1 Introduction
Agriculture is sensitive to water availability, with climatic extremes leading to substantial yield losses (Lesk et al.
2016). Droughts constrain crop productivity and pose an ongoing threat to global food production which is likely to increase due to climate change, population growth and pressure on water resources (Daryanto et al.
2017; Leng and Hall
2019). Building agricultural drought resilience is of crucial importance to global food security and human well-being (Carrão et al.
2016; Challinor et al.
2014). However, many factors limit effective adaptation, including informational, attitudinal and behavioural barriers (Howden et al.
2007). Understanding how to overcome these barriers and increase the adaptive capacity of farmers must be prioritised to support agricultural development, crisis prevention and vulnerability reduction (Lipper et al.
2014). As such, research has increasingly focused on how farmers’ understandings and perceptions of climate change determine their adaptation decisions (Deressa et al.
2011; Mertz et al.
2009; Sutcliffe et al.
2016), with studies addressing the cognitive processes that link climate perceptions with decisions to implement action (Grothmann and Patt
2005; Truelove et al.
2015).
The attitudinal factors that determine decisions to employ a protective measure in response to a threat (i.e. adaptation) have been modelled using the Protection Motivation Theory (PMT) conceptual framework. The PMT framework identifies two cognitive pathways which are believed to determine whether individuals decide to employ protective behaviour. These pathways are risk appraisal and adaptation appraisal. Risk appraisal combines an assessment of the potential severity of a negative impact with the likelihood that such an impact will occur. Adaptation appraisal combines an assessment of how effectively a proposed response would curtail the threat of the negative impact (response efficacy), with an assessment of the capacity of the individual to employ that particular response (self-efficacy), along with an assessment of the potential costs of doing so (Milne et al.
2000).
PMT was first employed in the 1970s within social psychology studies of health behaviours (Rogers
1975) and has since been highly influential across a range of disciplinary areas including business studies, computing and environmental science. It has been applied in relation to a wide array of different threats, including information security (Haag et al.
2021; Herath and Rao
2009), pollution risks (Wang et al.
2019) and the threat of extreme weather and climate change (Bubeck et al.
2012; Kuruppu and Liverman
2011; Grothmann and Patt
2005; Grothmann and Reusswig
2006). Most studies related to extreme weather have concentrated on flood protection, often in a European context (e.g. Babcicky and Seebauer
2017; Bamberg et al.
2017; Bradford et al.
2012; Poussin et al.
2014). Few studies have used PMT to explore drought adaptation in the Global South, although a body of research is building in Asian (e.g. Keshavarz and Karami
2016; Truelove et al.
2015) and African (e.g. Gebrehiwot and van der Veen
2015; Tabe-Ojong et al.
2020; Wens et al.
2021) contexts.
Knowledge and understanding of climate change are positively associated with farmers’ adaptation intentions (Ngo et al.
2020). Studies have noted the influence of different sources of information for risk and adaptation appraisal (Milne et al.
2000), including social networks (Babcicky and Seebauer
2017; Haer et al.
2016) and public institutions (Grothmann and Reusswig
2006), and the role of institutions in communicating climate information to influence local level responses (Dorward et al.
2020; Haer et al.
2016; Steynor et al.
2021). The provision of drought warnings, forecasts and advice can effectively build resilience to drought by mitigating impacts and leading to faster recoveries within farming communities in developing countries (Ewbank et al.
2019). Optimising the provision of this information is an increasingly key necessity for policy makers, climate services and agricultural development practitioners.
This paper aims to build understanding of the processes that determine adaptation to drought amongst farmers in Northern Thailand. In it, we explore and compare the roles that risk appraisal and adaptation appraisal play as determinants of past adaptations and their perceived success and as determinants of responsiveness to official adaptation recommendations and the desire to implement further adaptations in future. We also investigate the extent to which exposure to and perceptions of both formal and informal drought communications are associated with these adaptation outcomes.
Key questions:
1.
To what extent do PMT and drought communication variables influence adaptation outcome variables?
2.
Which determinants affect adaption and risk appraisal?
3.
What role do socioeconomic factors (age, education and wealth) play in determining access to drought communications?
Based on these questions, we discuss the extent to which PMT variables predict adaptation outcome variables and how institutional drought communications influence farmers’ motivations and actions to protect their production activities from the risk of drought. By using the PMT framework to understand the relative importance of farmers’ appraisals of risk and of adaptive responses within their agricultural decision-making and illustrating the influence of drought communications, the paper provides communication design recommendations that should increase farmers’ implementation of adaptations to reduce the negative impacts of drought on agricultural livelihoods.
3 Results
In the “
3” section, RQ1 presents the factors influencing the four adaptation outcome variables using individual regression models. In each case, the influence of the PMT and drought communication variables on the adaptation outcome variables is described, and other influential variables are identified. RQ2 explores factors affecting adaptation and risk appraisal, and RQ3 identifies relationships between adaptation, drought communications and socioeconomic characteristics (age, education and wealth).
3.1 RQ1: To what extent do the PMT and drought communication variables influence adaptation outcome variables?
3.1.1 Total adaptation count (regression model Ai)
Model Ai was significant (
F(8) = 12.232,
p = .000) and accounted for 37% of the variance in the total adaptation count scores (with an adjusted
R2 of .37). Of the PMT variables, only adaptation appraisal significantly predicted total adaptation count, risk appraisal did not. Of the adaptation appraisal components, the statement “Appropriate adaptation strategies exist for protecting my household farm from drought” was a stronger determinant than the aggregate “adaptation appraisal” variable; hence, only the “Appropriate…” statement variable was selected for inclusion in the model. Of the socioeconomic factors, only age was found to contribute significantly, with a negative relationship evident between age and the number of adaptations employed. In terms of drought communication variables, the count of types of drought information received significantly predicted total adaptation counts, as did agreement with the information efficacy statements, “early enough”, “accurate”, “relevant” and “best format”, and the number of farmers that information had been received from. Whilst the relationship between total adaptation count and the statements “relevant” and “accurate” were positive, the relationships with the statements “early enough” and “best format” were negative, suggesting that as the number of adaptations undertaken increased, agreement with these particular statements declined (see Table
2).
Table 2
Linear model of predictors of total adaptation count (model Ai, adjusted R2 = .37, F(8) = 12.232, p =.000)
(Constant) | 1.140 | 1.394 | | 0.818 | 0.415 | −1.615 | 3.896 |
“Appropriate adaptation strategies exist for protecting my household farm from drought” | 0.839 | 0.145 | 0.386 | 5.774 | 0.000 | 0.552 | 1.126 |
Age group | −0.712 | 0.230 | −0.206 | −3.089 | 0.002 | −1.167 | −0.256 |
Count of different drought monitoring information types received | 0.478 | 0.138 | 0.254 | 3.453 | 0.001 | 0.204 | 0.751 |
“The information provided arrives early enough in the season to be useful to my household farm” | −0.527 | 0.168 | −0.233 | −3.140 | 0.002 | −0.859 | −0.195 |
“The information provided is usually relevant to the farming in the household farm activities” | 0.605 | 0.253 | 0.207 | 2.393 | 0.018 | 0.105 | 1.105 |
“I am confident that the information provided is accurate” | 0.722 | 0.224 | 0.243 | 3.217 | 0.002 | 0.278 | 1.165 |
“The information is provided in the best format for farmers to access and understand it easily” | −0.845 | 0.261 | −0.273 | −3.240 | 0.001 | −1.360 | −0.329 |
Information was received from how many farmers | 0.171 | 0.066 | 0.173 | 2.572 | 0.011 | 0.040 | 0.302 |
3.1.2 Proportion of adaptations perceived as successful for longer term drought vulnerability reduction (regression model Aii)
Analysis showed that the drought communications variables did not improve model Aii, and this group of variables was excluded. The final model, which incorporated age, wealth, adaptation appraisal, animal husbandry and land within a supported irrigation zone, was significant (
F(5) = 5.132,
p = .000) and accounted for 16% of the variance in the proportion of adaptations perceived as successful. All included variables significantly determined the proportion of adaptations perceived as successful, with negative relationships pertaining with wealth and animal husbandry, indicating that wealthier farmers and those that raised livestock tended to perceive a lower proportion of their past adaptations as successful than other farmers (see Table
3).
Table 3
Linear model of predictors of proportion of adaptations perceived to successfully reduce vulnerability to future droughts (model Aii, adjusted R2 = .16, F(5) = 5.132, p = .000)
(Constant) | 0.231 | 0.174 | | 0.186 | −0.112 | 0.575 |
Age | 0.136 | 0.047 | 0.221 | 0.004 | 0.044 | 0.228 |
Wealth | −0.013 | 0.005 | −0.182 | 0.017 | −0.023 | −0.002 |
Adaptation appraisal | 0.075 | 0.026 | 0.222 | 0.004 | 0.025 | 0.126 |
Household raises animals | −0.135 | 0.054 | −0.184 | 0.014 | −0.242 | −0.028 |
Household has land in official irrigation zone | 0.14 | 0.057 | 0.179 | 0.016 | 0.027 | 0.252 |
3.1.3 Respondent adapted their agricultural practice in response to receiving a drought warning (regression model Aiii)
Model Aiii, which used binary logistic regression to investigate decisions to adapt in response to official advice, was significant (
χ2 (5) = 36.573,
p = .000) and accounted for 29% of the variance. Both fruit production and rice production were found to interact significantly with having received financial drought compensation, although when both fruit and rice were included in the same model, the significance of the rice × compensation interaction was lost (Table
4). Nevertheless, this result points to a potentially important relationship between past receipt of drought compensation and behavioural responses to drought information amongst producers of these crops, which reduced the likelihood that households would have changed something about their agricultural or livelihood activities in response to receiving drought information. However, given the cross-sectional nature of our data, we cannot discern whether failing to adapt in response to advice came before or after receiving compensation. The other significant determinants (all of which increased the likelihood that households would indicate having changed something in response to the information they received) were the number of droughts that the household had experienced over the last 10 years, the respondent’s adaptation appraisal score and if the respondent had received weather forecast information (Table
4). Amongst these variables, having received a weather forecast had an odds ratio of 3.15, meaning that the odds of households adapting in response to official advice were 3.15 times greater amongst households that had received weather forecast information than amongst those that had not. For every additional instance of drought the household reported experiencing in the last decade, the odds that they had adapted in response to a drought warning were increased by 1.41, and for every unit increase in adaptation appraisal, the odds increased by 1.56.
Table 4
Binary logistic regression results showing predictors of whether the household changed anything in response to receiving a drought warning (model Aiii). R2 .21 (Cox & Snell) and .285 (Nagelkerke). Model χ2 (5) = 36.573, p = .000)
(Constant) | −0.562 | 0.366 | 2.36 | 0.124 | 0.57 | | |
Received financial drought compensation × fruit production | −1.182 | 0.433 | 7.454 | 0.006 | 0.307 | 0.131 | 0.716 |
Instances of drought in last 10 years | 0.346 | 0.149 | 5.411 | 0.02 | 1.413 | 1.056 | 1.891 |
Adaptation appraisal | 0.44 | 0.196 | 5.055 | 0.025 | 1.552 | 1.058 | 2.277 |
Received weather forecast information | 1.148 | 0.52 | 4.874 | 0.027 | 3.153 | 1.138 | 8.74 |
Although not found to contribute significantly to the logistic regression, growers that indicated having received information via a social network messaging group were statistically more likely to have changed something in response to official advice (χ2 = 3.991, p = .046), as were growers that had been educated to high school level or above (χ2 = 7.780, p = .005).
3.1.4 Respondent indicated a desire to implement further adaptations (regression model Aiv)
Model Aiv was significant (
χ2 (4) = 26.164,
p = .000) and accounted for 23% of the variance. Within the model, the odds that a farmer would indicate wanting to make further drought adaptations were significantly increased (by 1.44) for each additional drought adaptation that they had previously implemented (Table
5). The odds also increased if they indicated lacking money for implementing adaptations, feeling well informed about drought risks or perceiving drought information to be in the right format for farmers. The odds were lowered amongst farmers that indicated stronger agreement that the drought information received had been relevant (Table
5). Whilst this last finding was unexpected, a potential explanation could be that a perceived lack of relevant drought information caused farmers to delay implementing their desired adaptations.
Table 5
Binary logistic regression results showing predictors of “Are there any further changes you would like to implement on the farm, but that you did not yet?” (model Aiv). R2 .17 (Cox & Snell) and .229 (Nagelkerke). Model χ2 (4) = 26.164, p = .000)
Constant | −5.852 | 2.027 | 8.334 | 0.004 | 0.003 | | |
“I have not got enough money to implement the necessary adaptation strategies” | 0.626 | 0.212 | 8.689 | 0.003 | 1.869 | 1.233 | 2.833 |
Total number of adaptations | 0.365 | 0.111 | 10.869 | 0.001 | 1.44 | 1.159 | 1.789 |
“I consider myself to be well informed about drought risks” | 0.633 | 0.264 | 5.746 | 0.017 | 1.883 | 1.122 | 3.158 |
“The information provided is usually relevant to the farming in the household farm activities” | −0.959 | 0.409 | 5.512 | 0.019 | 0.383 | 0.172 | 0.853 |
“The information is provided in the best format for farmers to access and understand it easily” | 1.037 | 0.394 | 6.935 | 0.008 | 2.82 | 1.304 | 6.102 |
Whilst risk appraisal was not a significant predictor of a desire to implement further adaptations in this model, a t-test revealed that farmers that indicated a desire to adapt further had significantly higher risk appraisal scores than farmers that indicated that they did not wish to implement any further changes (x ̅ = 11.9 vs x ̅ = 11.2, t = 1.985, p = 0.049).
3.2 RQ2: Which determinants affect adaption and risk appraisal?
3.2.1 Adaptation appraisal
Multiple significant associations were found using Pearson’s correlation coefficient between adaptation appraisal and the drought information communications variables. In particular, confidence in implementing adaptation strategies correlated with the number of types of climate information that had been received (r = .23, p = .04), the perception that climate information had helped the farmer prepare in the past (r = .40, p = .001) and that the information was provided in the best format for farmers (r = .36, p =.001).
Using multiple regression, adaptation appraisal scores were significantly determined by education (Beta .271, p = .001), the total number of adaptations that had been undertaken in the past (Beta .220, p = .005), the proportion of past adaptations perceived as successful (Beta .213, p = .007) and the drought information efficacy score (Beta .199, p = .013). The model was significant (F, 10.478; p = .000), and these four variables accounted for a fifth of the variance in adaptation appraisal (adjusted R2 = .21).
Wealth correlated significantly with adaptation appraisal (r = .207, p = .006), but did not contribute significantly to the regression model due to high semi-partial correlation with education. Older farmers (aged over 60) were also found to have significantly lower adaptation appraisal scores (mean = 0.27 vs 0.63 for younger farmers, t = −2.161, p = .032).
Adaptation appraisal was lower if a respondent perceived there to be higher costs associated with undertaking adaptations. Respondents’ perceptions of adaptation costs were measured via their scores for the availability of time, energy and finance for implementing adaptations. Whilst correlation between the availability of time and energy for implementing adaptations was high (.61, p = .000), there were no significant associations between financial capacity to adapt and these other factors, suggesting that financial constraints operate independently for a large proportion of the respondents.
Spearman’s rank tests identified significant weak positive correlations between age and agreement that the respondent could not implement adaptations due to insufficient energy and time (ρ (rho) =.20, p = .009 and ρ (rho) = .16, p = .032, respectively). Conversely, education correlated negatively with all three types of perceived adaptation cost: energy (ρ (rho) = −.25, p =.001), time (ρ (rho) = −.32, p = .000) and money (ρ (rho) = −.24, p = .002).
3.2.2 Risk appraisal
Farmers’ risk appraisal was not statistically associated with any of the climate information variables, apart from a weak positive association with how many other farmers the information was received from (r = .183, p = .016). Additionally, those that received information at a village meeting had significantly stronger mean agreement that their household was likely to be negatively affected by drought in the next 5 years, than those who received the information by other means (mean = 3.98 vs 3.59, t = 2.655, p = .021).
Growers that experienced drought in the most recent 3 years before the survey (in January 2020) displayed significantly higher agreement with the risk statements “The impacts of drought on my household farm are likely to be severe” (mean = 4.02, t = 2.951, p = .004) and “The impacts of drought on my farm are worse than before” (mean = 4.25, t = 5.316, p = .000) than those that did not have such recent experience of drought (means = 3.44 and 3.29, respectively). A positive correlation was observed with the total number of economic and crop drought impacts indicated by the farmer (r = .262, p = .000), and a weak negative correlation was observed between risk appraisal and adaptation appraisal (r = −.169, p = .026). These findings suggest that direct experience of drought influences risk appraisal, with more recent experiences having a more pronounced effect.
The three variables, years since last drought experience (Beta −.216, p = .011), number of farmers climate information was received from (Beta .185, p = .027) and household engages in animal husbandry (Beta .139, p = .096), resulted in a significant model (F = 5.721, p = .001) with an R2 of .096.
3.3 RQ3: What role do socioeconomic factors (age, education and wealth) play in determining access to drought communications?
Statistical tests demonstrated the role of socioeconomic factors in determining access to drought warnings. Older farmers reported receiving drought communications through significantly fewer channels (mean = 1.0 vs 1.3 for farmers aged under 60, t = −2.883, p = .004) and tended to report receiving drought communications from a smaller number of other farmers (although p = .059). Respondents in older age groups were less likely to have received drought warnings that were accompanied by advice on what to do (Fisher’s exact test = 14.998, p = .003), as were respondents with lower education levels (Fisher’s exact test = 33.219, p = .000).
Wealth was also a factor, with the highest mean household asset and land wealth scores found within households that received both a drought warning and advice on what to do (10.0), compared to lower scores amongst those that only received warnings (8.9), and those that received neither (5.4), F(2172) = 5.416, p = .005.
Adaptation appraisal scores followed the same pattern, with the lowest mean scores in households that received no warning or advice (−0.85), mean scores of 0.36 in households that only received the drought warning and mean scores of 0.73 in households that received both the warning and advice, F(2169) = 15.419, p = .000. Respondents that received drought warnings accompanied by adaptation advice were significantly more likely to have changed something as a result, than those that received the drought warning alone (67.5% compared to only 25.0%, χ2 = 24.696, p = .000).
5 Conclusion
The study supports protective behavioural changes (in this case, implementation of drought adaptations) being more closely linked to adaptation appraisal than risk appraisal. Further, it has indicated that an inverse relationship can pertain between these PMT constructs. Both institutional and local drought communications contribute in important ways to adaptation actions and perceptions, but institutional communications appear to operate more through pathways leading to adaptation via adaptation appraisal. Conversely, local communications are of greater significance for risk appraisal, although the study has not found evidence to support a direct pathway from risk appraisal to adaptation. Institutional drought communications should include adaptation advice to promote adaptation behaviour amongst agricultural producers, with drought communications framed in positive terms, emphasising the efficacy of adaptation recommendations and the feasibility of implementing them for the producers themselves. Policy makers should employ measures that support farmer participation in both local and institutional drought communication networks.
The study shows that farmers who are older, less educated or poorer are less likely to adapt to drought due to perceived costs and reduced exposure to and assimilation of drought information. More research is needed on how to shape and supply institutional drought and adaptation communications to meet the needs of these segments of agricultural populations more effectively. Whilst financial support may sufficiently address adaptation constraints for some, evidence of time and energy constraints for, particularly, older farmers indicate a need for different targeting of adaptation support for these farmers, including the provision of social safety nets in cases where adaptation is improbable. Evidence that those practicing animal husbandry perceive greater levels of drought risk and lower levels of adaptation success than other agricultural producers also suggests a need for specific targeted support.
Understanding adaptation as a continuous (rather than finite) process, characterised by feedbacks between experience and appraisal, means that efforts to build drought resilience must go beyond providing assistance only when drought is forecast or causing measurable impacts. Efforts to reduce drought vulnerability should provide long-duration, continuous institutional support for resilience building. Such support needs to engender pro-adaptive mindsets across all sections of agricultural communities by improving social and institutional drought communication networks, ensuring that organisations are receptive to climate and adaptation information fed back by farmers and developing habits across the board of engaging in ongoing climate, drought and adaptation dialogue.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.