Research part a on Deeper Water: Exploring Barriers and Opportunities for the Emergence of Adaptive Water Governance Social influence shapes adaptive water governance: empirical evidence from northwestern Pakistan

. Social-ecological change has placed unprecedented stress on water resources throughout the world. This has driven water users to employ a diverse range of adaptation strategies and necessitates new governance structures, such as adaptive water governance (AWG), which have the capacity to manage resources in the midst of uncertainty and complexity. As such, AWG has the potential to support household adaptation strategies; however, little empirical work has been done to identify the factors that facilitate the emergence of AWG. To address this gap, we conducted a household survey of 448 households in northwestern Pakistan, a post-conflict, water-scarce area where adaptive governance is needed to support rural livelihoods in the midst of numerous socioeconomic and environmental transformations. Indeed, we found that households in our study area perceived a range of changes to the water system, including but not limited to declines of fish populations, decreased quality and amount of river water, and an increase of local tourism. Respondents reported a range of adaptation strategies including increasing agricultural inputs, planting new crop varieties, and changing their domestic water supply system. In some cases, households employed these adaptation strategies despite economic barriers, and although many were willing to go against friends’ and community leaders’ opinions to adapt, and they were less likely to counter the opinions of family members. This reveals that households negotiate multiple factors in their decisions to adapt to social-ecological change; as such, there is a great need for flexible and collaborative governance systems such as AWG to support this complexity in household adaptation decision making. Further, we argue that the varying roles of social influence should be considered to align governance structures with household decision-making processes. Thus, we suggest that AWG will be more likely to emerge when decision makers involved in water management draw on existing informal institutions and cross-sectoral collaboration to reflect the complex ways water users adapt to social-ecological change.


INTRODUCTION
Rural livelihoods and the water systems on which they depend are under increasing stress and uncertainty from compounding social-ecological change (Foley et al. 2005, Vörösmarty et al. 2010, Mekonnen and Hoekstra 2016. This necessitates flexible and resilient governance systems that support adaptation to these changes. Adaptive water governance (AWG) has been presented as a governance system that creates the conditions to respond to uncertainty and address interactions between biophysical resources and social institutions (Huitema et al. 2009, Rijke et al. 2012. Thus, AWG is increasingly prescribed to support resilient water systems in the midst of increasing uncertainty and demand (Pahl-Wostl et al. 2012, Baird et al. 2016. Adaptive water governance draws on the principals of adaptive governance broadly, which is defined as a flexible, learning-based approach to ecosystem management based on multi-scalar collaboration between diverse networks of institutions (Akamani 2016). Thus, collaboration, public participation, and social learning have been identified as important components of AWG (Folke et al. 2005, Bodin and Crona 2009, Stein et al. 2011, Chaffin et al. 2016.
Household adaptive capacity is shaped in part by the governance structures found in the multi-scalar landscape in which adaptation occurs (Ivey et al. 2004, Elrick-Barr et al. 2014, Lockwood et al. 2015, Burnham and Ma 2018. In terms of AWG, previous studies have frequently shown that AWG provides opportunities to integrate water users' diverse values in management decisions (Bark et al. 2012, Huitema et al. 2009). It has also been posited that water users' adaptation is shaped in part by their trust in other water users, their trust in the water management institutions, and access to public participation in water governance, which are all key components of AWG (McCord et al. 2018). As such, this literature indicates that AWG can support water users' adaptative capacity in the midst of complex social-ecological change (Bayard et al. 2007, Karanja Ng'ang'a 2016, Li et al. 2017. In terms of household adaptation, socioeconomic demographic variables such as education level and age of household head, household size, income, and land ownership have each been identified as determinants of adaptation in many studies that focus on the relationships between access to social, political, financial, and natural capital and the decision to adapt (Croppenstedt et al. 2003, Below et al. 2012, Asfaw et al. 2019. At a household level, income, labor, and access to capital remain salient in adaptation decisions , Fosu-Mensah et al. 2012, Esham and Garforth 2013, Tessema et al. 2013, Antwi-Agyei et al. 2014). In addition to socioeconomic demographics, a growing body of adaptation literature has focused on how the informal institutions of social influence shape what adaptations are seen as acceptable (Curry et al. 2015, Cleaver and Whaley 2018, Porras et al. 2020). Social influence, or the ways in which behavior is shaped by descriptive and subjective norms in the local context, has been shown to affect household decision making in a variety of contexts such as natural resource management, adoption of conservation behaviors, changing livelihood strategies, and managing environmental risks and threats (e.g., Fielding 2008, Abrahamse and Steg 2013, Dang et al. 2014, van Valkengoed and Steg 2019. In some cases, social influence can be "more powerful than cost or considerations such as convenience or effectiveness" (Wolske et al. 2020:202) and is thus, another important factor that influences the learning and collaboration that is central to AWG. In fact, scholars have argued that a better understanding of social influence will support the emergence of AWG by responding to local needs and concerns (Kerner andThomas 2014, Porras et al. 2020). Thus, considering the role of social-economic demographics and social influence in adaptation decision making will allow governance structures to align with the informal institutions and social-economic context on the governance area (Cleaver andWhaley 2018, Porras et al. 2020).
We build upon this complex literature on AWG and adaptation to social-ecological change to examine the specific contexts and conditions that may foster or inhibit adaptation to socialecological change and the emergence of AWG among water users. Specifically, our study is situated in Pakistan, one of the most water-stressed countries in the world (Hofste 2019) where adaptive governance is needed to support rural livelihoods in the midst of numerous socioeconomic and environmental transformations. Specifically, in the northwestern province of Khyber Pakhtunkhwa (KP), the Swat and Kabul rivers supply water for irrigation, wild catch fishing and aquaculture, domestic use, hydropower, and domestic tourism. However, urbanization, industrialization, and agricultural intensification have increased pollutants in the water system (Ullah et al. 2013), and it is projected that climate change will increasingly stress Pakistan's water supply due to changes in precipitation and glacier behavior, causing increasing frequency of extreme events such as floods and droughts (Hussain and Mumtaz 2014). Finally, armed conflict displaced an estimated 2 million people around our research area in 2008, and in 2010, a massive flood further resulted in widespread destruction of infrastructure in the KP province (DDMU 2015). More recently, however, the area has seen development and rehabilitation, including the revitalization of the tourism industry (Hye and Khan 2013) as evidenced by an estimated 4.45 million tourists visiting sites around the Swat Valley in 2018 (World Bank 2019).
Nationally, water governance in Pakistan is guided by the 2018 National Water Policy as well as the 2005 National Environmental Policy, and the 2012 National Climate Change Policy (Government of Pakistan 2018). Water governance has largely been decentralized and a number of institutions are involved in the provision of domestic and irrigation water. Provincial irrigation and drainage authorities (PIDAs) and area water boards (AWBs) manage irrigation water, whereas farmers' organizations (FOs) and water user associations (WUAs) are responsible for rehabilitating irrigation canals and collecting water fees from users who receive water on a rotating schedule (FAO 2011). In terms of domestic water governance, public health engineering departments (PHEDs) are responsible for installing infrastructure and managing operation and maintenance in rural areas. Tehsil municipal authorities (TMAs) manage water supply operation and maintenance in urban areas although in some cities this responsibility has been delegated to Water and Sanitation Agencies (WASAs). It is important to note that governmentoperated infrastructure does not reach all residents in KP; thus, self-provisioning of water is common. NGOs also play a role in the installation of water systems in KP and often create or facilitate local governance structures to accompany the water infrastructure installed (Cooper 2018, Government of Pakistan 2018. At the same time, there has been concern that the current water governance system hinders communities' ability to adapt to complex changes in the area. This is due in part to conflict across spatial scales (i.e., provincial, regional, and international) and sectors (i.e., industry and agriculture) in addition to the inability of current structures to respond to changes in water availability (Yang et al. 2014). Thus, there is a need for flexible and collaborative AWG structures to support resilience to water stress and household adaptive capacity in KP.
Ecology and Society 27(3): 37 https://www.ecologyandsociety.org/vol27/iss3/art37/ As McCord et al. (2018) pointed out, although there is much research on water governance and household adaptation separately, little work has been done to link these topics. As such, our research sits at the intersection of examinations of AWG and household adaptation to social-ecological change and aims to: (1) better understand the role of socioeconomic and demographic characteristics and social influence in water users' adaptation strategies; and (2) inform the incorporation of social influence considerations in AWG to support household adaptation. As mentioned earlier, we understand adaptation actions to be employed in response to various social-ecological changes and stressors (Carr 2008, Forsyth and Evans 2013, Manuel-Navarrete and Pelling 2015, Burnham and Ma 2016, thus we do not isolate adaptation to water stress from adaptation to other forms of social-ecological change in our study. Rather, we acknowledge both the centrality of water in river-based communities alongside the complexity of adaptation decision making. This leads us to examine multiple adaptation actions and livelihood strategies rather than focusing solely on those that are directly reliant on the water system.

Research area
We conducted an in-person survey in three communities in the KP province in northwestern Pakistan: Madyan, Jehangira, and Landakay (see Fig. 1 and Table 1). Madyan and Landakay are located in the Swat District and Jehangira is located in Nowshera District. These communities were chosen to represent the livelihood strategies associated with the Swat and Kabul river ecosystems, including irrigated agriculture, wild catch and aquaculture fisheries, and tourism. Table 1 presents data at the district (i.e., Swat and Nowshera) level due to limited data availability at the community level. In summary, Jehangira (Nowshera District) is located on the Kabul River. It is closer to the urban center of Peshawar, and households in that area are less reliant on agriculture as a livelihood strategy compared to Swat District. Madyan and Landakay (Swat District) are located on the Swat River and are relatively farther away from urban centers than Nowshera, and households there are more reliant on agriculture than those in Nowshera.

Data collection
We surveyed 448 self-identified household heads in the 3 communities in the spring and summer of 2019.The survey collected data on household demographics, livelihood strategies, water management, perceived social-ecological changes, adaptation strategies, and perceived barriers to these adaptation strategies. The survey was designed based on findings from qualitative interviews conducted by a team of interviewers, including the first author, in the research area in 2018 and on a review of literature on key components of water governance (i.e., public participation, social networks, perception of management, etc.) and household adaptation (i.e., perceptions of socialecological change, commonly adopted adaptation strategies, barriers to adaptation, etc.).
Specifically, we asked respondents to indicate if their household adopted one of the 17 adaptation strategies listed in the survey in the past 10 years. The adaptation strategies included were based both on the aforementioned interviews as well as on existing literature documenting types of adaptation strategies (Agrawal andPerrin 2009, Burnham andMa 2016). This time period was chosen to include changes associated with the flood of 2010 and to allow respondents to use that significant event as a reference point. Because AWG functions in the context of complex socialecological changes (Huitema et al. 2009, Rijke et al. 2012) and we define adaptation to social-ecological changes in the context of multiple stressors, it is important for us to consider a wide range of adaptation strategies that households may adopt in our research area. Therefore, in addition to commonly documented adaptation strategies such as changing agricultural inputs and water supply, we also included decreased time fishing as an adaptation strategy. In fact, it was one of the most commonly reported strategies from our qualitative interviews when we asked local community members about how they had responded to social-ecological change. Specifically, decreased time fishing was described by some as a step toward seeking other livelihood strategies and diversifying their livelihoods, a strategy that has https://www.ecologyandsociety.org/vol27/iss3/art37/ long been studied as adaptation to social-ecological change (e.g., Smit and Skinner 2002, Osbahr et al 2008, Agrawal and Perrin 2009. That is, decreasing temporal investment in one livelihood strategy is an integral component of a households' construction and maintenance of a diverse portfolio of activities (Alobo Loison 2015). Decreasing time fishing was also described in terms of environmental management, another established type of adaptation (Burnham and Ma 2016). That is, local community members decreased the time they spent fishing to reduce overfishing, which they perceived to be contributing to the diminishing fish population. We also included migration as one of the possible adaptation strategies that respondents could check. Again, this emerged from our qualitative interviews as an important adaptation strategy. It has also been discussed in the adaptation literature. Both Agrawal and Perrin (2009) Fedele et al. 2020). Therefore, given how we define adaptation to social-ecological change, we included migration as one of the adaptation options. Additionally, in our survey, we only asked about livelihoodspecific adaptation strategies (i.e., increase agricultural inputs) to respondents who indicated engagement with the that livelihood (i.e., crop production). We asked all respondents about adaptation strategies (migrate, change domestic water supply) that are not specific to a particular livelihood.
The survey was piloted in similar communities in the neighboring Islamabad Capital Territory and questions were revised based on this process. In-person survey administration has been found to be most appropriate for rural areas such as our study communities and is conducive to longer or more complex survey designs (Neuman 2010). Therefore, the survey was conducted in-person by five enumerators who were graduate students or postdoctoral researchers from the University of Peshawar using a random walking sampling strategy. In this strategy, enumerators choose a random starting location, a direction to walk, and a number of houses to skip (e.g., sample every fifth house on the street) before they selected a house to conduct an in-person survey (Himelein et al. 2016). Although selection bias is a concern for using this method (Himelein et al. 2016), it is still considered an effective collection method in places like our research area where household records are limited, and internet and phone services vary, making other sampling strategies infeasible (Himelein et al. 2016). All survey data were collected using the Survey Solutions platform with handheld tablets. This study was approved by both Pakistani research authority at the University of Peshawar and the Purdue University Institutional Review Board. We also obtained local leaders' permission for data collection in each community.

Data analysis
Responses to questions about perceived social-ecological changes were analyzed using polychoric principal component analysis (PCA), which allowed us to reduce a large number of correlated variables into uncorrelated composite variables, called principal components (PCs), with a minimal loss of information (Field 2009, Pallant 2013. Standard PCA is designed for continuous variables. For our categorical variables, we calculated the polychoric correlation matrix and then conducted a PCA using the polychoric correlations (Kolenikov and Angeles 2009). To test the suitability of our data for PCA, we conducted both the Kaiser-Meyer-Olkin (KMO) test and Bartlett's test of sphericity. The KMO test measures the sampling adequacy for each variable in the model and shows the proportion of variance among variables that might be common variance. The closer the value is to one, the greater sampling adequacy; however, values above 0.5 are commonly considered acceptable (Hair et al. 2010). Bartlett's test of sphericity is used to test the null hypothesis that the correlation matrix is an identity matrix, which would indicate that the variables are unrelated and not appropriate for a PCA (Bartlett 1951). In our study, the KMO value was 0.64 and the Bartlett's test of sphericity was significant (p < 0.05). Hair et al. (2010) suggested that KMO values between 0.5 and 0.7 are mediocre but acceptable. Therefore, both tests indicate that PCA is an appropriate data reduction strategy for our data (Field 2009).
Generally, principal components with an eigenvalue of one or greater should be retained in a PCA (Kaiser 1958, Yong andPearce 2013). Principal component loadings of 0.50 or greater indicate a strong association among the survey items used to generate a particular PC (Cronbach 1951). This value of 0.5 has been used in several studies on adaptation decision making (e.g., Marshall andMarshall 2007, Burnham andMa 2017). In this study, polychoric PCA was performed on 448 complete observations and 15 survey items measuring perceived socialecological changes. The results are shown in Table 2. Five PCs had an eigenvalue greater than one and allowed for practical interpretation of the meaning of the PCs. Based on the associated survey items, the five PCs were labeled as tourism changes, agricultural changes, groundwater changes, solid waste changes, and fish changes ( Table 2). The scree plot of eigenvalues confirms our decision to retain these five PCs (Fig. 2). As a measure of scale reliability, Cronbach's alpha was calculated for each PC; 0.61 or higher is often considered moderately acceptable (Taber 2018), thus, these five PCs were retained. Because perceiving decreases in river water quality did not load significantly on derived PCs, it was left to be included as an individual variable in subsequent analyses.  After this PCA, we used a binomial logistic regression model to examine respondents' adoption of adaptation strategies as a dichotomous outcome (Comoé andSiegrist 2015, Mase et al. 2017). In this model the probability of the dependent variable is represented as: The model is represented as follows: where P is the probability of a household adopting an adaptation strategy and 1-P is the probability of a household not adopting an adaptation strategy. β 0 is the intercept, β 1 , β 2 … and β k are regression coefficients of the explanatory variables of X 1 , X 2 , X 3 , ..., X k . A greater-than-one odds ratio indicates a positive relationship (i.e., the odds of adopting an adaptation strategy increases as the explanatory variable increases), a less-than-one odds ratio indicates a negative relationship (i.e., the odds of adopting an adaptation strategy decreases as the explanatory variable increases), and an odds ratio of one indicates no relationship between the adoption of an adaptation strategy and the explanatory variables (Hosmer and Lemeshow 2000).
The response variables in our models measure if a respondent's household had adopted one of the following five adaptation strategies: increase agricultural inputs, migrate, decrease time fishing, change domestic water supply, or change crop variety. The response variable takes the value 1 if a respondent's household had adopted an adaptation strategy in the past 10 years and 0 otherwise (Table 3). These 5 adaptation strategies (out of the list of 17) were reported by 10% or more of our respondents. Following other studies on adaptation actions (Bryan et al. 2009), we focus on these five to allow for a more nuanced discussion of most commonly reported adaptations in this research area.
The explanatory variables in our models measure household socio-demographic characteristics, perceived social-ecological changes, and perceived economic and non-economic barriers to each adaptation strategy (Table 4). Following previous research on barriers (Croppenstedt et al. 2003, Dang et al. 2014, De Jalón et al. 2015, Marie et al. 2020, we used binary variables to measure the presence of barriers to adaptation. We did not ask respondents to report the number of times they considered each barrier in the last 10 years, nor were we able to use a Likert scale to measure the importance of a particular barrier due to the difficulty of recalling specific frequencies over a long period of time (Bound for each regression. The VIF scores were all below four, the rule of thumb criterion for multicollinearity (Hair et al. 2010).
Following methods from Wheeler et al. (2013), we modeled each adaptation strategy separately because of the heterogeneity of the adaptation strategies (Table 5). Nagelkerke's R² is used to evaluate the goodness of fit of the model (Nagelkerke 1991). Acceptable values for R² vary by discipline and specific method used (Moksony and Heged 1990). The Nagelkerke's R² values for our models are similar to those reported in previous studies of adaptation decision making (Comoé and Siegrist 2015, Truelove et al. 2015, Marie et al. 2020. All data analyses were conducted using Stata 16 and 17.

Socio-demographic and economic characteristics of survey respondents
An overview of the explanatory variables in the models are presented in Table 4. Overall, the socio-demographic and economic characteristics of our survey respondents seemed to be comparable to the broader population in our study area (Government of Pakistan Statistics Division 2010). The average respondent was 39.4 years old with 8.4 years of formal education. The mean household size was 10 persons, and 58% of our respondents reported living in households with a joint family structure, defined as "multi-generational families with two or more married children" (Ruggles 2010). This is slightly larger than the mean household size in KP (7.3 persons), likely due in part to our rural sample in which households are often larger than in urban areas (Pakistan Bureau of Statistics 2017). Our respondents' average farm size was 3.2 hectares (ha), which is larger than KP's average of 1.5 ha (Government of Pakistan Statistics Division 2010). However, more recent studies of agriculture in KP have reported large variation in farm size; Ullah et al. (2015) reported a mean of 2.38 ha and Khan et al. (2020) reported a mean of 7.02 ha. Our results fall within this variation of land ownership across the province. Our result could also indicate an increase in land ownership because the economy in the research area has stabilized after violent conflicts and disasters during the past decade (Hye and Khan 2013). Overall, 14.5% of our respondents reported that a household member held a leadership position in the community currently or in the past. Finally, our respondents owned an average of 2.6 head of cattle, included because it represents a source of wealth in the area (Ali and Rahut 2018).

Perceived social-ecological changes, adaptation strategies adopted, and barriers to adaptation action
We asked respondents what social-ecological changes they had observed in the past 10 years in their community. There was nearly universal agreement that river water quality had decreased (95%) and solid waste on both the shore (93%) and surface (90%) of the water had increased. Similarly, a majority of the respondents agreed that there had been a decrease in both fish numbers (89%) and size (78%). Other changes reported were related to the revitalization of the tourism industry in the area: 47% of the respondents reported an increase in tourists, 48% reported an increase in hotels, and 50% reported an increase in restaurants. Decreasing agricultural yields were reported by 36% of the respondents. Respondents also reported a decrease in ground water quality (31%) and quantity (37%), as well as changes in domestic water quantity (37%) and quality (46%).
We also asked respondents how they adapted to the aforementioned perceived social-ecological changes. Table 3 includes all adaptation strategies that were reported by at least 10% of our respondents and used in the model. Increasing agricultural inputs (35%) and having a household member migrate to another place for work (30%) were the most frequently reported strategies. Other respondents reported decreasing time spent fishing (21%), changing their domestic water supply (12%), and changing their crop variety (10%). This reflects other work from this region that reports households' using new crop varieties in response to market demand and changing water availability (Ali and Erenstein 2017, Khan et al. 2020, Nixon et al. 2022, shifting their water supplies from public to private sources and from surface to groundwater (Qureshi 2011, Nixon et al. 2022, and increasing their use of fertilizers and pesticides in response to deceasing agricultural yields (Ullah et al. 2018).
In terms of perceived barriers associated with decisions to employ each adaptation strategy, financial cost was the most frequently reported barrier by respondents (92%), followed by the time (64%) and labor (58%) needed to employ the adaptation. Going against friends' (13%) and leaders' (6%) opinions about an adaptation strategy were cited less frequently than going against family opinions (43%) as a barrier that the respondents had to overcome to employ an adaptation strategy. The least frequently reported barriers to adaptation were the need to go against cultural traditions (18%) and taking on increased uncertainty (16%) as a result of employing an adaptation strategy.

Factors influencing households' decisions to employ adaptation strategies
Results of the logistic regression models are shown in Table 5. First, of all socio-demographic variables included in the model, household income and cattle ownership, were the only two significant variables that predicted the adoption of adaptation strategies. The odds ratio for household income was 3.09 (p < 0.001); that is, with all other variables held constant, the odds of households with an annual income of more than 30,000 PKR adopting migration as an adaptation strategy is three times the likelihood of households with lower income adapting through migration. Additionally, heads of cattle owned had a positive relationship with both increasing agricultural inputs and changing crop varieties. Specifically, an increase of one head of cattle owned was associated with a 42% increase of the odds of respondents changing their crop varieties (p = 0.033) and a 62% increase of the odds of respondents increasing agricultural inputs (p = 0.007).
Second, of the variables measuring perceived social-ecological changes, perceiving agricultural changes was the only significant variable in one model. Specifically, for every one-unit increase of a respondent's perception of agricultural changes, the odds of respondents adapting by increasing their agricultural inputs increased 14.37 times (p < 0.001) when all other variables were held constant. Finally, our model results also showed how perceptions of barriers to adaptation were significant in several models that predicted respondents' adaptation decisions. Specifically, our results showed that perceiving financial barriers had a significant positive relationship with respondents' households changing their domestic water supply (odds ratio = 1.63; p < 0.001), increasing agricultural inputs (odds ratio = 6.02; p < 0.001), and engaging in migration (odds ratio = 4.58; p = 0.015). Perceiving increased time needed to employ the adaptation was positively associated with respondents' households changing domestic water supply (odds ratio = 10.34; p = 0.003) and migration (odds ratio = 2.12; p = 0.033). Perceiving increased labor needs as a barrier was positively associated with changing domestic water supply (odds ratio = 3.92; p = 0.019), migration (odds ratio = 2.19; p = 0.016), and increasing agricultural inputs (odds ratio = 11.86; p = 0.013).
Perceiving increased uncertainty as a barrier was also positively associated with decreasing time fishing (odds ratio = 11.43; p = 0.039).
In terms of the perceiving the influence of others' opinions as a barrier to adaptation, the opinions of community leaders as a barrier of adaptation were positively associated with decreasing time spent fishing (odds ratio = 21.56; p < 0.001), and friends' opinions were positively associated with migration (odds ratio = 2.52; p = 0.045). Perceiving their family members' opinions as a barrier was positively associated with respondents' households https://www.ecologyandsociety.org/vol27/iss3/art37/ decreasing time fishing (odds ratio = 7.17; p = 0.030). In contrast, family members' opinions had a negative relationship with respondents' households engaging in migration (odds ratio = 0.20; p < 0.001) and changing their domestic water supply (odds ratio = 0.37; p = 0.018).

DISCUSSION
Governance structures are vital in the support of household adaptation to social-ecological change; thus, this work aims to inform the emergence of AWG to respond to the complexity and nuance of household decision-making processes. Although adaptation strategies and social-ecological contexts will vary, this case study presents an example of barriers to adaptation to socialecological change and the need for AWG among water users in an area experiencing climate change, rapid urbanization, and tourism development. In particular, we illustrate that AWG is more likely to take place when decision makers involved in water management (1) consider the complex ways water users respond to multiple social-ecological stressors, (2) implement crosssectoral collaboration to reflect this complexity, and (3) draw on existing institutions as they relate to social influence.

Perceiving social-ecological changes does not necessarily result in adaptation
Our results extend previous research findings by demonstrating the complex relationships between perceptions of socialecological changes and decisions to adapt. In our study, perceived changes in the agricultural system (in terms of water supply and crop yield, for example) were positively associated with increasing agricultural inputs as an adaptation strategy. Similar behaviors have been observed in other agricultural contexts as well, showing that farmers who perceive risks to their production are often more likely to adapt or want to adapt (Mase et al. 2017, Azadi et al. 2019.
It is noteworthy that perceiving other social-ecological change was not significantly associated with adoption of adaptation strategies. For example, perceiving changes in surface water and groundwater did not affect respondents' water use even though many respondents did rely on these as sources of domestic water.
These results confirm what has been suggested in previous research, that is, that perceiving social-ecological changes does not necessarily lead to adopting adaptation strategies (Bryan et https://www.ecologyandsociety.org/vol27/iss3/art37/ al. 2009, Fosu-Mensah et al. 2012. Rather, it has been shown that even when individuals perceive climatic changes, lack of access to land, information, and finances may limit their capacity to adapt (Bryan et al. 2009(Bryan et al. , 2013. Further, perceiving climatic changes may not significantly influence adaptation because households adapt their livelihoods to multiple compounding social, economic, and environmental stressors rather than climate alone (Burnham and Ma 2018, Hoque et al. 2018, Galappaththi et al. 2020, Lenaiyasa et al. 2020, Erwin et al. 2021. Indeed, nonclimatic factors such as economic, political, and social stressors are often identified as more significant drivers of adaptation decisions than climate (Mertz et al. 2010, Tucker et al. 2010, Yaro 2013. Our results echo these findings by highlighting how perceived social-ecological changes have varied influence on households' decisions to adapt as they navigate the stressors and risks associated with multiple social-ecological changes simultaneously. Therefore, although our study shows that AWG emerged in response to growing water stress (Folke et al. 2005, Boltz et al. 2019, Pahl-Wostl 2019, it also supports previous work arguing that the promotion of AWG requires not only responses to multiple social-ecological stressors encountered by water users but accounting for the complexity of decision making in these contexts (Huitema et al. 2009, Rijke et al. 2012). This means that rather than focusing solely on water supply, decision makers who want to facilitate AWG will need to consider changes across sectors (i.e., changes to crop yields or agricultural market patterns) that may drive water demand and the priorities of water users. Polycentric governance, or governance that includes multiple governing bodies, is a key aspect of AWG; however, these studies on polycentric water governance have often focused on multi-scalar collaboration (e.g., local, regional, national governance bodies; Huitema et al. 2009, Engle and Lemos 2010, Pahl-Wostl et al. 2012, Chaffin et al. 2016. We argue that collaboratively addressing multiple drivers of household adaptation through cross-sectoral institutions will further promote AWG. For our research areas, this suggests a need for the PIDAs and AWBs to work closely with the KP Agricultural Department to facilitate cross-sectoral coordination and collaboration that more accurately reflects the factors that influence adaptation decision making at the household, community, and regional scales.

Household socio-demographic characteristics shape adaptation decisions
Our research identified two socio-demographic household characteristics that shape adaptation decisions. First, respondents' household income was positively associated with using migration as an adaptation strategy. This may be explained by the fact that labor migration often requires a large initial investment to support the relocation of a household member (Mendola 2008). As such, households with higher income may be more able to employ this adaptation strategy than those with lower household income. Additionally, heads of cattle owned by a household had a positive significant relationship with both increasing agricultural inputs and changing crop varieties as adaptation strategies. Cattle represent a form of wealth and often serve as a safety net in rural areas around the world, including our research area (Ali and Rahut 2018). As such, households with more cattle may feel less risk or be more able to invest in agricultural improvements, as shown in other studies of adaptation in agricultural settings (e.g., Wood et al. 2014, Rapsomanikis 2015. Together, our findings provide further evidence from the context of rural Pakistan that echo numerous studies showing that households' financial capital, including livestock ownership, supports their ability to adapt to socialecological change perhaps in part due to their means to both employ adaption action and recover from unintended outcomes (e.g., Bryan et al. 2009, Below et al. 2012. In terms of the implications for governance, this nuanced understanding between socio-demographic characteristics and adaptation informs targeted support to increase households' adaptive capacity. In particular, AWG will be more likely to emerge when decision makers involved in water management consider and attend to disparities in multiple sources of capital that may be hindering adaptive capacity. It is worth pointing out that beyond household income and cattle ownership, no other socio-demographic variables were statistically significant in our models, including age, education, and joint family structure, as suggested in other similar studies in the Global South , Below et al. 2012, Asfaw et al. 2019). However, the trends are similar to other findings (i.e., household size and education of the head have a positive odds ratio with increasing agricultural inputs and migration). This may be because we analyzed adaptation types separately and these indicators have different influences on these specific adaptations. Our research context may also influence these findings; for example, the education of the household head may not be an appropriate indicator of the education of the household, especially given past conflict and instability that might have hindered education and thus limited variation in attainment (Hye and Khan 2013, DDMU 2015. Therefore, an increase in access to education or younger household members' education might make education a more accurate predicator of adaptation. The insignificance of our sociodemographic predictors may also suggest that the perceived barriers to adaptation may be a more important factor in adaptation decisions. Thus, our findings indicate that although water users' multiple sources of capital (i.e., income, livestock ownership, etc.) shape household adaptive capacity, non-economic factors should also be considered for their influence in whether household engage in adaptation.

The differentiated role of perceived economic and non-economic barriers in shaping household adaptation decisions
In addition to socio-demographic factors and perceived socialecological changes, our results show that respondents' perceived barriers to adaptation also influence their decisions to employ some adaptation strategies. Our results show that perceiving economic barriers (finances, time, and labor) had a significant positive relationship with households changing their domestic water supply, increasing agricultural inputs, and engaging in migration as adaptation strategies. In other words, respondents reported employing these adaptation strategies even though they perceived them to be economically costly. Although time, labor, and financial factors can be significant factors that decrease the likelihood of employing adaptation actions (Silvestri et al. 2012, Tambo and Abdoulaye 2012, Tessema et al. 2013, Kuang et al. 2019, our results reveal that they do not necessarily decrease the likelihood of adaptation; rather, there may be other factors (e.g., perceived benefit of the adaptation, variation in magnitude of the cost etc.) that mitigate economic barriers and motivate adaptation Ecology and Society 27(3): 37 https://www.ecologyandsociety.org/vol27/iss3/art37/ decisions. In particular, our results show that the perception of economic barriers varies across adaptation types and that in some cases, economic investments may be the expected and accepted costs of adaptation, thus having no effect on adaptation decisions.
In addition to the economic barriers to adaptation, we also examined the influence of family members', friends', and leaders' opinions on adaptation actions. In our study, respondents appeared to be largely willing to adapt in ways that go against their community leaders' opinions, for example, to decrease time spent fishing. They also seemed willing to go against their friends' opinions to migrate. They were less willing, however, to adapt in ways that went against their family members' opinions, particularly in the case of changing domestic water supply and engaging in migration. This reveals that social influence does sway adaptation decisions, but more specifically, that it can have differing influence across actions and relationship types. Indeed, in a meta-analysis on the impact of social influence on conservation behavior, Abrahamse and Steg (2013) showed that the degree to which social influence impacted behavior varied by group identification. That is, social influence has a larger impact on individuals who strongly identified with their social group in comparison to those with weaker group identification. Fielding et al. (2008) found similar trends in their study of sustainable agriculture practices: individuals with strong group identification are predominately influenced by others in their social group, while those with weaker group identification are influenced by behaviors both in and outside of their social group. Similarly, a message is known to be more persuasive if it comes from sources that are similar to and liked by the recipient of the message in both consumer behavior (Pornpitakpan 2004) and energy use (Wolske et al. 2020). Therefore, it follows that in our research, family members had a strong influence on adaptation decisions likely due in part to respondents strong group identification with family members.
More specifically, our results may be related to the specific cultural context of northwestern Pakistan where it has long been shown that familial ties are a fundamental social structure (Ahmed 1980, Lindholm 1982; therefore, opinions of family members may be more important than opinions of others when making adaptation decisions. More importantly, our results reveal that the extent to which social influence from various actors or sources affects decision making may shift based on the decision in question and the different contexts in which decisions are made. Overall, however, our work indeed confirms that social influence sways adaptation decisions, and that in our context, this influence often comes from family rather than formal leadership. In the context of AWG, this points to the need for governance regimes to address barriers to adaptations beyond providing information or financial assistance. In particular, careful consideration of the sources of information, not just information content will ensure that these individuals are trusted by the water users.
More broadly, collaborative natural resource decision making has been documented in multiple contexts through the Global South and North (e.g., Selvaraju et al. 2005, McCabe et al. 2010, Iles et al. 2020. Given this prevalence, we posit that AWG will emerge in conditions that align governance structures with the existing institutions of social influence and decision making. Our research also adds specificity to our understanding of social influence by revealing that social influence may differ in the way it drives various adaptation strategies. That is, in our research, water users were unlikely to change their domestic water supply if their family disagreed with this adaptation, but did decrease time spent fishing even if their family disapproved. Adaptive water governance has proven to be a promising mechanism to include nuanced drivers of decision making (Bark et al. 2012, Chaffin et al. 2014, which makes it well-suited to incorporate these varied social influences in governance of adaptation. Furthermore, our research highlights the additional work needed to understand how social influence from various groups impacts adaptation decisions. Much work has been done to show how individuals' adaptation decisions are shaped by their family members, friends, and neighbors (Dang et al. 2014, van Valkengoed and Steg 2019); however, little has been done to tease out how the influence of these groups may vary. At the same time, it is important to keep in mind that there may be other possible explanations regarding the difference in how family members' opinions shape households' decisions to employ different adaptation strategies. For example, it could be because of the weights given to different family members with different opinions about an adaptation strategy (e.g., the effect may be different if a sister disagrees rather than a father), therefore, further work is needed to examine the nuanced influence of various family members in adaptation decision making.
Nevertheless, our work does inform AWG's incorporation of existing leaders and information sources in governance structures. Leadership in AWG has been shown to be a key component of water systems' adaptive capacity (Hurlbert and Diaz 2013). In particular, Ayre and Nettle (2017) highlighted the need for multiscalar leadership. Our work builds on this understanding by illustrating the influence of informal leaders on household adaptation decisions, and in that way, points to the need for the inclusion of informal leaders in this multi-scalar structure. For water governance in our research area, this means that entities like PIDAs, AWBs, and WUAs should consider ways to include informal leaders in both their decision making and information distribution to align their structures with the informal institutions of the area.

CONCLUSION
Our research illustrates that influences on adaptation decisions go beyond the barriers of finances, time, and labor, which have been a strong focus of many adaptation studies. In contrast, in our study context, water users were unlikely to engage in an adaptive practice if they perceived that an adaptation would go against their family members' opinions, but in many cases employed adaptation actions in spite of economic barriers. Therefore, we argue that although it is important to help rural households remove economic barriers in their adaptation decision-making processes, it is also important to recognize that social influences shape what adaptation strategies individuals and households are willing to employ. In some cases, this influence may be a more powerful driver of decisions than economic barriers. Adaptive water governance will therefore emerge when informal leaders are included in both decision making and information dissemination to align governance with existing institutions in our research area. In addition, our study reveals that multiple perceived social-ecological changes drive adaptation