Toward a unified understanding of climate anxiety: Examining measures of climate and eco-anxiety

PLOS One | https://doi.org/10.1371/journal.pone.0339562 December 29, 2025 8 / 19 For covariates—including risk perception of climate change–related health risks, self-efficacy to show individual emission-reducing behavior, readiness for climate action, and neuroticism—we used established models from the litera- ture. All correlations between constructs were estimated at the latent factor level, meaning that they were disattenuated for measurement error [26]. Latent factors were identified by standardizing them (i.e., fixing their variance to 1). Models were estimated using robust maximum likelihood (MLR) estimation, and statistical significance was assumed at ɑ = .05. Analyses were conducted using R (v4.3.1; R Core Team 2023) employing the packages lavaan (v0.6.16 [54]), semTools (v0.5.6 [55]), ggridges (v0.5.4 [56]), viridis (v0.4.2 [57]), ggplot2 (v3.4.3 [58]), maditr (v0.8.3 [59]), and dplyr (v1.1.3 [60]). Results First, we calculated descriptive measures for the CAS and HEAS. For both measures, items were not normally distributed, showing pronounced floor effects (https://osf.io/wvbd3/, S1 and S2 Figs). Measurement models of climate and eco-anxiety Using confirmatory factor analyses, we tested several models frequently discussed in the literature for both the CAS and HEAS. The fit indices for these models are presented in Tables 1 and 2. Additionally, in Figs 1 and 2, scheme plots illus- trate the tested models for CAS and HEAS, respectively, while Fig 3 presents scheme plots for joint models incorporating both measures.

CAS measurement model. For the CAS, we tested a one-factor solution as a simple g-factor model, as well as two two-factor solutions proposed by Clayton and Karaszia [6] and Wullenkord and colleagues [9]. The model by Clayton and Karaszia [6] distinguishes between cognitive-emotional impairment (items 1–8) and functional impairment (items 9–13). In contrast, Wullenkord and colleagues [9] proposed a structure based on behavioral symptoms (items 1–4, 7, 9, 11–13) and Table 1. Overview of measurement models for the Climate Anxiety Scale (CAS). Model X2(df); p CFI TLI RMSEA SRMR X2-difference test (df) to g-factor model; p CAS: g-factor χ² (65) = 538; p < .001 .962 .954 .054 .025 CAS: two factors; Clayton and Karaszia (2020) χ² (64) = 517; p < .001 .963 .955 .053 .025 χ² (1) = 9; p = .0021 CAS: two factors; Wullenkord et al. (2021) χ² (53) = 401; p < .001 .971 .964 .050 .022 χ² (12) = 78; p < .0011 Note. X2-difference tests were calculated only for nested models. 1The X2-difference test included a correction applied to compensate for possible distor- tions caused by non-normal data or complex model structures, resulting in different results from the simple difference between both models’ X2. https://doi.org/10.1371/journal.pone.0339562.t001 Table 2. Overview of measurement models for the Hogg Eco-Anxiety Scale (HEAS). Model X2(df); p CFI TLI RMSEA SRMR X2-difference test (df) to g-factor model; p HEAS: g-factor χ² (65) = 682; p < .001 .949 .938 .060 .035 HEAS: higher-order model χ² (61) = 286; p < .001 .985 .981 .033 .023 χ² (4) = 119;

p < .0011 HEAS: four factors χ² (59) = 213; p < .001 .992 .989 .025 .017 χ² (6) = 169; p < .0011 Note. X2-difference tests were calculated only for nested models. 1The X2-difference test included a correction applied to compensate for possible distor- tions caused by non-normal data or complex model structures, resulting in different results from the simple difference between both models’ X2. https://doi.org/10.1371/journal.pone.0339562.t002