The research sample consisted of 172 higher education teachers from Romania. Participants were recruited using non-probabilistic sampling strategies, specifically convenience sampling and snowball sampling.

The target population of the present study consists of higher education teachers working in academic institutions located in regions exposed to recurrent natural hazards. This professional group was selected due to its relatively homogeneous educational background, high cognitive demands, and institutional responsibilities related to risk awareness, information dissemination, and adaptive functioning in crisis contexts. Focusing on higher education teachers allows for the examination of cognitive and psychological resilience mechanisms within a population characterized by advanced cognitive resources and structured professional roles.

Given the non-probabilistic sampling strategy employed, the findings are not intended to be generalized to the broader population or to all occupational groups. Instead, the results are analytically generalizable to similar professional populations with comparable cognitive profiles and exposure contexts. This approach supports theory-driven inference regarding psychological mechanisms of resilience, while acknowledging limitations in statistical representativeness and external validity.

Non-probabilistic sampling strategies, including convenience and snowball sampling, are widely used in exploratory and applied psychological research when access to a clearly delimited population is limited and the primary research objective is to examine underlying psychological mechanisms rather than to generate population-level prevalence estimates (Etikan et al., 2015; Bryman, 2016; Creswell and Plano Clark, 2018). Such approaches are considered methodologically appropriate for theory-driven studies focusing on mediation and moderation processes within relatively homogeneous professional groups, where analytical generalization and the investigation of process-level relationships are prioritized over statistical representativeness.

Convenience sampling allowed the inclusion of participants who were readily accessible and willing to take part in the study, while snowball sampling facilitated the recruitment of additional respondents belonging to the same professional population. This combined strategy ensured adequate sample size and internal coherence, while maintaining relevance to the targeted cognitive and experiential profile. Although non-probabilistic sampling limits statistical generalizability, it is considered appropriate for theory-driven studies investigating mediation and moderation processes within relatively homogeneous groups (Hayes, 2022; Pallant, 2020).

For evaluating the constructs included in the research hypotheses, the Resilience to Natural Disasters Questionnaire was used, designed to assess risk perception and individual resilience in disaster contexts, specifically fires and floods. The questionnaire comprises two context-specific versions (fire and flood) that share an identical structural framework, while the formulation of the items has been tailored to the specific context of each hazard. The number of items and response formats are identical across versions, ensuring measurement comparability.

Risk perception was measured using 10–11 items that captured awareness, cognitive and emotional assessments, and preventive behaviors, with a 5-step Likert scale (1 = “never”, 5 = “always”). Individual resilience was assessed through four components: psychological, cognitive, adaptive, and institutional, each measured by four items on the same scale. In addition, procedural knowledge related to fire or flood response was assessed using 10 dichotomous items (yes/no), reflecting familiarity with hazard-specific operational steps.

Exploratory factor analysis (EFA) (Table 1) confirmed a consistent and reliable factor structure for all six scales of the instrument, with KMO values ranging from 0.720 to 0.813 and significant Bartlett tests (p < 0.001), indicating the adequacy of the data for dimensional reduction.

The number of factors extracted ranged from one to four, reflecting the theoretical complexity specific to each construct. The total variance explained, ranging from 42.29% to 62.30%, falls within the accepted psychometric limits for multidimensional instruments. Cronbach's α coefficient values, ranging from 0.791 to 0.899, indicate high internal consistency and robust conceptual stability for each scale. Overall, these results support the construct validity and internal reliability of the instrument for research on risk perception, procedural knowledge, and psychological resilience in emergency contexts (Field, 2024).

Following these psychometric evaluations, the instrument was considered appropriate for application in the main stage of the study, with the corresponding results presented in the next section.

The data collection was conducted online by distributing the „Resilience to natural disasters„ questionnaire in electronic format to teachers in higher education institutions. Participants were informed about the overall purpose of the study, the anonymity and confidentiality of their responses, and were asked to express informed consent before completing the questionnaire.

The average completion time was about 15–20 min. The answers were collected in a single session, with no time limit imposed, and participants were free to pause at any time. No material rewards were offered for participation.

The data was centralized and coded in SPSS (version 28) (BM Corp. IBM SPSS Statistics for Windows (Version 28.0), 2025), being used exclusively for scientific purposes.

To test the proposed hypotheses, a series of conditional process models were specified using the PROCESS macro for SPSS (v4.2; Hayes, 2022). PROCESS Model 4 was employed to examine simple mediation effects, testing whether cognitive resilience functions as an explanatory mechanism linking predictor variables to psychological resilience. PROCESS Model 7 was used to test moderated mediation, allowing the assessment of whether the indirect effects of risk perception vary as a function of contextual moderators, namely prior disaster experience and hazard type. In addition, PROCESS Model 1 was applied to examine simple moderation effects, testing whether prior disaster experience conditions the direct relationship between risk perception and adaptive resilience. Indirect effects were estimated using bias-corrected bootstrapping procedures with 5,000 resamples, providing robust confidence intervals. Together, these models provide an integrated analytical framework suitable for theory-driven research that simultaneously examines direct, indirect, and conditional effects.

Hypothesis 1 (H1). The effect of previous experience (X) on psychological resilience (Y) is transmitted indirectly through cognitive resilience (M), and the intensity of this indirect effect varies depending on the type of hazard (W: flood vs. fire). In other words, the mechanism of mediation through cognitive resilience is contingent upon the nature of the emergency, indicating a moderate mediation effect.

The hypothesis assumes a moderated mediation model (PROCESS Model 7), in which previous experience (X, coded 0 = without, 1 = with) influences psychological resilience (Y) through cognitive resilience (M), and this indirect effect is examined as a function of hazard type (W: 0 = fire, 1 = flood). Accordingly, a moderated mediation analysis (Model 7; Hayes, 2018) was conducted to test the mediating role of cognitive resilience in the relationship between prior disaster experience and psychological resilience, while assessing whether this indirect effect varies across hazard contexts (fire vs. flood) (Table 2).

The results show that previous experience significantly predicts cognitive resilience [b = 0.721, SE = 0.122, t = 5.892, p < 0.001, 95% CI (0.480, 0.963)], indicating that individuals with prior disaster exposure tend to mobilize higher levels of cognitive resources for emergency management. In addition, hazard type has a significant direct effect on cognitive resilience (b = 3.047, p < 0.001), suggesting that flood-related scenarios are associated with stronger cognitive activation compared to fire-related contexts. However, the interaction between experience and hazard type is not statistically significant (b = −0.064, p = 0.794), indicating that the relationship between prior experience and cognitive resilience does not vary as a function of hazard type.

On the outcome pathway (Y), previous experience exhibits a robust positive direct effect on psychological resilience (b = 0.814, p < 0.001), confirming its protective role. Cognitive resilience, in turn, shows a significant negative association with psychological resilience (b = −0.156, p = 0.025), suggesting that heightened cognitive engagement in coping strategies may be accompanied by a temporary reduction in perceived emotional resources, potentially reflecting cognitive overload or excessive preoccupation under stress.

Analysis of conditional indirect effects indicates that prior experience influences psychological resilience through cognitive resilience in both the fire scenario [b = −0.118, 95% CI (−0.253, −0.001)] and the flood scenario [b = −0.108, 95% CI (−0.270, 0.001)]. However, the index of moderated mediation is not statistically significant [Index = 0.010, 95% CI (−0.093, 0.110)], indicating that the magnitude of the indirect effect does not differ significantly across hazard types.

Taken together, these findings support the mediating role of cognitive resilience in the relationship between prior disaster experience and psychological resilience, while providing no empirical evidence that this mediation mechanism is contingent upon hazard type.

Hypothesis 2 (H2). The relationship between risk perception (X) and psychological resilience (Y) is mediated by cognitive resilience (M), and this mediation path is moderated by the type of emergency (W: floods,/fire).

The detailed results of the moderate mediation analysis are summarized in Tables 3–5, and the integrated interpretation of the model is presented below.

Moderate Mediation Analysis—PROCESS Model 7 (Hayes, 2022) (Table 3) confirms hypothesis H2, which posits that the relationship between risk perception (X) and psychological resilience (Y) is mediated by cognitive resilience (M), and the intensity of this mediation pathway is contingent upon the type of emergency (W).

The results show that risk perception has a significant positive effect on cognitive resilience [B = 0.0994, SE = 0.010, t = 10.17, p < 0.001, 95% CI (0.080, 0.119)], suggesting that a more acute perception of risk activates cognitive coping mechanisms such as planning, anticipating, and rationalizing the crisis. Also, the significant interaction between risk perception and the type of situation (X × W) [B = 0.1097, SE = 0.020, t = 5.61, p < 0.001, 95% CI (0.071, 0.148)] indicates that the relationship between risk perception and cognitive resilience is influenced by the nature of the threatening event, being stronger in conditions perceived as severe or unpredictable (such as fires).

In the final model, risk perception has a significant positive effect on psychological resilience [B = 0.1687, SE = 0.011, t = 14.97, p < 0.001, 95% CI (0.146, 0.191)], while cognitive resilience has a significant adverse effect on the same variable [B = −0.3950, SE = 0.052, t = −7.58, p < 0.001, 95% CI (−0.498, −0.292)]. This combination suggests the existence of a compensatory mechanism between the cognitive and affective components of resilience, in which an overactivation of cognitive control can temporarily diminish the emotional resources involved in psychological adaptation.

The results presented in Table 4 indicate that both indirect effects are negative and statistically significant, with their magnitude varying depending on the type of event. In the case of floods (W = 0), the indirect effect is −0.0176 [Boot SE = 0.0067, 95% CI: (−0.0313, −0.0050)], indicating a weak but consistent mediation of cognitive resilience between risk perception and psychological resilience. This suggests that in more predictable and controllable situations, the cognitive component of resilience plays a moderating role, facilitating risk processing without significantly affecting overall psychological balance.

In contrast, in the case of fires (W = 1), the indirect effect is −0.0609 [Boot SE = 0.0095, 95% CI (−0.0810, −0.0435)], of much higher magnitude, indicating an intensified cognitive mediation. This result highlights the fact that, in situations perceived as more severe, unpredictable, and with a low degree of personal control, individuals activate more complex cognitive resources to manage stress and uncertainty, albeit at a temporary higher psychological cost, as reflected in the decrease in the general level of psychological resilience.

The clear difference between the two conditions validates the existence of a mediation conditioned by the situational context, confirming that cognitive resilience more strongly mediates the relationship between risk perception and psychological resilience in situations perceived as threatening and unstable. This result is aligned with recent studies on contextual models of resilience (Bonanno, 2021; Kalisch et al., 2019; Chen et al., 2024) who claim that cognitive adaptation processes are more active and demanding under conditions of acute stress, influencing emotional balance and overall psychological performance.

In theoretical terms, this result suggests that cognitive resilience functions as a compensatory mechanism, allowing functionality to be maintained in the face of extreme threats, but can result in a temporary decrease in affective resources. On the other hand, in more controllable situations, cognitive involvement is effective and proportionate, contributing to a balanced adaptation.

Therefore, the analysis confirms the existence of a robust conditional indirect effect, which validates the moderate mediation model proposed in the H2 hypothesis. Risk perception influences psychological resilience in part through cognitive resilience, and the magnitude of this relationship depends on the nature of the emergency.

This finding reinforces the idea that psychological resilience is a context-sensitive process, shaped by the interaction between cognitive assessments and the degree of unpredictability of the environment, while emphasizing the importance of calibrating post-disaster psychological interventions according to the specifics of the situation (e.g., stimulating cognitive self-regulation in unpredictable events and training emotional strategies in controllable ones).

As highlighted in Table 5, the obtained moderate mediation index [Index = −0.0433, Boot SE = 0.0100, 95% CI (−0.0651, −0.0254)] is negative and significant, which indicates the presence of genuine moderate mediation, according to the criteria proposed by Hayes (2022).

The results confirm that the magnitude of the indirect effect of risk perception on psychological resilience, mediated by cognitive resilience, varies significantly depending on the nature of the emergency. More specifically, the data suggest that in conditions perceived to be more severe and unpredictable, such as fires, the mediated effect is more pronounced, reflecting more intense cognitive engagement (Bakic and Ajdukovic, 2021) in adaptive regulation processes (Mouguiama-Daouda et al., 2026). On the other hand, in more controllable situations, such as floods, cognitive mediation has a lower intensity, which indicates that psychological adaptation is supported to a greater extent by direct affective and behavioral mechanisms (Wu et al., 2025).

Therefore, the results in Table 5 reinforce the empirical evidence supporting the validity of the hypothetical H2 model and confirm that the interaction between cognitive and contextual factors determines how risk perception influences psychological adaptation. The results support the interpretation that cognitive resilience acts as a dynamic, flexible, and context-dependent mechanism, whose effectiveness varies according to the severity and unpredictability of the threatening situation (Lomeli-Rodriguez et al., 2025).

This result is consistent with recent literature approaching resilience from a contextual and multidimensional perspective (Bonanno, 2021; Southwick and Charney, 2022; Kalisch et al., 2019). According to these approaches, resilience is not a fixed feature, but an adaptive process that involves the complex interaction between cognitive resources, emotional state, and situational threat assessment.

From an applicative perspective, the results highlighted in Table 4 emphasize the importance of differentiating post-disaster psychological interventions according to the specific context. In situations characterized by a high degree of uncertainty and danger, such as fires, interventions should aim to strengthen cognitive control and self-regulation strategies to reduce cognitive overload and facilitate emotional recovery processes. Conversely, in more predictable events, such as floods, interventions can emphasize developing emotional strategies, regaining a sense of control, and restoring emotional balance. Thus, the significant moderate mediation index confirms that psychological resilience does not depend exclusively on the perception of risk, but also on the way in which cognitive mechanisms are activated and calibrated in relation to situational particularities.

Overall, these results offer an integrative perspective on the process of coping with stress and how individuals modulate their cognitive and emotional responses to manage critical situations effectively. Additional results confirm the dynamic, contextual, and multidimensional nature of the resilience process, demonstrating that the impact of risk perception on psychological resilience is not linear, but rather depends on the interaction between cognitive regulatory processes and the situational specificity of the threat.

From an applicative perspective, they provide an empirical basis for the development of differentiated psychological interventions, adapted to the type of disaster, which strengthen cognitive skills in situations of high uncertainty and favor emotional regulation in the most predictable ones.

Hypothesis 3 (H3). Previous experience with natural disasters (floods, fires) predicts the level of individual resilience (psychological, cognitive, adaptive, institutional), so people who have gone through at least one event show different levels of resilience compared to people without similar experiences.

The Kolmogorov–Smirnov and Shapiro–Wilk tests were applied separately to participants who had and those who had not experienced natural disasters.

As shown in Table 6, all four dimensions of resilience—psychological, cognitive, adaptive, and institutional—show significant deviations from the normal distribution (p < 0.05), indicating nonparametric distributions of the data.

In the inexperienced group, the Shapiro–Wilk test showed significant values for psychological resilience (W = 0.855, p = 0.006), cognitive resilience (W = 0.892, p = 0.029), adaptive resilience (W = 0.761, p = 0.000), and institutional resilience (W = 0.898, p = 0.037).

Also, for the experienced group, all four dimensions show significant deviations from normal (all p < 0.05).

The results confirm that the assumption of normality is not met; therefore, the analyzed distributions are nonparametric. To test the H3 hypothesis—regarding the differences in the level of resilience according to previous experience with natural disasters (Table 7)—the Mann–Whitney U nonparametric test was used, suitable for comparing two independent groups under the conditions of a nonparametric distribution of data (Field, 2024; Pallant, 2020).

The interpretation of the results in Table 7 highlights statistically significant differences between people with and without previous experience of natural disasters in three of the four dimensions of resilience.

Analysis by the Mann–Whitney U nonparametric test showed that people who have experienced at least one flood, fire, or other critical event have significantly higher levels of psychological resilience (U = 778.00, Z = −3.57, p < 0.001), cognitive resilience (U = 794.00, Z = −3.49, p < 0.001), and adaptive resilience (U = 360.00, Z = −5.58, p < 0.001) compared to participants without similar experiences.

The negative values of the Z statistics indicate that the average ranks are higher among the experienced group, suggesting a stronger ability to manage stress and adapt to challenging contexts.

For the institutional resilience dimension, the observed difference between groups did not reach the threshold of statistical significance (U = 1158.00, Z = −1.75, p = 0.081), although the trend of values indicates a similar direction to that of the other dimensions.

This result suggests that, unlike personal resilience (psychological, cognitive, adaptive), which is directly strengthened by the experience of coping with extreme situations, institutional resilience may depend more on the organizational framework, community support, and external structures involved in disaster management.

Overall, the results confirm the H3 hypothesis and support the idea that previous experience with natural disasters contributes to the development of internal coping resources, cognitive flexibility, and emotional stability.

Participants who have experienced such events appear to exhibit a greater capacity for psychological regulation and a more effective cognitive attitude in the face of stress, which confirms the importance of controlled exposure to complex situations in the development of individual resilience.

Hypothesis 4 (H4). Previous experience with natural disasters (floods, fires) moderates the relationship between risk perception and adaptive resilience, so people with direct experiences more effectively transform risk perception into adaptive behaviors.

Moderation Analysis (Model 1, PROCESS v4.2) (Hayes, 2022) (Table 8) was used to test the hypothesis that previous experience with natural disasters (such as floods and fires) modulates the relationship between risk perception and adaptive resilience.

The results indicate that the overall model is statistically significant, F(3, 168) = 66.61, p < 0.001, accounting for 54.3% of the variance in adaptive resilience (R² = 0.5433). The high value of the multiple correlation coefficient (R = 0.7371) indicates that the set of variables included in the model consistently contributes to the prediction of the level of adaptation.

The primary effect of risk perception on adaptive resilience is positive and statistically significant [b = 0.1066, p = 0.014, 95% CI (0.0219, 0.1913)], suggesting that as individuals perceive the risk associated with natural disasters more intensely, they exhibit higher levels of psychological and behavioral adaptive capacity. In other words, risk perception serves as a cognitive mechanism for activating adaptive strategies, thereby strengthening the capacity for flexible and solution-oriented responses.

The direct effect of previous experience on resilience is not significant [b = −1.6120, p = 0.444, 95% CI (−5.7586, 2.5346)], indicating that mere exposure to extreme events does not automatically cause an increase or decrease in adaptive resilience. However, the interaction term between risk perception and experience [b = 0.0774, p = 0.093, 95% CI (−0.0130, 0.1678)] is marginally significant, and ΔR² = 0.0078 shows a modest but theoretically relevant additional contribution of the interaction to the model. This suggests the existence of a moderating trend, according to which the effect of risk perception on adaptive resilience is stronger for people who have had direct experiences with natural disasters, compared to those who have not experienced such events.

In psychological terms, the results indicate that previous experience acts as a factor in amplifying the relationship between risk perception and adaptive behaviors. Individuals who have been exposed to crises seem to transform cognitive risk assessments more effectively into adaptive responses, using previous experiences as learning and adjustment resources. In contrast, people without similar experiences tend to show a weaker relationship between risk perception and resilience, due to the lack of relevant behavioral or emotional landmarks.

The H4 hypothesis is partially confirmed. Previous experience with natural disasters does not have a direct effect on adaptive resilience but influences the direction and intensity of the relationship between risk perception and psychological adaptation. This finding supports theoretical models of experiential resilience (Bonanno, 2021; Norris et al., 2008) according to which exposure to stressful events can strengthen cognitive and behavioral coping mechanisms when accompanied by a realistic and constructive perception of risk.

Hypothesis 5 (H5). Previous experience with natural disasters is positively associated with psychological resilience through cognitive resilience, indicating a mediating role of cognitive resilience in this relationship.

Simple Mediation Analysis (Model 4, PROCESS v4.2) (Hayes, 2022) (Table 9) was used to examine whether cognitive resilience explains the mechanism by which previous experience with natural disasters influences psychological resilience. The overall model is significant, F(2, 169) = 41.05, p < 0.001, accounting for 33.7% of the variance in psychological resilience (R² = 0.3375). This value indicates a moderate-to-high association between prior experience and cognitive and psychological resilience.

The results show that previous experience with extreme events (floods, fires) has a positive and significant effect on cognitive resilience [b = 2.32, p < 0.001, 95% CI (1.03, 3.60)]. This suggests that people who have experienced natural disasters develop more intense cognitive coping mechanisms, such as flexible thinking, positive reappraisal, and solution-oriented planning. The pathway between cognitive resilience and psychological resilience is also significant and positive [b = 0.69, p < 0.001, 95% CI (0.53, 0.86)], confirming that adaptive cognitive functioning supports emotional balance and the ability to self-regulate in the face of stress.

The direct effect of experience on psychological resilience remains significant [b = 1.30, p = 0.019, 95% CI (0.21, 2.38)], indicating that experience has an independent impact on emotional balance, but to a lesser extent than the total effect. The indirect effect (a × b) of experience on psychological resilience, via cognitive resilience, is significant 9b = 1.61, BootSE = 0.50, 95% CI (0.67, 2.66)]. Since the bootstrap confidence interval does not include a value of zero, a mediation effect is confirmed. Thus, previous experience influences psychological resilience partly by mediating cognitive resilience, which indicates partial mediation.

These results confirm the H5 hypothesis and support the theoretical model, which posits that difficult experiences can serve as factors for learning and strengthening cognitive coping strategies. Cognitive resilience serves as an explanatory mechanism between experience and psychological balance, meaning that people exposed to traumatic events not only develop emotional reactions of resistance but also cognitive structures that facilitate the reinterpretation of stressful situations and maintain a sense of internal control.

From a theoretical perspective, these results support the processual models of resilience (Bonanno, 2021; Norris et al., 2008; Tugade and Fredrickson, 2004). According to this, psychological adaptation is a dynamic process in which direct experience activates cognitive resources that modulate the emotional response. In this context, cognitive resilience acts as an “adaptive filter”, transforming the impact of negative experiences into opportunities for personal growth and strengthening emotional balance.

Previous experience with natural disasters contributes to psychological resilience, primarily through the development and activation of cognitive mechanisms. The results underscore the importance of incorporating cognitive dimensions into post-disaster intervention programs to facilitate adaptive reinterpretation processes and promote long-term mental health.