This review worked to include all searchable studies that have conducted research within the sociohydrological framework to provide a thorough and systematic examination of the types of data used and the methodology for overcoming the inherent spatial-temporal scale challenges within interdisciplinary research. Thus, the primary studies eligible for inclusion in this systematic review needed to meet the following criteria: (a) the study is conducted within the sociohydrological framework as defined by Sivapalan et al. (2011) in which the study considers both humans and water in a coupled system, (b) the study is performing analysis or implementing a model through a case study, (c) the study is from a peer-reviewed journal article.

The primary studies were obtained from searching commonly used economic, social science, and science databases to include EBSCO, ERIC, Google Scholar, ProQuest, SCOPUS, Taylor and Francis, Web of Science, and Worldcat. A search string was determined through an iterative process to identify studies that used any of the search terms in the title, abstract, or keywords. The final search string was the following: (“socio-hydrology ORsociohydrology OR sociohydrological OR sociohydraulic OR human flood interactions OR coupled human-water systems OR human-water dynamics”). The search was further limited to the years 2011–2021 since it was assumed that the new study of sociohydrology was coined by Sivapalan et al. (2011). The search was not limited by geographic location but was limited to full-text studies available in English. The official search ended March 2021 and returned 4,606 studies (Table 1).

Due to the nature and limitations of the search, in particular, the need to have the studies in English, reduces and underrepresents the sample of collected studies. This is an inherent drawback in most systematic reviews, thus even though this search was systematic, the primary studies in the search may still be biased in terms of variations and methodology for dealing with differing spatial-temporal scales within the sociohydrological framework. While this does not necessarily detract from the importance and application of the findings, it is critical to note that this limits the scope of the work and the extent to which the results can be generalized.

From the initial search of databases, a three-phase process was used to screen primary studies that met all eligibility criteria (Figure 1). In pairs, coders independently read the title and abstract for all the studies obtained in the original search using the search string and specified years. Each pair checked if the studies were conducting analysis within the sociohydrological frameworks by determining if both human and water components were included in the study to some degree. It was also important to distinguish between studies that were discussing sociohydrology and those that were proposing a framework or model compared to the studies that were actually implementing a framework or model. This distinction was made because the review only wanted to consider how differing spatial-temporal scales were used throughout the studies, thus a study had to actually be conducting sociohydrological analysis. For example, Di Baldassarre et al. (2015) proposed a novel framework for analyzing the dynamic interactions and feedbacks between flooding and societies but did not implement the framework through a case study. Additionally, Sivapalan (2015) discussed the benefits and challenges of endogenizing humans into hydrologic systems by broadening the hydrologic science field to include perspectives of both social and natural scientists but did not conduct data analysis. Thus, studies were only retained if they mentioned conducting analysis in sociohydrology or coupled human-water systems. Studies were also marked as “maybe” if they discussed sociohydrology or if it was unclear from the title or abstract if they met the criteria. Each pair checked if there were studies that were retained by one coder but not the other and discussed reasons for including or excluding studies. If there was still uncertainty, the group as a whole discussed the study and came to a consensus decision. A designated coder also reviewed the studies and removed any duplicates. The majority of the 4,606 studies examined did not discuss coupled human-water systems within the sociohydrology framework.

In phase two, 463 studies were marked “included” and 411 studies were marked “maybe” totaling 874 studies left for additional screening. Using Rayann (Ouzzani et al., 2016), pairs of coders independently assessed whether each study fit the eligibility criteria outlined above by reading the title, abstracts, introduction, and methods. The coders discussed any discrepancies and made exclusion decisions together and if needed, the group as a whole resolved any disagreement between coders. In this phase, articles were excluded if two-way feedbacks with both social and hydrological components were not discussed. Studies were also excluded if they were published as a book, thesis, dissertation, or if there was not a digital full-text in English. After screening, there were 152 studies left for further analysis.

Coders were trained together on how and what to code using two studies as examples. Each coder then independently coded the 152 eligible studies using a common coding schema (Table 2). Some of the key information included publication year, spatial and temporal extent and resolution(s), geographic location(s) of the study, data source(s), human and water components, typology, and methodological approach. Disagreements or concerns were resolved through group discussions and a consensus was met. As an additional check two coders were designated to review the coding for consistency and accuracy across all studies. Throughout the screening and coding process, coders met regularly and had weekly check-ins with the group.

The non-scale themes were identified and analyzed to determine the commonalities of the types of data and variables used to understand the sociohydrological process. There was not a clear temporal trend for water components addressed in the primary studies, rather the trend was more related to an increase in articles published under the umbrella of sociohydroly since 2011. The first non-scaler theme is water body component that describes the hydrological process considered in the primary study. Eight water components were identified and coded: water resource management, flood management and stormwater structures, farming and irrigation management, groundwater, water quality, reservoir and lakes, drought, and coastal zones. The most common water component was water resource management (39%; Table 3). This was not surprising because water resource management was defined very broadly to include any study that considered an area's water system as a whole instead of focusing on a specific water or hydrologic component. In other words, if a primary study did not fit into the other seven components then water resource management was coded. It should be noted that the percentage of the primary studies does not add to 100% given several of the articles featured multiple components. For instance, Iwanaga et al. (2018) covered four components, reservoir, water resource management, groundwater, and irrigation, in their integrated models and scenarios based on stakeholders in the basin region. Water resource management, reservoir, and drought often overlapped, as did groundwater and irrigation. Additionally, water components imbue a social-use connotation, such as irrigation and water resource management, as well as an implied social response, such as drought, flooding, or tsunamis in coastal areas.

In addition to water components, social components were also identified throughout the primary studies. This was important because the overreaching goal of sociohydrology is to no longer treat the human factor as a stationary external force, but to consider humans and their actions as a part of the water cycle dynamics (Sivapalan et al., 2011). As previously noted, modeling across and within global scales is challenging (Pande and Sivapalan, 2017). The same is true when scaling social science indicators. As hard as it is to capture social components spanning macro, meso, micro, and psycho-social levels, it is even harder to delineate social components due to their interrelation. In many cases authors use multiple of each, and/or indexes, scales, or theories which incorporate various, non-exclusive components. Due to this circumstance, social components were identified qualitatively. Nevertheless, five overarching social components were identified: socioeconomic, vulnerability/health, memory, demography, and behavior.

There exists a wide range of social in sociohydrology, the most common was socioeconomic indicator, included in 20% of the studies. Although, this social component featured a vast range of variables. For instance, Wang et al. (2019) approximated social by assessing population and Gross Domestic Product (GDP) yearly (basin scale) whereas Wilhelmi and Morss (2013) measured it using physical/language abilities and access to resources. Sun et al. (2019) measured socioeconomic with population, regional GDP, proportion of the secondary and tertiary industry, labor compensation per capita, and proportion of urban residents. Luo and Zuo's (2019) socioeconomic index included per capita GDP, per capita net income of rural residents, per capita disposable income of urban residents, tertiary industry share of GDP, population density, urbanization rate, Engel's coefficient, number of hospital beds per 10,000 people, per capita urban road area, and residents' satisfaction with the environment. Huber et al. (2019), in addition to households, economy, water supply, etc., included tourism. In addition, socioeconomic factors are sometimes explicitly used to approximate inequalities in water distributions or effects from conservation efforts. For instance, Breyer et al. (2018) found outdoor water conservation for reservoirs at the city scale produced, among other effects, downward redistribution of water along a socioeconomic gradient. In other words, as a result of the conservation, people with lower socioeconomic status had less water access than those with higher status.

Social vulnerability generally measures how individuals and communities are able to withstand shocks, including natural hazards. Cian et al. (2021) analyzed Northeast Italy's frequent flooding every 5 years using the flood vulnerability index, or FVI, which the authors claim to allow for a dynamic and adaptable assessment of vulnerability, necessary in preparation for the worst of climate change. Human development index (ZamanZad-Ghavidel et al., 2021). Hazarika et al. (2016) Dhemaji in the Upper Brahmaputra floodplain, indigenous knowledge and adaptations has allowed floodplain dwellers to be able to coexist with floodplain dwellers. Vulnerability can be tied to human resource depletion or ecosystem interference beyond the natural rate or environment (Liu et al., 2014). Related to vulnerability is human health. Keys and Wang-Erlandsson (2018) considered child malnutrition as an indicator of food security with moisture recycling. Baker et al. (2015) utilized social health risk, instead of the more widely-used contaminant concentration rate, as an optimization variable improves water management decisions aimed at maximizing social well-being.

Several authors assessed the social component via social memory, or the shared histories and experiences of groups of people in a place. In addition to the commonly-used variables of community awareness or sensitivity, Yu et al. (2017) add institutions for collective action and connections to an external economic system to their assessment of flood memory. Grames et al. (2016) include social memory and the sensitivity of social awareness which feedback on flood adaptation within the timespan of economic agents' decision-making.

Demography was frequently used as a control or endogenous variable (Braden et al., 2014; Roobavannan et al., 2017; Breyer et al., 2018; Ferdous et al., 2018; Garcia and Islam, 2018; Sarmento Buarque et al., 2020). A majority of sociohydrological research containing issues of scale and feedbacks do not utilize demographic variables to featurein analysis the latent structural sinequalities that operate at the center of the human-water nexus (Sanderson, 2018). Studies that did include these structural inequalities featured the typical demographic categories of race, age, sex, gender. Demographic factors and their influence on behavior were also examined. For example, Huang et al. (2020) find demographics and flood risk perception do not have direct impacts on protective coping behaviors with flooding, but are mediated by flood risk knowledge and flood risk attitude. If demographics are used, they are included in indexes with more traditional socioeconomic variables income, population, GDP, and so forth (e.g., Huber et al., 2019).

Finally, human behavior was the last social component identified throughout the primary studies. There are no universally accepted laws of human behavior as there are for physical systems. Therefore, the variables and methods which measure behavior are wide-ranging. One aspect of behavior was public participation, such as capacity building, participatory activities, and workshop discussions (Kotir et al., 2017; Re et al., 2017; Bojórquez-Tapia et al., 2019; Bradford et al., 2020). These were conducted with surveys, interviews, workshops, or data from previous studies. Public participation is generally considered to signal bottom-up socio-behavioral change over time. A socio-behavioral measure considered to have greater top-down determination is policy preference, choice, action, or intention. In fact, 21% of the studies included variables representing a socio-behavioral dimension including policies or a policy (Liu et al., 2015; Srinivasan, 2015; Di Baldassarre et al., 2016; Garcia and You, 2016; Giuliani et al., 2016; Grames et al., 2016; Bakarji et al., 2017; Farjad et al., 2017; Gonzales and Ajami, 2017; Haeffner et al., 2017; Han et al., 2017; Khan et al., 2017; Kotir et al., 2017; Noël and Cai, 2017; Re et al., 2017; Sanderson et al., 2017; Essenfelder et al., 2018; Keys and Wang-Erlandsson, 2018; Robinne et al., 2018; Abebe et al., 2019a,b; York et al., 2019; Bradford et al., 2020; Du et al., 2020; Wilfong and Pavao-Zuckerman, 2020; Bou Nassar et al., 2021; Hanus, 2021; Oneda and Barros, 2021; Savelli et al., 2021; ZamanZad-Ghavidel et al., 2021).

Studies not only examine the outward-facing socio-behavioral measures, but those more inward, deeply-seated, and traditionally sociological. These include value systems, attitudes (Ciullo et al., 2017; Huang et al., 2020; Bou Nassar et al., 2021), beliefs (e.g., Souza et al., 2021) and norms (e.g., Braden et al., 2014). The authors consider perceptions (Elagib et al., 2017; Devkota, 2020) behavioral because they are often conceptualized and operationalized in ways that assume that by perceiving something in a particular light, the agent's behavior or action will follow accordingly.

The data source types identified and coded for the primary studies were primary, secondary, simulated, and combinations thereof (Table 4). Primary data source is defined as data collected by the authors. Secondary data sources includes data that the authors retrieved from another source, such as local, regional, and state-based sources, however, most came from governmental surveys, such as the USDA (Dziubanski et al., 2020), Census Bureau (Sanderson et al., 2017), CIRD (Camargo, 2017), international organizations (ZamanZad-Ghavidel et al., 2021) which often provided economic and social variables. Simulated data was defined as data that the authors simulated for the study. The most common data source type throughout the primary studies was secondary data (47%) followed by primary and secondary (30%). There were 21 studies which included simulated data (e.g., Garcia and You, 2016; Gonzales and Ajami, 2017).

Following the identification of the data type source, eight unique data types emerged from the review: remote sensing, meteorological, survey (interview), survey (site investigation), survey (government), archaeological/historical, hydrologic, and maps. The results show that hydrologic data was included in 63 studies (41% of total), by far the most (Table 5). This is not surprising given 24% of the journals in this review were primarily hydrological (Table 6), with 20 alone coming from Hydrology and Earth System Sciences. Examples of hydrologic data included in the primary studies were stream gauge data (e.g., Garcia et al., 2016), soil moisture (e.g., Den Besten et al., 2016), groundwater and irrigation rates (e.g., Tellman et al., 2018; Gholizadeh Sarabi et al., 2021), and hydropower (e.g., Huber et al., 2019). Maps were defined as those not derived from remote sensing and were used the least (9%). In most cases, maps were used in addition historical or archeological studies (e.g., Nüsser et al., 2012; Baker et al., 2015; Chang and Huang, 2015; Caprario and Finotti, 2019; Cian et al., 2021; Gholizadeh Sarabi et al., 2021). Where historical and archaeological data was only used in 10% of the primary studies, by showing how human and water systems co-evolve over time, from ancient Rome (Di Baldassarre et al., 2017) to West Africa (Coutros, 2019) to the Tarim River basin in Western China and beyond (Liu et al., 2014).

Typology and methodological approaches were coded in this review to understand what types of methods were used to combine social and hydrological variables within the sociohydrological framework. Three typologies were coded based upon the types of sociohydrology analysis identified in Pande and Sivapalan (2017). These include process, historical, and comparative. These typology codes were used to identify how the methods and tools are used, including what types of questions they help answer about a system. Additionally, the following methodological approaches were identified throughout the primary studies: established modeling, system dynamic models, agent-based models, and synthesis analysis. The methodological codes provide insight into the diversity of tools and methods being used in sociohydrological research, as well as drawing attention to approaches unique to the developing field. Typologies were coded based upon the types of sociohydrology analysis identified in Pande and Sivapalan (2017). These include historical, comparative, and process sociohydrology. Each of these approaches varies from traditional hydrologic studies or modeling because they endogenize social factors into their analysis. Sociohydrologic approaches focus less on calculating flow rates and finite numerical outcomes and focus more on understanding system the nature of system evolution. Specifically, historical sociohydrology was defined as having the aim to understand a coupled system from its immediate or distant past and often leverages historical accounts of specific events. A major emphasis in historical sociohydrology is understanding the major events that forced a system to evolve to its current state. Examples of these major events include, but are not limited to, wars, regime changes, major infrastructure projects, policy changes, droughts, and other severe weather events. Comparative sociohydrology was defined as a study that compares different systems across space or time, with the goal of understanding how different systems respond to similar events, or testing if certain generalizations can be made about interactions in larger human-water systems. Lastly, process sociohydrology was defined as trying to understand and hypothesize about the nature of observed processes. Process sociohydrology often asks “what if” questions to understand how a system would respond to certain changes. Such changes can come in the form of new policies, the addition of new stakeholders, reductions in resource availability, or technological improvements. While models are not specifically required for sociohydrology analysis, they greatly assist process sociohydrology. The most common typologies were process and historical, with process being the most common in the year 2020 (Figure 2). Comparative studies peeked in 2019 with 4 studies.

Next, methodological approaches was coded to identify how methods or tools are used. Firstly, established models are models which have been replicated, used, and modified across studies and over time. Established models were identified in 22% of the primary studies and examples include MIKE SHE/MIKE 11 (Farjad et al., 2017), SWAT (Baker et al., 2015; Martin et al., 2016; Khan et al., 2017; Essenfelder et al., 2018), and ISBA-MODCOU (Medeiros and Sivapalan, 2020). By nature, these models are most familiar to traditional hydrologists but are modified to incorporate additional social parameters to study system feedbacks.

In contrast, t system dynamic models represented frameworks developed specifically for sociohydrologic analysis (e.g., Den Besten et al., 2016; Sanderson et al., 2017; Gholizadeh Sarabi et al., 2021). System dynamic models represent top-down feedback approaches utilizing multi-loop, non-linear structures. They can be as simple as a causal loop, or as complex as a system of differential equations. While these models should mimic general interactions between stakeholders and resources, they generally do not require extensive calibration (Ding et al., 2018). System dynamic models were used in 42% of the primary studies and were useful to studies spanning the global scale (e.g., Elshafei et al., 2014; Hou et al., 2019; Liu et al., 2019; Hossain and Mertig, 2020). Hou et al. (2019) examined the social water cycle fluxes across 39 countries assessing total water use of countries and evolution mechanisms. Liu et al. (2019) looked at global virtual water use across 44 nations and regions. Hossain and Mertig (2020) considered water footprint or “virtual water consumption” by looking at energy, world position, beef consumption, and urbanization.

Agent-based modeling (ABM) simulates the interactions between water systems and one or multiple agents, which can be conceptualized as individuals or collective groups (Akhbari and Grigg, 2013). Rather than using causal loops or differential equations to describe interactions, ABMs rely on computational information streams and predetermined rules. The predetermined rules in ABMs govern how each agent will respond to model scenarios and interact with other agents. While ABM is similar to system dynamic models in that they model interactions, the authors established separate categories for these approaches because ABM is considered a bottom-up approach to modeling sociohydrological processes while system dynamic models are considered top-down models (Ding et al., 2018). ABM appeared in 13% of the primary studies and has been used to analyze farmer's agency shaping their financial situation and their water conservation (Pouladi et al., 2019) and can be used as a decision support tool for watershed management (Huber et al., 2019). ABM are flexible and dynamic and can incorporate established models. For instance, Khan et al. (2017) develop an ABM framework that includes a SWAT model to simulate the impacts of water resource management decisions that affect the food–water–energy–environment (FWEE) nexus at a watershed scale.

Finally, synthesis analyses integrate varying perspectives, data types, and approaches to provide a narrative of the system. Synthesis analyses generally attempt to describe the current state of a system by analyzing individual components and making inferences about how the states of those components are linked. While not always the case, this approach is often qualitative in nature, and as a minimum requirement, incorporates some form of qualitative data in the analysis. For instance, Haeffner et al. (2017) examined the historical, biophysical, economic, and cultural relations embedded in the development of urban water infrastructure in an urban Mexico watershed. Vollmer et al. (2018) created the Freshwater Health Index that integrates governance mechanisms, ecosystem dynamics, and stakeholder perceptions to better engage stakeholders in addressing multiple freshwater demands. In addition, Zhao and Mo (2020) provide a comprehensive analysis of the Holocene hydro-environmental evolution in Jianghan-Dongting Basin in China by studying sediment cores.

Variables used for studying feedbacks were coded qualitatively. This was done because of the complexity and variety of variables that the primary studies used to quantitatively or qualifiedly describe the sociohydrology interactions. A few notable examples of feedbacks used are Bojórquez-Tapia et al.'s (2019) Analytic Network Process-situated feedback loop which involved synchronous and parallel updating of the geographic data river discharge time series (streamflow/inflow to reservoir). Roobavannan et al. (2017) mimicked and explained feedback between various subsystems of the human-water system, water availability, agriculture, environment health, manufacturing and services, technology, human population, and derived reservoir outflow from variation in water storage. Puzyreva and de Vries (2021) examined historical exposure to flooding, the community response and preparedness, evolution of risk acceptance, community engagement, perceptions of flooding causes, and how they feedback one another. Towler et al. (2019) drought feedbacks were influenced by science, technology, as well as historical lessons learned and management strategies. Robinne et al. (2018) included education capacity in their social variables. Baker et al. (2015) included media coverage and comments into their feedbacks. These are but a few of the feedback variables and mechanisms—creative, useful, and potentially transformative—found throughout the primary studies.

Temporal scale was investigated through both the temporal extent of a study and the temporal resolution at which the sociohydrological processes in the study was analyzed. Temporal extent was determined using the following categories: millennia, century, decadal, yearly, and event-based. A study was categorized as “millennia” if the temporal extent of the sociohydrological processes was over the course of one or multiple millennia. A study was categorized as “century” if the process was analyzed over the course of one or multiple centuries but less than a millennium. All other temporal extents followed this classification method with the exception of event-based studies. Event-based temporal extents referred to studies that did not specify a time frame within which analysis took place, but instead analyzed sociohydrological processes for the duration of a specific event or events. This was most common for studies that used simulated data to examine sociohydrological processes. For example, multiple studies simulated flood events to better understand pathways for city development (Viglione et al., 2014) or observe interactions between flood risk and behavior (Ridolfi et al., 2020a). Alternatively, some studies used models to simulate changes in a sociohydrology system over the course of the implementation of a specific policy (Du et al., 2020) or flood management strategy (Albertini et al., 2020). In these cases, temporal extent is not explicitly stated, but there is an implicit understanding that these events occur over a definite amount of time.

Results of coding showed primary studies analyzing sociohydrological processes over decadal (28%), yearly (30%), and event-based (18%) temporal extents to be the most common (Table 7). This trend could be indicative of the availability of necessary data to analyze human-water systems. For example, multiple studies reviewed relied on social data from either the Australian census (Elshafei et al., 2014, 2015; Roobavannan et al., 2017; Wei et al., 2017), which takes place every 5 years, or the United states census (Sanderson et al., 2017), which takes place on a decennial basis, providing reliable quantitative social data for studies investigating sociohydrologic processes over multiple decades. Moreover, data necessary to model hydrological processes is more abundant at these time scales as well, such as yearly water budgets (Keys and Wang-Erlandsson, 2018), water footprints (Souza et al., 2021), and data from Resource Bulletins that provide yearly data on water use and reservoir flows (Hou et al., 2019; Li et al., 2019; Luo and Zuo, 2019; Sun et al., 2019). Studies analyzing processes over centuries or millennia are more limited by data availability and often rely on archeological data (Pande and Ertsen, 2014; Wei et al., 2017; Lu et al., 2018; Zhao and Mo, 2020; Puzyreva and de Vries, 2021) or are limited to specific geographical areas that have well-documented hydrological and social data (Di Baldassarre et al., 2017). Examining the frequency of studies using these different temporal extents over time revealed that decadal and yearly temporal extents steadily increased as publications in sociohydrology increased (Figure 3). However, in 2020 studies employing decadal and yearly temporal extents decreased, and event-based studies increased.

Temporal resolution was also coded by recording the overall temporal resolution for a study or individually recording the temporal resolutions of the different data used in a study. Temporal resolution varied from groundwater level data collected every 15 min (O'Keeffe et al., 2020) to sociohydrological interactions over 6,000-year periods (early-Neolithic, middle-Neolithic, and late-Neolithic; Zhao and Mo, 2020). However, this variable is not reported quantitatively as the majority of studies reviewed did not provide sufficient information for coders to confidently identify temporal resolution. Lack of clear, explicitly stated temporal resolution emerged as a trend among the sociohydrology studies reviewed and was especially prominent among studies that used solely simulated data. Further, in the event that temporal resolution was reported for each data type used in a study, explanation of how these data types of differing temporal resolutions were integrated was often not present.

Similar to temporal scale, spatial scale for each primary study was examined through spatial extent and spatial resolution. Spatial extent was determined by identifying the spatial boundaries of a study as they fit into one of the following categories: global, international, national, regional, state, county, city, social network, watershed, or water body network. Primary studies categorized as having a “regional” spatial extent include studies that were analyzing sociohydrological processes at a spatial extent that did not conform to political boundaries (nation, state, county) but also did not belong to the water or network based categories (watershed, social network, water body network). For example, Sun et al. (2019) examined sociohydrological processes in a region covering 31 provincial-level administrative areas in China to better understand virtual water trade and water consumptions patterns. A study was categorized as having the spatial extent of a “water body network” when the study boundaries were determined by a specific geographical area surrounding a water body or system, such as an irrigation system (Khalifa et al., 2020), floodplain (Ferdous et al., 2018), or aquifer (Towler et al., 2019). A study was coded as having the spatial extent of a “social network” when the study area was defined by a social structure but did not conform to any of the more common social boundaries, such as county or state. For example, Bojórquez-Tapia et al. (2019) defined their study area by the sewage system in Mexico City. As a sewage system is socially constructed, this spatial extent qualified as a social network. Lastly, some primary studies were coded with more than one spatial extent as they analyzed areas of varying spatial extent for the purpose of comparison. For example, Akhter et al. (2021) compared sociohydrolocial processes at the national, state, and county level to gain a better understanding of flood risk management in floodplains across the United States.

The results of coding for spatial extent showed that the most common were watershed (37%), regional (24%), and city (23%; Table 8). Coders also observed that specific watersheds were of particular interest to researchers. For example, multiple primary studies focused on the Heihe River Basin in China (Lu et al., 2016; Wang et al., 2019; Du et al., 2020) and the Murrumbigee River Basin in Australia (Roobavannan et al., 2017, 2018, 2020). Further, studies analyzing sociohydrological processes within watersheds often prioritized water resource management, especially as it concerned the sociohydrological processes of irrigation (Roobavannan et al., 2017; Sanderson et al., 2017; Nüsser et al., 2019; Ogilvie et al., 2019; O'Keeffe et al., 2020) and drought (Di Baldassarre et al., 2017; Pouladi et al., 2019; Medeiros and Sivapalan, 2020). Studies analyzing sociohydrological processes at the city level were often focused on water resource management as it related to stormwater (Tellman et al., 2018; Oneda and Barros, 2021), wastewater infrastructure (Faust et al., 2017; Tellman et al., 2018; Souza et al., 2021), and human water consumption (Li et al., 2019; Savelli et al., 2021).

Watershed, regional, and city spatial extents being the most common among primary studies is also important for understanding the lens through which sociohydrology studies are framed. One of the key challenges of sociohydrology is appropriately integrating water systems that do not conform to political boundaries, and human systems that do not naturally align with hydrologic boundaries or networks. Reconciling these differences to model human-water systems or synthesize data to better understand sociohydrological processes requires careful consideration of the spatial extent of the study. Studies using watershed spatial extents inherently prioritize hydrological processes in the human-water system. Similarly, studies using city level spatial extents prioritize social processes in the human-water system. As regions can be defined both physically and socially, the popularity of this spatial extent potentially demonstrates the general flexibility of not using specific political or hydrological boundaries. Additionally, examining the use of these spatial extents over time, it can be observed that the use of regional spatial extents peaks in 2017, while the use of the watershed and city spatial extents steadily increases, peaking in 2020 (Figure 4).

Spatial resolution was also coded for by recording the overall spatial resolution for a study or individually recording the spatial resolutions of the different data used in a study. However, similar to the data gathered for temporal resolution, spatial resolution was not reported consistently or clearly enough in the primary studies reviewed for the coders to confidently identify the spatial resolution of each study and is therefore not reported quantitatively. Qualitatively, though, the coders observed that spatial resolution was reported more frequently for raster data than it was for vector data. In studies that used both vector and raster data for their modeling or analysis, this is especially notable given vector and raster data are inherently incompatible and must be manipulated to reach compatibility. Converting raster to vector or vector to raster to create compatible data layers comes with its own trade-offs (Congalton, 1997), such as loss of data quality due to aggregation or lessened reliability of data due to interpolation. However, transforming and manipulating data within the same data format to adjust for different scales still requires careful consideration of how the data and its subsequent interpretation and representation will be affected. Without explicit reporting of the spatial scales of these data sources or how these data sources are being manipulated within a model or series of analyses, it is difficult to critically reflect on scale issues in sociohydrology, as having access to this information is key to understanding the implications of scale issues on methodological approach and study results.

There were few studies that engaged in full transparency when reporting, though. For example, Wilhelmi and Morss (2013) conducted a case study of Fort Collins, Colorado, to gain a better understanding of interactions between social vulnerability and extreme precipitation events. The authors listed the data being used in the social vulnerability index in a table and included spatial and temporal scale information for each data source. Further, the authors explained how precipitation data was measured, manipulated, and ultimately integrated with the social vulnerability measures of sensitivity and coping capacity. In this study, a combination of tables, figures, and text explanation were used to explicitly communicate spatial and temporal scale, data manipulation as it related to scale, and data integration, allowing the reader to critically interrogate the results of the study.

The spatial distribution of the countries where the primary studies reviewed took place revealed high frequencies of studies in the United States, Australia, and China (Figure 5). This is undoubtedly partially due to the authors limiting their systematic literature search to English-only journal articles. However, this could also be indicative of a focus in sociohydrology literature on areas that are facing difficult social and water situations. For example, Australia has been experiencing extreme droughts and water shortages that are having severe effects on agriculture and farmers (Di Baldassarre et al., 2017; Roobavannan et al., 2017). A focus on sociohydrology issues in Australia is thus not surprising. Lastly, data availability plays an important role in the distribution of sociohydrology papers reviewed in this study.

There is a risk of coding error, as official intercoder reliability scores were not calculated. That said, the authors aimed to minimize error by randomly checking each other's work ex post facto for accuracy and detail. It is acknowledged that selective reporting of complete studies, such as publication bias or “outcome reporting bias” within individual studies, may occur in our methods and reporting results (Moher et al., 2009). The implications of these biases may be unclear, yet there is little dispute that selective outcome reporting typically occurs in the context of systematic reviews (Moja et al., 2005). To minimize biases, Moher et al.'s (2009) PRISMA checklist was used. Future systematic reviews could adapt this scale for those in the natural or social sciences, to more rigorously check for, and alleviate, these biases. However, such a re-tooling is outside the scope of this review.