This study recruited 959 Chinese university students (69.7% female; M = 19.60, SD = 1.40) from southeastern and central China through convenience sampling, who were allocated to four subsamples for item analysis, exploratory factor analysis, confirmatory factor analysis, and test–retest reliability assessment.

Three measures were included in this validation study: The Hogg Climate Anxiety Scale (HCAS; Hogg et al., 2024), Climate Change Anxiety Scale (CCAS; Clayton & Karazsia, 2020) and Short State Anxiety Inventory (SSAI; Spielberger et al., 1983; Marteau & Bekker, 1992). The measures were translated into Chinese using back-translation by a professional English instructor, following standard cross-cultural adaptation procedures. Discrepancies were resolved under the guidance of a senior psychometrics expert and an expert panel reviewed the items for semantic accuracy and cultural relevance. A pilot test with 30 students confirmed clarity before minor revisions were made. These are described next.

The research protocol was reviewed and approved by the Ethics Committee of Hanshan Normal University. All participation was voluntary, with informed consent obtained from each student. The students were in four samples for each of the major validation tasks.

Sample 1 was used for item analysis. Using convenience sampling, 220 responses were collected online via the Wenjuanxing platform, of which 202 were valid (valid response rate: 91.8%). The sample included 53 males and 149 females, with a mean age of 19.40 years (SD = 1.51).

Sample 2 was used for exploratory factor analysis (EFA). Similarly recruited through convenience sampling and the Wenjuanxing platform, this sample comprised 310 valid responses out of 335 collected (valid response rate: 92.5%). Participants included 84 males and 226 females, with a mean age of 19.97 years (SD = 1.51).

Sample 3 was used for confirmatory factor analysis (CFA), criterion-related validity, and reliability analysis. Data were obtained via both online administration (Wenjuanxing) and on-site surveys. Of the 488 questionnaires distributed, 447 were valid (valid response rate: 91.6%). This sample consisted of 154 males and 293 females, with a mean age of 19.34 years (SD = 1.19).

Sample 4 was used to assess test–retest reliability. One hundred students were randomly selected from Sample 2 and were retested one month after the initial survey, resulting in 77 matched pairs. This subsample included 18 males and 59 females, with a mean age of 19.09 years (SD = 0.81).

As a token of appreciation, participants received a small gift upon completing the questionnaires.

Statistical analyses were conducted using SPSS 27.0, Mplus 8.3, Amos 28, and R 4.4.3. Specifically, Item analysis was performed on Sample 1 in SPSS 27.0. Exploratory Factor Analysis (EFA) was conducted on Sample 2 using principal feature extraction with varimax rotation to examine the scale’s factor structure.

Confirmatory Factor Analysis (CFA) and Validity Assessment: CFA was performed on Sample 3 in both Mplus 8.3 and Amos 28 to evaluate model fit, convergent validity, and discriminant validity. Reliability was assessed via Cronbach’s α, McDonald’s ω, split-half reliability, and test–retest reliability; criterion-related validity was also examined.

Network Analysis: Network modeling of climate anxiety items was conducted in R. Each item was represented as a node, with statistical direct associations depicted as edges. Centrality indices and network stability were calculated. The Network Comparison Test assessed structural differences across subgroups.

During manuscript preparation, ChatGPT-4 (OpenAI, San Francisco, CA, USA) was used solely for language polishing and grammar checks to improve readability. All AI-generated suggestions were critically reviewed and finalized by the authors. The research design, methodology, data analysis, and conclusions were developed independently without AI assistance. No confidential or personally identifiable data were entered into the tool.

The Sample 1 item analysis results (see Table 1) indicated that the correlation coefficients between each item of the HCAS and the total scale score ranged from 0.567 to 0.787 (p < 0.001). The extreme group method was employed, whereby participants were ranked based on their total HCAS scores, with the top 27% and bottom 27% forming the high-scoring and low-scoring groups, respectively. Independent samples t-tests were used to examine the differences between these high and low groups for each item; all p-values were less than 0.001. Furthermore, the Cronbach’s α coefficient for the total scale did not increase when any individual item was removed. These findings suggest that all items of the HCAS possess good discriminatory power.

Reliability analyses conducted on Sample 3 demonstrated strong internal consistency for the HCAS. The total scale yielded a McDonald’s ω of 0.936, and the composite reliability values for the four subscales were 0.910, 0.820, 0.854, and 0.892, respectively. Test–retest reliability was 0.716, while split-half reliability reached 0.869. Collectively, these findings confirm that the HCAS possesses excellent internal consistency and satisfactory temporal stability.

Measurement invariance across gender (male = 1, female = 2) was evaluated using Sample 3. As summarized in Table 6, all models—including configural, metric (weak), scalar (strong), and strict invariance—exhibited good fit indices (ΔCFI < 0.01, ΔTLI < 0.05, ΔRMSEA < 0.01), supporting strong invariance across gender groups.

To further investigate potential gender differences, a Network Comparison Test was conducted to assess invariance in network structure between male and female participants (van Borkulo et al., 2022). As shown in Figs. 4 and 5, the results indicated no significant overall structural differences (p = 0.366). Global network strength was similar for males (6.020) and females (6.229), with a non-significant difference of 0.209. Only two edges (a3–c1 and c1–d3) differed significantly between genders, these involved affective symptoms, behavioral symptoms, and personal impact anxiety, but the effect sizes were negligible. Such minor differences may be attributable to gender-related socialization in emotional or behavioral expression but did not indicate any systematic structural bias. Taken together, the results of measurement invariance and network comparison analyses support the structural equivalence and fairness of the HCAS across gender groups, reinforcing its validity as a reliable and unbiased tool for assessing climate anxiety among university students.

Item analysis revealed moderate to high item-total correlations, with all items displaying significant discrimination between high and low scorers. The HCAS demonstrated excellent internal consistency across the total scale and its four subscales, as well as satisfactory test–retest reliability, indicating stable temporal properties.

Both EFA and CFA confirmed a stable four-factor structure, with item-factor assignments mirroring the original instrument, thereby supporting cross-cultural construct equivalence. Indices of convergent and discriminant validity all met or exceeded recommended thresholds, further reinforcing the scale’s construct validity.

In terms of criterion validity, the HCAS total score was significantly correlated with both the CCAS and the “Presence of Anxiety” subscale evidencing good external validity. Moreover, regression analyses showed that the HCAS had stronger predictive power for general anxiety than the CCAS, suggesting that climate anxiety, as assessed by the HCAS, constitutes a related but distinct construct from generalized anxiety. This aligns with Cameron and Kagee’s (2025) assertion that climate anxiety is closely linked to, yet distinguishable from, traditional anxiety and depression.

Multi-group confirmatory factor analysis demonstrated that the HCAS achieves configural, metric, scalar, and strict invariance across gender, indicating equivalent factor structure, loadings, and intercepts for both male and female respondents. This supports the instrument’s suitability for valid cross-group comparisons.

Additionally, network analysis revealed no significant differences in symptom interconnectivity or global network strength between genders. The convergence of evidence from both factor and network models underscores the fairness, reliability, and generalizability of the HCAS for use in diverse populations.

Network analysis supplemented traditional psychometric evaluation by visually and statistically validating the HCAS’s internal structure. Items clustered tightly within dimensions, confirming the coherence of the four-factor model. Bootstrapped analyses confirmed the robustness of the network structure across different sample sizes. The absence of gender-based structural differences further highlights the scale’s stability and broad applicability. Importantly, identifying central symptoms within the network provides new directions for the development of targeted psychological interventions for climate anxiety.

Centrality metrics identified a3 (“I feel particularly worried about climate change”), a2 (“When I think about climate change, I can’t stop or control my worrying”), and c3 (“Thinking about climate change makes it difficult for me to fully focus on my work or study”) as the most influential nodes, these results suggest that interventions could focus on managing excessive worry and improving attentional control through cognitive-behavioral and mindfulness-based strategies, thereby reducing climate anxiety and supporting daily functioning.

Compared to existing tools such as the CCAS, the HCAS demonstrates clear strengths in structural clarity, cultural adaptability, and predictive validity. Its multidimensional approach—encompassing affective symptoms, rumination, behavioral symptoms, and personal impact anxiety—offers a nuanced understanding of the cognitive-behavioral mechanisms underlying climate anxiety. Network analysis reinforced these domains and highlighted central symptoms consistent with Freeston et al.’s (2020) transdiagnostic model of intolerance of uncertainty, which emphasizes the interplay between cognitive disruption and functional impairment. The rigorous translation and back-translation process ensured semantic and conceptual equivalence, while measurement invariance testing confirmed fairness across gender groups—crucial for promoting psychometric equity in non-Western contexts.

Notably, the HCAS outperformed the CCAS in predicting general anxiety. While the HCAS significantly predicted the presence of anxiety, the CCAS did not, highlighting the unique utility of the HCAS in both research and applied settings. This scale holds promise for integration into university mental health early warning systems, for use in climate resilience education, and as a guide for targeted interventions.

Despite the strong psychometric performance of the HCAS, several limitations should be acknowledged. First, the sample consisted primarily of university students from southeastern and central China, limiting generalizability. Future research should include more diverse populations, such as ethnic minorities, rural residents, and individuals from other regions. Second, the cross-sectional design precludes analysis of temporal dynamics or causality. Longitudinal studies are warranted to examine changes in climate anxiety around specific events (e.g., typhoons, heatwaves) and to clarify causal pathways. Third, although the HCAS is effective in identifying core symptoms, its application in intervention studies remains limited. Future research should explore its use in developing and evaluating interventions that foster resilience, adaptive coping, and proactive engagement with climate challenges. Fourth, this study did not assess participants’ direct exposure to extreme climate events, which may influence climate anxiety. Without such data, it is unclear whether individual differences reflect personal experiences with severe weather or broader climate concerns. Future research should include measures of extreme weather exposure to better understand its role in climate anxiety.