local worker and the expertise, information or tools needed to act. Decision support and professional networks may shorten this distance if they leave the local worker better able to respond.

Authority distance is the separation between a person affected by a decision and the institution able to explain, change or remedy it. Automated systems may enlarge this distance when local staff cannot understand or override a recommendation.

Relational distance is the loss of familiarity, trust and contextual knowledge when a service becomes impersonal. It is especially important in health, education, welfare and other settings where people disclose sensitive information or depend on continuing relationships.

These distances can move in opposite directions. An online assistant may reduce access distance while increasing authority distance. Remote clinical support may reduce capability distance and preserve relational continuity when a local nurse remains the patient’s point of contact. The effect is produced by service design, staffing and governance, not by the model alone.

This leads to three possible service relationships. In an extension model, AI supports local professionals and institutions. In a substitution model, it replaces a local post, office or route to human judgment. In an extraction model, island data and decisions move into systems governed elsewhere, without a corresponding increase in local capability. A single programme can contain all three. The purpose of the distinction is to identify which relationship dominates in practice.

3. Healthcare at a distance

Health provides the clearest case for extending scarce expertise and the clearest warning against mistaking information for care. Small island systems must often decide which capabilities can be maintained locally, which can be shared across islands and which require overseas referral. Meessen et al. (2024) connect these choices to small-system fixed costs, limited specialist staff, migration, transport and exposure to shocks. Suzana et al. (2018) similarly examine the role of importing health services in small island states. Digital and AI tools enter an existing geography of referral and dependence; they do not begin from a blank institutional field.

3.1 Tuvalu shows what connectivity can change

Tuvalu’s experience is important precisely because it is not an AI case. In 2020 the country began installing Very Small Aperture Terminals at health facilities on remote outer islands. Borgelt et al. (2022) found that the connections enabled regular communication between outer-island health workers and the country’s hospital, supported clinical decision-making, reduced some domestic and overseas referrals, and expanded supervision and professional education. The study also found that reliability depended on electricity and other services outside the health sector. Its authors describe digital health as a tool rather than a panacea.

This is the service foundation on which many claims about AI depend. A diagnostic model cannot extend specialist support when the connection fails, a device cannot be powered or local staff cannot use the output. The Tuvalu case also shows why a locally present professional remains central. The technology made the nurse’s work more connected; it did not make the nurse unnecessary. An island-centred AI project should start from this pattern of capability extension.

3.2 Vancouver Island demonstrates bounded augmentation

Nanaimo Regional General Hospital began piloting AI-assisted magnetic-resonance imaging software in 2022, tested an updated version in 2023 and then adopted the software permanently. Island Health reports that the software can improve image quality while reducing scan times for some examinations. It states that scans previously taking ten to fifteen minutes can, in some cases, be completed in about five minutes, and that the software has since been introduced at hospitals in Victoria, Campbell River and the Comox Valley (Island Health, 2025).

This is an operational AI case, but its limits matter. The evidence is an institutional account rather than an independent comparative evaluation. The system supports image acquisition; it does not replace the scanner, technologist or radiologist. Island Health explicitly stresses human oversight because image anomalies and missing information still require expert detection. The case therefore illustrates bounded augmentation within a comparatively well-resourced island health authority. It cannot be assumed to transfer unchanged to small systems with fewer machines, less technical support or unreliable connectivity.

3.3 Trinidad and Tobago demonstrates readiness, not deployment

Addison, Pooransingh and De Freitas (2026) surveyed six of Trinidad and Tobago’s nine County Medical Officers of Health about AI-enabled early-warning systems. All six identified infectious disease and flooding as priority applications. Respondents also identified budget, technical-capacity and data constraints. The authors recommend a phased approach with local validation, human oversight, interoperable data, clear accountability and communication that reaches underserved and low-connectivity communities.

The study is useful because it reports the views of senior public-health practitioners in a twin-island SIDS. It does not demonstrate that an AI warning system has been deployed or that it improves outcomes. Its small, specialised sample also precludes broad generalisation. Read accurately, the case shows how island institutions can define priorities and conditions before procurement. That is a form of agency. It is stronger than importing a ready-made system whose data requirements and alert thresholds were developed elsewhere.

3.4 Health services require a minimum viable presence

AI may make specialist knowledge more available, but it should not turn locally present care into an optional overhead. Research on the recruitment and retention of doctors on remote islands identifies island background, island-based medical education and quality of life as relevant factors (Yamauchi et al., 2025). Research in Carriacou and Petite Martinique also shows that healthcare acceptability depends on context and patient experience, including travel and waiting times (Brizan-St. Martin et al., 2023). Together, these findings warn against evaluating an AI service only through throughput or model accuracy.

Every island health system needs to define a minimum viable presence: the staff, equipment, medicines, authority and emergency capacity that must remain locally available. AI can then be assessed by whether it makes that presence more capable. The World Health Organization’s guidance on AI for health places autonomy, safety, transparency, accountability, inclusiveness and sustainability at the centre of governance (WHO, 2021). In an island context, these principles require a practical answer to a further question: who is present when the connection or prediction fails?

4. Education and specialist knowledge

Island education systems face a similar tension. Digital tools can widen subject access and support teachers who work at professional distance from specialist colleagues. They can also encourage a model in which generic content replaces local curriculum, teaching relationships and cultural knowledge.

The Solomon Islands show how far the material prerequisites remain from the rhetoric of AI. Horihau, Chesal and Suiramo (2025) describe more than 900 islands, expensive electricity and internet access, uneven connection outside Honiara and provincial centres, limited devices, maintenance problems and a shortage of local technical support. Data systems and school connectivity are developing, but the authors emphasise continuing support, funding and teacher training. The case is not evidence of widespread classroom AI. It is evidence that claims about AI access are premature where schools still lack dependable power, equipment and support.

Hawaiʻi provides a different example. The state Department of Education publishes guidance for employees and students, makes student use conditional on school approval and a responsible-use form, and provides Gemini and NotebookLM to staff (Hawaiʻi State Department of Education, n.d.). This is operational provision at system level, accompanied by rules. The public page does not, however, show whether effects differ among Oʻahu and the neighbour islands, whether Hawaiian-medium education has distinct requirements, or how dependence on a single commercial ecosystem is assessed. Those are questions for evaluation rather than reasons to dismiss the programme.

UNESCO’s guidance on generative AI in education argues for a human-centred approach, data protection, institutional validation and age-appropriate use (Miao and Holmes, 2023). For island schools, human-centred provision should also be place-centred. A useful system would help teachers adapt material, connect learners with specialist subjects, work in local languages and reduce professional isolation. A harmful substitution would use the availability of a chatbot to excuse an unfilled teaching post or withdraw a course.

The decisive unit of policy is therefore not access to a tool. It is the educational relationship around the tool. Teachers need time to evaluate outputs, authority to reject unsuitable material, training connected to their subjects, and routes to shape systems around local language and curriculum. Students need privacy, clear rules, accessible alternatives and human help. AI can expand the resources available to a small school, but it cannot assume the public responsibilities of the school.

5. Government without a local office

Public administration is moving rapidly toward AI-assisted search, drafting, classification and virtual assistance. The OECD reports that 67 per cent of OECD countries used AI to improve public-service design and delivery in the evidence it reviewed. It identifies possible gains from automating routine tasks, allocating resources, supporting decisions and making information continuously accessible (OECD, 2025). The same review records risks involving accountability, explainability, data quality, privacy and overreliance on recommendations.

Singapore’s Pair suite offers a substantial small-island-state example. Since mid-2023, Open Government Products has rolled out tools for secure chat, meeting notes and search across public records. A joint account by the Pair team and UNDP reported more than 60,000 registered users, more than 20,000 weekly active users and an estimated 46 per cent saving in time spent on administrative tasks by September 2024 (Ong and Chew, 2024). These figures are reported by the programme and should not be treated as an independent evaluation. The institutional design is nevertheless instructive: a small public team developed tools around government tasks, maintained privacy controls, tested models and built staff guidance alongside adoption.

Singapore demonstrates how AI may release public servants from some routine work while keeping them inside the service. It does not answer the archipelagic question. A dense city-state with strong digital capacity differs from a jurisdiction in which residents cross water to reach an office and staff on smaller islands exercise delegated discretion. In the latter setting, an efficiency programme may close the local office because a chatbot is available. Access distance then falls for routine queries while relational and authority distance increase for difficult cases.

The problem is particularly acute in welfare, immigration, housing and health administration, where a person may need explanation and discretion rather than information retrieval. The United Nations Special Rapporteur on extreme poverty warned that the “digital welfare state” can obscure decision-making, intensify surveillance and make human interaction harder to obtain (Alston, 2019). The issue is not confined to fully automated decisions. OECD analysis of a Swedish employment decision-support system found that caseworkers formally made decisions but were instructed primarily to follow the model’s recommendation, with difficult override procedures (OECD, 2025). Human presence without practical authority is not meaningful human review.

An island-centred administrative service should therefore measure more than transactions completed online. It should record whether people can reach a responsible officer, whether outer-island staff can override or escalate a case, whether decisions are explained in locally intelligible terms, and whether a non-digital route remains available. The physical office may change function, but its remaining role should be determined by need and democratic access rather than by the existence of an automated channel.

6. Transport, emergencies and environmental services

AI can support systems whose scale extends far beyond a single settlement. Satellite-image classification, predictive maintenance, demand forecasting and hazard detection are relevant to islands because land, sea, transport and emergency services are closely connected. Regional cooperation can also make specialist capability affordable.

The Pacific Community’s Digital Earth Pacific programme uses artificial intelligence with Earth-observation data to identify and monitor environmental change, including land cover and mangroves (Pacific Community, 2025). Its regional form is significant. Pacific islands gain access to shared data infrastructure and technical capability that many could not maintain separately. The arrangement may support local planning, blue-carbon work and disaster preparation without requiring every government to reproduce the same platform.

Regional provision does not remove questions of authority. Island agencies and communities need to understand classification errors, influence the categories used, add local observations and decide how sensitive data are shared. A model that identifies mangrove extent may be technically accurate while missing customary use, tenure or culturally important boundaries. Regional capacity should create a stronger relationship between island knowledge and technical analysis, rather than turn islands into data suppliers for an external product.

Continuity is also a service requirement. Franken et al. (2022) show that many small and island jurisdictions have limited submarine-cable redundancy and that failures can produce cascading social and economic effects. The underlying network data in that study should not be read as a current cable inventory, but the service-design lesson remains. An AI-supported ferry schedule, emergency dashboard or public-information assistant must have a low-bandwidth or offline fallback. Backups should be tested under the conditions in which the service will be needed, including power loss, damaged landing infrastructure and simultaneous disruption across several islands.

The Trinidad and Tobago public-health study makes the same point from another direction. Respondents valued dashboards, emergency-service integration and alerts, yet the authors emphasise human oversight, local validation and communication for low-connectivity populations (Addison, Pooransingh and De Freitas, 2026). Prediction is only one part of an early-warning system. A useful warning must reach people, be trusted, connect to local response capacity and lead to action.

7. Algorithmic centralisation

Island services have long been centralised for understandable reasons. Specialist staff, hospitals, courts, universities and administrative departments cannot always be duplicated across small populations. AI changes the cost and reach of central services, but it may also deepen a pattern in which expertise and authority accumulate in the capital or outside the jurisdiction.

Algorithmic centralisation occurs when a digital system makes remote delivery easier for the provider while removing capability from the place where the service is used. It can take several forms:

• local staff enter data but cannot inspect, adapt or challenge the system;
• an outer-island office loses decision-making powers because recommendations are generated centrally;
• a commercial platform becomes the only route to a public service;
• performance measures reward transaction speed but ignore unresolved cases, travel, trust and exclusion;
• service continuity depends on a single cable, cloud region or vendor; or
• data produced by island communities improve external systems without building local institutional capability.

Centralisation is not inherently wrong. A shared specialist team may improve safety and consistency. A regional platform may be more resilient and affordable than several small systems. The relevant question is whether functions are centralised with transparent responsibility and a deliberate distribution of capability. A system may centralise computation while decentralising usable expertise to local workers. Conversely, a system may appear geographically distributed because residents access it from home while all practical authority has moved farther away.

This distinction matters for procurement. Governments should require suppliers to explain where data are processed, how models change, what local staff can override, how records can be exported, what happens during loss of connection, and how the service can be transferred at the end of a contract. These are service questions as much as technical ones.

8. The Island AI Services Test

The proposed test is a framework for design, procurement and evaluation. It should be applied to a specific service and island geography. A national score would conceal the distributional questions the test is meant to expose.

8.1 Evidence status comes first

Before applying the substantive test, reviewers should classify the evidence. The cases in this paper demonstrate why.

This classification should be published with procurement and evaluation records. It prevents a proposal from being described as an achievement and an institutional report from being mistaken for independent evidence.

8.2 Measure capability as well as convenience

Transaction speed is easy to count. Local capability is harder but more important. Evaluators should ask whether staff acquire transferable skills, whether local institutions can maintain essential functions, whether island-based professional networks become stronger and whether the system reduces avoidable referrals without reducing appropriate care. A faster service that leaves no one able to act during an outage has traded visible convenience for hidden fragility.

8.3 Define the right to a person

Some decisions require a human route because they affect rights, safety or access to essential support. Other interactions require a person because the user cannot navigate a digital service, needs discretion or must explain circumstances the system does not represent. The right to a person should specify response times, escalation authority and accessible channels. It should not be satisfied by placing an employee behind an interface without the power to change the outcome.

8.4 Publish island-level results

National averages are especially misleading in archipelagos. Evaluation should report uptime, response time, error, uptake, escalation and user experience by island where privacy permits. It should examine whether the service reduces travel, supports local employment and works during disruption. Communities should participate in defining success before deployment and in reviewing evidence afterwards.

9. Principles for island-centred AI services

Eight principles follow from the analysis.

1. Begin with the service problem. Identify the actual journey, delay, shortage or coordination failure before selecting an AI system.
2. Extend local capability. Prefer uses that help island workers and institutions do more, learn and retain authority.
3. Protect minimum viable presence. Define the people, facilities and human routes that technology must support rather than replace.
4. Keep authority reachable. Assign responsibility, enable local override and provide an effective route to explanation, correction and appeal.
5. Design for disconnection. Test low-bandwidth, offline, power-loss and vendor-exit arrangements as part of normal assurance.
6. Govern island data. Apply privacy, security and public-record duties alongside collective and Indigenous authority where relevant.
7. Measure the archipelago. Publish geographically disaggregated results and investigate uneven effects between principal and smaller islands.
8. Count institutional effects. Evaluate workforce retention, training, local procurement, dependency and resilience, not only speed and cost.

These principles favour neither localism at any price nor centralisation by default. Some specialist functions should be shared. Some routine interactions should be automated. The boundary should be set through evidence, public purpose and continuing accountability.

10. Conclusion

AI can bridge parts of the water. It can move information, interpretation and specialist support more quickly than people or equipment can travel. Tuvalu’s digital-health experience shows how connection can strengthen outer-island professionals. Island Health’s imaging programme shows a bounded use in which software supports staff and equipment already in place. Singapore’s Pair suite shows how a small state can develop tools around public-sector work. Hawaiʻi and Digital Earth Pacific show the importance of system-wide rules and regional capacity. Trinidad and Tobago shows the value of defining local priorities and safeguards before deployment.

None of these cases supports the claim that distance has disappeared. Services still depend on electricity, cables, devices, maintenance, skilled workers and institutions able to answer for decisions. The most consequential distance may be institutional rather than geographic: the gap between a person and whoever can explain, alter or remedy an outcome.

The proper aim is selective proximity. AI should bring knowledge and administrative capacity closer where physical duplication is impossible, while preserving the people and institutions that make services trustworthy and resilient. It should reduce avoidable journeys without making every human encounter exceptional. It should strengthen local professionals instead of converting them into data-entry points for distant systems. It should make regional cooperation more usable without making responsibility harder to locate.

An island service has been improved when residents can obtain help sooner, local workers can act with greater capability, failures can be managed locally and decisions remain open to explanation and challenge. If AI produces convenience by withdrawing presence, authority and resilience, the water has not been bridged. The service has simply moved farther away.

References

Addison, L., Pooransingh, S. and De Freitas, L. (2026) ‘Artificial intelligence-enabled early warning systems for public health preparedness: perspectives of senior public health leaders in a Small Island Developing State’, Frontiers in Public Health, 14, 1931303. https://doi.org/10.3389/fpubh.2026.1931303

Alston, P. (2019) Extreme Poverty and Human Rights: Note by the Secretary-General. United Nations General Assembly, A/74/493. https://docs.un.org/en/A/74/493

Borgelt, K., Siose, T.K., Taape, I.V., Nunan, M., Beek, K. and Craig, A.T. (2022) ‘The impact of digital communication and data exchange on primary health service delivery in a small island developing state setting’, PLOS Digital Health, 1(10), e0000109. https://doi.org/10.1371/journal.pdig.0000109

Brizan-St. Martin, R., St. Martin, C., LaFoucade, A., Mori Sarti, F. and McLean, R. (2023) ‘An empirically validated framework for measuring patient’s acceptability of health care in Multi-Island Micro States’, Health Policy and Planning, 38(4), pp. 464–473. https://doi.org/10.1093/heapol/czad012

Franken, J., Reinhold, T., Reichert, L. and Reuter, C. (2022) ‘The digital divide in state vulnerability to submarine communications cable failure’, International Journal of Critical Infrastructure Protection, 38, 100522. https://doi.org/10.1016/j.ijcip.2022.100522

Hawaiʻi State Department of Education (n.d.) ‘Artificial Intelligence’. https://hawaiipublicschools.org/student-programs/artificial-intelligence/

Horihau, J., Chesal, K. and Suiramo, P. (2025) ‘From paper to AI: Connecting education to the grid in the Solomon Islands’, World Education Blog, 29 August. https://world-education-blog.org/2025/08/29/from-paper-to-ai-connecting-education-to-the-grid-in-the-solomon-islands/

Island Health (2025) ‘The Cutting Edge of Medical Imaging: How AI Improves MRIs’, 22 July. https://medicalstaff.islandhealth.ca/news-events/mri-cutting-edge

Jennings, A. (2026a) AI-slands: Towards an Island-Centred Approach to Artificial Intelligence. AI-slands Working Paper No. 1. Institute for Northern Studies, University of the Highlands and Islands. https://ai.insularworlds.co.uk/working-paper-1.html

Jennings, A. (2026b) AI and Island Cultures: Preservation, Extraction and the Right to Refuse. AI-slands Working Paper No. 2. Institute for Northern Studies, University of the Highlands and Islands. https://ai.insularworlds.co.uk/working-paper-2.html

Jennings, A. (2026c) Small Islands, Big Decisions: AI Governance, Dependency and the Right to Shape the Digital Future. AI-slands Working Paper No. 3. Institute for Northern Studies, University of the Highlands and Islands. https://ai.insularworlds.co.uk/working-paper-3.html

Meessen, B., Ancia, A., Gill, D., LaFoucade, A., Lalta, S., Sandoval, G. and Waqa, G. (2024) ‘When one size does not fit all: aid and health system strengthening for Small Island Developing States’, Health Policy and Planning, 39(Supplement 1), pp. i4–i8. https://doi.org/10.1093/heapol/czad089

Miao, F. and Holmes, W. (2023) Guidance for Generative AI in Education and Research. Paris: UNESCO. https://doi.org/10.54675/EWZM9535

OECD (2025) Governing with Artificial Intelligence: The State of Play and Way Forward in Core Government Functions. Paris: OECD Publishing. https://doi.org/10.1787/795de142-en

Ong, B. and Chew, G. (2024) ‘Pairing with AI for Public Sector Impact in Singapore’, United Nations Development Programme Singapore Global Centre, 26 September. https://www.undp.org/policy-centre/singapore/blog/pairing-ai-public-sector-impact-singapore

Pacific Community (2025) ‘How SPC’s Digital Earth Pacific harnesses Artificial Intelligence to revolutionise our understanding of the environment’, 13 May. https://www.spc.int/updates/blog/dynamic-story/2025/05/how-spcs-digital-earth-pacific-harnesses-artificial-intelligence

Suzana, M., Walls, H., Smith, R. and Hanefeld, J. (2018) ‘Achieving universal health coverage in small island states: could importing health services provide a solution?’, BMJ Global Health, 3, e000612. https://doi.org/10.1136/bmjgh-2017-000612

World Health Organization (WHO) (2021) Ethics and Governance of Artificial Intelligence for Health: WHO Guidance. Geneva: WHO. https://www.who.int/publications/i/item/9789240029200

Yamauchi, M., Watanabe, J., Kawazuma, Y. and Kotani, K. (2025) ‘Factors Related to Recruitment and Retention of Doctors on Remote Islands: A Systematic Review’, Tohoku Journal of Experimental Medicine, 267(3), pp. 273–280. https://doi.org/10.1620/tjem.2025.J048