Disaster preparedness paradox: Evidence of lower public readiness for extreme rainfall compared to super tropical cyclones in Hong Kong

The 2023 Extreme Rainfall Event, triggered by the remnant low-pressure trough of Typhoon Haikui (no. 2311), was a catastrophic event for Hong Kong, setting multiple historical rainfall records and causing severe city-wide disruption [31]. Typhoon Haikui landed in Taiwan on 3 September and Fujian again on 5 September, over 800 km away from Hong Kong, prompting only the lowest tropical cyclone warning signal in Hong Kong. However, from the night of 7 September to the morning of 8 September 2023, the Hong Kong Observatory (HKO) recorded an unprecedented hourly rainfall of 158.1 millimeters, the highest since records began in 1884 [30,31]. The Black Rainstorm Warning Signal, meaning extremely heavy rain, exceeding 70 millimeters per hour across the city, was in force for 16 hours and 35 minutes, the longest duration since the warning system’s introduction in 1992 [31]. The event overwhelmed urban infrastructure, resulting in widespread consequences: more than 140 injuries were reported, and authorities received over 60 reports of flooding and 200 reports of landslides [31]. The torrential rain caused dramatic scenes of flooding in MTR stations, such as Wong Tai Sin Station, and shopping malls, leading to the partial suspension of the Kwun Tong Line and leaving vehicles submerged on major roads like Lung Cheung Road [30,31].

This study employed a comparative cross-sectional design to investigate differences in the general public’s disaster preparedness for two distinct major hazards in Hong Kong: Super Typhoon Saola and the 2023 Extreme Rainfall Event. Data was collected via an online questionnaire survey following both events, enabling a direct comparison of risk perception, hazard knowledge, risk information seeking, and preparedness behaviors between the two hazards. This methodological approach allows for the identification of hazard-specific variations in the public’s readiness, providing critical insights for targeted disaster risk management and policy development.

First-hand data were collected through an online questionnaire survey (Google Forms) in Hong Kong from 10 September to 31 October 2023. This data collection period commenced within two months of the occurrence of the two selected hazard events. Non-probability sampling was utilized, targeting Hong Kong citizens aged 18 or above holding a valid Hong Kong Identity Card, whether permanent or non-permanent residents. A pilot study was conducted beforehand to ensure that the questionnaire was valid, reliable, and effective. The survey was distributed via the affiliated university’s internal platforms and locally popular social media channels, including WeChat groups, WhatsApp groups, Facebook, and LinkedIn. After removing respondents not meeting the selection criteria, the final sample size consisted of 502 participants.

The study received ethical approval from the Secretary of Survey and Behavioral Research Ethics Committee at The Chinese University of Hong Kong, in accordance with the Declaration of Helsinki (Ref. SBRE‐22‐0240, approved in December 2022).

Initial descriptive statistics characterized the sample’s sociodemographic profile, including individual factors, family economic and housing conditions, and healthcare needs. To compare disaster preparedness and individual disaster risk management between Super Typhoon Saola and the 2023 Extreme Rainfall Event, paired-sample tests were employed; specifically, McNemar’s test for categorical outcomes and the Wilcoxon Signed-Rank Test for continuous, non-normally distributed variables, with a significance level of p<0.05. Furthermore, multivariate linear regression models were constructed to identify the significant influencing factors—including risk perception, hazard knowledge, risk information seeking, and sociodemographic variables—on the level of preparedness for each hazard. All statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS, Version 28.0.1.0).

The study sample (N=502) was characterized by a majority of female participants (64.7%) and a relatively young demographic profile, with 51.0% aged 21–40 and 29.5% aged 18–20 (Table 1). The cohort was highly educated, with 86.3% having attained a postsecondary qualification, and predominantly unmarried (80.9%). In terms of socioeconomic and housing conditions, family income was distributed across low (30.3%), middle (37.5%), and high (32.3%) groups. A majority lived in private housing (59.2%), though rental tenancy was common (62.7%). Most participants were long-term residents of their community, with over half (52.4%) living in the same area for more than ten years, primarily in the New Territories (56.2%). Regarding family healthcare needs, a minority of households included children (28.7%) or elderly people (16.5%), while 19.9% reported a member with a chronic disease and 6.6% had a member with a disability.

The analysis of preparedness measures for Super Typhoon Saola and the 2023 Extreme Rainfall Event revealed a statistically significant difference in public readiness between the two distinct hazards. Descriptive statistics indicated a consistently higher prevalence of preparedness actions for Super Typhoon Saola than the 2023 Extreme Rainfall Event across most measures. This pattern was confirmed by McNemar’s tests, which identified significant differences (p<0.001) for fundamental actions such as buying more food, reducing going out, checking doors/windows, securing outdoor objects, and checking official updates. While measures like clearing drainage outlets, living in risk areas, and monitoring specific hazards showed no significant difference, relocating vehicles and monitoring employer/school notifications were significantly different at p<0.05. Critically, the overall disaster preparedness score, analyzed using the Wilcoxon Signed-Rank test, was significantly higher for Typhoon Saola (Mean=0.68, SD=0.18) compared to the 2023 Extreme Rainfall Event (Mean=0.65, SD=0.22; p<0.05).

Moreover, the analysis of preparedness by population subgroups (Table 2) revealed significant sociodemographic and socioeconomic disparities in readiness for both Super Typhoon Saola and the 2023 Extreme Rainfall Event. A consistent pattern emerged, indicating that vulnerability and resources shaped preparedness levels. For both hazards, married individuals, non-local-born residents, those with shorter community tenure (<10 years), and families with children demonstrated significantly higher preparedness scores (p<0.05). A pronounced socioeconomic gradient was evident, as individuals in the lowest (<HKD 20,000) and middle (HKD 20,000–39,999) income brackets were significantly better prepared than their high-income (≥HKD 40,000) counterparts (p<0.05). Notably, the disparity by gender was significant only for the rainfall event, with females reporting higher preparedness than males (p<0.01). Conversely, factors such as age, education (for the rainfall event), housing type, accommodation tenure, and the presence of elders or members with chronic disease or disabilities did not yield significant differences in preparedness.

The Wilcoxon Signed-Rank tests revealed significant differences in individual disaster risk management between Super Typhoon Saola and the 2023 Extreme Rainfall Event. While overall risk perception was marginally higher for the extreme rainfall event (p<0.01), the public perceived a significantly greater future occurrence possibility for super typhoons (p<0.001). A critical divergence was observed in the appraisal of consequences: the perceived impacts on individuals and families were higher for the rainfall event (p<0.01), yet the impacts on Hong Kong as a whole were judged to be significantly more severe for the typhoon (p<0.001). This suggests a nuanced risk assessment where personal threat is distinguished from collective societal disruption. Correspondingly, objective hazard knowledge was significantly higher for the typhoon (p<0.001). Although traditional information channel utilization remained consistent across both hazards (p=0.700), the use of new information channels was significantly more prevalent during the typhoon (p<0.05) (Table 3).

Multilevel linear regression models revealed distinct and shared influencing factors on disaster preparedness for Super Typhoon Saola and the 2023 Extreme Rainfall Event, explaining 12.0% and 9.8% of the variance in preparedness scores, respectively (both models p<0.001, Table 4). A key finding was the differential role of gender, where being female was a significant positive predictor of preparedness for the extreme rainfall event (β=0.108, p<0.05) but not for the super typhoon. Conversely, being married was a significant factor only for super typhoon preparedness (β=0.134, p<0.05). Notably, non-local-born status was a robust, significant predictor for both hazards (Typhoon: β=0.118, p<0.05; Rainfall: β=0.114, p<0.05), suggesting migration background consistently enhances readiness. Socioeconomically, a higher household income (≥HKD 40,000) was associated with significantly lower preparedness for both events (β=-0.135, p=0.016). Contextual vulnerability also played a role, as residing near a river significantly increased preparedness for the super typhoon (β=0.108, p<0.05), while longer community tenure (>10 years) significantly decreased it (β=-0.105, p<0.05). Most critically, risk information seeking emerged as the strongest shared driver; utilization of traditional information channels was a highly significant positive predictor for both the super typhoon (β=0.168, p<0.001) and the rainfall event (β=0.135, p<0.01). This underscores that beyond sociodemographics, proactive information-seeking through established channels is a fundamental component of disaster preparedness across hazard types (Table 4).

The consistent pattern of greater preparedness for Super Typhoon Saola highlights how hazard predictability, scale, and established response frameworks drive mobilization. Tropical cyclones, characterized by multi-day forecasts and society-wide warnings, elicit concrete preparatory actions across food provisioning, home fortification, and information seeking. In contrast, extreme rainfall events — defined by abrupt onset, localized impact, and ambiguous triggers — struggle to translate heightened risk perception into tangible readiness. These findings confirm previous assertions [13,15] that temporal and spatial uncertainty serve as formidable barriers to adaptive behaviors, and underscore the fatal flaw of assuming cross-hazard preparedness equivalency.

The paradox observed—higher personal risk assessment for extreme rainfall but lower practical preparedness—highlights that psychological salience does not automatically translate into protective action, echoing the well-documented intention–behavior gap in disaster settings [34,35]. Cognitive biases such as optimism bias [36] and normalcy bias [37] may lead individuals to underestimate their vulnerability or delay action despite recognizing high risks. Moreover, ambiguity in onset and a lack of clear preparedness cues for extreme rainfall exacerbate decisional inertia, as outlined by Lindell & Perry [38], who show that perceived efficacy and clarity of recommended actions are critical for mobilization. These findings underscore the necessity for disaster strategies that are psychologically informed, actively bridging the gap between risk awareness and preparedness behavior.

Moreover, the collective framing of tropical cyclone risk (large-scale societal disruption) arguably leverages institutional communication and social norms more effectively than the fragmented, individual experience of extreme rainfall events. This pattern was evident in the lower preparedness scores and in specific actions such as checking windows, securing outdoor objects, and reducing mobility found in Hong Kong, which were significantly less likely during the rainfall event. These results reinforce the necessity to recognize typological differences in hazard manifestation, a crucial step to avoid the “one-size-fits-all” fallacy in disaster readiness campaigns.

The findings strongly corroborate the critical role that socioeconomic and demographic factors play in shaping disaster preparedness. The unexpected inverse relationship between household income and preparedness and extreme rainfall recasts the common assumption that greater economic resources confer higher adaptive capacity, which has been observed for both super tropical cyclones and extreme rainfall. Wealthy individuals may have access to insurance and financial resources that give them a false sense of security, potentially leading to complacency in their preparedness for extreme weather events [39]. Instead, those in low and middle income brackets—often facing greater physical vulnerability and fewer options for external support—demonstrated higher readiness, echoing emerging evidence of adaptive strategies among the economically marginalized.

Migratory status also emerged as a consistent facilitator of preparedness, suggesting that non-local-born residents may be more attuned to disaster communication, possibly reflecting experiences or social networks from prior exposure in different regions. The non-local-born residents’ adaptation to new environments necessitates a proactive approach to disaster preparedness. They usually recognize the importance of preparedness as a vital component of successful adaptation in a new locale, making them more inclined to invest in personal safety measures [40].

Therefore, the findings reveal a clear manifestation of the disaster preparedness paradox, wherein awareness and resources fail to translate into adequate preparedness [41]. Specifically, local-born and high-income respondents demonstrated significantly lower preparedness levels, suggesting that entrenched complacency and an optimism bias, potentially born of a history of successful past coping, may suppress protective action despite high objective risk and economic capacity.

Moreover, some predictors of preparedness were not universal but hazard-specific, reflecting the distinct temporal and spatial dynamics of each threat. Factors such as marital status, long community living duration, and proximity to a river emerged as significant only for super tropical cyclone preparedness. This pattern underscores the anticipatory nature of tropical cyclone response: individuals with greater social responsibilities (marriage), long-term local knowledge of community-specific flood risks (proximity to a river) are more motivated and equipped to undertake the complex, multi-step preparations required for a large-scale, forecasted event. Individuals with greater social responsibilities, such as married residents, tend to translate their sense of obligation to protect their loved ones and community members into proactive disaster preparedness measures, including establishing emergency plans and maintaining supplies [42]. Familiar with their surroundings are more likely to recognize the significance of flooding risks and invest time and resources to prepare adequately [43]. This awareness of specific local conditions can lead to more effective preparation strategies tailored to community needs. However, individuals with deep-rootedness in the community (longer community living duration) are less likely to adopt preparedness measures against super tropical cyclones. This confirms the previous findings that individuals with extensive local knowledge might underestimate the likelihood of disasters or view them as part of their daily life, leading to complacency [38]. Their familiarity with the environment can result in overconfidence that diminishes the perceived necessity for proactive measures.

Furthermore, gender differences were hazard-specific, with women showing greater preparedness for extreme rainfall but not for tropical cyclones, highlighting that gendered risk appraisal may vary by hazard type and warrants targeted communication interventions. Women may develop a more nuanced understanding of specific risks associated with extreme rainfalls due to personal experiences in their communities, such as flooding or managing children during hazardous weather, leading to a higher motivation for preparedness [44]. Research indicates that past experiences related to flooding can create stronger emotional connections, driving individuals to adopt preparedness measures [45]. This emotional resilience supports proactive preparedness for extreme rainfalls and flooding. In contrast, super tropical cyclones in Hong Kong elicit broad community mobilization, reinforced by government-led warnings and standardized, societal preparedness routines [23]. In these situations, approaching risk is highly visible, information is disseminated uniformly, and mass preparations are encouraged. Such collective framing leads both men and women to similar behaviors, as gender differences are mitigated by the shared societal response mechanisms. Last but not least, the lack of significant effects for age, education, and health-related vulnerability further highlights the nuanced, context-dependent drivers of disaster preparedness, breaking with common risk management assumptions.

Most strikingly, information-seeking via traditional information channels emerges as the strongest shared determinant of disaster preparedness on both super tropical cyclones and extreme rainfall events, outstripping even socioeconomic predictors. The significant positive association between traditional information channels and disaster preparedness, contrasted with the non-significant role of new channels, corroborates earlier findings from Hong Kong [23]. This consistent pattern, observed across a five-year interval that included a global pandemic and rapid digitalization, underscores the enduring authority and reliability vested in official, established media during super tropical cyclones and extreme rainfall events.

This also supports calls for strengthening trusted communication infrastructures and adapting real-time messaging to hazard characteristics. The persistently higher use of new media during tropical cyclones suggests that digital integration is increasingly relevant but must be matched by reliable, timely alerts for sudden-onset events like extreme rainfall.

These findings necessitate a strategic shift away from uniform public education and toward hazard-specific, population-tailored initiatives. Key recommendations include: (1) designing anticipation-driven interventions for tropical cyclones, leveraging structured warning periods for mass mobilization; (2) creating ultra-fast, localized awareness and response mechanisms for extreme rainfall, focusing on clear triggers, actionable guidance, and inclusive outreach; (3) targeting high-income and long-term resident populations with customized campaigns, counteracting complacency or false security typically associated with resource abundance; (4) empowering adaptive groups such as non-local-born and lower-income populations as resilience champions, drawing on their demonstrated engagement; (5) intensifying risk information dissemination across both traditional and digital platforms, with increased clarity on the distinct nature and requirements of super tropical cyclones versus rainfall event preparedness.

Although this study provides valuable evidence on hazard-specific preparedness in Hong Kong, the issue of sampling bias presents a significant limitation that warrants further consideration. The substantial overrepresentation of female (64.7%) and highly educated (post-secondary degree, 86.3%) respondents may have significantly shaped the observed preparedness patterns, as both females and higher education levels are repeatedly shown to correlate with higher disaster risk perception and adaptive behaviors [44,46]. This sampling structure could therefore lead to inflated estimations of public readiness, particularly for hazards where women and highly educated individuals historically engage more readily in preparedness activities [47]. Moreover, all ten preparedness measures are commonly taken for both typhoons and extreme rainfall in Hong Kong, as extreme rainfall events often coincide with strong winds. However, certain actions, such as checking windows and securing outdoor objects, may be more typhoon-specific, indicating the need for more hazard-specific measures in future research.

Furthermore, another major limitation of this study is its reliance on self-reported preparedness behaviors, which, despite being collected within two months of the hazard events, remain susceptible to recall bias and social desirability effects that may inflate or deflate actual readiness upon the specific hazards [48]. This may also influence the observed preparedness paradox. Additionally, the cross-sectional design precludes causal inference and limits understanding of temporal relationships between risk perception and preparedness [49]. Future research should prioritize longitudinal and experimental approaches, expand sampling diversity, and integrate psychometric models to unpack the interplay between risk perception, social norms, and adaptive action.