The Influential Role of Sociocultural Feedbacks on Community-Managed Irrigation System Behaviors During Times of Water Stress
Abstract
Sociohydrological studies use interdisciplinary approaches to explore the complex interactions between physical and social water systems and increase our understanding of emergent and paradoxical system behaviors. The dynamics of community values and social cohesion, however, have received little attention in modeling studies due to quantification challenges. Social structures associated with community-managed irrigation systems around the world, in particular, reflect these communities' experiences with a multitude of natural and social shocks. Using the Valdez acequia (a communally-managed irrigation community in northern New Mexico) as a simulation case study, we evaluate the impact of that community's social structure in governing its responses to water availability stresses posed by climate change. Specifically, a system dynamics model (developed using insights from community stakeholders and multiple disciplines that captures biophysical, socioeconomic, and sociocultural dynamics of acequia systems) was used to generate counterfactual trajectories to explore how the community would behave with streamflow conditions expected under climate change. We found that earlier peak flows, combined with adaptive measures of shifting crop selection, allowed for greater production of higher value crops and fewer people leaving the acequia. The economic benefits were lost, however, if downstream water pressures increased. Even with significant reductions in agricultural profitability, feedbacks associated with community cohesion buffered the community's population and land parcel sizes from more detrimental impacts, indicating the community's resilience under natural and social stresses. Continued exploration of social structures is warranted to better understand these systems' responses to stress and identify possible leverage points for strengthening community resilience.
Key Points
- Applied a sociohydrologic model that includes sociocultural feedbacks to socioeconomic and biophysical dynamics
- In addition to water availability and economics, community outcomes were influenced by community cohesion factors
- Demonstrated the importance of social cohesion in mediating a community's sensitivity to natural and social stresses
1 Introduction
Changing precipitation patterns and warmer temperatures, coupled with increasing population growth and changing lifestyles, are expected to increase pressure on water resources and, consequently, competition between resource users (Perrone & Hornberger, 2014). In recognition of the strong coupling of human and water systems, an integrated approach to understanding the changing connections between water and society is an important focus of current and ongoing research (Montanari et al., 2013). Drawing from diverse disciplines and related domains of coupled natural and human systems and socio-ecological systems, sociohydrology research focuses on the two-way feedbacks between social and water systems with an emphasis on endogenizing human behavior (Pande & Sivapalan, 2017).
Given the complex interactions between human and physical systems, interdisciplinary approaches are necessary prerequisites for sociohydrological studies (Schmidt, 2008). Sociohydrological studies have provided valuable insights into general as well as (apparent) paradoxical societal behaviors by integrating management dynamics into hydrological analyses (Nalbantis et al., 2011). For example, Di Baldassarre et al. (2013) demonstrated that when levees are affordable, risks shifted from low-impact, frequent flooding to high-impact, infrequent flooding within a community due to reductions in societal awareness and associated politics. Studies incorporating social dynamics have also explained how lack of irrigation access can drive increases in farmer suicides in an otherwise booming economy (Pande & Savenije, 2016) and how agro-economic policies could drive farmers toward planting water-intensive crops even during drought periods (Gunda et al., 2017).
One of the challenges facing sociohydrological studies, however, is the incorporation of certain social dimensions of the interactions (Pande & Sivapalan, 2017). The quantifiable nature of economics and land use patterns has facilitated integration of these variables within interdisciplinary water studies (Flint et al., 2017). However, interactions between human and water systems are also strongly mediated by relatively intangible concepts of community values, norms, and behaviors (Sivapalan et al., 2012). Although the important role of institutions in mediating the impacts of environmental degradation has been highlighted by researchers (Kandasamy et al., 2014; White, 2013), little attention has been given to how community values and consequent influence on community interactions can impact overall community behaviors in modeling studies. Generally, community values and behaviors have been conceptualized implicitly as changes in water allocation policies (Liu et al., 2015) or explicitly as general awareness or memory of hazards (Di Baldassarre et al., 2013) and as sensitivity to changes in economic and ecosystem services (Elshafei et al., 2014). However, these approaches do not probe deeply into how a community's social structure (informed by its values and experiences) could dynamically impact its response to physical stresses.
Understanding such social aspects is particularly important for community-managed (or traditional) irrigation systems, which can outperform state-only run systems in terms of productivity (Frey et al., 2016). This is, in part, due to the communities' social structures that allow for adaptive and resilient behavior during periods of water scarcity (Burchfield & Gilligan, 2016; Fernald et al., 2013). One such community system in the United States is the acequia, in which water diverted from snowmelt-fed streams is communally managed to support local livelihoods (Fernald et al., 2015). Despite arid conditions in the American Southwest, sociological studies find that acequias have been able to survive for centuries due to the connectivity of the local landscape and people, both between irrigators within a community and between neighboring communities (Arellano, 2014; Rodriquez, 2006). For example, contrary to the prevailing prior appropriation doctrine of the American West (which dictates that future water rights are prioritized by the order in which the water source was used historically), allocation decisions within the acequias are made through sharing agreements, reflecting these communities' cultivated values of equity in water sharing during times of both plenty and scarcity (Fernald et al., 2012). Rules and customs dictating water distribution are supported by both collective knowledge and a sense of mutualism among acequia members; in this context, mutualism is defined as the feelings of collective well-being (including social identity, pride of place, and maintenance of historical traditions) held by individual community members (Arellano, 2014; Rivera et al., 2014; Rodriquez, 2006). Local acequia traditions and management practices, such as the annual ditch cleaning, help create and sustain mutualism within the community (Arellano, 2014; Rivera et al., 2014; Rodriquez, 2006).
The unprecedented environmental and social pressures facing these and similar irrigation communities globally have raised concerns about their continued survival (Fernald et al., 2015). Concurrent with social pressures of demographic shifts, increased economic opportunities outside the traditional community structure, and increasing demands for land and water by downstream and non-agricultural users, acequias are facing environmental pressures associated with increased drought prevalence and increasing temperatures (Diaz & Eischeid, 2007). Climate change studies in the region have characterized possible changes in streamflow patterns (Cayan et al., 2010; Elias et al., 2015), impacts to agriculture and domestic water users (Hurd & Coonrod, 2012), and these users' perceptions of community preparedness (Mayagoitia et al., 2012). The actual dynamics of acequias under modified streamflow patterns, however, have not been studied using a modeling approach. Given the intricate linkages between nature and culture (i.e., community practices) in acequias, combined hydrological and social analyses are critical to understanding the impact of community values and consequent norms and practices on community responses to environmental changes, and for identifying possible tipping points that could result in community fragmentation (Cox & Ross, 2011; Rango et al., 2013; Turner et al., 2016).
The aim of this study is to assess the response of communities with communally-managed irrigation systems (specifically, acequias) to physical impacts posed by climate change, such as decreased streamflow and earlier timing in peak flow. Acequias have water sharing agreements with nearby communities, so the total water available to the communities is influenced by both natural and social conditions. Therefore, we consider changes in available irrigation water due to both: 1) streamflow changes in the upland areas under different climate scenarios (i.e., upstream pressures) and 2) neighboring communities' water demands (i.e., downstream pressures). We achieve these aims by linking an established acequia model to an upland hydrology model; the general acequia model structure (which includes qualitative, sociocultural dynamics) was developed with insights from multiple disciplines (including hydrology, agricultural economics, sociology, and anthropology) and community stakeholders (Turner et al., 2016). The general acequia model was calibrated using data for the Valdez acequia (on the Rio Hondo in New Mexico; Figure 1) and scenarios were simulated under different runoff conditions using a counterfactual trajectories approach (Srinivasan, 2015).
Location of upland subwatershed and Valdez acequia in northern New Mexico.
Although water scarcity has been linked to the collapse of some civilizations, it has also been linked to increased cooperation in others (Cetin, 2014; Kuil et al., 2016; Pande & Ertsen, 2014). The persistence of acequias for hundreds of years suggests a high intrinsic adaptive capacity of the communities' management cultures to an array of environmental and social pressures. Therefore, we hypothesize that the acequia community structure, which focuses on developing social cohesion, reduces the community's sensitivity to changes in streamflow patterns. In addition to informing acequia adaptation, our analysis explores the value of actively incorporating social structures (informed by norms and behaviors) in sociohydrological studies to understand the sensitivity and response to stress of community-managed irrigation systems, which are present throughout the world (Burchfield & Gilligan, 2016; McManus et al., 2012; Thapa et al., 2016).
2 Study Area and Methods
2.1 Study Area
Acequias have evolved over 10,000 years from the deserts of the Middle East (where they first originated) to southern Spain and eventually, the Spanish colonies that occupied present-day New Mexico and southern Colorado (Arellano, 2014). Derived from the Arabic word as-sāqiya meaning water conduit, the word “acequia” can refer to both the physical infrastructure (i.e., ditch that diverts water from the natural stream) and the community who manages and governs the ditch (Fernald et al., 2012); for clarity, we restrict our usage of the term “acequia” to the community and refer to the physical infrastructure as the “ditch.” Acequias generally divert water from upland streams (using gravity-fed ditches and head gates) to support their agricultural livelihoods. Similarly to other long-lasting agricultural systems, acequias have adapted their practices over time to sustain their way of life (Fleming et al., 2014; Rivera et al., 2014). For example, as more income opportunities became available outside the farm, acequias shifted away from subsistence agriculture to cattle-based production (Fernald et al., 2015). Generally, families in acequias depend more on wage labor than agriculture-based income, but active involvement in agricultural activities within the community helps sustain mutualism (Fernald et al., 2012).
Currently, over 800 acequias are present in northern New Mexico, some of which are over 300 years old (Fernald et al., 2012). The Rio Hondo, a tributary of the Rio Grande in northern New Mexico that drains 185 km2, supplies water to 11 acequia systems, irrigating over 2,870 acres of farmland (Fleming et al., 2014). We focus our analysis on the Valdez acequia, which is the most upstream community on the river and thus, the first user of the snowmelt water on the Rio Hondo. The primary crops grown in this region are alfalfa, hay, orchards, grains, and vegetables including beans, corn, and green chile (Sabu, 2014). Similar to other acequia systems, the Valdez acequia elects three commissioners and a mayordomo to help maintain the ditch. The mayordomo coordinates the annual ditch cleaning and water releases during the irrigation season, and also resolves disputes between community members. The community members along the acequia (known as parciantes) are responsible for the mayordomo's salary and for contributing to the ditch cleaning and other maintenance activities (Rodriquez, 2006). In addition to marking the beginning of the irrigation season, the annual ditch cleaning provides an opportunity for the parciantes to address necessary repairs and discuss streamflow projections for the season (Fernald et al., 2012). Participating in such communal activities and local religious events renews the parciantes' connection to the community and the land, thereby helping to maintain their mutualism (Rivera, 1998).
Acequias link members along their shared ditches, but are also linked to other acequias that depend on diversions from the same stream. Water sharing between acequias is negotiated through agreements, administered by the New Mexico Office of the State Engineer, that specify the amount of water available for a given community (rather than at the individual level typical of prior appropriation allotments) (Fernald et al., 2012). Ditch headgate closures are required in order to comply with the State Engineer adjudications that stipulate how much streamflow diversion is allowable. Currently, the Valdez acequia is allowed to divert up to 22% of headwater streamflow (Sabu, 2014).
Streamflow for the Valdez acequia originates from the upland subwatershed, which occupies a little over half (96 km2) of the Rio Hondo watershed (Figure 1). The upland subwatershed is defined as the region of area that drains to the United States Geological Survey (USGS) gage 8267500 (name: Rio Hondo near Valdez) on the Rio Hondo, where discharge measurements have been collected since 1935 (USGS, 2015c). The upland subwatershed of the Rio Hondo lies above 2,300 meters in elevation, is underlain by unweathered bedrock, and is predominantly forested (MRLC, 2015; NRCS, 2015; Raheem et al., 2015). A majority of the upland subwatershed is designated as wilderness areas, where the use of power tools is legally prohibited (Frutz, 2015); therefore, only very limited management currently occurs in the upland forests of the Rio Hondo (Frutz, 2015). No fires have been observed within the Rio Hondo watershed since at least 1984 (MTBS, 2015). In addition to being a source for streamflow, the upland subwatershed provides valuable grazing areas that support cattle production by the community members. These areas are used to feed cattle in the summer months while stored or purchased feed is used during the winter months (Turner et al., 2016). Grazing permits for the upland subwatershed are issued by the U.S. Forest Service (USFS).
2.2 Model Descriptions
The study area is represented by two models, an upland model that focuses on the hydrology of the upland subwatershed (hereafter referred to simply as “upland model”) and a valley model that focuses on the intricate connections between the hydrological, biophysical, socioeconomic, and sociocultural systems in the Valdez acequia (hereafter referred to simply as “valley model”) (Figure 2). The structure and linkages of the valley model were informed by insights from multiple disciplines and community stakeholders (Fernald et al., 2012). Specifically, team members from a range of disciplines (including hydrology, livestock and rangeland science, agronomy, economics, rural development and planning, and cultural anthropology) engaged with parciantes, community leaders, and cooperative extension service agents over multiple months through workshops, phone calls, and meetings to capture the dynamics specific to acequia systems (Guldan et al., 2013). This participatory approach enabled the translation of community experiences and sociocultural information into the modeling framework (Lane, 2014). The valley model focuses on a single acequia, so regional dynamics (including crop prices and demands of neighboring acequias and other water users), which are outside the model scope, are treated as exogenous conditions. In contrast to the valley model, the structure for the upland model was strongly guided by input parsimony while capturing the primary hydrological processes (e.g., precipitation partitioning, evapotranspiration, baseflow, and saturation excess runoff) that influence streamflow at the outlet (supporting information Text S1 and Figure S1). Both the upland and valley models are built on a system dynamics platform (Powersim Studio 10 Expert) with a monthly time step (see supporting information and Turner et al. (2016) respectively). This platform was selected because the icon-based, graphical interface provided an intuitive, visual environment that facilitated communication between diverse disciplines and stakeholders (Tidwell et al., 2004; Tidwell & Van Den Brink, 2008). A complete description (including governing equations) for the valley model is provided in Turner et al. (2016) and for the upland model in the supplementary information. Therefore, we limit our description of the two models here to the aspects most pertinent to understanding the results.
Influence diagram summarizing models' structure and linkages. The shapes indicate whether a variable is an exogenous input (parallelogram and diamond) or process within a model (oval). A majority of the variables are simulated in the valley model, except for the streamflow from the upland subwatershed, which are simulated in the upland model, outlined in the blue, dashed box. The text color of the variable indicates the general dynamics they capture: interconnecting variables (black), sociocultural dynamics (purple), biophysical (green), socioeconomic (orange), sociohydrological (light blue), and hydrological (dark blue). The colors of the shapes indicate the variables modified for the different scenarios: upland pressures simulated by changes in the climate conditions (filled in yellow) and the additional downstream pressures simulated by changes in headgate closures (filled in gray). In the interest of readability, not all linkages in the models are presented in the influence diagram (e.g., return flow from irrigation and impact of population demographics on domestic pumping). For a full list of equations governing model linkages, see Turner et al. (2016) for the valley model and the supplementary information for the upland model.
As mentioned previously, the amount of water available to an acequia is influenced by the total amount of water available in a stream as well as social dynamics regarding headgate closures and diversion rates. The total water available in the stream is simulated in the upland model using a lumped approach that characterizes the upland subwatershed into individual hydrologic response units (HRUs) (supporting information Table S1 and Figure S2); the HRU method provides sufficient discretization of watershed properties while maximizing computational efficiency (Aragon, 2008). Monthly precipitation and temperature data derived from Bureau of Reclamation data sets, which have a 1/8th degree resolution (BoR, 2014), are the primary inputs into the upland model. During the irrigation season (Apr-Oct), discussions between neighboring acequias and neighbors within an acequia are undertaken to ensure everyone gets their share of water as needed, potentially leading to headgate closures. In any given month in the model, if the headgate is closed, no water is diverted into the irrigation ditch. Otherwise, the streamflow is diverted into the irrigation ditch based on the allotments in the flow agreement (i.e., 22% for the Valdez acequia).
In the valley model, water delivered to the ditch may be used for crop irrigation or may seep from the canal to recharge the shallow groundwater, supporting both riparian vegetation and pumping for domestic needs (Fernald et al., 2012). Irrigation water influences crop yields by meeting crop water demands. Crop yields are determined using production functions for different crops (FAO, 2012; Glover et al., 1997), which take into account the crops' consumptive irrigation requirements calculated using the Hargreaves method with temperatures in the valley (Hargreaves et al., 2003). In addition to crop yields, crop profitability for a parciante in the valley model is influenced by commodity prices, crop selection, and land use decisions that determine the amount of land in production and size of the parciante's parcel. Agricultural land in the Valdez acequia can be transferred between states of production and fallowness depending on agricultural economics, time management, and rural community dynamics (Figure 2).
The impact of water availability and economic policies on crop selection and farmer livelihoods have been well established (Elshafei et al., 2015; Gunda et al., 2017), but community insights indicate that time management is an important variable that connects the socioeconomic and sociocultural dynamics in acequias (Turner et al., 2016) (Figure 2). Specifically, stakeholder insights indicated that with more time in agriculture, parciantes shift away from producing hay and alfalfa toward higher valued commodities such as green chile and other vegetables, which require more labor to cultivate and maintain (supporting information Figure S3). In addition to crop profits and financial performance, time spent in agriculture is also driven by mutualism, which has a positive impact on parciantes' willingness to spend time in agricultural activities, which in turn reinforces participation in communal activities within the acequia (Turner et al., 2016). Thus, the ‘time spent in agriculture’ variable links market-based (e.g., crop profits and financial performance) and sociocultural-based (e.g., mutualism and community participation) dynamics in the model.
Grazing dynamics also influence overall agricultural profits of parciantes. Hay produced in the valley, which is influenced by irrigation water supplied by the ditch, is stored for feeding livestock during winter months. During times of decreased forage production, purchasing feed reduces parciantes' cropping profits and increases the pressure to seek alternative ventures, due to the higher opportunity costs of time in agriculture; parciantes have a general preference to avoid debt. Consequently, overall livestock inventories may fluctuate seasonally. Cattle herd size management and grazing profits are also impacted by upland grazing, which is influenced by drought conditions and policies governing forage access on the surrounding public lands. Based on regression analysis of historical data, grazing quality was estimated to be a function of rainfall and drought (as measured by the Palmer Drought Severity Index, calculated using the MATLAB tool provided by Jacobi et al., 2013). Access to forage is influenced by changes in USFS grazing policies, which determine the maximum stocking rates of livestock producers in the valley model. Changes in allocated USFS grazing permits were modeled with step functions that reflect historical policy changes.
In addition to hydrological, biophysical, and socioeconomic variables, community behavior in the valley model is strongly influenced by acequia demographics (including population size, intergenerational land transfers, emigration of established parciantes, and immigration of newcomers from non-acequia families) and associated sociocultural dynamics (Figure 2). In particular, mutualism was included in model formulation due to its important role in maintaining and strengthening acequia identity, culture, and management. Although qualitative in nature, mutualism has been recognized as a key characteristic of acequias that contributes to their longevity over time and their adaptability to changing economic, political, or environmental conditions. Attempting to model mutualism and its effect on acequia systems began with Martínez-Fernández et al. (2000), becoming more complex with Turner et al. (2016) and Tenza et al. (2017). Within the valley model, the Valdez mutualism variable is dynamically driven by mayordomo leadership and community participation, the latter of which is influenced by community demographics, land use decisions, and time spent in agriculture variables (Figure 2); community demographics are influenced by incentives to stay or leave the acequia, including the opportunity cost of earning wages elsewhere. As the opportunity cost of agricultural activities increases within the acequia, parciantes may sell portions of their land in the model, providing newcomers an opportunity to move into the community. The introduction of newcomers causes a lag in community participation (and consequently mutualism) in the model that accounts for the time it takes these individuals to acclimate to local practices (Figure 3). Mutualism, in turn, has a positive effect on parciantes' decisions to work in agricultural activities, which creates feedback loops affecting agricultural profits, land use decisions, and eventually other factors that impact community participation variables in the model (Turner et al., 2016). Following stakeholder input, we assume that the mutualism variable in the model is influenced by community behavior over the last 36 months (i.e., three years without any community participation input would lead to complete reduction in mutualism), which was corroborated during model calibration against real-world behaviors (Turner et al., 2016). The impact of sociocultural variables, for which little quantifiable data exist, on community behavior was evaluated using sensitivity analyses by Turner et al. (2016).
Summary of socioeconomic dynamics surrounding acequia mutualism in the valley model, developed using stakeholder insights. The “+” symbol on each causal link indicates that changes (either an increase or decrease) in the tail variable creates a change in the same direction for the variable on the arrow head, while the “−” indicates the opposite. The “+” symbols within each feedback loop represents reinforcing (or positive) feedback, while the “−” symbol represents balancing (or negative) feedback processes. The arrows with double lines represent dynamics with significant time delays.
2.3 Model Historical Calibrations
The upland model, which runs continuously from 1959 to 2099, was calibrated using USGS flow data from 1959 to 1998. Soil depths and hydraulic conductivity values in the upland hydrology model were verified using built-in optimization tools in the Powersim software. Correlation between simulated runoff from the upland model and observed flow data is 0.78 (Figure 4). The simulated cumulative runoff over the 30 years is 87.3% of the cumulative runoff recorded at the gage site; the cumulative distribution function (CDF) of the simulated runoff has a shorter tail than the CDF of the USGS gage data (supporting information Figure S4). At a monthly scale, the median runoff during May is lower in the upland model simulations while the median runoff during July is lower in the USGS data (supporting information Figure S5).
Runoff comparison between simulated and observed data during the historical period.
The valley model calibration was conducted using readily available data from 1969 to 2009 for agricultural profitability, cattle inventory, community size, and land use, drawing from an array of data sources (supporting information Table S2) (Turner et al., 2016). For wage rates and commodity prices, regional and national averages were used whenever local economic data were unavailable (supporting information Figures S6 and S7) (Turner et al., 2016). Correlations between the historical data and simulated variables in the socioeconomic data range from 0.41 to 0.99 with Theil inequality statistics for the mean, variance, and covariance spanning from 0.01 to 0.86 (Table 1). Historical records (e.g., individual mayordomo agreements in the headwaters) do not include whether the ditch headgate was open or closed each month. Therefore, in order to replicate the process of gate closures made by mayordomos during times of water sharing, we used a random number generator (between 0 and 1) along with an irrigation month threshold to simulate the headgate closures; if the random number was greater than the threshold, then the headgate was open, else the headgate was closed, eliminating acequia ditch flow. Calibration results indicate that historically, the Valdez headgate was closed 25% of the time during the irrigation season.
| Correlation (r) | Theil inequality statistics | |||
|---|---|---|---|---|
| Community variables | Um | Us | Uc | |
| Agricultural profitability | 0.41 | 0.08 | 0.10 | 0.81 |
| Cattle inventory | 0.55 | 0.01 | 0.13 | 0.86 |
| Land Use: agricultural | 0.87 | 0.13 | 0.60 | 0.26 |
| Land use: built | 0.99 | 0.58 | 0.33 | 0.08 |
| Community size | 0.95 | - | - | - |
- Note. Theil inequality statistics for accuracy include percentages of mean errors due to differences in mean (Um), variance (Us), and covariance (Uc) between observed and predicted values. Theil values for community size are not shown due to lack of sufficient data points.
2.4 Scenarios
The changing climate is expected to influence not just flow timing but also overall water availability in streams. Therefore, it is critical that acequia analyses consider changes from both the upland subwatershed and downstream influences from other water users. For the different climate pathways and gate closure scenarios, we compare 30-year historical time period (1969–1998) outputs (e.g., streamflow from the upland model or agricultural profitability from the valley model) to a 30-year “near future” period (2019–2048) and a 30-year “far future” period (2069–2098).
The HadGEM2-ES global circulation model (GCM) is one of the most accurate climate models for the American Southwest (van Riper et al., 2014). Therefore, downscaled precipitation and temperature data from the HadGEM2 GCM (BoR, 2014) were used to simulate streamflow under future climate change conditions in the upland model. Since the 2°C is increasingly becoming an unachievable goal (Friedlingstein et al., 2014), we use representative concentration pathways (RCP) 4.5 and RCP 8.5 to represent the lower and higher emission possibilities respectively; using RCP 4.5 to represent the lower end of emissions ranges is consistent with the approach taken in the U.S. National Climate Assessment (Melillo et al., 2014). Relative to the historical period, temperatures are consistently higher in the future across all months whereas precipitation magnitude and variability differs by month (supporting information Table S3 and Figures S8–S10).
The streamflow generated by the upland model, information regarding upland drought conditions, and valley temperatures (for each pathway) is input into the historically-calibrated valley model to assess associated impacts in the community. Following the “counterfactual trajectories” approach, the economic exogenous variables reflect historical conditions (e.g., external jobs availability and increasing wage rates over time; supporting information Figure S6) to highlight how community dynamics might have evolved had the Valdez acequia experienced the streamflow patterns expected under climate change with the headgate closure frequencies maintained at 25% of months, consistent with historical conditions; maintaining the gate closures at historical rates allows us to isolate the physical impacts driven by climate change. In other words, we are “reimagining the past,” rather than trying to predict rapidly changing social futures, which could evolve in unforeseen ways at decadal scales, preventing accurate modeling (Sivapalan et al., 2014; Srinivasan, 2015).
Although the valley model focuses on a single acequia, we expect that any upstream streamflow reductions would be further magnified downstream. Since streamflow changes are expected to decrease in some parts of the year and increase in others, we assume that water users downstream of the Valdez acequia would likely apply pressure on the headwater acequias to reduce the frequency of streamflow diversions into their ditches (historically estimated at 25% closures, as described above). We model these additional downstream pressures by considering increases to the headgate closure rates in the valley model to 50% and to 75% of the irrigation months. Under each scenario, we assumed that when the headgate was open, the irrigation diversions into the Valdez acequia were maintained at historical allotment levels of 22%. In summary, the scenarios considered in the analysis are two climate pathways (RCP 4.5 and RCP 8.5) for three headgate closure rates (25% (i.e., historical rates), 50%, and 75%) in two time periods (near future and far future).
3 Results
Relative to historical streamflow, streamflow under climate change is lower across both pathways, with RCP 8.5 in the far future period exhibiting the greatest differences (Figure 5 and supporting information Table S4). In the near future period, upland streamflow decreases are most notable for June and July (supporting information Table S4), which results in a shift of peak streamflow from June to May under both pathways (Figure 5). In the far future period, a similar peak shift from June to May is present under RCP 4.5, but under RCP 8.5, the peak shifts even earlier to April (Figure 5).
Simulated monthly streamflow under climate change using the upland model. Relative to the historical period, peak flow occurs earlier in the year due to higher spring runoff and lower summer runoff under the two RCPs.
Historically, there was a surplus of flows (defined as the difference between accumulated flows in the ditch and consumptive irrigation requirements of the crops) in the Valdez acequia (supporting information Figure S11) but this surplus decreases across all future scenarios (supporting information Figure S11). Under both RCPs, agricultural profitability is consistently higher in the Valdez acequia under the historical, 25% gate closure rates but much lower at higher gate closure rates (Figure 6a and supporting information Figure S12a); crop production profits follow similar patterns while grazing profits show little variability across the different gate closure scenarios (supporting information Figure S13). Monthly comparison of acequia ditch flow shows increases in applied irrigation water between April and June relative to the historical period in the 25% gate closures scenario (supporting information Figure S14). In the 25% gate closures scenario, vegetable production is greater and feed purchases for cattle lower than in the historical period (supporting information Figures S15 and S16). The parciante parcel size (i.e., working farm size) and percentage of area being used for agriculture both increase under the 25% gate closures scenario and decrease under the other two gate closure scenarios (Figure 6b and supporting information Figure S17, respectively). In contrast, the community size and percentage of newcomers in the community patterns both decrease under the 25% gate closures scenario and increase under the higher gate closure rates (Figure 6c and supporting information Figure S18, respectively). The amount of time spent in agriculture by parciantes and the community's mutualism are relatively stable under both climate pathways and across all three gate closure scenarios (supporting information Figures S19 and S20, respectively).
Comparison of (a) agricultural profitability per parciante parcel, (b) parciante parcel size, and (c) community size under the different climate pathways and head gate closure scenarios. Points indicate the normalized difference in mean from the historical, calibrated values and the bars represent +/- 1 standard deviation, normalized relative to the standard deviation of the historical values. Relative to the 50% and 75% gate closure scenarios, agricultural profitability and mean parcel size are larger while community size is smaller in the 25% gate closure scenario.
4 Discussion
4.1 Climate Change Impacts
The general streamflow patterns observed under the RCPs (e.g., lower flows overall and relative increase in flow during the spring months) are consistent with patterns observed by Elias et al. (2015) for the Upper Rio Grande Basin and USGS (2015a) for the Upper Rio Grande subbasin for both the Hadgem2-ES and mean GCMs. Therefore, we have high confidence that the streamflow patterns generated by the upland model capture the general dynamics expected under climate change. Using these projected streamflow estimates, we generated counterfactual trajectories in the historically-calibrated valley model to understand possible climate change impacts in the Valdez acequia.
Under the 25% gate closure scenario, the streamflow under the climate change scenarios had a relatively positive impact on the agricultural profitability of the Valdez acequia (Figure 6). Parsing agricultural profitability into the crop and livestock components highlights that the increased profitability was derived from crop profits (supporting information Figure S13). Historically, the irrigation ditch flow for the small headwater acequia was much greater than the water demand for crops (i.e., evidence of surplus irrigation flows, supporting information Figure S11). Although the flows are reduced under the RCPs, there is still a surplus under the 25% gate closure scenario (supporting information Figure S11). Thus, crop production, which is modeled as a function of the available water and consumptive irrigation requirement, would not have been adversely impacted by the lower streamflow in the climate change pathways.
At a monthly scale, shifts in streamflow allowed for greater irrigation applications earlier in the season when they are generally more critical (supporting information Figure S14). This dynamic coupled with the rising commodity prices observed over the simulation horizon (supporting information Figure S7) improved crop profitability and thus, total agricultural profits per parciante parcel (Figure 6a); the implications of some of these economic assumptions are explored in section 4.2. Increased profitability was positively reinforced through increased time in agriculture (Figure 3), which drives vegetable production (supporting information Figure S3), a more valuable crop on a per unit basis than less time-intensive crops such as grass hay and alfalfa (supporting information Figure S15). Increased profitability reduced the pressure for parciantes to leave the acequia (either short- or long-term) by leasing or selling land, leading to slightly increased parcel sizes (Figure 6b). With more land remaining in agriculture, fewer opportunities arose for newcomer introductions (since less land was available), leading to a slightly lower community size (Figure 6c). Similar directional relationships were shown in the sensitivity analyses performed by Turner et al. (2016). Generally, our model results indicate that the Valdez acequia's adaptive practices (i.e., switching between crops) are likely adequate to allow the community to persist and thrive had climate change-expected streamflow patterns occurred with the historical gate closure rate of 25%.
Although earlier peak flows could provide some benefits to headwater acequias in the form of increased early-season soil moisture, the overall reduction in streamflow could create new conflicts with users downstream. Generally, there are at least some downstream users with priority rights over the headwater communities; additionally, urban municipal stakeholders could apply pressure such that the headwater acequias would divert water into their respective ditches less often. We assume this exogenous policy response to elevated water stress would occur in the form of increased closure frequencies of the acequia headgate (from 25% to 50% and 75%), effectively restricting irrigation and allowing water to pass downstream. We expected the inclusion of regional dynamics that amplify water stress (from an exogenous, downstream pressure) would create significant shifts in current acequia function, both socioeconomically (e.g., agricultural profitability) and socioculturally (e.g., mutualism). Unlike the business-as-usual policy of 25% gate closures, we did observe major negative shifts in agricultural profitability with increasing gate closures (Figure 6a), driven by both decreases in vegetable and hay production from restricted irrigation applications (supporting information Figure S11) and increasing purchased feed costs necessary to maintain livestock levels (supporting information Figure S16). Reduced crop production and grazing quality were driven by water availability and drought respectively in the modeling approach, reflecting the dominant factors expected to influence crop and grazing yields in the southwestern US (Blanc et al., 2017; Reeves et al., 2017). Lower profitability incentivized parciantes to spend slightly less time in agricultural activities and sell fractions of their land, giving newcomers opportunities to move into the community (supporting information Figure S18), both of which shrank parciante parcel size and increased the overall community size (Figures 6b and 6c).
The increase in community size despite decreases in agricultural profitability (a paradoxical behavior from a traditional economic perspective) was driven by acequia mutualism, which showed little change across the scenarios (supporting information Figure S20). Mutualism, an important acequia characteristic, is created over time within the community and feeds back to nearly all sectors of acequia activities, influencing economic and environmental decisions, religious rituals, newcomer acclimation to the community, and effective leadership transitions among others (Rivera, 1998; Rodriquez, 2006). Given mutualism's central link in the socioeconomic behavior (i.e., its influence on parciantes' time spent in agriculture; Figure 3) and decreasing profits, we expected to see much greater variation in the modeled acequia's sociocultural dynamics over time. However, introduction of newcomers to the community balanced the emigration losses from existing parciantes, effectively maintaining the community size under both RCPs. Even though reduced agricultural profitability is reflected in the lower parcel size (Figure 6b), total land in agricultural production mostly shifted from irrigator to irrigator or irrigator to newcomer (depending on the specific social shifts in the community) (Figure 3). This helps explain why little to no variation was seen in total irrigated acreage in the model (supporting information Figure S17), despite impact of reduced annual profitability on parciante choices regarding time spent to agriculture and land allocation between crops. The time spent in agriculture by the newcomers in the model (supporting information Figure S19) positively influenced community participation, contributing to maintaining mutualism and a stable land base while they became acclimated to the community (Figure 3); historical accounts have documented the important role newcomers have played in helping maintain acequia practices (Crawford, 1993; Rivera, 1998; Rodriquez, 2006). Sensitivity analyses of the endogenous valley model structures indicate that time spent on agriculture is a key variable mediating the outcomes stemming from both biophysical and socioeconomic variable changes (Turner et al., 2016); the importance of time together and shared experiences on influencing a range of social dynamics, including civic engagement and marital success, has been well studied (Guldner & Swensen, 1995; Shah et al., 2002). These social dynamics helped counter the impacts from parciantes' departures on community cohesion in the model. In turn, a positive net rate of change in mutualism was maintained, effectively buffering against greater impacts (to the community size, land use, and ongoing participation of members in community activities) from reduced agricultural profitability.
A failure to observe any abrupt changes in the sociocultural dynamics of the Valdez acequia indicates that the adaptive capacity of the internal governing and management structure of the Valdez acequia would have likely been adequate to enable the community's continued existence even if climate change-expected streamflow patterns had occurred in the past. These results are not surprising given that acequias' structure has been influenced by and reflects these communities' survival for hundreds of years through a multitude of natural and social shocks (Rivera et al., 2014; Rodriquez, 2006). The simulations indicate that the social responses to physical stresses (i.e., 50% and 75% gate closure scenarios) are more likely to adversely impact acequia agricultural socioeconomics than the physical stresses themselves (i.e., 25% gate closure scenario), a finding consistent with previous studies (Cox, 2014; Turner et al., 2016; Wise & Crooks, 2012). Even under the harsh gate closures simulated, feedbacks associated with members' participation in acequia activities enabled the community to continue to stay together because their livelihoods are driven more by mutualism than by profit. This dynamic highlights the importance of understanding social cohesion and its contributions to a community's sensitivity to stress, an important sociohydrological state variable identified by Elshafei et al. (2014). The relative insensitivity of the Valdez acequia to hydrological and related economic pressures has important implications for community resilience, indicating the members could stay together long enough to develop new adaptation strategies. In particular, changing crops, changing time spent on the farm, selling a portion of their land, or allowing newcomers into the community all reflect historical practices in these communities that have helped them sustain their way of life (Crawford, 1993; Fernald et al., 2012; Rivera, 1998; Rodriquez, 2006).
Although changes in physical and social variables both generate similar pathways of change, greater uncertainties are likely stemming from the latter given the complexities with predicting social behavior. Community participation (the primary indicator of mutualism), which had significant impacts on agricultural profitability, amount of irrigated land, and residential land use variables in sensitivity analyses (Turner et al., 2016), is almost universally held in high regard by the parciantes and identified for ongoing protection (Mayagoitia et al., 2012). These sentiments have motivated the recent emergence of local and regional acequia associations that engage in collective water rights adjudication on behalf of the acequias (Cox, 2014) and help address the social drivers that are contributing to parciantes leaving acequias (Cox, 2014; Wise & Crooks, 2012).
Sociohydrological research endogenizes human decisions, thereby increasing our understanding of feedbacks between physical and social water systems (Sivapalan et al., 2012). Modeling studies in this domain, however, have mostly overlooked cultural values, which may be more predictive of behavior than utility maximization of economic risk (Caldas et al., 2015; Sivapalan & Blöschl, 2015). This research highlights the value of incorporating and understanding role of dynamically-embedded sociocultural processes (such as mutualism in acequias or more generally, social cohesion) on community resilience. Similar to institutions, however, social cohesion is not a simple panacea (Ostrom & Cox, 2010). Social cohesion in acequias is strongly dependent on the time spent together by the community members in agricultural activities, but cohesion in other community-managed irrigation systems may be oriented around kinship (such as in Sri Lanka (Leach, 1961; Murray & Little, 2000) or around religious activities (such as in Bali (Lansing & Miller, 2005). Additionally, the mechanisms through which social cohesion could influence community dynamics may also vary from crop selection (observed in acequias) to fluidity of land tenure (observed in Sri Lanka) and other modified landscapes (observed in Bali). Unraveling the diverse structure of both formal and informal social checks and balances, especially within self-governed communities, may provide insights into differences in community behaviors (such as cooperation vs dispersion during times of stress) in space and time (Kuil et al., 2016; Pande & Ertsen, 2014). Although such social information is challenging to capture quantitatively within models (Levy et al., 2016; Pande & Sivapalan, 2017), engagement with stakeholders could help address these issues and provide powerful input for calibration of variable linkages and overall model structure (Lane, 2014; Tidwell & Van Den Brink, 2008). In addition to providing a vehicle for engaging stakeholders in structured dialogues, modeling discussions with stakeholders may also strengthen the community fabric through continued education and communication of hydrological processes (Fernald et al., 2012; Guldan et al., 2013). Given the vulnerabilities of community-managed irrigation systems and other traditional communities to climate change and the potential of our model decisions to shape the world around us (IPCC, 2014; Lane, 2014), continued exploration of sociocultural dynamics are important to help better inform adaptation strategies to the unprecedented physical and social pressures facing these communities.
4.2 Model Limitations and Lines of Future Investigations
As with any model, decisions regarding model construction involve a number of tradeoffs (between generality, realism, and precision) that influence findings (Troy et al., 2015). Within the valley model, the parciantes have agency to decide how much land is in production, decide which crops to plant, and whether or not to leave the community. Factors that are not in direct control of the acequia members (e.g., commodity prices and wage rates) are treated as exogenous factors in the valley model. Assumptions associated with the scenario approach (i.e., counterfactual trajectories), model simplifications associated with variable interactions (e.g., crop production in the valley model and vegetation disturbances in the upland model), focus on a single acequia, and data limitations all influence the interpretability and implications of the simulations.
The counterfactual trajectories approach allows us to simulate possible community impacts without having to forecast social futures. Although crop selection and external movement of parciantes for wage labor are captured in the historical data over the last few decades in the acequia (Figure 3), we assumed no changes occurred in the social exogenous factors, similar to approaches taken by other studies (e.g., Pande & Savenije, 2016). However, commodity prices and wage rates may have been different had the climate been indeed different. For example, regional crop prices could have been higher due to reduced crop production from regional water stress, which could have further incentivized parciantes to stay in agriculture assuming they had sufficient access to water resources. These economics could have also potentially increased barriers faced by the parciantes to market production and further influenced demographic shifts, which the valley model currently assumes are not significant. Future analyses will explore the use of global economic models (e.g., Env-Linkages (Château et al., 2014) to inform these exogenous socioeconomic conditions.
The focus on a single acequia limits our exploration of regional dynamics, including newcomer rates and the interactions between multiple water users. Currently, the valley model assumes there is no shortage of newcomers wanting entrance into the Valdez community, which is likely, from parciantes in downstream acequias looking to move into relatively more water secure upstream communities or from retirees and telecommuters who care little about agricultural profitability. However, in the absence of the newcomers or their successful acclimation into the community, the Valdez acequia could decrease in size and potentially disband altogether. Exogenous downstream pressures are manifested as changes in headgate closure rates in the model, but it is quite possible that different mechanisms are employed within the communities (Chen et al., 2016). Adaptation of these communities from subsistence-based agricultural to wage-based labor indicates the communities' abilities to adapt to changing economics. Although agricultural profits are not the primary source of income in acequias, continued involvement in these practices are intricately connected to mechanisms (e.g., annual ditch cleaning) that promote mutualism in the community (Fernald et al., 2012). So it is quite possible that the communities' social interactions could enable its evolution to organize around a different form of agriculture altogether to help maintain their cohesion. Further discussions with stakeholders is necessary to understand critical thresholds for community fragmentation and to inform the development of future social trajectories in modeling scenarios (Guldan et al., 2013).
Although reduced water availability is a prominent factor affecting crop yields in the southwest, crop profits could also be adversely impacted by other climate change impacts (such as CO2 fertilization, increased vapor pressure deficits, and heat stress) (Deryng et al., 2014; Kiniry et al., 1998), which are outside of the current model scope. We also assume no changes in the distribution of vegetation in the upland subwatershed under the climate change scenarios, which does not reflect pine mortality rates in southwest forests (Hurteau et al., 2014); the increased aridity and warmth has also made the vegetation more susceptible to disturbances such as bark beetles and forest fires (Bentz et al., 2010; Dennison et al., 2014). Therefore, the streamflow changes simulated by the upland model are most likely on the conservative end of potential impacts. An additional biophysical simplification is regarding land quality in the valley model. Although land use decisions reflect parciantes' movement in the community, we assume that the productivity of agricultural fields is not impacted by land degradation (such as erosion and increased salinity) (Webb et al., 2017) or increased frequency of high-intensity precipitation expected under climate change (Nearing, 2001). Therefore, increases in agricultural production estimates are most likely overestimated.
Upland plant community changes can also be influenced by grazing management decisions (e.g., overgrazing, fire suppression activities, and response and recovery to wildfires). When a fire occurs, for example, the USFS adjusts grazing allotments to allow for system recovery; adjustments can last anywhere from 1 year for small fires to up to 6 years for larger fires (Frutz, 2015). Vegetation changes would likely alter forage quantity and quality allotted to parciantes, and thus, influence stocking rates and profitability of livestock production. Including such linkages in future simulations will expand the exploration space of sociohydrologic feedback loops and how regional resource management policies can impact the social and natural systems in which they are embedded.
The historical calibration of the valley model was challenging due to the lack of sufficient social data at relevant spatial and temporal scales (Levy et al., 2016). Even when certain data were available during the historical period, model simulations of these variables are influenced by dynamics for which data are lacking (e.g., mutualism). So for the available data sets, an acceptable level of both precision (using correlation measures) and accuracy (using Theil statistics) were sought between historical data and model simulations in order to build confidence in the valley model's ability to capture the overall dynamics of the Valdez community. For the historical period, the Valdez acequia model showed strong correlations between the simulated and observed data for community size and land use variables (Table 1). Cattle inventory and agricultural land use showed both a high degree of correlation and strong accuracy. Although model precision for agricultural profitability was the lowest among measured variables, it captured the dominant trends well, with Theil inequality levels of 0.08 (mean), 0.10 (variance), and 0.81 (covariance). In general, the Theil statistics of the historical simulations in the valley model meet or approach acceptable model evaluation standards (Sterman, 2000), and where they do not (e.g., the land use and community size variables), evaluations were impacted by limited data. Overall, the statistics are similar to those of previous valley models that captured both observed acequia dynamics and exhibited acceptable levels of sensitivity needed to observe and analyze alternative scenarios (Turner et al., 2016). Verification for future scenarios could not be conducted (especially the high gate closure rates) since the valley model was exposed to stresses much larger than those captured in the calibration data sets of the Valdez acequia.
The lack of social data at a fine resolution is one of the primary reasons that the seasonal agricultural flows are simulated at a monthly time step. Although interannual variability in flow (stemming from increased precipitation variability; supporting information Figure S9 and Table S3) did not greatly impact mutualism dynamics, a monthly time step prevents us from looking at smaller decision dynamics that could impact trajectories of societal change (Ertsen et al., 2014). Additionally, the shorter CDF tail and underestimated June flows in the upland model historical outputs most likely arise from storm dynamics (and subsequent infiltration excess runoff) not being adequately addressed at the monthly-scale resolution of the model. The valley model discrepancies could potentially be addressed by developing additional data sets at a higher temporal resolution, an approach that showed higher degrees of precision and accuracy in replicating historically-observed patterns for a comparable acequia model in Alcalde, NM (Turner et al., 2016).
Lastly, our historical understanding of the community only spans five decades, which does not capture the full range of social dynamics that have influenced and modified the acequia community structure over its long history. Therefore, continued engagement with diverse stakeholders is warranted to improve our understanding of the linkages between the physical and social systems in community-managed irrigation systems, to refine our models to reflect the observed realities across multiple temporal and spatial scales, and to explore the range of future social scenarios important to the stakeholders. Although the models are missing potentially important physical and social dynamics, our analysis provides insight into plausible pathways of adaptation to climate-related water stress and demonstrates that the vulnerability of a system to stress should not be viewed solely in the context of the physical changes threatening a community.
5 Conclusion
Although technological and institutional interventions have received substantial attention, societal resilience has been linked to intrinsic cultural dynamics in societies as well. Social structures, especially in long-standing, community-managed irrigation systems around the world, often reflect the communities' experiences with a multitude of natural and social shocks. This study is an important first step in understanding the contribution of sociocultural dynamics to the intrinsic adaptive capacity of acequias and similar communities. Using a modeling approach that is informed by multiple disciplines and stakeholders, we effectively demonstrated the importance of social cohesion on mediating a community's sensitivity to hydrological and economic stresses. Although we expected reduced streamflow levels would lead to reduced crop production (and subsequently, agricultural profitability), we found that earlier peak flows, combined with adaptive measures of shifting crop selection, allowed for greater production of higher value crops and led to fewer people leaving the acequia. However, when exposed to downstream pressures (expressed as additional headgate closures), agricultural profitability decreased but feedbacks associated with newcomer introduction and mutualism buffered the community from more detrimental impacts to its population and parcel sizes. Constructed primarily using community participation variables, mutualism refers to the feelings of collective well-being that have been recognized as an important acequia characteristic. Mutualism and its influence on time spent by community members in agriculture served as an important, dynamic feedback variable connecting the hydrological and socioeconomic components in the Valdez acequia. The mechanisms through which social cohesion influence community behavior, however, can vary depending on the community; for example, kinships or religious entities associated with social cohesion may dominate feedbacks to land tenure and other community dynamics in some parts of the world. Continued exploration of these diverse social structures and associated feedbacks in sociohydrological modeling studies is needed to increase our understanding of the influential role of sociocultural dynamics in determining the resiliency of community behaviors.
Acknowledgments
This study was funded by NSF grants (CNH-1010516, WSC-1204685, and DGE-0909667) and through funds provided by the College of Agriculture, Natural Resources, and Human Sciences at Texas A&M University-Kingsville. The authors would like to thank Sam Fernald, Andres Cibils, George Hornberger, Geoff Klise, Jose Rivera, Sylvia Rodriguez, Murugesu Sivapalan, and Mark Stone for their guidance. The upland model and streamflow data are accessible through CUAHSI's Hydroshare repository. Valley model and data can be obtained by e-mailing Benjamin L. Turner: [email protected]. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International, Inc., for the U.S. Department of Energy's National Nuclear Security Administration under contract DE-NA-0003525.
