Synthesizing dam-induced land system change

Abstract

Dam construction and operation modify land systems. We synthesized 178 observations of dam-induced land system changes from 54 peer-reviewed case studies. Changing extents of forests (23%), agricultural land (21%), and built-up areas (11%) were reported frequently, alongside alterations in land use intensity (23%). Land cover changes were mostly related to hydropower and multi-purpose dams, while irrigation dams dominantly caused land use intensity changes. While a significant share of the changes was caused by reservoir flooding (29%), indirect effects which interact with societal and environmental systems (42%) were of utmost importance. We suggested the distance to the dam and the time since commissioning as potential controls for the direction of land system changes. Our insights provide opportunities for future inductive investigations across large populations of dams at regional to global scales and highlight that multi-disciplinary research perspectives are imperative for the production of generalizable knowledge.

Electronic supplementary material

The online version of this article (10.1007/s13280-018-01144-z) contains supplementary material, which is available to authorized users.

Introduction

Globally, tens of thousands of dams and reservoirs provide societal and economic benefits, such as irrigation water for nearly one million square kilometers of agricultural land (ICOLD 2017) and 3000 terawatt hours of electricity per year (Zeng et al. 2017). The growing number of dams also causes a plethora of severe environmental and socio-economic side-effects with global-scale implications, including modifications of hydrological cycles (Rosenberg et al. 2000), hydroclimatic shifts (Destouni et al. 2012; Jaramillo and Destouni 2015), increasing greenhouse gas emissions (Barros et al. 2011), as well as negative implications for human health (Keiser et al. 2005), livelihoods (Kirchherr et al. 2016), and food security (Fung et al. 2019). Furthermore, dams modify land systems, which encompass all processes and activities related to the human use of land, including its social and ecological benefits and unintended consequences (Verburg et al. 2013). Land submergence during reservoir flooding or repercussions from social or environmental side-effects can, for instance, alter the spatial patterns or intensity of land use (WCD 2000; Tortajada et al. 2012). However, a globally consistent assessment of the effects of dams on land systems has been missing to date.

The effects of dams on land systems can be divided into direct and indirect effects. Direct effects of dams refer to land cover or land use changes caused by reservoir flooding, which encompass changes in area extent and modification within the land use and land cover categories, such as changes in the intensity of land use. Due to their spatially and temporally restricted nature, the direct effects can be captured with current methodological toolkits. For instance, remote sensing allows to detect the extent of reservoir inundation and the associated losses of natural ecosystems, built-up area (e.g., settlements, infrastructure), and agricultural land.

Indirect effects relate dams and land systems through more complex sequences of causes and effects. Indirect effects of dams are diverse, non-static, can occur in proximate or distal locations, and often unexpectedly. These effects are difficult to monitor and isolate with current methods, and are therefore only weakly understood. For instance, dams improved irrigation water availability globally (Biemans et al. 2011), thereby increasing the reliability and profitability of agricultural production, which can foster agricultural expansion and intensification (Thomas and Adams 1999; Fraiture et al. 2014). Novel livelihood strategies and economic opportunities emerge from such changes in agricultural production systems (Gordon and Meentemeyer 2006; Casado et al. 2016), which in turn alter land ownership structures (Loker 2003) or increase the competition for land and water resources (Thomas and Adams 1999; Delang et al. 2013). Synthesizing these manifold repercussions of dam construction on land systems is a key for better understanding of dam-induced land system changes.

Empirical evidence concerning the impact of dams on land systems is currently limited to case study findings. Synthesizing case study evidence on the causes and effects of dam-induced land system change allows to attain a better understanding of the frequencies and characteristics of such interactions (van Vliet et al. 2016; Magliocca et al. 2018). In this context, meta-studies integrating quantitative and qualitative observations of land system change can help to advance theory building, identify knowledge gaps for future scientific inquiry, and inform policy (Magliocca et al. 2015; van Vliet et al. 2016). As dam construction is part of the option space to meet the increasing demands for hydropower and irrigation water (You et al. 2011; Zarfl et al. 2015), integrating and combining evidence of dam-induced land system change will help to advance empirical insights and provide stimuli for future research.

Dam-induced land system changes are characterized by complex causal chains, which involve a variety of land system components, actors, and processes. This complexity requires disentangling the causal chains of dam-induced land system change. Decomposing causal chains into single unidirectional links, defined as causal effects (Meyfroidt 2016), allows for synthesizing findings across heterogeneous study sites and methodologies and thus facilitates a systemic understanding of the observed phenomena.

Isolating the causal effects of dams on land systems from other drivers or determinants of land system change, such as changing market forces or climate change, is challenging. Counterfactual study designs allow for setting up quasi-natural experiments, which render them valuable for common research questions in land system science (Meyfroidt 2016; Butsic et al. 2017). Given the abundance of suitable datasets to support such applications, counterfactuals are cost-efficient and readily applicable tools for isolating the effects of dams on land systems (Braatne et al. 2008). Investigating the types of counterfactuals used in the existing literature can help to reveal suitable study designs and common analytical gaps to assess the effects of dams across diverse disciplinary perspectives, data types and coverages, study site contexts, and geographic scales.

This study aims at providing a descriptive overview of the current state of research concerning dam-induced land system change. We synthesized the literature of land system changes associated with the construction and operation of dams, the spatial distribution of such changes, as well as the datasets and counterfactual methodologies employed to isolate dam-induced effects. Our main research questions are:

• Which types of dam-induced land system change were commonly reported in the scientific literature?

• What are the most common processes and directionalities of dam-induced land system changes?

• Which types of data and counterfactual methodologies were used for attributing land system change to dam construction or operation?

Materials and methods

Case selection

We gathered all available peer-reviewed studies, which focused on land system change associated with the construction or operation of dams. We searched research articles published in English since 1950 and listed in the ISI Web of Science (last accessed on Feb 1, 2018). We restricted our analysis to publications whose titles contain the keywords “dam*” or “reservoir*” and at least one of the following: “land*”, “chang*”, “ecosystem*”, “settle*”, “agricult*”, “irrigat*”, “crop*”, “flood*”, “pasture*”, “grassland*”, “urban*”, “wetland*”, whereas the asterisk acts as placeholder for additional characters.

The search resulted in 3193 publications, which formed the literature corpus used for subsequent refinements. Based on the abstracts, we categorized the literature regarding its thematic scope, thereby excluding unrelated articles, which focused, e.g., on dam engineering (n = 485), modeling of dam failure (n = 238), effects of climate change and land use on water quality and quantity in man-made reservoirs (n = 194), oil and gas extraction (n = 150), geology and quaternary sciences (n = 124), and other unrelated fields. We focused on observed interactions of dam construction and operation and land systems and therefore also excluded studies that modeled the potential impacts of planned infrastructures, and studies that simulated test floods or the effects of dam removal or collapse. Our final literature corpus consisted of 54 case studies (all references are listed in Table S1).

Study site characteristics

We compared dam attributes, as well as socio-economic and biophysical context of the 54 case studies against a global sample (n = 8444) of dams (FAO 2015). We used Mann–Whitney U tests to assess differences in the statistical distribution of eleven attributes, namely latitude, longitude, elevation, mean annual temperature, total annual precipitation, commissioning year of the dam, reservoir storage capacity, travel time to major cities as a proxy for remoteness, population density, amount of irrigation water applied in 0.5° grid cell, and conveyance losses (see Supplementary Material Table S3 for details). We further compared the shares of the four most frequent purposes of the dams (irrigation, hydropower, flood control, water supply) across the selected observations and all dams registered in a non-spatial global database of large dams (ICOLD 2017).

Synthesis of case study evidence

We extracted direct and indirect interactions between dam construction or operation and the land system (from here on called observations) from the 54 case studies. Each observation represents one link of a causal chain that relates dam construction or operation (cause) with a change in one or multiple land system components further down the causal chain (effect). Each observation thus consists of a cause, a directionality (increase, decrease, or general alteration), and a resulting effect in the land system or a related system component (e.g., agricultural land extent, population displacement, or changing water seasonality). To avoid replications, observations must be unique combinations of a study area, a cause, directionality, and an effect (van Vliet et al. 2015). In the case of multiple dams or study areas being subject to one analysis, or one study site being studied multiple times, similar findings were counted once, whereas new findings were accounted for separately.

All extracted observations were thematically coded. We classified causes and effects into nine broad land system categories, including seven land use and land cover types (agriculture, bare land, grassland, forests and woody vegetation, wetlands, water, built-up) as well as land use intensity and land demand. While effects on the land use and land cover types exclusively relate to changing extents, effects on land use intensity include alterations of management strategies such as irrigation or fertilization, the cultivation of new crop types or varieties, increasing productivity in terms of multi-cropping, or yield improvements. Dam-induced effects on land demand include increasing land prices, increasing competition over land resources, and changes in land ownership. All aspects not directly affecting the land system categories per se were thematically aggregated into the categories of environmental change and societal change. Environmental changes include alterations in hydrological regimes, sediment load, river morphology, floral and faunal habitat, or local climate. Societal changes comprise population displacement, conflicts, changing livelihoods, alterations in water resource management, or changes in policy (see Table S2 for further examples).

We visualized all observations using chord diagrams, which allow for illustrating the overall system, while highlighting the most frequent interactions between individual system components. Note that feedbacks (i.e., when cause and effect were present within the same category) can be artifacts of the thematic aggregation, e.g., when population displacement (societal change category) caused shifts in livelihood strategies (also societal change category).

Spatial scales, data sources, and counterfactuals

For each article, we recorded the scale of analysis. Local scale analyses cover study sites of limited spatial extent, such as single villages. Regional scale analyses are characterized either by observations across multiple isolated study sites or wall-to-wall coverage of larger regions. National scale accounts for analyses of a whole country, larger cross-boundary areas, and continental scale covers several countries or entire continents. Further, we recorded the study location and stratified the observations by UN world regions.

We documented the start and end of the study period. In cases where multiple datasets were used, we recorded the earliest observation as the start and the latest as the end. We further recorded the types of data used. These include interviews and qualitative data, quantitative field data, including measurements or surveys, geospatial data, remote sensing data, and combinations thereof.

We classified the employed counterfactual study design of the case studies into temporal counterfactual (pre- vs. post-dam or sequences of revisits in the post-commissioning phase), spatial counterfactual (upstream vs. downstream; progressive downstream; dammed vs. free-flowing), spatio-temporal counterfactual (spatial and temporal combined), or no counterfactual (Braatne et al. 2008). Manipulative strategies (modifications of river flow, test floods, or dam removal) or biophysical modeling also qualify as potential counterfactuals. However, the first did not occur in the screened literature and the latter was excluded in our case selection.

Results

Study site characteristics

We compared dam attributes, as well as socio-economic and biophysical characteristics of the 54 case study locations to a global sample of dams. The examined case studies did not significantly differ from the global sample in case of most attributes (p < 0.001), except for latitude, mean annual temperature, commissioning year, and storage capacity (Supplementary Material Table S3). Relative to the global sample, the dams listed in our selected cases were located in lower latitudes, and warmer regions, were commissioned later, and had larger storage capacities. Unsurprisingly, the full variability of the global sample was rarely covered by the dams contained in our case studies (Fig. 1). For instance, our sample failed to include dams located in remote locations, in high elevations, or in areas of high population densities. Furthermore, our case study sample contained a larger share of dams with irrigation function and a lower share of hydropower and flood control dams as compared to the global sample (Table 1).

Fig. 1.

Comparison of selected dam attributes, socio-economic and biophysical characteristics of the case studies (red points) against a global sample of dams (blue points)

Table 1.

Shares of dominant purposes for the dams selected from the case studies compared to the global sample of dams

Purpose (single or multiple)	Selected cases (%)	Global sample (%)
Irrigation	55.2	34.7
Hydropower	27.1	16.5
Water supply	31.2	12.9
Flood control	2.2	12.4
Multi-purpose	15.9	16.7

Systematic view of dam-induced land system changes

We extracted a total of 332 observations from the 54 case studies. Of these, 178 observations affect the land system categories, whereas the remaining 154 relate to changes in non-land categories. Of the 178 observed land system changes, 29% were caused by reservoir flooding and affected five of the nine land system categories (Fig. 2a), most frequently forests and natural vegetation, agricultural land, built-up area, and grasslands. Another 29% of the observations directly resulted from dam construction or operation (Fig. 2b), such as changes in water surfaces, agricultural land, forests, built-up land, or wetlands. The notion of “direct” describes the logic in which causal effects were formulated in the case studies, and does not necessarily reflect the actual causal pathway. For instance, dam-induced increases in built-up or agricultural land are in most cases a consequence of resettlement schemes, and hence not directly caused by dam operation.

Fig. 2.

Chord diagrams visualizing a the effects of reservoir flooding on land systems, b dam-induced changes in the land system, c dam-induced changes in society or environment which further affect the land system, and d all effects between dams, society, environment, and land systems. Width of the connections represents frequency and arrow heads indicate the direction of the observed causal effects

Indirect effects on land systems accounted for 42% of the observations and can be further categorized according to their cause being a dam-induced environmental (12%) or societal change (12%; Fig. 2c), or a change in one of the land system categories (18%; Fig. 2d). Specifically, dam-induced changes in the environment most frequently led to alterations in forests, land use intensity, and agricultural land. Dam-induced societal change mainly affected land use intensity, built-up areas, and agricultural land extent. Dam-induced land system changes caused secondary effects in the land system, such as dam-induced changes in agricultural land extent which affected land use intensity. Land use intensity changes triggered additional repercussions, for instance if the introduction of irrigation allowed for the cultivation of more productive crop varieties. Altogether, we registered 24 unique types of indirect effects caused by previous dam-induced alterations in land systems.

Besides land system changes, the literature provided evidence of 154 causal effects in the non-land categories, 52% of which related to environmental and 44% to societal impacts, as well as 4%, in which dam operation itself was affected (Fig. 2c). Dams caused most of these social and environmental effects, which in turn led to secondary effects in the other categories, including effects of environmental changes on societies (10%), and societal changes affecting the environment (3%). Additionally, initial changes triggered repercussions in the same category, for instance where resettlement schemes lead to changes in income structure. These repercussions occurred similarly frequent in the environmental (10%) and societal change (7%) category (Fig. 2).

Directionality of land system changes

Land system changes operate in different directions, which unfold in increasing or decreasing extents of land use and land cover types, but also increases or decreases in land use intensity, or land demand (Fig. 3). We accounted for the differences in directionality by stratifying the 178 effects on individual land system categories by their direction (Fig. 4). Changes with unspecified or unclear directionalities (n = 17) were omitted from these analyses.

Fig. 3.

Counts of decreases (blue), and increases (red) of specific land system components

Fig. 4.

Counts of decreases (blue), and increases (red) of specific land system components, stratified by hydropower, irrigation, and multi-purpose functions

The most frequently reported land use and land cover changes were losses of forests and riparian vegetation (15%), almost exclusively due to flooding (10%), conversion to agriculture or built-up areas (3%), and hydrological changes (1%). In contrast, downstream alterations of the hydrological regimes led to woody vegetation succession (7%). Losses of agricultural land (11%) occurred due to flooding (9%) and conversion to built-up land (2%). Increases of agricultural land (8%) occurred due to the displacement of submerged agricultural land and agricultural expansion in previously uncultivated areas.

Management-related increases in land use intensity were observed (11%), but also decreasing land use intensity (6%) was reported. Losses of built-up area, including settlements and transportation infrastructure, were found (5%), alongside expansion of housing areas due to resettlement schemes (7%). Moreover, dam construction has led to grassland loss (6%), specifically due to reservoir flooding (4%) and land conversions (2%), and to decreasing wetland areas (4%). Changes in land demand were rarely assessed in our case study sample but in the few reported cases overall increasing demand dominated (2%) over decreases (1%).

A stratification by reservoir purpose (Fig. 4) revealed that hydropower and multi-purpose dams often negatively affected the extent of forests and agricultural land, while the majority of studies investigating irrigation dams reported increasing land use intensity. Differences across groups of commissioning year were less pronounced (Supplementary Material, Fig. S3), while stratification of storage capacity revealed a tendency for agricultural land loss and increasing land demand for reservoirs with storage capacities above 1000 Mm³ (Supplementary Material, Fig. S4). Land use and land cover changes occurred similarly over time after dam commissioning (Supplementary Material, Fig. S5). Contrastingly, land use intensity and land demand increased frequently within initial years after dam commissioning, while decreases dominated from 30 years or more after dam commissioning.

Processes of dam-induced agricultural change

Our synthesis revealed frequent interactions between dams and agricultural land extent and land use intensity. Dam-induced agricultural changes were characterized by differing directionalities and are highly interlinked with societal and environmental changes caused by dam operation, e.g., resettlement schemes or changing flood regimes. Disentangling the nature of these changes allowed us to identify recurring processes in the context of dam-induced agricultural change, which we classified into six process categories.

• Agricultural land loss (n = 20) was primarily related to submergence during reservoir flooding, affecting both rainfed (Zhao et al. 2013; Keken et al. 2015) and irrigated croplands (Shao et al. 2005; Delang et al. 2013). Moreover, resettlement caused the loss of agricultural lands due to conversions from agricultural to built-up land (Cao et al. 2011; Jafari and Hasheminasab 2017).

• In this context, agricultural abandonment (n = 2) occurred on cultivated floodplains following dam-induced alterations of hydrological regimes, which lead to reduced flooding and thus reduced suitability for cultivation (Thomas and Adams 1999) or as a consequence of resettlement schemes and subsequent population displacement following dam construction (Main 1990).

• Agricultural expansion (n = 14) was often reported in vicinity of newly constructed settlements, partly on less suitable lands, such as hillslopes (Shao et al. 2005; Wiejaczka et al. 2017). Additionally, agricultural land expanded in natural ecosystems or marginal lands (Thomas and Adams 1999; Loker 2003; Gordon and Meentemeyer 2006).

• Land use displacement (n = 6) occurred in cases when agricultural land loss was compensated for through agricultural expansion elsewhere. For example, the loss of agricultural land due to submergence was offset by expanding agricultural land around newly constructed settlements (Loker 2003; Yang et al. 2014). Partly, land use displacement occurred over long time frames, e.g., where the adoption of alternative management strategies resulted in the displacement of rainfed and irrigated croplands decades after the initial submergence of agricultural land (Thomas and Adams 1999).

• Land use intensification (n = 20) related to alterations of agricultural management were typically induced by the construction or improvement of irrigation systems (Ashraf et al. 2007; Strobl and Strobl 2011) and subsequent adoption of more productive crop varieties (Main 1990; Thomas and Adams 1999), increasing cropping frequency (Loker 2003; Ashraf et al. 2007; Fraiture et al. 2014), or higher livestock densities (Khlifi et al. 2010).

• Land use dis-intensification (n = 10) led to decreasing agricultural productivity, e.g., caused by reduced river flow (Blanc and Strobl 2014; Jafari and Hasheminasab 2017) or low agricultural suitability of the newly cultivated arable lands (Loker 2003; Shao et al. 2005).

Stratification by world regions

We identified Asia as a hotspot of research on dams and land change (Table 2). 14 of the 54 studies examined dams in China, of which ten were dedicated to the Three Gorges Dam, the world´s largest hydropower dam by installed capacity.

Table 2.

Number of studies, causal effects, and effects on land systems per world region

World Region	Studies	Causal effects	Effects on land systems
Africa	7	58	42
Asia	23	169	81
Australia	1	0	0
Europe	4	25	15
Latin America	11	59	32
North America	8	21	8

A stratification of all observations by world region revealed regional patterns of study foci (Fig. 5). The role of environmental changes was pronounced in study sites in North America, whereas studies on the African continent widely focused on societal changes and their interactions with agricultural land extent and land use intensity. Changing land demands were studied solely in Latin America and Africa. Changes in built-up area were in the spotlight in Europe and Asia.

Fig. 5.

Chord diagrams of causal effects stratified by world region

Regions with few studies of dam-induced land system change, such as Europe and North America, showed a smaller number of unique causal effects and thus affected fewer land system categories. This trend suggests that an increasing amount of research can help to unveil the complexity of dam-induced land system changes.

Data types, methods, and counterfactuals

Changes in the environment, society, and land systems were observed using diverse data types, study periods, and scales of analyses. Data were most frequently sourced using a combination of methods (n = 29), but some studies relied solely on remote sensing (n = 12), field measurements (n = 6), qualitative data (n = 4), or statistical and geospatial data (n = 3; Fig. 6). Study periods ranged from single points in time up to 113 years with an average of 27 years. The scales of analyses ranged from local (n = 12), to regional (n = 39), national (n = 1) and continental (n = 2).

Fig. 6.

Number of case studies across data and counterfactual types

Concerning counterfactual types, 25 studies used temporal comparisons, of which 20 compared phases of pre- and post-dam commissioning and two observed multiple periods in the post- commissioning phase (Fig. 6). Only two studies pursued spatial comparisons, by comparing sites with different levels of hydrological alteration and with increasing distance to the impoundment. Combined spatial and temporal counterfactuals were used in 19 studies, dominated by comparisons of dammed and undammed sites before and after commissioning. A total of eight studies did not use any counterfactual. Accordingly, most of the observed land system changes were found through regional scale analyses using temporal counterfactuals or combined spatial and temporal counterfactuals (also see Supplementary Material, Figs. S1 and S2).

Discussion

We synthesized the peer-reviewed literature focusing on dam-induced land system changes. Our study design allowed for incorporating quantitative and qualitative findings, which permitted systematic insights into the types, directions, and frequencies of specific change processes. The case studies included in our analysis suggest that dam construction and operation substantially affected the extent and intensity of agricultural land use as well as the extent of natural vegetation. Indirect effects shape a diverse array of land system change processes with societal and environmental consequences and thus play a significant role in the context of dam-induced land system change. The analysis allowed us to identify a set of commonalities and methodological limitations that help to guide future research activity concerning dam-induced land system changes.

Systemic complexity calls for additional research

Despite the global abundance of dams, scientific literature on dam-induced land system change is sparse. The majority of the cases we considered examined dams in Asia but the mere 24 studies that matched our criteria dwarf compared with the 33,000 large dams that are currently operational in Asia (ICOLD 2017). The low number of studies may in part be due to the fact that we constrained our analyses to studies published in English, listed in the ISI Web of Science, and to those with an explicit focus on land. We thereby omitted other dam-induced changes with relevance for the land system, as for instance shifting evapotranspiration rates (Jaramillo and Destouni 2015), changes in rural poverty (Duflo and Pande 2007), or food production (Orr et al. 2012). The low number of studies contained in our meta-study also reflects that research about the interactions of dams and land systems has only recently gained traction, particularly with the advent of dam critics following the release of the Report of the World Commission on Dams in the late 1990s. This was reflected by the lack of studies published before 1990.

Globally, the frequency of direct effects was outweighed by indirect effects, which relate dams and land system components through multiple processes, partly with strong environmental and societal implications. World-region stratification demonstrated that the amount of research activity and the associated diversity of disciplinary perspectives, study sites, methodologies, and time frames studied, co-determined the complexity of the observed system by raising the number of unique causal effects. Future research on dam-induced land system change is thus essential for improving our understanding of social and environmental costs and benefits of dams and reservoirs.

Potential controls for effect directionality

Dam-induced land system changes partly operated in contradicting directions. We repeatedly encountered two factors that potentially affected the direction of dam-induced effects on land systems.

First, we observed the time since dam commission as a controlling factor for some land system changes. Temporal variations in the strength and direction of land system changes partly led to the inversion of land change processes with increasing operational lifetime of the dam. For instance, forest cover losses were observed immediately after dam construction, while forest succession at later stages of the operational phase led to increasing forest cover (Wiejaczka et al. 2017). Similarly, dam construction had negative effects on agricultural production systems immediately after commissioning, whereas multi-decadal adaptation processes of the population resulted in improved agricultural productivity, even compared to pre-commissioning levels (Thomas and Adams 1999). These temporal non-linearities highlight the need for continuous monitoring of the land systems in post-commissioning phases to study the sequences of effects and feedbacks in a systematic manner.

Secondly, the spatial configuration of the study region appeared as a controlling factor for effect directions. For instance, the magnitude of riparian vegetation changes was found to be negatively related to the distance to the impoundment (Gordon and Meentemeyer 2006) or elevation above the reservoir water level (Kellogg and Zhou 2014). While some studies reported cropland productivity increases downstream of dams, with no changes in the vicinity of the reservoirs (Strobl and Strobl 2011), others challenge this finding (Fraiture et al. 2014). Other studies suggested dam-induced land use changes to occur mostly in upstream watersheds and predominantly within a distance of three kilometers to the dam (Zhao et al. 2013). Contrastingly, our findings reveal process-specific variations of distances, as for instance local changes in riparian vegetation communities (Azami et al. 2004) and tree mortality occurred hundreds of kilometers downstream of the impoundment (Assahira et al. 2017). Thus, variations of observed effects might occur in relation to the distance and the relative location to the dam (e.g., upstream or downstream). Unfortunately, a lack of detailed study site descriptions in relation to the investigated dam hampered an empirical analysis of these patterns. However, our insights into spatial and temporal controls are in line with earlier meta-synthesis efforts on the social impacts of dams (Kirchherr et al. 2016).

Methodical limitations and challenges

In line with other synthesis studies in land system science, the occurrence frequency of a studied process was seen as indicative of its existence and relevance in the system (Geist and Lambin 2001; van Vliet et al. 2015). We can thereby learn which processes are well documented and identify knowledge gaps about processes that warrant further research. Due to the diversity of data types and sources, as well as strengths and caveats of the methods employed, the level of confidence associated with each observation varies. Specifically, the attribution of indirect effects over long time frames remains a challenge, as multiple external factors might co-determine the process of interest.

The observations included here represent causal effects in that a change in factor X causes a change in outcome Y, while evidence for causal mechanisms, which explain specifically how the cause produces an effect (see Meyfroidt 2016 for terminology), was scarce. We thus suggested an array of agricultural change processes to provide a framework which allows for categorizing the processes underlying and following dam-induced agricultural changes.

Evaluating the robustness and suitability of the applied methodological toolkits is beyond the scope of this study. However, the majority of case studies included in this synthesis relied on counterfactual study designs. Typically, these included the use of time series to enable comparison of land systems before and after dam commissioning, mostly at the regional scale. While the dominance of regional analyses suggests that this spatial scale is appropriate for analyzing dam-induced land system change, it also points to a dearth of studies which examine dam-induced land system changes at smaller geographic scales.

The knowledge generated from synthesis efforts such as ours can fertilize theory development, yield policy recommendations, or serve for improving the calibration of process-based land change models (Magliocca et al. 2015). Compared to a global sample of dams, the dams included in our meta-study were relatively similar in terms of dam attributes and socio-economic and biophysical characteristics, but failed to capture the global variability of most attributes. Due to the relatively small number of studies, different methodologies, and scales of analysis included in the presented study, our findings are context-dependent and currently do not support the production of generalizable knowledge (Magliocca et al. 2018). We therefore call for reporting detailed descriptions of study site geographies (Margulies et al. 2016), as well as quantitative estimates of land system change in future studies, independent of the methods employed. This will foster the production of generalizable knowledge from case study synthesis. Further, current advances in statistical toolkits, such as counterfactuals, and increasing availability of geospatial data on dam locations (Lehner et al. 2011), forest loss (Hansen et al. 2013), or water surfaces (Pekel et al. 2016) bear potential for analyzing dam-induced land system change across larger populations of dams up to the global scale. This will grant novel opportunities for knowledge production using inductive approaches.

Conclusions

Dam-induced changes in agricultural systems, including changes in the extent of cultivated land and in land use intensity, are globally relevant phenomena that deserve further investigation on smaller geographic scales. We presented a descriptive representation of the current state of research about the effects of dam construction and operation on changes in land systems. To do so, we relied on observations from various scales, contextual settings, and disciplinary angles. Time-series analyses that trace land system changes in regular intervals on a regional scale have in the past shown potential to capture dam-induced land changes and the spatial and temporal variability therein. To enable and improve future synthesis efforts that quantitatively capture the processes in dam-induced land system change, we encourage researchers to report a detailed description of the study site location, period since dam commissioning, and, wherever possible, standardized quantities of change such as estimates of land cover change or changes in agricultural production.

Inferring the causal interactions between dams and land systems is methodically challenging and reliant on the abundance of suitable data and methodological toolkits. Considering the rapid growth of openly accessible data sets and the increasing connectedness between research disciplines, the emergence of innovative methods will hopefully guide novel research opportunities in the nexus of water, land, food, and energy, thus shedding new light on the effects of dams on land systems.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Biographies

Philippe Rufin

is a Ph.D. candidate at the Geography Department of Humboldt- Universität zu Berlin, Germany. His research interests include the development and application of remote sensing and modeling approaches for capturing and understanding man-made land system changes.

Florian Gollnow

is a Postdoctoral Fellow at the National Socio-Environmental Synthesis Center at Maryland University. His research focuses on understanding the impact of policies for nature conservation on dynamics of land use changes and ecosystem preservation.

Daniel Müller

is a senior research associate at the Leibniz Institute of Agricultural Development in Transition Economies (IAMO). His research aim is to advance the understanding of land system changes, including their multiple repercussions on human welfare, food production, greenhouse gas emissions, and biodiversity.

Patrick Hostert

is the head of the Geomatics Lab at the Geography Department of Humboldt- Universität zu Berlin, Germany. His research focuses on land system science and remote sensing.