Flood wellbeing is measured as three domains, not one outcome
The paper treats flood-related wellbeing as a multidimensional condition rather than a single psychological outcome. It builds a bespoke instrument organised into three domains: affective–cognitive distress appraisal (BE−), diminished perceived efficacy and agency (DO−), and perceived relational–institutional connectedness (CONNECT+). Each domain is scored across multiple indicators, so a single yearly value aggregates indicator-level evidence rather than one survey question.
This matters because the three domains answer different questions. BE− captures distress appraisal such as anxiety and hopelessness; DO− captures reduced coping capacity, impaired daily functioning, and constrained resources; CONNECT+ captures social support, community assistance, and institutional trust. A reader who collapses them into one "flood wellbeing" number would lose the paper's central comparison, since the study's findings depend on how these domains move together and on which pairs correlate most strongly.
An LLM judge scores discourse against strict evidence rules
The pipeline processed approximately 222,566 discourse records from Meta Content Library. Of these, 13.4% passed the flood-wellbeing relevance and evidence gate and were retained for scoring, 85.8% were excluded as non-relevant, and a further 0.8% contained insufficient evidence for wellbeing assessment. The low retention rate reflects the instrument's stringent requirement that observable evidence of flood-related wellbeing constructs be present before any score is assigned.
Scores run from 0 to 3, but zero has a specific meaning: it is assigned only when no admissible evidence — including direct statements, descriptive content, tonal or pragmatic cues, figurative language, framing, relaying, silence, or satire — can be identified for an indicator. A zero therefore reflects absence of evidence in the discourse record, not absence of the underlying construct, and is treated as missing data. The lowest substantive score is 1, and zero-valued indicators are excluded from activation analyses. Among retained records, score 1 dominates across all indicators, meaning low-intensity evidence is the most common form of expression. The indicators for anxiety, low efficacy, and institutional trust contain no observations at the maximum score of 3.
Flood stressors activate all three domains at once
The paper reports that climate stressors are positively associated with all three wellbeing domains, which it interprets as direct activation. The strongest activation appears for BE−, with moderate correlations with precipitation (r = 0.60) and flood-event frequency (r = 0.58). DO− also correlates positively with precipitation (r = 0.24) and flood frequency (r = 0.42). CONNECT+ shows weaker but consistently positive associations with precipitation (r = 0.15) and flood frequency (r = 0.23), and its strongest association is with named-storm activity (r = 0.44).
The temporal pattern reinforces this. Domain scores rise concurrently during flood-active windows, including 2015–2016 with Storms Desmond, Eva, and Frank and prolonged Shannon-basin flooding; 2020 with Storms Ciara, Dennis, and Ellen and the Skibbereen/Bandon flooding; 2023 with Storm Babet and the Midleton flooding; and 2024–2025 with multiple severe storm events. In each window, BE−, DO−, and CONNECT+ rise together rather than moving in opposing directions. The authors note that higher connectedness, support, and institutional trust do not coincide with lower distress or efficacy impacts at the aggregate population level; instead, protective resources tend to increase alongside adversity. They read this as more consistent with resource mobilisation and altruistic-community dynamics than with the classical stress-buffering expectation that social resources suppress distress under elevated exposure.
Two mediation pathways appear in the correlation structure
Beyond direct activation, the domain–domain correlations suggest two interconnected resilience pathways. The first is an appraisal–efficacy pathway, represented by the strong positive association between DO− and BE− (r = 0.60). Under this pathway, flood exposure activates efficacy-related deficits — reduced coping capacity, impaired daily functioning, constrained access to resources — which are in turn associated with heightened stress, anxiety, and hopelessness. The authors connect this to transactional stress theory and disaster self-efficacy research, in which perceived coping capacity is a key determinant of emotional adjustment under disaster exposure.
The second is a social-buffering pathway, represented by the positive association between CONNECT+ and DO− (r = 0.50), together with an indirect linkage between CONNECT+ and BE− through DO−. This suggests that periods of greater efficacy-related adversity are also periods when social connectedness, protective support, and institutional trust become more salient in public discourse. The direct CONNECT+ to BE− association is observable but substantially weaker (r = 0.25) than the DO− to BE− and CONNECT+ to DO− relationships. The authors interpret this as evidence that institutional and social resources operate mainly through efficacy and agency — shaping whether people perceive they can cope and maintain functioning — rather than as an immediate reduction of affective distress in discourse.
What the correlations do and do not establish
All reported relationships are Pearson correlations between aggregated wellbeing domain scores and climate reference series, and between the wellbeing domains themselves. They describe co-movement at the aggregate population level over 2012–2025, not individual-level causal effects. The paper's language is careful on this point: it describes pathways, associations, and mediation structures surfacing in discourse, and it frames the correlation structure as evidence for co-activation rather than proof of causal direction.
The discourse-based design also carries a specific boundary. Because the instrument requires observable evidence before scoring, and because zero scores are treated as missing rather than as low wellbeing, the measures capture how flood-related wellbeing is expressed in public posts, not the prevalence of wellbeing states in the Irish population. The retention rate of 13.4% means the scored corpus is a filtered subset of flood-related discourse, and the dominance of score 1 means most retained evidence is low-intensity. Readers should treat the domain scores as discourse salience and expression measures, not as survey-equivalent prevalence estimates.
The weak CONNECT+ link to distress is the key comparison
The most decision-relevant comparison in the paper is not whether flood stressors activate wellbeing domains — all three show positive associations — but how the protective domain connects to the others. CONNECT+ correlates with DO− at r = 0.50 and with BE− at only r = 0.25, while DO− correlates with BE− at r = 0.60. The authors argue that this ordering suggests institutional and social resources are expressed in discourse primarily through perceived efficacy and agency, and only indirectly through affective burden.
For readers assessing flood adaptation and resilience, this reframes what to look for. The paper's evidence points to perceived coping capacity and agency as the domain through which social and institutional protective resources are most strongly linked to flood-related appraisal, rather than to a direct suppression of distress. The named-storm result adds a second comparison: CONNECT+ is most strongly associated with named-storm activity (r = 0.44) rather than with precipitation or flood-event frequency, which the authors suggest may reflect the high public visibility of named storms and the media, emergency communication, and community mobilisation they attract.
Where these findings transfer and where they stop
The evidence covers flood-related public discourse in Ireland from 2012 to 2025, drawn from Meta Content Library, and the instrument was built for that context. The findings describe how flood stressors, distress appraisal, efficacy deficits, and relational–institutional connectedness co-move in that discourse corpus. They do not establish that the same correlation structure holds in other countries, in survey-based wellbeing measures, or at the individual level, and the paper does not claim that.
The paper's own contribution is threefold: a bespoke theory-informed instrument for unobtrusive discourse analysis, identification of activation and social–ecological resilience mediation pathways in Irish flood discourse, and a multi-indicator, multi-domain temporal dataset for 2012–2025. The dataset is positioned to support longitudinal research on wellbeing dynamics, disaster experience, climate adaptation, social–ecological resilience, and public responses to flooding. The most useful takeaway for readers is the paper's specific priority: when studying flood wellbeing over time, track distress appraisal, efficacy and agency, and relational–institutional connectedness as separate but co-activated domains, and treat the efficacy–agency domain as the main channel through which social and institutional resources appear in discourse.
Sources used in this answer
Assessing flood-related wellbeing from public discourse in Ireland using large language model as judge, 2012–2025
Flood stressors directly activate affective–cognitive distress appraisal, diminished perceived efficacy and agency, and perceived relational–institutional connectedness in Irish flood discourse from 2012 to 2025, with appraisal–efficacy and social-buffering mediation pathways structuring how these domains co-move.