The importance of seasonal flow timing for riparian vegetation dynamics: a systematic review using causal criteria analysis


Joe Greet, Department of Resource Management and Geography, University of Melbourne, Burnley Campus, 500 Yarra Boulevard, Richmond, VIC 3121, Australia. E-mail:


1. Whilst it is widely recognised that a natural flow regime is important for sustaining riverine ecosystems, the relative importance of the various components of flow regime for riparian vegetation dynamics is poorly understood. We sought to determine the current extent of knowledge on the importance of seasonal flow timing for riparian plants by conducting a systematic review of the literature using causal criteria analysis.

2. Using a definition of ‘riparian’ that included riverine, wetland and floodplain systems, we found sufficient evidence to provide strong support for the existence of causal relations between seasonal flow timing and a number of riparian plant processes, namely rates of waterborne dispersal (hydrochory), germination and growth, as well as riparian community composition. There was insufficient evidence to infer a causal relationship between flow timing and the reproduction or survival of riparian plants.

3. Thus, we argue that seasonal flow timing is important for many of the processes that generate and sustain riparian vegetation communities. River regulation, and/or flow management aimed at restoring ecological values, should consider flow timing and its implication for riparian flora. Because of regulation, many of the rivers of south-eastern Australia have inverted seasonal flow patterns. Whilst direct evidence of the effects of this inversion on the flora of these rivers is lacking, the results of our causal analysis allow us to predict how these plant communities may have been affected.

4. However, these predictions must be treated with caution because of the reliance of some of the causal analyses on wetland studies. For riverine flora, further research is particularly needed on the effects of seasonal flow timing on hydrochory, survival and reproduction.

5. Causal criteria analysis provides a defensible and efficient means for assessing the extent of evidence for or against ecological hypotheses of this kind. In this case, systematic review of the literature provided strong evidence to support a number of causal links between seasonal flow timing and riparian vegetation dynamics, whilst also efficiently identifying knowledge gaps.


Riparian vegetation is a key component of riverine ecosystems, fulfilling many important ecological functions including the provision of food and habitat, the moderation of stream water temperatures, the filtering and cycling of sediments and nutrients, and the stabilisation of stream banks (Richardson et al., 2007). A natural flow regime is considered of central importance in sustaining the ecological integrity of riverine ecosystems (Poff et al., 1997), and it is generally accepted that the alteration of natural flow regimes through river regulation will affect the processes that sustain riparian vegetation communities (Bunn & Arthington, 2002; Nilsson & Svedmark, 2002). However, the relative importance of the various components of the flow regime – magnitude, frequency, duration, timing and rate of change – for riparian vegetation dynamics is poorly understood.

For riparian vegetation, the importance of seasonal timing is particularly poorly researched (Roberts, 2002; Poff & Zimmerman, 2010). Moreover, much of the existing research focuses on wetland flora, with far fewer studies of riverine plants. With riparian management being one of the most common forms of river restoration (Palmer et al., 2005; Brooks & Lake, 2007), greater knowledge of the importance of seasonal flow patterns for riparian vegetation dynamics is required. The focus of this review is to assess the current extent of this knowledge.

An understanding of the importance of streamflow timing is particularly relevant for lowland rivers in south-eastern Australia. The regulation of many of these rivers has inverted their seasonal flow patterns, such that high flows in winter and spring have been replaced by high flows in summer and autumn to provide water for irrigation (Walker, 1985; Maheshwari, Walker & McMahon, 1995; Reid & Quinn, 2004). To date, studies of the effects of this regulation on plants (e.g. Walker et al., 1994; Blanch, Ganf & Walker, 1999; Blanch, Walker & Ganf, 2000) have not specifically focussed on flow timing. Although many studies have been conducted in Australia on the effect of seasonal timing on wetland vegetation (e.g. Britton & Brock, 1994; Nielsen & Chick, 1997; Robertson, Bacon & Heagney, 2001; Warwick & Brock, 2003), studies on the importance of flow timing for riverine plants are sparse (Roberts, 2002). Accordingly, we attempt to use this review to make predictions concerning the probable effects of seasonal flow inversion on the flora of south-eastern Australian rivers.

Systematic review using causal criteria analysis

Previous reviews on the importance of flow regime for riparian vegetation dynamics (Bunn & Arthington, 2002; Nilsson & Svedmark, 2002; Poff & Zimmerman, 2010) have been largely descriptive and have not assessed the extent and quality of evidence for causal relationships. In this review, we adopt the causal criteria analysis method as described by Norris et al. (2008) to provide a framework for the systematic review of evidence from the literature.

Causal criteria analysis was originally developed in epidemiology, where a lack of experimental data can result in a weak ability to infer causal relationships (Hill, 1965; Susser, 1991). It has since become a well-established method within epidemiology (Weed, 1997), with a number of analyses appearing in the health sciences literature every year (e.g. McLaren et al., 2010; Roffey et al., 2010). The method tests pieces of evidence against a series of criteria and can be used to build an argument for causality through the collective strength of a number of pieces of weak evidence. Causal criteria analysis is different from but complementary to meta-analysis – an analytic tool with which ecologists are likely to be more familiar. Meta-analysis is concerned with synthesising studies to determine an overall effect size associated with some ‘treatment’ (Sutton & Higgins, 2008). However, like any statistical analysis of observational data, meta-analyses should not be used to make claims about causality (Greenland, 1998). In contrast, causal analysis provides a consistent, transparent and logical method to identify likely causes (Suter, Norton & Cormier, 2010). The rigourous statistical approaches of meta-analysis can provide valuable inputs to a causal analysis by better quantifying some of the individual criteria than is commonly the case (David, 1999). Indeed, recent epidemiological studies have explicitly combined the two (Malin, Morris & Khan, 2010).

Like epidemiology, ecological studies often suffer from improper randomisation and replication of treatments, which weakens their ability to draw strong inferences (Downes et al., 2002). Over the last decade, there has been a growing interest in applying the causal criteria to ecological questions (Beyers, 1998; Fabricius & De’Ath, 2004; Adams, 2005; Plowright et al., 2008; Suter et al., 2010). Using causal analysis, we attempt to answer the question: what is the importance of seasonal timing for riparian vegetation dynamics?


The causal criteria analysis method described by Norris et al. (2008) applies an eight-step framework to evaluate the strength of evidence for one or a number of putative cause–effect relationships. The eight steps are an iterative process that can be roughly divided into three sections. We provide a brief outline of the actions required for each step, as well as the outcomes of those actions for the Problem formulation and Literature review sections below. A full description of the method is provided in Norris et al. (2008). The outcomes of the final section, Weighting evidence and judging causation, which constitute the main findings of the systematic review, are detailed in the Results.

Problem formulation

The initial steps of the framework (Steps 1–4, 6) are concerned with developing the questions that will be assessed by the analysis. These steps require the framing of a broad question (Step 1) and the placement of that question in a geographical context (Step 2). The development of a conceptual model of the potential cause–effect linkages (Step 3), and the documentation of relevant potential quantifiable causes and effects (Step 4), implicitly consider the causal criteria of ‘plausibility’ and ‘coherence’ (Hill, 1965) in that the questions being asked by the analysis need to be consistent with known principles (Suter et al., 2010). Following a review of the literature (Step 5, see below), the questions, hypotheses and conceptual model may be refined in an iterative process (Step 6).

Steps 1–2.  Initially, our concerns related to the flow regulation of many rivers in south-eastern Australia and the possible impacts of this regulation on riparian vegetation. The natural flow pattern of these temperate rivers is largely seasonal, with high flows in winter/spring and low flows in summer/autumn. Regulation to provide water for agricultural needs has resulted in a reversal, or inversion, of these seasonal flow patterns downstream of the impoundments. High flows are now experienced in summer/autumn and low flows in winter/spring. Nonetheless, for our analysis, we sought to develop general principles regarding the relationship between flow timing and vegetation dynamics, as these provide for predictions of vegetation responses to changing flow conditions that can be broadly applicable (Merritt et al., 2010b). Hence, the broad question addressed by this analysis was: What impact does the seasonal inversion of flow patterns have on riparian vegetation dynamics?

Steps 3–4.  The proposed conceptual model is based on the principle that aquatic species have evolved life-history strategies primarily in direct response to natural flow regimes (Bunn & Arthington, 2002). Many of the life-history strategies of riparian plants, such as dispersal by water (hydrochory), germination, growth, survival and reproduction, are considered to be adapted to seasonal flow patterns (Naiman & Decamps, 1997; Nilsson & Svedmark, 2002). Thus, we hypothesised that rates of, and opportunities for, recruitment of riparian plants are reduced by seasonal flow inversion (Fig. 1).

Figure 1.

 A conceptual model of the relationship between the life-history stages of riparian plants and natural and inverted seasonal flow regimes. For the natural seasonal flow pattern (solid line), the following generalised pattern is indicated: the seed release period of many riparian plants coincides with peak flow; after waterborne dispersal (hydrochory), these seeds are deposited during receding flows in conditions appropriate for germination; these germinants are able to grow and later reproduce during the growing season (spring – early summer). Under an inverted seasonal flow pattern (dashed line), the synchrony of these relationships is lost. For example, riparian plants could find themselves inundated during the growing season.

For the causal analysis, seasonal flow inversion (or seasonal timing) is defined as the cause. The quantifiable effects relate to the life-history stages of riparian plants shown in the model: hydrochory, germination, growth, reproduction and survival, as well as the composition of the riparian vegetation community itself. The initial hypotheses were that seasonal flow inversion would result in changes to rates of all these processes and to the composition of the riparian vegetation community. The causal linkages tested by the analysis are illustrated in Fig. 2.

Figure 2.

 Putative causal relationships between seasonal flow timing and riparian vegetation dynamics to be tested by the causal analysis.

Step 6.  A literature search failed to reveal any studies that directly assessed the effect of a change in seasonal flow timing on rates of hydrochory. Therefore, this question was instead posed as two separate questions. First, do rates of hydrochory vary seasonally (where the cause is season and the effect is rates of hydrochory)? The implication here is that if rates of hydrochory are seasonal and seasonal flow patterns are altered, then rates of hydrochory will be affected. The second question relates to the importance of hydrochory for recruitment: Do rates of hydrochory (cause) affect community composition (effect)? The effect of hydrochory on community composition was also considered because the link between hydrochory and recruitment was contentious within the literature identified.

Following these steps, the specific hypotheses to be addressed became:

With regard to hydrochory:

  • H1. Rates of hydrochory vary with season
  • H2. Rates of hydrochory affect community composition, in terms of:
    •   a. species richness
    •   b. species composition

With regard to recruitment processes:

  • H3. Seasonal timing affects germination, in terms of:
    •   a. species richness
    •   b. species composition
  • H4. Seasonal timing affects rates of growth
  • H5. Seasonal timing affects rates of reproduction
  • H6. Seasonal timing affects rates of survival

Whilst overall,

  • H7. Seasonal timing affects community composition, in terms of:
    •   a. species richness
    •   b. species composition

Literature review

Step 5.  The causal criteria analysis framework (Norris et al., 2008) requires that a systematic and documented method for retrieving literature be employed to reduce any subjectivity and bias of the reviewer. The reviewer records their search strategy and also justifies how the relevance or non-relevance of studies was determined (for details of our search strategy see online Data S1). For relevant studies, the evidence is extracted, so that it may be evaluated and weighted. Evidence extraction involves recording information on whether the study findings support the hypothesis, the type of experimental or survey design used and the number of replicates. This information is used to weigh the evidence in Step 7.

Any study investigating the effect of seasonal timing on flora within a fluvial context, regardless of climate, geomorphology, etc., was considered. Furthermore, rather than restrict the analysis to the few studies of riverine flora, we included studies of ‘riparian’ flora; studies of riverine, wetland and floodplain systems were all considered. Given the similarities of these habitats and their plant assemblages, we considered it likely that the floras of these environments would behave similarly.

Weighting evidence and judging causation

Step 7.  For each relevant study, the quality of the evidence is evaluated in terms of these three study quality attributes:

  • (i) study design type
  • (ii) number of independent sampling units used as controls
  • (iii) number of (potentially) impacted independent sampling units

Studies in which the error terms are well controlled (e.g. BACI designs) attract greater weighting than less rigourously controlled designs (e.g. only impact locations sampled). Studies with more than one impact location also have more weight, as a larger number of impact locations leads to a better estimate of the range of dynamics experience by impact locations and reduces the likelihood of observing a spurious result (Norris et al., 2008). Control locations are important for improving inferential power (Downes et al., 2002), and increasing numbers of these also contribute to the weight of a given study (Table 1).

Table 1.   Weights applied to study design types, number of independent control sites and number of independent impact sites to calculate an overall study weight for each relevant study (Norris et al., 2008)
Study attributeWeight
Study design type
 After impact only1
 Reference/control vs impact no before; before vs after no reference/control2
 Gradient response model3
Number of independent control sites
Number of independent impact sites

For each piece of evidence identified from a study, the weights for each of the three attributes are summed to give an overall study weight. For example, for a Before vs. After study with one control and one impact location, the overall study weight would be 2 (design) + 2 (control site) + 0 (no additional impact sites) = 4. The weights (and the threshold discussed below) reflect previously elicited expert opinions about the number of consistent results from high- and/or low-quality studies that one needs to see before being confident of a causal link (R. H. Norris, pers. comm.). These default weightings can be adjusted to suit the particular circumstances of a review, but the user should document any changes and justify them.

Step 8.  Once the evidence has been assembled and weighted, it is combined to assess the level of support for a causal relationship. The Norris et al. (2008) causal analysis method relies principally on the causal criterion ‘consistency of association’– the repeated observation of an association between putative cause and effect under different conditions and assessed using different methods (Hill, 1965). For this purpose, the sum of study weights for evidence in favour of the hypothesis is compared against the sum of weights against the hypothesis. A default threshold of 20 summed study weight points defines the point at which sufficient evidence for (or against) the hypothesis exists to give strong support for (or against) the existence of a causal relationship. As with the study weights themselves, these thresholds can be varied by the user, but such changes should be documented and justified.

A collection of pieces of evidence in favour of the hypothesis with a total sum of study weights of 20 or more is defined as providing strong support for an association between the cause and effect. Against this, a combined study weight of 20 or more for studies that do not show support for the hypothesis indicates a lack of consistency for the cause and effect relationship. A number of combinations of evidence for and against the hypothesis are therefore possible (Table 2). The Norris et al. (2008) algorithm also considers dose–response relationships for strengthening arguments for causality, but we did not consider these, as the main cause in our review (seasonal flow inversion) is categorical.

Table 2.   Possible causal criteria analysis outcomes and their interpretation. Adapted from Norris et al. (2008)
Evidence in favour of hypothesisEvidence not in favour of hypothesisConclusion
≥20<20Support for hypothesis
<20≥20Support for alternate hypothesis
<20<20Insufficient evidence
≥20≥20Inconsistent evidence

The interpretation of ‘support for hypothesis’ is clear. ‘Support for alternate hypothesis’ may include the null hypothesis of no effect or may indicate a directional effect opposite to that expected. In this study, none of the hypothesis tests are explicitly directional, and so we use the familiar term ‘null hypothesis’. ‘Insufficient evidence’ may indicate a real gap in the literature or may be able to be remedied by a deeper literature review. A finding of ‘inconsistent evidence’ suggests a logical inconsistency at some point in the process, and Norris et al. (2008) suggests returning to the conceptual model to attempt to determine why inconsistent results are being found.


A total of 35 relevant studies were identified, from which 65 pieces of evidence were extracted. The individual study weights are provided in online Data S2. Table 3 lists these studies, organised by the respective hypotheses for which they provided evidence. Some studies provided evidence for more than one hypothesis and/or causal linkage. The summed weights of the studies in support of each hypothesis, and those not in support, are provided in Table 4 along with the outcome of the causal analysis for each hypothesis.

Table 3.   The relevant studies identified, listed under the respective causal linkages for which they provided evidence
Putative causal linkage (cause – effect)Support for hypothesisSupport for null hypothesis
  1. S, species richness; C, species composition; R, study of riverine flora; W, study of wetland flora; F, study of floodplain flora.

Season – hydrochoryAndersson & Nilsson (2002)R, Boedeltje et al. (2004)R, Gordon & van der Valk (2003)W, Gurnell et al. (2006)R, Gurnell et al. (2008)R, Middleton (2000)W, Moggridge et al. (2009)R, Moggridge & Gurnell (2009)R, Riis (2008)R, Schneider & Sharitz (1988)W, Skoglund (1990)R, Tabacchi et al. (2005)RRiis (2008)R
Hydrochory – community composition (S)Andersson et al. (2000b)R, Gurnell et al. (2006)R, Jansson et al. (2005)R, Leyer (2006)R, Merritt et al. (2010a)RBissels et al. (2004)F, Gerard et al. (2008)F, Holzel & Otte (2001)F, Merritt et al. (2010a)R, Riis (2008)R, Rosenthal (2006)F
Hydrochory – community composition (C)Andersson et al. (2000b)R, Gerard et al. (2008)F, Gurnell et al. (2006)R, Jansson et al. (2005)R, Leyer (2006)R, Merritt et al. (2010a)R, Rosenthal (2006)FHolzel & Otte (2001)F, Riis (2008)R
Timing – germination (S)Baldwin et al. (2001)W, Britton & Brock (1994)W, Warwick & Brock (2003)WCapon (2007)F
Timing – germination (C)Baldwin et al. (2001)W, Britton & Brock (1994)W, Capon (2007)F, Gerritsen & Greening (1989)W, Noe (2002)W, Warwick & Brock (2003)WNone
Timing – growthBaldwin et al. (2001)W, Budelsky & Galatowitsch (2004)W, Reily & Johnson (1982)R, Robertson et al. (2001)W, Stromberg & Patten (1990)R, Warwick & Brock (2003)WBaldwin et al. (2001)W, Bonilla-Warford & Zedler (2002)W, Stromberg & Patten (1990)R
Timing – reproductionGerritsen & Greening (1989)W, Warwick & Brock (2003)WBonilla-Warford & Zedler (2002)W
Timing – survivalBudelsky & Galatowitsch (2004)W, van Eck et al. (2006)FNone
Timing – community composition (S)Robertson et al. (2001)WNielsen & Chick (1997)W
Timing – community composition (C)Beauchamp & Stromberg (2007)R, Budelsky & Galatowitsch (2004)W, Nielsen & Chick (1997)W, Robertson et al. (2001)W, Toner & Keddy (1997)R, van Eck et al. (2006)FNone
Table 4.   Causal criteria analysis outcomes for the various hypotheses tested
HypothesisEvidence in favour of hypothesisEvidence not in favour of hypothesisConclusion
  1. Summed study weights over the threshold dividing ‘High’ and ‘Low’ levels of support are printed in bold typeface. Abbreviations used are as for Table 3.

Season – hydrochory513Support for hypothesis
Hydrochory – community composition (S)1928Support for null hypothesis
Hydrochory – community composition (C)306Support for hypothesis
Timing – germination (S)168Insufficient evidence
Timing – germination (C)320Support for hypothesis
Timing – growth3015Support for hypothesis
Timing – reproduction98Insufficient evidence
Timing – survival120Insufficient evidence
Timing – community composition (S)88Insufficient evidence
Timing – community composition (C)420Support for hypothesis

We found sufficient evidence to support the hypothesis that rates of hydrochory vary with season (H1). There were 12 studies reporting evidence for such a causal relationship, whilst only one study provided evidence in support of the null hypothesis (Table 3).

The test of the effect of rates of hydrochory on community composition provided divergent results for the two hypotheses. For the hypothesis that increased hydrochory leads to increased species richness (H2a), the evidence was conflicting. In this case, we found support for the null hypothesis, but also considerable evidence in favour of the hypothesis for which the summed weight was 19 – just below the threshold value (Table 3). In contrast, we found sufficient evidence to support the hypothesis that rates of hydrochory affect species composition (H2b), and little evidence against.

For the putative causal linkages involving seasonal timing, we found the evidence to be either sufficient to provide support for a causal relationship or insufficient to draw any conclusions (Table 4). Overall, the number of indentified relevant studies that addressed these hypotheses was relatively low (15 in total) and more supported the various hypotheses (14 studies) than did not (five studies). Most (nine studies) of these were of wetland flora, and of the rest, four related to riverine flora and two to floodplain flora.

With regard to rates of germination, we found support for the hypothesis that seasonal timing affects species composition (H3b), but insufficient evidence to support a link between timing and species richness (H3a). We found support for the hypothesis of a causal relationship between seasonal timing and rates of growth (H4). However, there was little evidence either way for a link between seasonal timing and rates of survival or reproduction (H5, H6). Whilst we found sufficient evidence for a relationship between seasonal timing and species composition of the riparian vegetation community (H7b), there was insufficient evidence to support causality between seasonal timing and species richness (H7a). Figure 3 illustrates the causal criteria analysis outcomes for the various putative causal linkages.

Figure 3.

 Causal criteria analysis outcomes for the putative causal relationships tested.


It has long been suggested that the timing of flow peaks affects the processes that generate and maintain riparian vegetation (Naiman & Decamps, 1997; Poff et al., 1997). Our review divided this hypothesis into a series of specific hypotheses testable by causal analysis. Below, we interpret the outcomes of our systematic review, before returning to our initial question and examining what we can predict with regard to the effects of seasonal flow inversion on riparian flora. As this study is one of the first published studies to apply the causal criteria analysis method of Norris et al. (2008), we conclude with a critical appraisal of the advantages and disadvantages of the approach to answering ecological questions of this kind.

Role of hydrochory

H1: Rates of hydrochory vary with season. Result: Support for hypothesis.  We did not find any studies that directly assess the impact of altered seasonal timing of flow on rates of hydrochory. This is a clear knowledge gap affecting our ability to predict the ecological effects of seasonal flow inversion and meant that we had to approach this question indirectly.

A number of studies have addressed the impact of flow regulation, and dams in particular, on hydrochory (Andersson, Nilsson & Johansson, 2000a; Merritt & Wohl, 2002, 2006; Jansson et al., 2005; Brown & Chenoweth, 2008). These have reported divergent results; some have shown dams to drastically reduce the numbers of propagules moving downstream (Merritt & Wohl, 2006; Brown & Chenoweth, 2008), whilst others have not found any reduction (Jansson et al., 2005).

A large number of studies have, however, explored temporal patterns of hydrochory (Table 3). Almost all of these studies found patterns of hydrochory to be seasonal. Seasonal patterns of hydrochory have been reported across a range of climates, in both regulated and unregulated systems, and using a range of techniques. Only one study reported that patterns of hydrochory were not seasonal, but this result was for vegetative propagules only. In the same study, patterns of hydrochorous seed dispersal were found to be seasonal (Riis, 2008). Whilst the dispersal patterns of vegetative (e.g. plant fragments) and generative (e.g. seeds) propagules may well differ (Leyer & Pross, 2009), many of the above-mentioned studies either failed to differentiate between the two or focused specifically on seeds, precluding a comparison of these processes.

Many of these studies also reported that rates of hydrochory were correlated with discharge (e.g. Middleton, 2000; Boedeltje et al., 2004; Riis, 2008). Similarly, flood events have often been reported as responsible for peaks in hydrochory or propagule deposition (Tabacchi et al., 2005; Rosenthal, 2006; Vogt, Rasran & Jensen, 2006). It might be argued that rates of hydrochory are seasonal only because discharge rates are also seasonal. However, in a number of studies, considerable increases in rates of hydrochory were found to follow increases in discharge, but only during particular times of year (Schneider & Sharitz, 1988; Moggridge & Gurnell, 2009; Moggridge, Gurnell & Mountford, 2009).

The timing of seed release of riparian plants also appears to be important in shaping the seasonal nature of hydrochory. Many studies have reported that the timing of seed release of some riparian/wetland plant species is sensitively linked to seasonal flow patterns (Staniforth & Cavers, 1976; Skoglund, 1990; Kubitzki & Ziburski, 1994; Mahoney & Rood, 1998; Pettit & Froend, 2001; Stella et al., 2006). It is argued that these plants are adapted to release their seeds during receding high flows to facilitate seed dispersal and deposition in appropriate germination sites (Fenner, Brady & Patton, 1985; Nilsson et al., 2010). Indeed, a number of studies in the analysis recorded peaks in dispersal rates that were related to seasonal patterns of seed release (Boedeltje et al., 2004; Moggridge et al., 2009). Similarly, seasonal patterns of hydrochorous dispersal observed within tidal systems have been attributed to the seed production periods of marsh species (Wolters, Garbutt & Bakker, 2005; Hopfensperger & Baldwin, 2009). Thus, the seasonal patterns in hydrochory observed are probably best described as a combined result of the timing of seasonal patterns of seed release and the timing and magnitude of hydrological events (Schneider & Sharitz, 1988; Moggridge et al., 2009).

Finally, it should be noted that, whilst rates of hydrochory have been shown to be seasonal across a wide range of climates, we are not aware of any temporal studies of hydrochory within Australia. Semi-arid and temperate inland Australian rivers are some of the most hydrologically variable and unpredictable systems in the world (Puckridge et al., 1998; Capon, 2007), and this could be expected to impact upon the synchrony of any relationships between season and hydrochory. Indeed, Pettit & Froend (2001) found that the relationship between the timing of seed release of the dominant overstorey species and hydrology was stronger when comparing a tropical river to a temperate one in Western Australia. However, seasonally adapted reproductive patterns have been reported for macrophytes inhabiting unpredictable wetlands within temperate regions of Australia (Froend & McComb, 1994; Rea & Ganf, 1994). Thus, whilst patterns of hydrochorous dispersal are seasonal across a range of climates, these patterns may be less defined in the rivers of south-eastern Australia.

H2: Rates of hydrochory affect community composition. Results: Support for null hypothesis (Richness), Support for hypothesis (Species composition).  There are two probable explanations for the inconsistent results in these assessments: the adoption of the broad definition of ‘riparian’ for the analysis and the divergent experimental approaches of the relevant studies.

A number of studies that have tested the role of hydrochory come from wet grassland restoration projects in northern Europe (Holzel & Otte, 2001; Bissels et al., 2004; Rosenthal, 2006; Gerard et al., 2008). Whilst limited dispersal has often been cited as the main obstacle to restoration, the reinstatement of flooding and input of propagules via hydrochory has not resulted in the recruitment of rare and target species into those degraded habitats (Holzel & Otte, 2001; Bissels et al., 2004; Gerard et al., 2008). However, two of these studies did provide support for a causal relationship between hydrochory and recruitment in terms of species composition (Rosenthal, 2006; Gerard et al., 2008). In contrast, riverine studies of the relationship between hydrochory and recruitment have generally reported a strong relationship between these two processes for both species richness and composition (Andersson, Nilsson & Johansson, 2000b; Jansson et al., 2005; Gurnell et al., 2006; Leyer, 2006). Moreover, Leyer (2006) found the importance of hydrochory to be greater at sites closer to the stream.

Another factor confounding this assessment is the divergent successional stages addressed by these studies. The wet grassland studies all focus on attempts to reintroduce species into established plant communities, whereas all but one of the riverine studies test for the role of hydrochory in colonisation. From the clearly divergent results of these two groups of studies, it could be argued that hydrochory is more important for recruitment within pioneer rather than established communities. This notion is supported by a recent long-term study of propagule dispersal and riparian vegetation development (Merritt, Nilsson & Jansson, 2010a). In that study, hydrochory was found to promote higher species richness during the initial 2 years of colonisation of bare river margins. However, environmental conditions and local propagule sources were more important in structuring riparian communities and maintaining species-rich sites in the longer term.

The importance of hydrochory in shaping riparian vegetation communities has also been suggested in a number of studies of boreal rivers (Nilsson, Gardfjell & Grelsson, 1991; Johansson, Nilsson & Nilsson, 1996; Jansson et al., 2000; Nilsson et al., 2002; Engstrom, Nilsson & Jansson, 2009). These studies typically reported links between vegetation demographics and dispersal strategy. However, they did not attempt to quantify or control for rates of hydrochory, and so were inappropriate for the formal causal analysis. Moreover, these studies largely relied on propagule floating ability or buoyancy as an indicator of dispersal strategy. Whilst buoyancy is a common trait of riparian species that undoubtedly facilitates dispersal by water (Kubitzki & Ziburski, 1994; Lopez, 2001; Boedeltje et al., 2003), an increasing number of studies have begun to question the importance of buoyancy for dispersal by water (Danvind & Nilsson, 1997; Andersson et al., 2000b; Leyer & Pross, 2009). Recently, the dispersal of non-buoyant seeds has been shown to be consistent with the transport of mineral sediments by river water and important for riparian vegetation dynamics (Gurnell, 2007; Gurnell et al., 2007, 2008; Nakayama et al., 2007; Markwith & Leigh, 2008; Chambert & James, 2009).

Recruitment processes

H3: Seasonal timing affects rates of germination. Results: Insufficient evidence (Richness), Support for hypothesis (Species composition).  A number of studies reported an effect of seasonal timing on the species composition of germinants from riparian seedbanks (Table 3) and the collective weight of this evidence provided a high level of support for a causal relationship (Table 4). All of the relevant studies were based on wetland seedbanks, except for one that examined floodplain seedbanks. In contrast, there was insufficient evidence to assess the effect of seasonal timing on the species richness of germinants. The results of studies that provided evidence for this hypothesis were varied. Warwick & Brock (2003) found greater numbers of species germinated in summer compared to autumn, but Britton & Brock (1994) found the opposite, with more species germinating in autumn and fewest in summer. Studies that did not provide evidence in favour of the hypothesis typically reported that responses to seasonal timing were not consistent among species. For example, Capon (2007) found that similar numbers of species emerged in response to summer or winter flooding, but that less abundant species differed in their response to the seasonal timing of inundation. Similarly, both Gerritsen & Greening (1989) and Noe (2002), neither of which studies tested for species richness specifically, reported that rates of germination were species-specific with regard to season.

H4: Seasonal timing affects rates of growth. Result: Support for hypothesis.  A number of studies provided evidence of a causal relationship between seasonal timing and growth, but the evidence provided by these studies was not always consistent (Table 2). For example, Baldwin, Egnotovich & Clarke (2001) reported an effect of flood timing on the stem lengths of a majority of freshwater marsh species, but not on their biomass.

In the two riverine studies relevant to this hypothesis, altered seasonal flow patterns were partly responsible for the reduced growth of riparian tree species following the regulation of rivers in south-western United States (Reily & Johnson, 1982; Stromberg & Patten, 1990). The most relevant of the studies to address this hypothesis because of its geographical context was conducted in mesocosm wetlands within the Barmah-Millewa red gum forests on the Murray River (Robertson et al., 2001). They found that the growth of macrophytes was greater following spring than summer floods and concluded that spring floods (currently lacking because of seasonal flow inversion) are critical for the growth of wetland macrophytes.

H5, H6: Seasonal timing affects rates of reproduction and survival. Results: Insufficient evidence (both).  We found only three studies that provided evidence regarding an effect of seasonal timing on rates of reproduction and two studies for an effect on survival (Table 3). None of these studies were based on riverine flora. All but one of these studies provided evidence in support of the respective hypotheses, but further studies are required to fill this knowledge gap before we can be confident of these conclusions.

H7: Seasonal timing affects community composition. Results: Insufficient evidence (Richness), Support for hypothesis (Species composition).  Seasonal timing of flooding has been shown to have a significant effect on the composition of riparian vegetation communities in a large number of studies (Table 2). In contrast, we only found two studies that investigated the effects of seasonal timing on species richness. These two studies were both wetland mesocosm experiments conducted along the Murray River (Nielsen & Chick, 1997; Robertson et al., 2001). Both of these studies recorded greater plant diversity for spring flood treatments compared to summer flood treatments, although this result was not statistically significant in the latter study.

Two studies that assessed this hypothesis reported that non-natural seasonal water patterns affected species composition by facilitating the establishment of exotic species. Beauchamp & Stromberg (2007) reported that water releases in summer created favourable recruitment opportunities for the invasive riparian trees Tamarix spp. Similarly, Budelsky & Galatowitsch (2004) found that the aggressive non-native perennial Phalaris arundinacea was able to dominate the native Carex stricta under temporally inverted flooding regimes. These findings are consistent with studies on the general effects of river regulation, in which altered flow regimes have been implicated in the increased vulnerability of riparian communities to invasion by exotic species (Decamps, Plantytabacchi & Tabacchi, 1995; Planty-Tabacchi et al., 1996; Richardson et al., 2007).

We identified a number of other studies that have linked altered seasonal timing with changes to riparian vegetation patterns. However, these were not included in the analysis as they did not formally test for such a relationship. Many studies have attributed the decline of riparian forests in the south-western United States to the loss of peak spring flows under regulated conditions. These flows are recognised as particularly important for the recruitment of native Populus spp. (cottonwoods) and Salix spp. (willows) (Fenner et al., 1985; Rood & Mahoney, 1990; Mahoney & Rood, 1998). In a recent study of wetlands along the River Murray, the prevalence of exotics was greatest in a reach subject to inverted seasonal flow patterns (Catford & Downes, 2010), again however, this relationship was not tested directly.

Implications for riparian flora

We are confident that seasonal timing is important for riparian vegetation dynamics. This broad conclusion is in agreement with general findings of previous reviews that have addressed this question (Nilsson & Svedmark, 2002; Poff & Zimmerman, 2010). However, causal criteria analysis has allowed us to assess the amount of evidence for and against various postulated causal relationships within this broader question, providing a far more rigourous assessment of the current state of knowledge.

There is strong evidence that the patterns of germination and growth of riparian plants, as well as the composition of riparian plant communities, are affected by seasonal timing. For germination and community composition, we can expect species composition, but not necessarily species richness, to be affected by changes in seasonal timing. This appears to be because the response of riparian plants to seasonal timing is species-specific in some cases. We would have liked to conduct the analyses presented in this review at a functional group level to find groups of plants with shared traits in relation to seasonal timing (‘vegetation flow-response guids’; Merritt et al., 2010b). However, the current state of knowledge within the literature precluded any such analyses, but we did find some evidence to suggest that exotic species may be advantaged by changes to natural seasonal timing. This has also been suggested by previous reviews (Nilsson & Svedmark, 2002; Richardson et al., 2007).

Despite the existence of some evidence that seasonal timing affects the rates of reproduction and survival of riparian plants, we did not find sufficient evidence to be confident of these conclusions. However, we believe further research would probably strengthen the case for such causal relationships.

In assessing support for causality, the causal criteria analysis framework adopts the convention that if there are multiple possible causal pathways linking cause to effect, then support for the hypothesis for one or more of these pathways is sufficient to infer support for the causal relationship overall (Norris et al., 2008). Under this convention, the fact that we found strong evidence that patterns of hydrochory are seasonal and, independently, that community composition is sensitive to rates of hydrochory implies further support for the hypothesis that seasonal flow patterns are important for riparian vegetation dynamics (Fig. 3). Again, however, this inference is confined to an effect on species composition, and not species richness, although the evidence suggests that an effect on species richness is more likely for pioneer communities and at sites closer to the stream.

We must treat the conclusions related to recruitment processes with some caution, however, as they rely largely on studies of wetland flora. We identified very few riverine studies that tested the identified hypotheses. Moreover, the adoption of the broad concept of riparian for the causal analysis may account for the divergent results of some assessments. For example, whilst all studies involving riverine plant communities found hydrochory to contribute to species richness, studies of wet floodplain habitats did not. This suggests that if a greater number of riverine studies were available and the scope of the analysis was limited to these studies, more consistent results would have been found. However, with the limited number of riverine studies available, if our review had not included wetland and floodplain studies, there would have been ‘insufficient evidence’ to assess almost every hypothesis. Clearly, the purported importance of timing for riverine plants requires further investigation.

Our causal analysis found that seasonal flow timing probably affects many of the life-history stages of riparian plants. We suggest, however, that the importance of seasonal timing for riparian plants is possibly greater for systems in which seasonal patterns are more distinct. Nonetheless, the general findings of our analysis allow us to make some predictions in regard to the rivers of south-eastern Australia. We conclude that, despite the lack of specific studies on the impact of inverted seasonal flow patterns on the flora of these rivers, these plant communities have almost certainly been affected as a result of this inversion. Specifically, patterns of hydrochory and plant growth, as well as the composition of the vegetation community, have probably been affected.

Assessing the utility of causal criteria analysis

Systematic reviews are far less common in ecology than in fields such as medicine (Pullin, Knight & Watkinson, 2009), and most ecological reviews are largely descriptive (but see Poff & Zimmerman, 2010). To our knowledge, this is the first published literature review in ecology that employs causal criteria analysis.

Ecologists may be more familiar with quantitative methods of synthesis such as meta-analysis. However, it seems that much of the information contained within the ecological literature is not amenable to meta-analysis, either because the ecological response metric is not quantitative or because insufficient information is provided in the publication. In our study, meta-analysis would not have been possible for a number of the hypotheses (H1, H2b, 3b and 7b) because of the non-quantitative nature of the response metric (e.g. species composition). For the remaining hypotheses, we assessed the studies included in our causal analysis for the presence of the summary data necessary to calculate a standardised effect size (Some subset of: treatment means, sd’s, N’s; SS, MS, d.f.; Gurevitch & Hedges, 2001; Levine & Hullett, 2002). We found that roughly two-thirds would be amenable to meta-analysis (see Data S3), a comparable result to a similar assessment of epidemiologic investigations (Bekkering, et al. 2008). Also, in common with Bekkering et al. (2008), we found that studies that reported an association (e.g. significant difference between treatments) were more likely to provide the necessary summary data, a practice that may threaten the validity of meta-analyses. Overall, meta-analysis could not have been used to calculate an ensemble effect size for any of the hypotheses we addressed, and in comparison, causal criteria analysis was able to include a far greater number of existing studies.

The causal analysis approach adopted for this review has a number of advantages. Its structured nature makes it highly transparent and repeatable. The process of refining questions and conceptual models in an iterative fashion led us to the adoption of clear and testable hypotheses. Clearly defined hypotheses help to direct literature searches efficiently. They also enable the identification of knowledge gaps within the literature and guard against the reviewer making conclusions based on insufficient evidence. Moreover, the conclusions we have drawn from this process are transparent and suffer from less bias than is often the case with descriptive reviews (Slavin, 1995). Overall, we found the approach amenable to the production of a logical and succinct review and believe that causal analysis can introduce rigour and efficiency into reviews in ecology.

There were some difficulties encountered using the method of Norris et al. (2008). Locating the specific information required to weight the studies was often time-intensive, especially as study designs are not always made explicit by authors. We experienced difficulties assigning study weights for some studies that did not fit neatly into the limited number of experimental designs for which weights have been defined. Also, the method as currently described does not make clear the appropriate treatment of inconsistent pieces of evidence from a single study. Overall, however, we believe the approach is a very good first step on the path to much more systematic treatment of evidence in ecological reviews.


We conducted a systematic review of the evidence from the literature on the importance of seasonal timing for riparian vegetation dynamics. We are confident that seasonal timing is important for a number of processes that generate and sustain riparian plant communities. In particular, causal analysis provided strong evidence to support causal links between seasonal timing and the patterns of hydrochory, germination and growth of riparian plants, as well as overall plant community composition. Given that alteration of flow regimes is typically confounded with other environmental factors, the findings of a general nature produced by this review are likely to indicate strong responses to flow alteration of both scientific and management interest (Poff & Zimmerman, 2010). We found insufficient evidence to infer causality regarding seasonal timing and reproduction or survival, and it is in these areas, as well as with the relationship between riverine plants and seasonal timing in general, that we encourage more research.

Overall, however, there is sufficient evidence to be confident that changes in the timing of seasonal flow patterns will affect the riparian vegetation of regulated rivers. Accordingly, flow management aimed at maintaining and/or restoring the ecological values of riverine ecosystems should consider flow timing and its implications for riparian vegetation dynamics. Finally, whilst the causal analysis method of Norris et al. (2008) has both advantages and disadvantages, it provides fertile ground from which the systematic assessment of evidence for causal relationships in ecology can grow.


Joe Greet is supported by an Australian Postgraduate Award and eWater student top-up scholarships. Angus Webb is funded by the eWater project ‘Ecological Management and Restoration’. Thanks to Evan Harrison and Beth Wallis for useful discussions on causal analysis and to Beth Wallis for reviewing an earlier version of this manuscript. Thanks also to Peter Dawson for help with Fig. 1. We also thank two anonymous reviewers for their valuable comments.