1. Introduction
Storms, droughts, heatwaves, and other severe weather events (SWEs) are well-recognized effects of anthropogenic climate change (Wuebbles et al. 2014; Wuebbles 2016). In forests, damage caused by SWEs can reduce the value of forest stands and compromise key forest ecosystem services such as carbon sequestration, habitat provision, and water filtration (Seidl et al. 2017; Millar and Stephenson 2015; Sánchez et al. 2021). People who manage their own forestland are in a unique position to observe these threats and take steps to reduce their impacts. For example, forest owners can thin forest stands to reduce competition for moisture and plant tree species that are better adapted to drier conditions, thereby making forest ecosystems more resilient to SWEs (Duveneck and Scheller 2016).
Individual and family forestland owners, also known as small woodland owners, manage nearly 40% of U.S. forestland nationwide (Butler et al. 2021a). How they adjust their management practices to reduce adverse impacts from climate change (i.e., adapt) is, therefore, consequential for forest ecosystems. A growing body of research suggests that small woodland owners make decisions to adapt to climate change based on their weather experiences, weather perceptions, and climate change beliefs (Blennow 2012; Ogunbode et al. 2019b; Fischer 2019; Fischer et al. 2022), yet less is known about the relative relationships between experience, perception, and belief. In particular, the influence of weather experiences on climate change belief remains poorly understood, as is the potential for climate change belief to influence experiences and perceptions of SWEs (Howe et al. 2019).
We investigate the interaction between experience of SWEs, perception of SWE trends, and belief in climate change among small woodland owners in the U.S. upper Midwest. This region is experiencing more frequent and severe heatwaves and droughts, and warmer winter temperatures than 50–100 years ago, as well as greater precipitation and heavier rainfall events (particularly in summer and autumn) (Handler et al. 2014b; Pryor et al. 2014; Swanston et al. 2018). These changes in temperature and precipitation patterns, which are expected to increase in the future under a range of climate scenarios, amplify other forest stressors, such as insects and diseases (Dale et al. 2001), making them relevant concerns for all forest owners (Swanston et al. 2018). Small woodland owners control 55% of forestland in the upper Midwest (Butler et al. 2021a) putting them in the position to take actions that can help to mitigate climate change in the long term by protecting forests to insure continued provision of ecosystem services such as water, habitat, and carbon sequestration (Malmsheimer et al. 2008).
We use structural equation modeling (SEM) of small woodland owner survey data to investigate the relationships between experiencing SWEs directly, perceiving that SWEs are part of larger trends, and believing that climate change is occurring. We have three specific research questions:
-
Do small woodland owners who have experienced SWEs perceive changing trends in SWEs and have higher levels of belief in climate change?
-
Do small woodland owners who perceive trends in SWEs have greater belief in climate change, or do small woodland owners who believe in climate change have greater perception of trends in SWEs?
-
Do the relationships between SWE experience, trend perception and climate change belief among small woodland owners differ for severe storms and droughts?
To investigate the first question, we consider SWE experience as a predictor of perceived changes in SWE frequency and severity (“trend perception”) and SWE experience as a predictor of climate change belief, with trend perception as a mediator. This approach emphasizes SWE experience as a key predictor of climate change belief. To investigate the second question, we consider climate change belief as a predictor of trend perception at the same time that trend perception is a predictor of climate change belief. This approach is consistent with the idea that individual beliefs shape perceptions and interpretations of SWE experiences at the same time that SWE experiences are accurately observed (Hornsey et al. 2016; Howe et al. 2019; Sambrook et al. 2021). To investigate the third question, we compare our results for severe storms and droughts to see if these events are perceived differently or are differentially associated with climate change.
We make three contributions to the literature with the results of our analysis. The first contribution is to better understand the relationship between SWE experience, SWE trend perception and climate change belief using a nonrecursive SEM that considers the potential for reciprocal effects between SWE trends and climate change belief. In a nonrecursive SEM two or more variables are both outcomes and predictors of each other, in contrast to recursive SEMs where effects are assumed to flow in only one direction (Paxton et al. 2011; Bollen 1989). The second contribution is increased understanding of how different types of SWE experiences relate to perceptions of local weather patterns and climate change belief. The third contribution is to the scarce literature on how small woodland owners believe in and perceive climate change in the context of their forestland.
2. Conceptual approach
There are two main conceptual approaches used to understand the relationship between SWE experiences and climate change belief: experiential learning and motivated reasoning. These two approaches, while not mutually exclusive, suggest two different directions to the relationship (Weber 2010) (Fig. 1). Experiential learning holds that experiencing SWEs makes people more likely to believe in climate change (Tversky and Kahneman 1973; Keller et al. 2006; Whitmarsh 2008). Motivated reasoning holds that believing in climate change makes people more likely to believe that they have experienced SWEs and to attribute those events to climate change (Druckman and McGrath 2019; Bayes and Druckman 2021; Ripberger et al. 2017; Myers et al. 2013; Broomell et al. 2017). The literature suggests that both approaches contribute to understanding the relationship between climate change belief and SWE experience (and trend perceptions, which fall somewhere between them) (Myers et al. 2013). However, very few studies have explicitly tried to consider both approaches at once, likely because doing so requires complex models that simultaneously include reciprocal effects like those shown in Fig. 1 (Marquart-Pyatt et al. 2015; Hornsey et al. 2016).
Conceptual diagram of the relationship between severe weather experiences/perceptions and climate change belief via experiential learning and motivated reasoning.
Citation: Weather, Climate, and Society 14, 4; 10.1175/WCAS-D-21-0176.1
The importance of understanding the dominant effect direction between climate change belief and SWE experience lies in better understanding what influences climate change belief and ultimately climate change adaptation. Given that there is empirical support for both experiential learning and motivated reasoning from past research (Myers et al. 2013; Hornsey et al. 2016; Howe et al. 2019; Sambrook et al. 2021) we do not expect our analysis to show that one approach is the clear winner in explaining the relationship. Rather, we believe that both forces are likely at work.
The implications of each conceptual approach can be usefully thought through by considering the extremes where one approach dominates. If experience is the most important factor influencing climate change belief, then, with enough experiences, people will eventually come to believe that climate change is happening and deliberately move toward adapting to it. If instead we find that climate change belief predicts a person’s experience of SWEs, then adaptation to climate change will depend on the person’s existing climate change beliefs and broader belief system (which includes political orientation). This means that those who believe in climate change will already be inclined to adapt, and those who do not believe in climate change will resist adapting, regardless of SWE experience.
In the context of climate change, experiential learning is the perspective that experience of SWEs is the dominant force in forming climate change belief and builds on the idea that personal experience (through availability of experiences to recall, and ease of recall) is a key driver of risk perceptions (Tversky and Kahneman 1973; Keller et al. 2006; Whitmarsh 2008). Personal experience of SWEs helps make climate change seem more “real” to individuals, thus making them more likely to believe climate change is real, is happening now, and that they should take personal actions and support policies to address it.
In the case of climate change, the concept of experiential learning is the basis for the expectation that experiencing SWEs or observing changes in local weather patterns makes people more likely to believe in climate change, though the empirical findings have been mixed (Howe et al. 2019). A complicating aspect of this literature is the range of experience measures used. Some studies use self-reported measures of SWE experience to predict climate change belief or climate change concern (Ogunbode et al. 2019b; Demski et al. 2017; Spence et al. 2011), others use external measures of experience (Konisky et al. 2016; Taylor et al. 2014; Fownes and Allred 2019), and still others use respondent location to assume experience (Bergquist et al. 2019; Carlton et al. 2016). Another group of studies focus on perceived and externally measured weather changes or trends over time rather than discrete SWE experiences (Sugerman et al. 2021; Taylor et al. 2014; Hamilton and Stampone 2013). Overall, research has found trends in SWEs or temperatures to be more important (larger in magnitude) predictors of climate change belief than direct experience with single SWEs (Hornsey et al. 2016). However, SWEs may have large effects on climate change belief in the short term (Konisky et al. 2016).
Motivated reasoning, specifically directionally or politically motivated reasoning, refers to the process through which individuals selectively interpret new information to confirm their existing beliefs (Druckman and McGrath 2019; Bayes and Druckman 2021; Ripberger et al. 2017). In the context of climate change, this occurs when climate change belief leads individuals to be more likely to attribute their experience of extreme events to climate change (Myers et al. 2013; Broomell et al. 2017; Druckman and McGrath 2019). The inverse effect is also possible, such that individuals with low climate change belief are less likely to attribute SWE experience to climate change (Druckman and McGrath 2019), though evidence suggests that they may still recognize the SWEs as being unusual (Ripberger et al. 2017; Broomell et al. 2017).
Studies of motivated reasoning and climate change belief have found a range of sociocultural factors, such as cultural values, religion, and political identity or orientation, to be important predictors of climate change belief and how experiences of SWE are interpreted (Goebbert et al. 2012; Price et al. 2014; Hornsey et al. 2016; Arbuckle 2017; Newman et al. 2018). For instance, previous research has found that peoples’ experiences and perceptions of local weather conditions and the relationship between local weather conditions and climate change belief are associated with their political orientation (e.g., Goebbert et al. 2012; Borick and Rabe 2014; L. C. Hamilton et al. 2018; Albright and Crow 2019; Druckman and McGrath 2019; Howe et al. 2019). Other studies have examined motivated reasoning by investigating the effect of climate change belief on perceptions of weather changes and found an association between those who believe in climate change and those who perceive changes in their local weather patterns (Howe and Leiserowitz 2013; Niles and Mueller 2016; Broomell et al. 2017; Howe 2018).
There is also evidence that different kinds of SWEs may be observed and interpreted differently, particularly for fast versus slow events and between events related to temperature versus precipitation. For instance, Howe et al. (2014) find that perceptions generally follow documented event locations, and that more severe and discrete events (hurricanes and tornadoes) are recalled more accurately than more diffuse and slower acting events (like droughts). Goebbert et al. (2012) find that political orientation and cultural type influenced perceived changes in temperature more than perceived changes in floods and droughts, which they attribute to greater politicizing of temperature anomalies as an indicator of climate change. Taylor et al. (2014) found that perceiving changes in hot and dry conditions, and in high rainfall conditions, had an effect on climate change belief, but that perceiving changes in colder conditions did not have an effect on climate change belief.
Examining the relative importance of experiential learning and motivated reasoning effects in a single model requires a complex modeling approach that allows multiple simultaneous outcomes (Marquart-Pyatt et al. 2015). Very few studies model the potential for perceptions/experiences and climate change belief to mutually reinforce each other, though some studies discuss the possibility (Marquart-Pyatt et al. 2015; Akerlof et al. 2013; Blennow et al. 2012; Hornsey et al. 2016). Hornsey et al. (2016) in particular discuss the importance of potential reciprocal effects between trend perceptions and climate change belief due to unavoidable conceptual overlap in operationalizing trend perception and climate change belief, and they note that “it cannot be ruled out that this would have inflated the relatively large positive correlation with climate change belief” (Hornsey et al. 2016, 623–624). This potential overlap and its implication for inflated estimates are the basis of our second research question.
The one study we are aware of that models reciprocal effects between climate change experience perceptions and climate change belief is that by Myers et al. (2013), who use panel data to test the fit of 6 models of the relationship between climate change belief and experience. They find that the best fitting models are the nonrecursive models that contain a feedback between experience and belief, and the single best model is the one with direct reciprocal effects between climate change experience and belief. van der Linden (2014) uses a reciprocal effects model to examine the relationship between climate change risk perceptions and negative affect of climate change, rather than climate change beliefs, and find that risk perception and negative affect have a significant, mutually reinforcing reciprocal relationship.
3. Data and methods
a. Study system
Our empirical context is severe weather events in forests of the upper Midwest, specifically the “Northwoods,” also known as the Laurentian Mixed Forest ecological province (Cleland et al. 2007), a mixed-forest ecosystem in the northern portions of Minnesota, Wisconsin, and Michigan (Fig. 2). Climate models suggest that more frequent and severe heatwaves, droughts, winter thaws and precipitation events will worsen in this area over the next century as a result of climate change (Duveneck and Scheller 2016; Handler et al. 2014a,b; Janowiak et al. 2014; Swanston et al. 2018). Small woodland owners, also known as family forest owners,1 are a subset of nonindustrial private forest owners, (Harrison et al. 2002), and include individuals, partnerships, family limited liability companies, trusts, and estates. Small woodland owners controlled approximately 55% of forestland in the north-central United States (Illinois, Indiana, Iowa, Michigan, Minnesota, Missouri, Ohio, and Wisconsin) at the time of our study (Butler et al. 2021a). Ten acres (1 acre = 0.4 ha) is commonly considered to be the minimum ownership size where traditional forestry management approaches are practical (Butler et al. 2021b). In the north-central United States, the average small woodland owner with 10 or more acres has 50 acres of forest, and 99% own less than 500 acres (Butler et al. 2021a). These small woodland owners are 64 years old on average; 81% are male, 99% are white, and 33% have a bachelor’s or advanced degree (Butler et al. 2021a). They own their forest property for a variety of reasons ranging from enjoyment of scenery to timber production (Butler et al. 2021a).
Map of the study area.
Citation: Weather, Climate, and Society 14, 4; 10.1175/WCAS-D-21-0176.1
Forest owners can significantly affect the climate and forest ecosystems through various management and mitigation strategies (Duveneck and Scheller 2016). Those who participate in climate change adaptation practices often believe that climate change is occurring (Blennow et al. 2012; Blennow and Persson 2009; Boag et al. 2018). However, there is little research on how experience of SWEs may or may not shape small woodland owners’ beliefs in climate change, with or without consideration of political beliefs and cognitive biases, especially in the United States. Limited research has been conducted on small woodland owners and their weather perceptions and climate change beliefs in the United States, but this research has found that the majority of these owners do believe in climate change (Boag et al. 2018; Khanal et al. 2016). Furthermore, there is no evidence that U.S. small woodland owners hold different climate change beliefs than nonforest owners in their area (Hamilton et al. 2016).
b. Sample and data collection
Our sample consists of owners of small, nonindustrial, private woodlands with 10 or more acres of forest (hereinafter referred to as “small woodland owners,” or just “owners”) in nine Northwoods counties that have experienced greater than average damage from SWEs. To identify the counties, we calculated z scores for each county in the Northwoods based on the relative area that had sustained forest damage from pests, storms, fire, and drought, and the number of heatwave days, winter ice days and heavy precipitation days (Table 1). We obtained these data from the U.S. Forest Service Insect and Disease Detection Survey (U.S. Forest Service 2018), the Monitoring Trends in Burn Severity (MTBS) database of wildfires (MTBS 2018), and the U.S. Climate Resilience Toolkit Climate Explorer (U.S. Federal Government 2015). We combined the county exposure z scores with GIS maps of relative extent of forest cover and small, nonindustrial, private forestland to select the three counties in each state that had the highest levels of stressor exposure (via the composite z scores) and extensive areas of small forestland ownership for which we could get tax lot-level ownership information (Fig. 2). The counties selected were Cook, Pine, and St. Louis Counties in Minnesota; Douglas, Burnett, and Iron Counties in Wisconsin; and Oceana, Missaukee, and Mackinac Counties in Michigan. To select the sample, we randomly cast points onto the nine county maps to generate a list of parcels. We obtained the names and addresses of the parcel owners according to public tax lot records from the Digital Map Products Co. We removed from the list any owners who appeared to be corporate, public, or institutional.
Stressor exposure measures and data sources. IDS = U.S. Forest Service’s Forest Health—Insect and Detection Survey; MTBS = Monitoring Trends in Burn Severity; USCE = U.S. Climate Resilience Toolkit Climate Explorer.
In winter and spring 2019, we mailed2 surveys to the 4472 selected owners in the nine counties (randomly drawn from the list generated) proportionately to the estimated number of small woodland owners in the county. The survey administration followed the tailored design method (Dillman et al. 2014) beginning with an announcement card, followed a week later by the survey, and a thank you/reminder card 2 weeks later. We sent a second survey to those who did not respond to the first survey. A cardstock bookmark printed with a tree species alphabet was included in the survey mailing and was the only form of incentive offered. The survey consisted of 31 questions over eight pages. The topics and framing of the survey were guided by previous qualitative focus groups in the Northwoods region (Fischer 2019; Fischer et al. 2022). Of the surveys sent, 467 were found to be ineligible (respondent reported land not forested, respondent is not the owner, owner is deceased, bad address/inaccurate ownership record), while 1255 valid responses were received (owned a parcel of at least 10 forested aces and answered 50% or more of the survey questions) yielding an effective overall response rate of 31%.3
c. Variables and model design
We focus our analysis on two outcome measures of interest: whether respondents agree that “climate change is happening” as our measure of climate change belief (cc_belief), and whether respondents believe that severe storm and drought events have changed or will change in the future as our measure of trend perception (storm_trend and drought_trend, respectively). Both of these variables are included as binaries (Table 2; see the appendix for additional detail on variable construction). We chose to use binary measures for all variables, even when originally measured on 4- or 5-point scales, such as the climate change measure and the trend perception components, to reconcile variables that were measured on different scales and to aid in the interpretation and comparison of the probit results. To look for differences between types of severe events and due to correlations between types of severe events, we ran separate models for severe storms and droughts.4 Our main explanatory variable of interest is having experienced forest damage from severe storms and droughts. Severe storms and droughts are causes of forest damage in the Northwoods according to past observation and future climate projections (Duveneck and Scheller 2016; Handler et al. 2014a,b; Janowiak et al. 2014; Swanston et al. 2018) about which landowners have also expressed concern (Ontl et al. 2018; Fischer et al. 2022) (storm_experience and drought_experience, respectively).
Variable names, definitions, and descriptive statistics. The range of all variables is 0–1, and n = 809. See the appendix for wording of survey questions used and measurement construction of all analysis variables.
Because environmental concern and belief in climate change decrease with age, are higher for women rather than men, and increase with education level (Hornsey et al. 2016; Shao and Goidel 2016; Hamilton and Stampone 2013) we control for age (age_75plus), education (college_degree), and gender (female) (Table 2). We also control for county political orientation (democratic_county), as the association between political orientation and climate change belief is well established (Borick and Rabe 2014; Konisky et al. 2016; Hornsey et al. 2016), and it was not assessed on the survey (Table 2). We used a county-level measure for political orientation as election results are readily available at this scale, and congressional districts in the rural areas where the vast majority of our respondents live are areas that cover multiple counties. We chose to use the results of the 2018 midterm elections as it was the closest election to our period of data collection. As we have constructed it, we view the Democratic county measure as indicating that the odds are favorable (50% or greater chance) that our respondent is a Democrat. We do not include income because of its correlation with education. Because democratic_county is a county-level variable, we do not include an explicit county or state control variable in the analysis. Democratic_county was found to be more influential in the analysis than state or county control variables, in addition to being more conceptually relevant.
We include in our models a measure of the importance of wood products as a reason for owning woodland (wood_products) as an instrumental variable for trend perception. An instrumental variable is an exogenous variable that is significantly related to one endogenous variable but not the other; an instrumental variable is required for a nonrecursive model to be identified (Paxton et al. 2011). Nonrecursive models are simultaneous equation regression models in which two or more of the endogenous variables affect one another directly or indirectly, and/or there are correlations among some of the disturbance terms of the endogenous variables (Paxton et al. 2011). We expect individuals who place priority on producing wood products to be more heavily invested in their forests than owners who do not, as previous research has found these owners to be more active forest managers (Joshi and Arano 2009; Khanal et al. 2017) and more concerned about threats to their forest (Fischer et al. 2014; M. Hamilton et al. 2018) than owners who do not prioritize wood products. As such, we expect them to pay more attention to conditions in their forest and any changes in patterns of severe events than would those for whom wood products are a less important reason for ownership.
In the first set of models, we consider both trend perception and climate change belief as outcomes of experience, with trend perception acting as a mediator between experience and climate change belief (model A in Fig. 3).5 We then consider a model where trend perception and climate change belief simultaneously influence each other, a nonrecursive model (Paxton et al. 2011) (model B in Fig. 3). We found no evidence of significant correlation between the errors of climate change belief and trend perception, so we do not include an error correlation in the models.
Path diagram of models, showing the paths included in the mediation model (model A) and the additional path that is included to make the initial version of the reciprocal effects model (model B).
Citation: Weather, Climate, and Society 14, 4; 10.1175/WCAS-D-21-0176.1
d. Analysis methods
For this analysis we use SEM using the lavaan package (0.6–6) in R. SEM is an appropriate method for estimating reciprocal relationships such as those that are expected to exist in the formation or maintenance of climate change belief (Marquart-Pyatt et al. 2015). SEM has been used in other studies about climate change belief (van der Linden 2014; Spence et al. 2011; Ogunbode et al. 2019b; Myers et al. 2013; Ogunbode et al. 2019a), but only Myers et al. (2013) and van der Linden (2014) use nonrecursive models.
Because of the use of binary endogenous variables, we use a diagonally weighted least squares (DWLS) estimator, specifically the weighted least squares means and variance adjusted (WLSMV) estimator in lavaan, with a probit link (Li 2016). The coefficients estimated by this method are probits, which we interpret using the latent response approach. The latent response approach pairs each observed binary endogenous variable with a unique latent continuous variable. The model coefficients are interpreted as linear effects of the independent variables on the latent endogenous variable(s) (Muthén et al. 2016), which is conceptually consistent with trend perception and climate change belief as concepts on a continuous scale of agreement and disagreement (Brenkert-Smith et al. 2015) even though we are using binary measures.
4. Results
Our analyzed sample consisted of 809 cases after cases with missing values on the analysis variables were dropped. Our analysis sample does have significantly fewer female respondents and respondents age 75 or older than our full sample (using two-sample t tests). There are also differences between the analysis sample and dropped cases on the storm experience and trend perception measures, with respondents who had storm experience and storm trend perceptions disproportionately dropped from the analysis sample due to having lower rates of missing values on the storm experience variable versus the drought experience variable. However, we felt that it was important to use the same sample for both the storm and drought analyses as one of our research questions is on the comparison between types of severe events. The analysis sample is 77% male and 16% are 75 or older. Approximately one-half of the respondents have a college degree, and about one-half live in Democratic counties (Table 2). Table 2 displays the descriptive statistics of the variables used in the analysis.
The correlations between analysis variables (Table 3) support the hypothesis that experience has an indirect effect on climate change belief that is mediated by trend perception. Climate change belief and trend perception have strong, positive correlations (0.50 for storms and 0.51 for droughts). Trend perception has strong, positive correlations with event experience (0.52 for storms and 0.64 for drought). At the same time, the correlations between climate change belief and the experience variables are small (0.13 for storms and 0.15 for drought). The correlations also do not show evidence of multicollinearity, as indicated by the small correlations between the exogenous variables (variables 4–10 in Table 3).
Tetrachoric correlation coefficients and results of likelihood ratio χ2 test (n = 809). The dagger, single asterisk, double asterisks, and triple asterisks indicate p < 0.1, 0.05, 0.01, and 0.001, respectively. Italicized numbers indicate correlations between variables that do not appear in the same model together.
a. Mediation models
To answer our first research question on how SWE experience influences trend perceptions and climate change belief, we use a mediation model (model A in Fig. 3). We find that owners who had storm experience perceived that severe storms were becoming worse (storm_trend), as compared with those who did not report storm experience (0.922; p < 0.001) (Table 4). We also find that owners who perceive storm trends are worsening have higher belief in climate change (0.476; p < 0.001). The indirect effect of storm experience on climate change belief via storm trend is positive (0.439; p < 0.001), but the direct effect is negative (−0.249; p < 0.1). Overall (the total effect), landowners who have had storm experience have higher climate change belief (0.190; p < 0.1) than those who have not had storm experience. In the storm mediation model, all the control variables, except female, have significant indirect effects on climate change belief, but none have significant direct effects. Having at least a college degree, residing in a Democratic county, and prioritizing wood products were all found to have positive and significant total effects on climate change belief. The model fit is very good, with a nonsignificant χ2, a root-mean-square error of approximation (RMSEA) of 0, and a comparative fit index (CFI) and Tucker–Lewis index (TLI) of 1.00 or above (West et al. 2012).
Results of mediation model of storm experience on climate change belief (n = 809). The dagger, single asterisk, double asterisks, and triple asterisks indicate p < 0.1, 0.05, 0.01, and 0.001, respectively. Structural equation modeling with the WLSMV estimator was used for the analysis; estimates are unstandardized probit coefficients. RMSEA, CFI, and TLI are defined in section 4a.
For droughts, we find that landowners with drought experience perceived that droughts were becoming worse (drought_trend) (Table 5), as compared with those with no drought experience (1.236; p < 0.001). We also find that landowners who perceive that drought trends are getting worse have higher belief in climate change (0.486; p < 0.001). The indirect effect of drought experience on climate change belief via drought trend is positive and significant (0.601; p < 0.001). Landowners who have drought experience have lower climate change belief when the effect is considered directly (−0.343; p < 0.1). The total effect of drought experience on climate change belief, is positive but not significant (0.257; p = 0.107). Among the control variables, having at least a college degree and residing in a Democratic county both have significant positive total effects on climate change belief. The effect of education on climate change belief is both direct and indirect, while the effect of residing in a Democratic county is direct. The model fit is very good, with a nonsignificant χ2, an RMSEA of 0, and a CFI and TLI above 1.00 (West et al. 2012).
As in Table 4, but for results of mediation model of drought experience on climate change belief.
The mediation models provide important information about the relationship between SWE experience and climate change belief but are missing a key piece of information about the relationship. A mediation model, like our model A, provides us with the estimated total effect of an exogenous variable on an endogenous variable, which is the same estimate we would get if we ran the model without mediation (e.g., left trend perception out of the model entirely), and lets us decompose the total effect into the direct and indirect portions (Paxton et al. 2011; Bollen 1989). In both of our mediation models the total effect of SWE experience on climate change belief is relatively small, and not statistically significant in the case of droughts, because the direct and indirect effects have opposite signs, which suppresses the size of the total effect (Bollen 1989; MacKinnon et al. 2000). We find that the indirect effect of SWE experience on climate change belief via trend perception is larger than the direct effect. This large indirect effect of SWE experience on climate change belief indicates the importance of trend perception as a mediator, but it leaves us with a rather confusing negative direct effect between SWE experience and climate change belief. However, the reciprocal effects models reveal that the size of this relationship is not as large, nor is the direct effect as important, as the mediation models indicate.
b. Reciprocal effects models
To answer our second research question about the presence of reciprocal effects between storm trend perception and climate change belief, we created a nonrecursive model that includes reciprocal effects between storm trend and drought trend and climate change belief (model B in Fig. 3). Our first specification of the reciprocal effects model included the same paths between the exogenous and endogenous variables as the mediation model. We were particularly interested in what would happen to the direct effect of storm experience and drought experience on climate change belief when the reciprocal effect was included, given their negative effect in the mediation models.
From the results of this initial reciprocal effects model, we find that the direct linear effect of storm experience on climate change belief and drought experience on climate change belief are very small and not significant (0.008; p = 0.985 and −0.085; p = 0.933, respectively). In respecifying the model, we thus restrict to zero (effectively leave out) the direct effect between storm experience and climate change belief and drought experience and climate change belief. We also exclude from the respecified storm model the direct effect of age being 75 or more and female on climate change belief because they also have a very small nonsignificant direct linear effect in the original reciprocal effects model (0.042; p = 0.887 and 0.058; p = 0.786, respectively). In respecifying the drought model, we also leave out the direct effect of age being 75 or more and female on climate change belief because of very small, nonsignificant direct linear effects in the original reciprocal effects model (−0.002; p = 0.992 and 0.019; p = 0.947, respectively). The model fit of both these initial models is unknown because they are exactly identified.
The respecified reciprocal effects model for storms, has very good model fit (Fig. 4), with a nonsignificant χ2 test, an RMSEA of zero, and the CFI and TLI both 1.0 or above. With the loop effect6 included (Table 6), we find that owners who have storm experience have higher storm trend perception in total (0.924; p < 0.001), and those with higher storm perception also have higher climate change belief (0.206; p < 0.1). The indirect/mediated effect of storm experience on climate change belief is positive (0.182; p < 0.1). The reciprocal effects between storm trend perception and climate change belief are both found to be positive (0.206; p < 0.1, and 0.252; p < 0.1, respectively). The loop effect is positive and significant (0.049; p < 0.01). Among the other explanatory variables (total effects), we find that owners having at least a college degree, those residing in a Democratic county, and those prioritizing wood products all have higher storm trend perception while those 75 or older have lower storm trend perception. Owners with at least a college degree and those residing in a Democratic county have greater climate change belief while those over age 75 have lower climate change belief.
Path diagram and raw coefficients of revised model B for storms. Structural equation modeling with WLSMV estimator was used for the analysis; estimates are unstandardized probit coefficients, with standard errors in parentheses. RMSEA, CFI, and TLI are defined in section 4a. Footnote a indicates that the noted text pertains to the variance of the latent continuous variable associated with the observed binary outcome, and footnote b indicates robust measures. The dagger, single asterisk, double asterisks, and triple asterisks indicate p < 0.1, 0.05, 0.01, and 0.001, respectively.
Citation: Weather, Climate, and Society 14, 4; 10.1175/WCAS-D-21-0176.1
As in Fig. 4, but for drought.
Citation: Weather, Climate, and Society 14, 4; 10.1175/WCAS-D-21-0176.1
As in Table 4, but for results of reciprocal effect model (respecified model B) of storm trend perception and climate change belief.
The respecified reciprocal effects model for droughts, has very good model fit (Fig. 5), with a nonsignificant χ2 test, an RMSEA of zero, and the CFI and TLI both 1.0 or above. With the loop effect (see footnote 6) included (Table 7), we find that owners with drought experience have higher drought trend perception (1.235; p < 0.001), and those with higher drought perception have higher climate change belief (0.229; p < 0.1). The indirect effect of drought experience on climate change belief via drought trend is positive (0.269; p < 0.1). For the reciprocal effects between drought trend perception and climate change belief, the effect of drought trend perception on climate change belief is positive (0.229; p < 0.1), while the effect of climate change belief on drought trend perception is slightly larger but is not statistically significant (0.235; p = 0.156). However, the loop effect of the reciprocal relationship between drought trend and climate change belief is positive and significant (0.051; p < 0.05). Among the other explanatory variables (total effects) shown in Table 7, we find that owners who have at least a college degree have higher drought trend perception than those without a college degree. Owners with at least a college degree and those residing in a Democratic county have greater climate change belief.
As in Table 4, but for results of reciprocal effect model (revised model B) of drought_trend and CC_belief.
The reciprocal effects models show us that the indirect effect in the mediation model is the primary relationship between SWE experience and climate change belief for both storms and droughts but that its size in the mediation model is inflated by the significant reciprocal relationship between trend perception and climate change belief that is not included. When we only include one path for the relationship between trend perception and climate change belief, as we do in the mediation model, it includes the combined effect of the two smaller bidirectional effects and makes the direct relationship between trend perception and climate change belief appear to be larger than it is.7 Because the relationship between trend perception and climate change belief is part of the indirect effect of experience on climate change belief, the indirect effect of experience on climate change belief is also inflated in the mediation model, which makes the direct effect of experience on climate change belief negative.
5. Discussion
In this paper, we sought to answer three questions about the relationship between SWE experiences, SWE trend perceptions, and climate change belief among small woodland owners. For our first research question, about the influence of SWE experience on trend perceptions and climate change belief, we find that small woodland owners with SWE experience have much higher perceptions that SWEs are becoming worse (trend perceptions) and have a greater belief in climate change relative to owners without SWE experience. We also find that the relationship between experience and climate change belief is indirect, mediated by the perception of severe storm and drought trends, though the size of the indirect effect is inflated in the mediation model. For our second research question, about the possibility of a mutual/reciprocal relationship between trend perception and climate change belief, we find that SWE trend perception and climate change belief are mutually related and reinforce each other, with climate change belief having a slightly larger effect on trend perception than trend perception has on climate change belief. In relation to our third research question, about the relationships between SWE experiences, trend perceptions and climate change belief when considering severe storms versus droughts, we do not find major differences in the relationships between SWE experience, trend perception and climate change belief between severe storms and droughts, though we do find some differences among the control variables.
Our finding of reciprocal effects between trend perception and climate change belief has important implications for interpreting past research and for the design of future research related to climate change belief and perception. Previous studies have found that perceiving changes in SWEs or local weather patterns has a larger effect on climate change belief than just experiencing SWEs (Taylor et al. 2014; Hornsey et al. 2016), which is consistent with the findings of our mediation models but not with the findings of our reciprocal effect models, from which we find the total effects of trend perception and SWE experience on climate change belief to be of similar size. It is possible that unmodeled reciprocal effects between local weather perceptions and climate change belief are inflating the estimates in these previous studies. Hornsey et al. (2016) recognize the possibility of reciprocal effects between conceptually similar measures, even when efforts are made to avoid it by making sure that the measures of weather perceptions or experiences are not framed in terms of climate change. Researchers should pay particular attention in future research to the potential for reciprocal effects between measures related to climate change, since we have demonstrated that this conceptual possibility is real and affects our conclusions about the relationships between our measures of interest.
The reciprocal relationship between trend perception and climate change belief that we find for both storms and droughts that are positive and significant indicate that there is an amplifying relationship between the two measures (Paxton et al. 2011). This reciprocal relationship indicates that both owners who perceive that SWE trends are changing have greater belief in climate change, and owners who believe in climate change have higher perceptions of changing trends in SWEs. The effect of climate change belief on trend perception is slightly larger than the effect of trend perception on climate change belief for both severe storms and droughts, but in both cases the magnitudes of the reciprocal effects are similar. The size of the total effect of climate change belief on drought perception is notable, even though it is not statistically significant. The total loop effect between climate change belief and drought perception is, however, statistically significant. This finding supports the idea that people who believe climate change is happening are more likely to ascribe their experiences of severe weather events to being part of a trend, as well as that people who notice a trend in severe weather events are more likely to believe in climate change.
The only notable differences we found between severe storms and droughts were in the control variables. We expected to find differences between the effects of severe storms and droughts, due to their acute versus gradual nature making changes in drought frequency and severity more difficult to notice than those of damage-causing storms [consistent with the findings of Howe et al. (2014)]. However, the overall pattern of effects and effect sizes that we find between severe storm and drought experience, trend perception and climate change belief are remarkably similar. A key point of difference between storms and droughts is in their relationship with political orientation. We find that respondents in Democratic-leaning counties are more likely to feel that storms are getting worse than in Republican-leaning counties, but we found no significant relationship between political orientation and the perception that droughts are getting worse. Previous research has found political differences in the perceptions of temperature-based weather changes and abnormalities (Goebbert et al. 2012; Broomell et al. 2017), but not related to storms and droughts (Goebbert et al. 2012), though this could be a difference between small woodland owners and the general population.
An interesting finding among the control variables is that education has the largest total effect on climate change belief of any predictor in our models. This finding is in contrast to Hornsey et al. (2016) who found political orientation and experience of local weather change to both have much larger effects on climate change belief than education. In relation to political orientation, this difference from the findings of Hornsey et al. (2016) could be explained by our use of a county-level measure of political orientation rather than an individual-level measure. Since we do not expect all of our respondents to align politically with the county majority, we might expect to find a smaller effect of political orientation on individual climate change belief than studies that use an individual measure of political orientation. We have already discussed the way that the effects of experiencing local weather changes on climate change belief might be inflated in other studies and thus be found to have a larger effect than education. Khanal et al. (2016) found education to be positively associated with climate change belief in their survey study of southern small woodland owners, though they did not consider the effects of SWE experience in their study.
Our results are consistent with previous findings that both experiential learning and motivated reasoning are involved in shaping climate change belief among the general population (Howe and Leiserowitz 2013; Akerlof et al. 2013; Howe 2018; Sambrook et al. 2021) but our study is one of the few to explicitly model the relative influences of these two conceptual approaches. Our results also support the idea that severe event experience is filtered through a meaning-making stage similar to that found by Ogunbode et al. (2019a,b). In our analysis, we find support for the idea that experience influences climate change belief only if it contributes to individuals’ perceptions that events of that type have changed over time and likely will continue to change. As Howe et al. (2019) point out, “[u]nderstanding the constructed nature of experience together with the importance of experiential processing in driving judgments and decision making suggests that it may be helpful to improve our understanding of how measured climate changes influence subjective beliefs about those changes” (Howe et al. 2019, p. 14).
Our results are also consistent with the limited and primarily qualitative literature on U.S. small woodland owner experiences with SWEs and their climate change beliefs. This literature has found that U.S. small woodland owners do notice SWEs in their woodlands, but that alone is not enough for them to be certain about climate change. For instance, Grotta et al. (2013), in focus groups with small woodland owners in Washington, Oregon, Idaho, and Alaska, found that small woodland owners referenced their own experiences and observations as an important information source for their climate change beliefs even as they expressed skepticism and uncertainty about climate science and information sources they perceived as biased. Similarly, Fischer et al. (2022) found that small woodland owner focus group participants in the upper Midwest are noticing greater frequency and severity of severe rain and wind storms, as well as overall warming trends in both summer and winter, but are also uncertain about the meaning and implications of their observations. Boag et al. (2018) found that most of the small woodland owners they interviewed in eastern Oregon felt that snowpack in their area had decreased over time regardless of their beliefs about climate change.
One of the major advantages of our design is that unlike some previous studies (e.g., Myers et al. 2013; Fownes and Allred 2019), our measures of experience, and trend perceptions, were not asked in the context of climate change (i.e., the question in no way framed the severe events as being related to climate change or required any acknowledgment or assumption of climate change to answer in the affirmative). We used this question design to minimize conceptual overlap between questions, as noted by Hornsey et al. (2016), but it was not sufficient to avoid circularity between trend perception and climate change belief as we found in our analysis.
One potential limitation of our study is that we used county-level rather than respondent-level data on political orientation, given its importance in predicting climate change belief in the literature. A question on political orientation was intentionally not included in the survey to avoid making the survey appear politically motivated or discouraging respondents from honestly answering the climate change belief questions. Another potential limitation is our use of cross-sectional data. We conceptualize the experience of SWEs as preceding the trend perception, but we cannot make a causal connection with our cross-sectional data since respondents are providing their recalled experience and trend perceptions at the same time. While not explicitly a limitation, we acknowledge that our focus on counties with greater than average damage from SWEs represents a best-case scenario for studying how climate change impacts are being noticed (or not). A similar analysis in a place with fewer SWEs may show a weaker relationship between experience and climate change belief.
A key question that follows from the present study is how the reciprocal relationship between trend perception and climate change belief relates to behaviors that make the individual or society more or less adapted to climate change. Building on the present model for instance, is climate change belief or trend perception (or both) important predictors of behavior? Is this effect mediated by risk perception? By perceived response or adaptation appraisal (Grothmann and Patt 2005)? Recent studies have found that climate change belief is much more predictive of climate change risk perception than perceptions of temperature change (Marlon et al. 2019), and that experiencing SWEs does not contribute significantly to climate change concern or willingness to support policies to fight climate change (Gärtner and Schoen 2021). What can we learn about these relationships between beliefs, experiences, risk perception and actions using models in which we consider the potential for mediation and reciprocal effects?
6. Conclusions
Our results support the findings of previous studies that both experiential learning and motivated reasoning are at play in shaping climate change belief (Myers et al. 2013; Hornsey et al. 2016; Howe et al. 2019), but our results provide more nuance in understanding the relationships between these forces than most other studies [Myers et al. (2013) being a notable exception]. We find that experience does have an important, though indirect, effect on climate change belief, which is evidence of experiential learning, the expectation that experiencing SWEs makes people more likely to believe in climate change (Tversky and Kahneman 1973; Keller et al. 2006; Whitmarsh 2008). We also find that Democratic county is at least as important a predictor of climate change belief as SWE experience; this is evidence of motivated reasoning, the concept that believing in climate change makes people more likely to believe that they have experienced SWEs and to attribute those events to climate change (Druckman and McGrath 2019; Bayes and Druckman 2021; Ripberger et al. 2017; Myers et al. 2013; Broomell et al. 2017). The mutually reinforcing relationship that we find between climate change belief and SWE trend perception supports our conclusion that both forces, experiential learning and motivated reasoning, are key aspects of climate change belief and the perception of changes in local weather patterns.
This study is one of the few studies on climate change beliefs that uses nonrecursive models to consider the presence of reciprocal effects between core concepts (Marquart-Pyatt et al. 2015; Hornsey et al. 2016). Such an approach provides important insights into the complex relationships and dynamics involved in climate change belief, while adding a cautionary note to studying related climate change concepts to avoid inflated effects (Marquart-Pyatt et al. 2015; Hornsey et al. 2016). Our findings suggest that while it is not unreasonable to assume that trend perception influences climate change belief or vice versa, the effect size calculated singly for the relationship will be larger than it should be, since the single effect is capturing effects that in fact go in both directions. We need more studies that consider the potential for reciprocal relationships to better identify and incorporate the nuances of the relationships between key concepts related to climate change belief.
The National Woodland Owners Survey uses the term family forest owner to describe forest ownerships by individuals and joint ownerships by multiple individuals, families, trusts, and estates (Butler et al. 2021b). Nonindustrial private forest owners are defined as private owners who do not operate mills (Harrison et al. 2002) and can include nontimber companies, conservation trusts, and clubs and recreational organizations, in addition to family forest owners.
We chose to use a mail survey for two reasons: 1) Mailing addresses were the only known way to select and contact a random sample of small woodland owners in our study area, because publicly available tax lot records do not include phone numbers or email addresses. 2) Mail is a common method used to study rural and older subpopulations, such as forest owners and farmers, many of whom do not have and/or may not use computers or the internet. Mail surveys are the method used by the National Woodland Owners Survey and the USDA Census of Agriculture to reach these populations. The greatest potential for bias using this method is that respondents who cannot read or who do not speak English will not be able to participate in our study. Given the demographic characteristics of our study population in our study area, we believe this potential for bias is minimal.
This rate is equivalent to response rate 2 as defined by the American Association for Public Opinion Research (https://www.aapor.org/AAPOR_Main/media/publications/Standard-Definitions20169theditionfinal.pdf). The county-level response rates ranged from 14% (Missaukee County, Michigan) to 41% (Cook County, Minnesota).
We initially considered models involving winter thaws and heatwaves in addition to severe storms and droughts. However, we found a significant interaction between experience and trend for winter thaws, and interaction tests for heatwaves did not converge, possibly because of low frequencies in our sample. Since our interest is in the potential for reciprocal effects, which is complicated by an interaction effect, we thus chose to focus our analysis on severe storms and drought, for which there is no significant interaction.
We also considered a mediation model with climate change belief acting as the mediator between experience and trend perception. We do not present the results of this model because the key relationship of interest—between trend perception and climate change belief—is extremely similar in its linear effect size regardless of the direction in which the effect is specified.
The loop effect can be thought of as the indirect effect of each of the endogenous variables on themselves (Paxton et al. 2011), and as a multiplier of the effects between variables included in the loop (Hayduk 1987). Thus, the estimates in Table 6 are the products of the raw coefficients in Fig. 4, and 1 + the loop effect; the same applies for Table 7 and Fig. 5.
Regression coefficients, including those in SEM, are not inherently directional; the researcher imposes the direction based on theoretical assumptions or known causal mechanisms (Kline 2012). A mediation model with all the same predictors for both endogenous variables would produce the same estimate for the path between the endogenous variables, regardless of in which “direction” we estimate the relationship.
Acknowledgments.
This work was supported by the USDA National Institute of Food and Agriculture McIntire–Stennis Program (1011135), USDA Forest Service Northern Research Station, University of Michigan Energy Institute, University of Michigan Graham Sustainability Institute, and University of Michigan Undergraduate Research Opportunity Program. We thank Bill Cook, Julie Crick, Stephen Handler, Trisha Gorby, Mike Reichenbach, Matt Russell, Eli Sagor, Mike Smalligan, Stephanie Snyder, and Kris Tiles for providing feedback and advice during the planning and interpretation stages of the research. We also thank Aniseh Bro and Leigh Mitchell for their vital work on the survey design and data collection.
Data availability statement.
Because of their confidential nature, supporting data cannot be made openly available. Details of the data and how to request access are available from corresponding author Fischer at the University of Michigan.
APPENDIX
Survey Questions and Creation of Analysis Variables
This appendix describes the source survey questions and construction of the variables used in the analysis. For each analysis variable we include a copy of the survey question or questions as they appeared to participants, the SAS code used to construct the measures used in the analysis, and additional description as needed.
Question 26 (Fig. A1) was used in construction of the variable “cc_belief.” The relevant SAS code is
-
if Q26a_CCReal in (1 2) then cc_real=0; if Q26a_CCReal in (3 4) then cc_real=1;
Survey questions discussed in this appendix.
Citation: Weather, Climate, and Society 14, 4; 10.1175/WCAS-D-21-0176.1
The “trend_perception” variables were constructed from three survey questions [questions 11, 12, and 14 (Fig. A1)]. For each of these questions, a binary measure was constructed that indicated greater perceived change on that dimension (change in past frequency, change in past damage/severity, and future change). The three binary measures were added together to create an index that indicates the number of dimensions that the respondent thought had/will change (0–3), and then this index was split to create the final binary measure used in the analysis indicating perceived change on two or three of the dimensions. Question 11 was reverse recoded so that 4 = more often and 5 = much more often. The SAS code for measures for storms and droughts that indicate respondent thinks that they are occurring more or much more often is
-
if Q11b_FreqStorm in (4 5) then freq_storm_inc=1;
-
else freq_storm_inc=0;
-
if Q11b_FreqStorm=. then freq_storm_inc=.;
-
if Q11d_FreqDrought in (4 5) then freq_drought_inc=1;
-
else freq_drought_inc=0;
-
if Q11d_FreqDrought=. then freq_drought_inc=.;
Question 12 was reverse recoded so that 4 = more damaging and 5 = much more damaging. The SAS code for measures for storms and droughts that indicate respondent thinks that they have become more or much more damaging is
-
if Q12b_SevStorm in (4 5) then sev_storm_inc=1;
-
else sev_storm_inc=0;
-
if Q12b_SevStorm=. then sev_storm_inc=.;
-
if Q12d_SevDrought in (4 5) then sev_drought_inc=1;
-
else sev_drought_inc=0;
-
if Q12d_SevDrought=. then sev_drought_inc=.;
Question 14 is used for measures for storms and droughts that indicate respondent thinks that they may get worse in the future. The SAS code is
-
if Q14b_WorsenStorm in (3 4 5) then worse_storm_inc2=1;
-
else worse_storm_inc2=0;
-
if Q14b_WorsenStorm=. then worse_storm_inc2=.;
-
if Q14d_WorsenDrought in (3 4 5) then worse_drought_inc2=1;
-
else worse_drought_inc2=0;
-
if Q14d_WorsenDrought=. Then worse_drought_inc2=.;
We then add together the three storm dimensions:
-
storm_inc_index = sev_storm_inc + freq_storm_inc + worse_storm_inc2;
add together the three drought dimensions:
-
drought_inc_index = sev_drought_inc + freq_drought_inc + worse_drought_inc2;
and divide the index measures to determine which respondents thought two or three of the event dimensions had/will get worse:
-
if storm_inc_index in (2 3) then storm_inc=1;
-
if storm_inc_index in (0 1) then storm_inc=0;
-
if drought_inc_index in (2 3) then drought_inc=1;
-
if drought_inc_index in (0 1) then drought_inc=0;
Another variable that was constructed was “experience.” Question 9 (Fig. A1) was coded such that if impacts were reported for an event then the “no impacts” question is equal to 0. If no response was given for an event, then all questions related to that event were considered to be missing. The SAS code is
-
if Q9b5_StormNoImpact=0 then storm_impact9=1;
-
if Q9b5_StormNoImpact=1 then storm_impact9=0;
-
if Q9b5_StormNoImpact=. then storm_impact9=.;
-
if Q9d5_DroughtNoImpact=0 then drought_impact9=1;
-
if Q9d5_DroughtNoImpact=1 then drought_impact9=0;
-
if Q9d5_DroughtNoImpact=. then drought_impact9=.;
Question 31 (Fig. A1) was used in construction of the variable “college_degree.” The SAS code is
-
if Q31_Education in (5 6) then ed_ba_plus=1;
-
else ed_ba_plus=0;
-
if Q31_Education=. then ed_ba_plus=.;
Question 27 (Fig. A1) was used in construction of the variable “age_75plus.” The SAS code is
-
age=2019-Q27_Birthyear;
-
if age>=75 then age_75plus=1; else age_75plus=0;
-
if Q27_Birthyear=. then age_75plus=.;
Table A1, along with survey respondents’ addresses, was used in construction of the variable “democratic_county.” Party assignments for each county were determined from results from the 2018 gubernatorial elections, except for states that did not hold gubernatorial elections in 2018, in which case results from the 2018 senate race were used instead. There are 10 cases for which the senate election result was used (see the table footnote), with one respondent from each county so indicated. Election results are from Politico.
Question 28 (Fig. A1) was used in construction of the variable “female.” The SAS code is
-
if Q28_Gender=2 then female=1;
-
if Q28_Gender=1 then female=0;
-
if Q28_Gender=. then female=.;
Democratic- and Republican-leaning counties of residence for survey participants in the analysis sample, as determined by survey mailing address.
Question 7 (Fig. A1) was used in construction of the variable “wood_products.” The SAS code is
-
if Q7e_ReasonProducts in (3 4) then ReasonProducts_high=1;
-
if Q7e_ReasonProducts in (1 2) then ReasonProducts_high=0;
REFERENCES
Akerlof, K., E. W. Maibach, D. Fitzgerald, A. Y. Cedeno, and A. Neuman, 2013: Do people “personally experience” global warming, and if so how, and does it matter? Global Environ. Change, 23, 81–91, https://doi.org/10.1016/j.gloenvcha.2012.07.006.
Albright, E. A., and D. Crow, 2019: Beliefs about climate change in the aftermath of extreme flooding. Climatic Change, 155, 1–17, https://doi.org/10.1007/s10584-019-02461-2.
Arbuckle, M. B., 2017: The interaction of religion, political ideology, and concern about climate change in the United States. Soc. Nat. Resour., 30, 177–194, https://doi.org/10.1080/08941920.2016.1209267.
Bayes, R., and J. N. Druckman, 2021: Motivated reasoning and climate change. Curr. Opin. Behav. Sci., 42, 27–35, https://doi.org/10.1016/j.cobeha.2021.02.009.
Bergquist, M., A. Nilsson, and P. Wesley Schultz, 2019: Experiencing a severe weather event increases concern about climate change. Front. Psychol., 10, –220, https://doi.org/10.3389/fpsyg.2019.00220.
Blennow, K., 2012: Adaptation of forest management to climate change among private individual forest owners in Sweden. For. Policy Econ., 24, 41–47, https://doi.org/10.1016/j.forpol.2011.04.005.
Blennow, K., and J. Persson, 2009: Climate change: Motivation for taking measure to adapt. Global Environ. Change, 19, 100–104, https://doi.org/10.1016/j.gloenvcha.2008.10.003.
Blennow, K., J. Persson, M. Tomé, and M. Hanewinkel, 2012: Climate change: Believing and seeing implies adapting. PLOS ONE, 7, e50182, https://doi.org/10.1371/journal.pone.0050182.
Boag, A. E., J. Hartter, L. C. Hamilton, N. D. Christoffersen, F. R. Stevens, M. W. Palace, and M. J. Ducey, 2018: Climate change beliefs and forest management in eastern Oregon: Implications for individual adaptive capacity. Ecol. Soc., 23, 1, https://doi.org/10.5751/ES-10355-230401.
Bollen, K. A., 1989: Structural Equations with Latent Variables. John Wiley and Sons, 514 pp.
Borick, C. P., and B. G. Rabe, 2014: Weather or not? Examining the impact of meteorological conditions on public opinion regarding global warming. Wea. Climate Soc., 6, 413–424, https://doi.org/10.1175/WCAS-D-13-00042.1.
Brenkert-Smith, H., J. R. Meldrum, and P. A. Champ, 2015: Climate change beliefs and hazard mitigation behaviors: Homeowners and wildfire risk. Environ. Hazards, 14, 341–360, https://doi.org/10.1080/17477891.2015.1080656.
Broomell, S. B., J.-F. Winkles, and P. B. Kane, 2017: The perception of daily temperatures as evidence of global warming. Wea. Climate Soc., 9, 563–574, https://doi.org/10.1175/WCAS-D-17-0003.1.
Butler, B., S. M. Butler, J. Caputo, J. Dias, A. Robillard, and E. M. Sass, 2021a: Appendix 1: Summary tables. Family forest ownerships of the United States, 2018: Results from the USDA Forest Service, National Woodland Owner Survey, USDA General Tech. Rep. NRS-199, 46–51, https://doi.org/10.2737/NRS-GTR-199.
Butler, B., S. M. Butler, J. Caputo, J. Dias, A. Robillard, and E. M. Sass, 2021b: Family forest ownerships of the United States, 2018: Results from the USDA Forest Service, National Woodland Owner Survey. USDA General Tech. Rep. NRS-199, 52 pp., https://doi.org/10.2737/NRS-GTR-199.
Carlton, J. S., A. S. Mase, C. L. Knutson, M. C. Lemos, T. Haigh, D. P. Todey, and L. S. Prokopy, 2016: The effects of extreme drought on climate change beliefs, risk perceptions, and adaptation attitudes. Climatic Change, 135, 211–226, https://doi.org/10.1007/s10584-015-1561-5.
Cleland, D. T., J. A. Freeouf, J. E. Keys, G. J. Nowacki, C. A. Carpenter, and W. H. McNab, 2007: Ecological subregions: Sections and subsections for the conterminous United States. USDA General Tech. Rep. WO-76D, 1 pp., https://doi.org/10.2737/WO-GTR-76D.
Dale, V. H., and Coauthors, 2001: Climate change and forest disturbances. BioScience, 51, 723–734, https://doi.org/10.1641/0006-3568(2001)051[0723:CCAFD]2.0.CO;2.
Demski, C., S. Capstick, N. Pidgeon, R. G. Sposato, and A. Spence, 2017: Experience of extreme weather affects climate change mitigation and adaptation responses. Climatic Change, 140, 149–164, https://doi.org/10.1007/s10584-016-1837-4.
Dillman, D. A., J. D. Smyth, and L. M. Christian, 2014: Internet, Phone, Mail, and Mixed-Mode Surveys: The Tailored Design Method. John Wiley and Sons, 509 pp.
Druckman, J. N., and M. C. McGrath, 2019: The evidence for motivated reasoning in climate change preference formation. Nat. Climate Change, 9, 111–119, https://doi.org/10.1038/s41558-018-0360-1.
Duveneck, M. J., and R. M. Scheller, 2016: Measuring and managing resistance and resilience under climate change in northern Great Lake forests (USA). Landsc. Ecol., 31, 669–686, https://doi.org/10.1007/s10980-015-0273-6.
Fischer, A. P., 2019: Adapting and coping with climate change in temperate forests. Global Environ. Change, 54, 160–171, https://doi.org/10.1016/j.gloenvcha.2018.10.011.
Fischer, A. P., J. D. Kline, A. A. Ager, S. Charnley, and K. A. Olsen, 2014: Objective and perceived wildfire risk and its influence on private forest landowners’ fuel reduction activities in Oregon’s (USA) ponderosa pine ecoregion. Int. J. Wildland Fire, 23, 143–153, https://doi.org/10.1071/WF12164.
Fischer, A. P., M. Russo, G. Powers, and J. G. Powers, 2022: Behavioral adaptation to climate change: New insights on psychosocial frameworks from the context of managed forests. Sustainability Sci., 17, 1023–1036, https://doi.org/10.1007/s11625-021-01085-9.
Fownes, J. R., and S. B. Allred, 2019: Testing the influence of recent weather on perceptions of personal experience with climate change and extreme weather in New York State. Wea. Climate Soc., 11, 143–157, https://doi.org/10.1175/WCAS-D-17-0107.1.
Gärtner, L., and H. Schoen, 2021: Experiencing climate change: Revisiting the role of local weather in affecting climate change awareness and related policy preferences. Climatic Change, 167, 31, https://doi.org/10.1007/s10584-021-03176-z.
Goebbert, K., H. C. Jenkins-Smith, K. Klockow, M. C. Nowlin, and C. L. Silva, 2012: Weather, climate, and worldviews: The sources and consequences of public perceptions of changes in local weather patterns. Wea. Climate Soc., 4, 132–144, https://doi.org/10.1175/WCAS-D-11-00044.1.
Grothmann, T., and A. Patt, 2005: Adaptive capacity and human cognition: The process of individual adaptation to climate change. Global Environ. Change, 15, 199–213, https://doi.org/10.1016/j.gloenvcha.2005.01.002.
Grotta, A. T., J. H. Creighton, C. Schnepf, and S. Kantor, 2013: Family forest owners and climate change: Understanding, attitudes, and educational needs. J. For., 111, 87–93, https://doi.org/10.5849/jof.12-052.
Hamilton, L. C., and M. D. Stampone, 2013: Blowin’ in the wind: Short-term weather and belief in anthropogenic climate change. Wea. Climate Soc., 5, 112–119, https://doi.org/10.1175/WCAS-D-12-00048.1.
Hamilton, L. C., J. Hartter, B. D. Keim, A. E. Boag, M. W. Palace, F. R. Stevens, and M. J. Ducey, 2016: Wildfire, climate, and perceptions in northeast Oregon. Reg. Environ. Change, 16, 1819–1832, https://doi.org/10.1007/s10113-015-0914-y.
Hamilton, L. C., M. Lemcke-Stampone, and C. Grimm, 2018: Cold winters warming? Perceptions of climate change in the north country. Wea. Climate Soc., 10, 641–652, https://doi.org/10.1175/WCAS-D-18-0020.1.
Hamilton, M., A. P. Fischer, S. D. Guikema, and G. Keppel-Aleks, 2018: Behavioral adaptation to climate change in wildfire-prone forests. Wiley Interdiscip. Rev.: Climate Change, 9, e553, https://doi.org/10.1002/wcc.553.
Handler, S. , and Coauthors, 2014a: Michigan forest ecosystem vulnerability assessment and synthesis: A report from the Northwoods Climate Change Response Framework project. USDA General Tech. Rep. NRS-129, 242 pp., https://doi.org/10.2737/NRS-GTR-129.
Handler, S., C. W. Swanston, P. R. Butler, L. A. Brandt, M. K. Janowiak, M. D. Powers, and P. D. Shannon, 2014b: Climate change vulnerabilities within the forestry sector for the midwestern United States. Climate Change in the Midwest: A Synthesis Report for the National Climate Assessment, J. A. Winkler et al., Eds., Island Press, 114–151.
Harrison, S., J. Herbohn, and A. Niskanen, 2002: Non-industrial, smallholder, small-scale and family forestry: What’s in a name? Small-Scale For. Econ. Manage. Policy, 1, 1–11, https://doi.org/10.1007/s11842-002-0001-3.
Hayduk, L. A., 1987: Structural Equation Modeling with LISREL: Essentials and Advances. The Johns Hopkins University Press, 405 pp.
Hornsey, M. J., E. A. Harris, P. G. Bain, and K. S. Fielding, 2016: Meta-analyses of the determinants and outcomes of belief in climate change. Nat. Climate Change, 6, 622–626, https://doi.org/10.1038/nclimate2943.
Howe, P. D., 2018: Perceptions of seasonal weather are linked to beliefs about global climate change: Evidence from Norway. Climatic Change, 148, 467–480, https://doi.org/10.1007/s10584-018-2210-6.
Howe, P. D., and A. Leiserowitz, 2013: Who remembers a hot summer or a cold winter? The asymmetric effect of beliefs about global warming on perceptions of local climate conditions in the U.S. Global Environ. Change, 23, 1488–1500, https://doi.org/10.1016/j.gloenvcha.2013.09.014.
Howe, P. D., H. Boudet, A. Leiserowitz, and E. W. Maibach, 2014: Mapping the shadow of experience of extreme weather events. Climatic Change, 127, 381–389, https://doi.org/10.1007/s10584-014-1253-6.
Howe, P. D., J. R. Marlon, M. Mildenberger, and B. S. Shield, 2019: How will climate change shape climate opinion? Environ. Res. Lett., 14, 113001, https://doi.org/10.1088/1748-9326/ab466a.
Janowiak, M. K., and Coauthors, 2014: Forest ecosystem vulnerability assessment and synthesis for northern Wisconsin and western Upper Michigan: A report from the Northwoods Climate Change Response Framework project. USDA General Tech. Rep. NRS-136, 247 pp., https://doi.org/10.2737/NRS-GTR-136.
Joshi, S., and K. G. Arano, 2009: Determinants of private forest management decisions: A study on West Virginia NIPF landowners. For. Policy Econ., 11, 118–125, https://doi.org/10.1016/j.forpol.2008.10.005.
Keller, C., M. Siegrist, and H. Gutscher, 2006: The role of the affect and availability heuristics in risk communication. Risk Anal., 26, 631–639, https://doi.org/10.1111/j.1539-6924.2006.00773.x.
Khanal, P. N., D. L. Grebner, I. A. Munn, S. C. Grado, R. K. Grala, J. E. Henderson, and M. K. Measells, 2016: Nonindustrial private forest landowner beliefs toward climate change and carbon sequestration in the southern United States. J. For., 114, 524–531, https://doi.org/10.5849/jof.15-033.
Khanal, P. N., D. L. Grebner, I. A. Munn, S. C. Grado, R. K. Grala, and J. E. Henderson, 2017: Typology of nonindustrial private forest landowners and forestry behavior: Implications for forest carbon sequestration in the southern US. Small-Scale For., 16, 419–434, https://doi.org/10.1007/s11842-017-9363-4.
Kline, R. B., 2012: Assumptions in structural equation modeling. Handbook of Structural Equation Modeling, R. H. Hoyle, Ed., Guilford Press, 111–125.
Konisky, D. M., L. Hughes, and C. H. Kaylor, 2016: Extreme weather events and climate change concern. Climatic Change, 134, 533–547, https://doi.org/10.1007/s10584-015-1555-3.
Li, C. H., 2016: The performance of ML, DWLS, and ULS estimation with robust corrections in structural equation models with ordinal variables. Psychol. Methods, 21, 369–387, https://doi.org/10.1037/met0000093.
MacKinnon, D. P., J. L. Krull, and C. M. Lockwood, 2000: Equivalence of the mediation, confounding and suppression effect. Prev. Sci., 1, 173–181, https://doi.org/10.1023/A:1026595011371.
Malmsheimer, R. W. , and Coauthors, 2008: Forest management solutions for mitigating climate change in the United States. J. For., 106, 115–117.
Marlon, J. R., S. van der Linden, P. D. Howe, A. Leiserowitz, S. H. L. Woo, and K. Broad, 2019: Detecting local environmental change: The role of experience in shaping risk judgments about global warming. J. Risk Res., 22, 936–950, https://doi.org/10.1080/13669877.2018.1430051.
Marquart-Pyatt, S., A. K. Jorgenson, and L. C. Hamilton, 2015: Methodological approaches for sociological research on climate change. Climate Change and Society: Sociological Perspectives, R. E. Dunlap and R. J. Brulle, Eds., Oxford University Press, 369–411.
Millar, C. I., and N. L. Stephenson, 2015: Temperate forest health in an era of emerging megadisturbance. Science, 349, 823–826, https://doi.org/10.1126/science.aaa9933.
MTBS, 2018: Fire occurrence dataset. USGS, accessed 19 July 2018, https://www.mtbs.gov/direct-download.
Muthén, B. O., L. K. Muthén, and T. Asparouhov, 2016: Regression and Mediation Analysis Using Mplus. Muthén and Muthén, 519 pp.
Myers, T. A., E. W. Maibach, C. Roser-Renouf, K. Akerlof, and A. A. Leiserowitz, 2013: The relationship between personal experience and belief in the reality of global warming. Nat. Climate Change, 3, 343–347, https://doi.org/10.1038/nclimate1754.
Newman, T. P., E. C. Nisbet, and M. C. Nisbet, 2018: Climate change, cultural cognition, and media effects: Worldviews drive news selectivity, biased processing, and polarized attitudes. Public Understanding Sci., 27, 985–1002, https://doi.org/10.1177/0963662518801170.
Niles, M. T., and N. D. Mueller, 2016: Farmer perceptions of climate change: Associations with observed temperature and precipitation trends, irrigation, and climate beliefs. Global Environ. Change, 39, 133–142, https://doi.org/10.1016/j.gloenvcha.2016.05.002.
Ogunbode, C. A., G. Böhm, S. B. Capstick, C. Demski, A. Spence, and N. Tausch, 2019a: The resilience paradox: Flooding experience, coping and climate change mitigation intentions. Climate Policy, 19, 703–715, https://doi.org/10.1080/14693062.2018.1560242.
Ogunbode, C. A., C. Demski, S. B. Capstick, and R. G. Sposato, 2019b: Attribution matters: Revisiting the link between extreme weather experience and climate change mitigation responses. Global Environ. Change, 54, 31–39, https://doi.org/10.1016/j.gloenvcha.2018.11.005.
Ontl, T. A., C. Swanston, L. A. Brandt, P. R. Butler, A. W. D’Amato, S. D. Handler, M. K. Janowiak, and P. D. Shannon, 2018: Adaptation pathways: Ecoregion and land ownership influences on climate adaptation decision-making in forest management. Climatic Change, 146, 75–88, https://doi.org/10.1007/s10584-017-1983-3.
Paxton, P., J. R. Hipp, and S. Marquart-Pyatt, 2011: Nonrecursive Models: Endogeneity, Reciprocal Relationships, and Feedback Loops. SAGE, 128 pp.
Price, J. C., I. A. Walker, and F. Boschetti, 2014: Measuring cultural values and beliefs about environment to identify their role in climate change responses. J. Environ. Psychol., 37, 8–20, https://doi.org/10.1016/j.jenvp.2013.10.001.
Pryor, S. C., D. Scavia, C. Downer, L. Gaden, L. Iverson, R. Nordstrom, J. Patz, and G. P. Robertson, 2014: Midwest. Climate Change Impacts in the United States: The Third National Climate Assessment, J. M. Melillo, T. C. Richmond, and G.W. Yohe, Eds., U.S. Global Change Research Program, 418–440.
Ripberger, J. T., H. C. Jenkins-Smith, C. L. Silva, D. E. Carlson, K. Gupta, N. Carlson, and R. E. Dunlap, 2017: Bayesian versus politically motivated reasoning in human perception of climate anomalies. Environ. Res. Lett., 12, 114004, https://doi.org/10.1088/1748-9326/aa8cfc.
Sambrook, K., E. Konstantinidis, S. Russell, and Y. Okan, 2021: The role of personal experience and prior beliefs in shaping climate change perceptions: A narrative review. Front. Psychol., 12, 669911, https://doi.org/10.3389/fpsyg.2021.669911.
Sánchez, J. J., R. Marcos-Martinez, L. Srivastava, and N. Soonsawad, 2021: Valuing the impacts of forest disturbances on ecosystem services: An examination of recreation and climate regulation services in U.S. national forests. Trees For. People, 5, 100123, https://doi.org/10.1016/j.tfp.2021.100123.
Seidl, R., and Coauthors, 2017: Forest disturbances under climate change. Nat. Climate Change, 7, 395–402, https://doi.org/10.1038/nclimate3303.
Shao, W., and K. Goidel, 2016: Seeing is believing? An examination of perceptions of local weather conditions and climate change among residents in the U.S. Gulf Coast. Risk Anal., 36, 2136–2157, https://doi.org/10.1111/risa.12571.
Spence, A., W. Poortinga, C. Butler, and N. F. Pidgeon, 2011: Perceptions of climate change and willingness to save energy related to flood experience. Nat. Climate Change, 1, 46–49, https://doi.org/10.1038/nclimate1059.
Sugerman, E. R., Y. Li, and E. J. Johnson, 2021: Local warming is real: A meta-analysis of the effect of recent temperature on climate change beliefs. Curr. Opin. Behav. Sci., 42, 121–126, https://doi.org/10.1016/j.cobeha.2021.04.015.
Swanston, C., and Coauthors, 2018: Vulnerability of forests of the midwest and northeast United States to climate change. Climatic Change, 146, 103–116, https://doi.org/10.1007/s10584-017-2065-2.
Taylor, A., W. B. De Bruin, and S. Dessai, 2014: Climate change beliefs and perceptions of weather-related changes in the United Kingdom. Risk Anal., 34, 1995–2004, https://doi.org/10.1111/risa.12234.
Tversky, A., and D. Kahneman, 1973: Availability: A heuristic for judging frequency and probability. Cognit. Psychol., 5, 207–232, https://doi.org/10.1016/0010-0285(73)90033-9.
U.S. Federal Government, 2015: U.S. Climate Resilience Toolkit Climate Explorer. Accessed 8 June 2018, https://crt-climate-explorer.nemac.org.
U.S. Forest Service, 2018: Insect & disease detection survey (IDS) data. Accessed 12 April 2018, https://www.fs.fed.us/foresthealth/applied-sciences/mapping-reporting/detection-surveys.shtml.
van der Linden, S., 2014: On the relationship between personal experience, affect and risk perception: The case of climate change. Eur. J. Soc. Psychol., 44, 430–440, https://doi.org/10.1002/ejsp.2008.
Weber, E. U., 2010: What shapes perceptions of climate change? Wiley Interdiscip. Rev.: Climate Change, 1, 332–342, https://doi.org/10.1002/wcc.41.
West, S. G., A. B. Taylor, and W. Wu, 2012: Model fit and model selection in structural equation modeling. Handbook of Structural Equation Modeling, R. H. Hoyle, Ed., Guilford Press, 209–231.
Whitmarsh, L., 2008: Are flood victims more concerned about climate change than other people? The role of direct experience in risk perception and behavioural response. J. Risk Res., 11, 351–374, https://doi.org/10.1080/13669870701552235.
Wuebbles, D. J., 2016: Setting the stage for risk management: Severe weather under a changing climate. Risk Analysis of Natural Hazards: Interdisciplinary Challenges and Integrated Solutions, P. Gardoni, C. Murphy, and A. Rowell, Eds., Springer, 61–80.
Wuebbles, D. J., K. Kunkel, M. Wehner, and Z. Zobel, 2014: Severe weather in United States under a changing climate. Eos Trans. Amer. Geophys. Union, 95, 149–150, https://doi.org/10.1002/2014EO180001.