Ecological responses to altered flow regimes: a literature review to inform the science and management of environmental flows

Introduction

The importance of a river’s flow regime for sustaining biodiversity and ecological integrity is well established (Poff et al., 1997; Hart & Finelli, 1999; Bunn & Arthington, 2002). Streamflow is viewed as a ‘maestro’ (Walker, Sheldon & Puckridge, 1995) or ‘master variable’ (Power et al., 1995) that shapes many fundamental ecological characteristics of riverine ecosystems. From a basic ecological perspective, extreme events such as high flows and low flows exert selective pressure on populations to dictate the relative success of different species, and patterns of variation in ‘sub-lethal’ flows can influence the relative success of different species and regulate ecosystem process rates (Resh et al., 1988; Hart & Finelli, 1999). The range and variation of flows over recent historical time, referred to as the natural flow regime (Richter et al., 1996; Poff et al., 1997), sets a template for contemporary ecological processes (Resh et al., 1988; Doyle et al., 2005), evolutionary adaptations (Lytle & Poff, 2004) and native biodiversity maintenance (Bunn & Arthington, 2002).

Flow regime varies geographically in response to climate (precipitation and temperature) and catchment controls on runoff (topography, geology, land cover, position in network). All flow regimes may be considered unique in some measure of their details; however, many authors have stressed the need to characterise the similarity among flow regimes to provide typologies that can support a priori predictions (e.g. ecological and evolutionary convergence under geographically disjunct selection regimes) and development of general principles for flow regime management (Arthington et al., 2006; see Poff et al., 2006 for a brief review).

Numerous case studies (and expert knowledge) provide the foundation of our scientific understanding that many types of flow alteration (e.g. magnitude, frequency, and timing) induce a variety of ecological responses (Bunn & Arthington, 2002). This, combined with the recognition that ubiquitous flow alteration threatens the biodiversity and ecosystem functions of rivers on a global scale (Postel & Richter, 2003; Nilsson et al., 2005;Dudgeon et al., 2006; Poff et al., 2007) has led to an accelerating interest in developing a general, quantitative understanding of aquatic ecosystem response to various types and degrees of flow alteration. Such understanding is needed to help support scientifically defensible guidelines for developing flow standards that could be applied to all streams and rivers, even those lacking baseline data (Arthington et al., 2006; see Poff et al., 2010).

A comprehensive synthesis of case studies has the potential to reveal generalised, possibly quantitative, relationships between ecological responses and specific types of flow alteration. Such a synthesis can advance general understanding in the science of environmental flows (Poff et al., 2003) and can help set priorities in the next phases of hydro-ecological research to support development and implementation of regional environmental flow standards (Poff et al., 2010). Reviews to date have been incomplete or inconclusive. For example, Poff et al. (1997) and Bunn & Arthington (2002) selectively reviewed the literature and clearly illustrated general ecological principles of flow regime alteration. In the most complete effort to date, Lloyd et al. (2003) examined 70 studies for relationships between hydrologic change and ecological or geomorphological change and reported that 87% of the studies documented changes in either or both of these variables in response to reduced flow volumes. However, a more focused quantitative analysis of only 14 studies found no evidence to reject the null hypothesis that ecological change was independent of hydrological change, nor was there any simple linear or threshold relationship between the size of ecological change and the size of the hydrological change.

Lloyd et al. (2003) identified several constraints that limit attempts to derive quantitative relationships from a literature review, including lack of control or reference sites for ‘unaltered’ conditions, other environmental changes (e.g. sediment flux, temperature) occurring at the impacted sites and inadequate historical information. Given that alteration of flow regimes is typically confounded with other environmental factors, we would not necessarily expect unambiguous relationships between single measures of flow alteration and ecological response (see Konrad, Brasher & May, 2008; Poff et al., 2010). However, any general relationships that did emerge would presumably indicate strong responses to flow alteration and thus be of both scientific and management interest. Lloyd et al. (2003) also suggested that a larger dataset that spanned a broader range of response types and hydrologic alterations might be able to detect more robust relationships.

With these considerations in mind, we had three major goals for the present paper. First, we aimed to conduct a more comprehensive review of the literature on ecological responses to alteration of natural flow regimes, focusing primarily on publications in the last 10 years, i.e. since the publication of Poff et al. (1997) who reviewed much of the older literature. Second, where possible, we wanted to assemble quantitative relationships published in these disparate studies to determine if any statistically supported patterns could be extracted between defined types of flow alteration and ecological response metrics. In particular we were interested in detecting any potential ‘threshold’ relationships that would be especially useful in a management context (see Arthington et al., 2006). Third, we wished to identify ‘gaps’ in information presented in the published papers that represent research needs and priorities to advance the development of robust flow alteration–ecological response relationships that can support development of science-based, regional flow guidelines, as proposed in the new framework called ELOHA – Ecological Limits of Hydrological Alteration (Poff et al., 2010).

Methods

Literature review and qualitative relationships

We extensively reviewed the literature and compiled a list of 165 papers that reported on either aquatic (in-channel) or riparian responses to flow regime alteration. Most of the studies were conducted in North America (89), with good representation from Europe (44) and Oceania (25), but with only a few from South America (4) and Africa (3). We logged responses as ecosystem, community, or population-level responses. Most papers (145) reported responses in terms of population or community change, and we partitioned them into six taxonomic categories: macroinvertebrates, fishes, riparian vegetation, aquatic primary producers, birds, and amphibians (Fig. 1).

Flow alteration often involves simultaneous modification of several components of the natural flow regime (magnitude, frequency, duration, timing and rate-of-change; Poff et al., 1997). Many papers reported on more than one altered flow component; however, for the purposes of our analysis we focused on the primary flow component reported for each study, i.e. the one that was given the most emphasis by the authors, either in terms of alteration or study design. Most papers (99) reported flow alteration primarily in terms of some metric of magnitude. Changes in duration (25), timing (16), frequency (16) and rate of change (5) were also reported, and four papers did not specify a flow component. Of the total number of studies that we reviewed, 70% focused only on flow modification. A subset of the 165 papers also mentioned alterations in other environmental drivers, including sediment (14%), temperature (11%), or sediment plus temperature (5%).

Most of the flow alterations reported in the reviewed studies were primarily the result of dams and impoundments (88 papers). Other flow alterations resulted from water diversions (17), groundwater abstraction (6) and levees (7). A few studies reported flow alterations from weirs, road construction or channelisation, but many did not report a source (32) or reported unspecified multiple sources of factors affecting the flow regime.

For all 165 papers we summarised the reported results ‘qualitatively’, i.e. we categorised the reported change in ecological metrics that were associated with the different kinds of flow alteration. This general analysis was conducted to provide a broad synopsis of the reviewed literature.

Results

Quantitative analysis

Quantitative estimates of macroinvertebrate response to flow magnitude alteration were available for 25 data points. Macroinvertebrate abundance and diversity both generally declined in response to alteration in flow magnitude, whether an increase or a decline (Fig. 2). Most measurements of flow magnitude represented large changes of near −100% or +100%, with few intermediate values. This lack of data points throughout the entire gradient of alteration makes it difficult to detect any threshold or nonlinear responses to changes in flow magnitude. Flow alteration was measured as either a change in total discharge or a change in base flow (i.e. low flow or drought conditions). There was no consistent difference in direction or magnitude of macroinvertebrate response in terms of source of flow magnitude alteration.

Fishes showed consistent negative responses to alteration in flow magnitude, whether measured by changes in abundance, population demographic parameters or diversity of assemblages (Fig. 3). As with macroinvertebrates, response of fishes tended to be measured in relation to higher values of flow alteration (−50% to −100% and +75% to +100%). The lack of data points in the more moderate ranges of flow alteration precluded estimation of any potential threshold response and limited inference on lower levels of alteration that may not have negative impacts on fish species and assemblages. The two specific types of flow alteration reported for these papers were changes in average discharge and short-term variation, for which both increases and declines in flow magnitude were reported.

Riparian responses did not demonstrate any consistent trends with respect to alteration in flow magnitude (Fig. 4). The studies we reviewed examined reductions in peak discharge (15 papers) and average flow (five papers); no increases in discharge magnitudes were reported. All of the riparian studies recorded changes in peak flows, so riparian responses in Fig. 4 can be associated with decreases in flood peaks, leading to reduction or elimination of overbank flooding. Two studies reported demographic responses that were both negative, but the small sample size precludes any general inference. Five of the studies reported increased values of riparian response metrics to flow reductions. These were mostly associated with riparian encroachment into the channel, increased species via terrestrialisation of the riparian community and an increase in herbaceous cover in the riparian understory.

Discussion

Our goal was to extract information from published papers and search for any consistent patterns in ecological response to flow regime change. We hoped that a careful examination of the literature would reveal quantitative relationships that could advance our general understanding of flow alteration–ecological response relationships and thus inform environmental flow management.

As expected, our qualitative analyses clearly demonstrated the many ecological consequences of flow alteration. Of the 165 papers reviewed, 152 (92%) of them reported negative ecological changes in response to a variety of types of flow alteration. Only 21 papers (13%) reported an increased value for ecological response metrics, and these often reflected shifts in ecological organisation such as increase in non-native species or non-woody plant cover on dewatered floodplains. Thus our findings broadly corroborate earlier summaries of ecological responses to flow regime alterations (Poff et al., 1997; Bunn & Arthington, 2002), and particularly those of Lloyd et al. (2003), who found that 56 of 65 (86%) studies examining ecological responses to flow modification documented ecological changes. Moreover, our findings suggest that larger changes in flow alteration are associated with greater risk of ecological change from pre-management conditions (2-4). Interestingly, no studies reported on primarily ecosystem functional responses (e.g. riparian production, nutrient retention), even though many ecological processes are clearly flow-dependent (Hart & Finelli, 1999; Doyle et al., 2005). This absence points to an obvious research gap in the environmental flows literature, one that perhaps reflects the often short-term, ‘snap-shot’ nature of biological sampling done to document ecological change in flow-altered systems.

Most of the studies included in our review (60%) examined flow alteration in terms of changes to some measure of flow magnitude (99 of 165 papers). This bias toward alteration in flow magnitude is similar to previous surveys. Lloyd et al. (2003) found 45 of 70 publications (64%) to report flow alteration in terms of abstraction or water withdrawals for irrigation. In the present paper we were able to extend the analysis of flow alteration to other important components of the natural flow regime, including frequency (16 papers), duration (25), timing (16) and rate-of-change (5). Previous review papers have not thoroughly documented the ecological responses to altered flow frequency and timing of peak flows, and the information in Table 1 (see also the online appendix) advances our general knowledge base for these kinds of flow alterations.

We were able to extract quantitative relationships between flow alteration and ecological responses of macroinvertebrates, fishes, and riparian vegetation from 55 papers. This represents a substantial improvement in sample size compared to Lloyd et al.’s (2003) effort, which yielded only 14 papers suitable for quantitative analysis. However, similar to their effort, we found almost 75% of the ecological data available to be reported as responses to alteration in some measure of flow magnitude. We were not able to extract any robust statistical relationships between the size of the flow alteration and the ecological responses, which varied among the different taxonomic groups. This probably reflects, at least in part, the paucity of data points in the low to middle range of flow alteration (0–50%), particularly for macroinvertebrates and fish.

Fish were the only taxonomic group to consistently respond negatively to changes in flow magnitude irrespective of whether the flows increased or decreased. Under reduced flows all 10 fish responses were negative and eight of these exceeded 50% (Fig. 3). Diversity showed a consistently large decline, especially where flow magnitudes exceeded 50% change. Together, these observations suggest that fish are sensitive indicators of flow alteration, at least as measured and reported by these studies. In contrast, no clear patterns emerged for responses of macroinvertebrates or riparian species to changes in flow magnitude. For riparian species, responses were recorded only for decreases in magnitude (measured as decreases in peak flows) and responses showed both increases and decreases; however, the increased riparian responses mostly reflected more non-woody vegetation cover on the floodplain or an increase in upland species.

One potential reason for the different responses among taxonomic groups may be the types of taxa that were examined and differences in species-specific responses to alterations in flow magnitude. Consistent patterns in responses of fish to altered flow magnitude may arise because the studies we reviewed primarily examined responses of native, sensitive and/or specialist taxa. Studies of other taxonomic groups, such as macroinvertebrates and riparian vegetation, included taxa that may be hypothesised to have conflicting responses to changes in flows, such as non-native, tolerant or generalist taxa, in addition to native, sensitive, and specialist organisms. We were not able to refine our analysis based on such differences among taxonomic groups or other fine details of species-specific responses. For example, riparian responses to reductions in peak flows may depend on organism physiology for tolerance to reduced soil moisture or to changes in soil chemistry (see Merritt et al., 2010). Knowledge of the autecology of the particular species reported in the studies could have allowed us to examine whether ecological responses were dominated by particular types of species, such as tolerant, specialist or non-native species. For example, it is well documented that riparian cottonwood tress (Populus species) suffer greatly diminished success following peak flow reductions compared to invasive saltcedar (Tamarix) in flow-altered rivers of the American West (e.g. Stromberg et al., 2007; Merritt & Poff, in press), a result not captured in our global-scale statistical analysis.

In general, while our analysis does provide some insight into the possible relative sensitivity of different ecological endpoints to alteration in flow magnitude, it does not yield unambiguous, transferable empirical relationships for developing quantitative guidelines to support regional environmental flow standards. This finding is similar to that reported by Lloyd et al. (2003), who with a smaller dataset found little quantitative support for either simple linear relationships or threshold relationships between ecological responses and flow magnitude alteration. Several possible reasons can be invoked to explain this.

First, as with all surveys of a broad literature, the individual studies composing the analysis were not designed specifically to address our hypothesis, i.e. that degree of ecological alteration is associated with degree of hydrological alteration. Despite the fact that we assembled a reasonably large set of papers in this review, a wide variety of ecological metrics and types of flow alteration was reported. Measures of flow modification were not consistently reported, and criteria for establishing reference conditions were variable. For example, some studies calculated a site-specific percent change in flow relative to a historical period, whereas others reported flow alteration relative to a spatial reference (such as a site upstream of a dam).

Another constraint in our analysis was the inability to place flow alteration in each study into a more specific environmental context. Aquatic and riparian species will respond to multiple hydrologic drivers, and these are often confounded. For example, a change in magnitude of high flows is often accompanied by a change in frequency, and either or both of these may be influencing biological response. We were unable to evaluate such multiple drivers in any consistent manner across all studies; indeed, these factors were generally unreported in the original publications. Likewise, other environmental characteristics, such as temperature regime or sediment-habitat factors (including hydraulic structure and dynamics) were not reported and could not be assessed in our retrospective analysis. Flow alteration is only one environmental factor to which riverine species respond (Poff et al., 1997; Bunn & Arthington, 2002) and incorporation of other environmental factors is important to advancing environmental flows science (Lloyd et al., 2003; Konrad et al., 2008; Olden & Naiman, 2010; Poff et al., 2009; Stewart-Koster et al., 2009), even when those factors cannot be rigorously accounted for statistically. For example, Konrad et al. (2008) used quantile regression to develop statistical relationships between macroinvertebrate metrics and flow characteristics across more than 100 sites that differed in terms of non-flow environmental features. Unfortunately, the shortage of data points in the low alteration range in our dataset prevented us from effectively using quantile regression to develop meaningful relationships between flow alteration and ecological responses.

A potentially significant limitation in our analysis was the inability to account for hydroclimatic and other regional differences among the study sites. Given the relatively small sample size, we did not stratify the studies according to natural flow regime type or geomorphic setting, although these factors are expected to influence ecological responses to hydrologic alteration (Poff & Ward, 1989; Arthington et al., 2006) and potentially offer a basis for transferring flow alteration–ecological response relationships geographically (see Poff et al., 2010 for an extended discussion). Thus, we would argue that our inability to find statistical relationships in a limited set of studies at the global scale does not necessarily imply such relationships would not be found by reviewing studies conducted over a smaller geographic extent or within particular stream types where environmental conditions (e.g. temperature, geomorphology) are more similar.

One conclusion from our analysis is that to develop quantitative relationships useful for supporting regional environmental flow efforts like ELOHA (Poff et al., 2010), synthesising existing literature at the global scale will not be sufficient. Indeed, several data gaps have been clarified by constraints inherent in our analysis. These include design of the original studies, variation in the reporting of mode and magnitude of flow alteration, inability to examine species-specific responses, lack of data points for the entire range of flow alteration and inability to distinguish among types of alteration within a flow component. These limitations point to the need for prescribed monitoring programs (e.g. following water development projects) in which data could be collected in a before-after-control-impact design (Underwood, 1994) to support statistical inference of ecological responses to flow alteration. Likewise, incorporating knowledge from ecosystem-scale experiments and monitoring associated with environmental flow releases could advance our understanding of ecological responses to flow alteration. Such experimental flow releases offer exceptional opportunities to learn about the interactions of flow alteration and other environmental drivers (King et al., 2010; Shafroth et al., 2010).

However, new efforts and approaches are also needed to gather information on flow alteration–ecological response relationships. One promising approach would be to design sampling programs that target existing flow-altered sites across gradients of flow alteration to allow general relationships between flow alteration and ecological response to be inferred (Arthington et al., 2006; Poff et al., 2010). Often, biological sampling is concentrated in smaller streams and areas that represent reference or good biological condition, and fewer sampling sites are available across a range of stream sizes and condition types. Similarly, monitoring of sites that have experienced flow alteration is generally not done and so learning opportunities are missed (Souchon et al., 2008; Webb et al., 2010). New data collection efforts directed at sampling a range of undisturbed to highly flow-altered sites would be useful in testing specific hypotheses that could inform environmental flow science and management. In addition, existing data could also be marshaled to test such hypotheses. For example, large biological datasets exist in several countries (and state or provincial jurisdictions within countries), including the United States Environmental Protection Agency’s Environmental Monitoring and Assessment Program (EMAP), United States Geological Survey’s National Water Quality Assessment (NAWQA), Europe’s Euro-limpacs program, Europe’s River Invertebrate Prediction and Classification System (RIVPACS), and the Australian River Assessment System (AusRivAS). These large databases, if analysed with an eye toward degree of flow alteration, carefully selected ecological response metrics, stream typology, and multiple environmental drivers, hold the potential to reveal important relationships and key information gaps needed to guide research and application in environmental flows science.