Land use intensification and global change have led to decreasing biotic diversity (Foley et al., 2005). Global species numbers show clear negative trends, though different levels of spatial scale show very different trends (McGill et al., 2015). Much of this biodiversity loss is caused by increasing homogenization of communities (i.e., biotic homogenization), and not necessarily by loss of local diversity (Dornelas et al., 2014). For the landscape scale this may imply that landscape diversity may decrease mainly due to the disappearance of differences between local communities (Smart et al., 2006), rather than locally detectable declining species numbers (McGill et al., 2015). As declines in biodiversity threaten the multifaceted functioning of ecosystems at both local and landscape scales (van der Plas et al., 2016), its conservation and restoration requires appropriate landscape-wide management strategies.

The process of ecological succession has long been acknowledged as a primary driver of biodiversity (Sousa, 1979). Different stages of ecological succession may exhibit different species richness levels. A classic example on forest succession shows higher species numbers in intermediate stages of succession (Odum, 1969; Whittaker, 1970). Furthermore, different successional stages may harbor very different sets of species. Thus, the landscape diversity is not a function of the local diversity alone, but also of the complementarity between stages present in the landscape. Hence, to maximize the diversity of a landscape, both the local diversity of successional stages, as well as the difference between communities of the different stages needs to be considered. This line of thought is well represented in the classical partitioning of the landscape diversity (γ) into a local component (α) and a turnover component (β) (Whittaker, 1960; Jost et al., 2010). Periodic resets of unidirectional succession through naturally occurring disturbance events in different parts of the landscape may contribute to the maintenance of a mosaic of successional stages (Sousa, 1984). In natural systems, examples of such disturbances include fire (Vandvik et al., 2005), scouring by peak river discharges (Tockner et al., 2000) or landslides (Walker et al., 1996). In absence of such natural dynamics, for example due to human interventions, there is a risk of loss of successional stage heterogeneity within the landscape (Baptist et al., 2004).

To obtain biodiversity within a landscape that has lost its natural dynamics due to human influence, management efforts need to be directed toward maintaining a landscape with a variety of successional stages present. Such management has been widely applied in a variety of different ecosystems (Baptist et al., 2004; Vandvik et al., 2005) and is known by different names, e.g., cyclic rejuvenation, rotational management and periodic ecosystem reset. Here we use the term cyclic rejuvenation through management (CRM), which is the practice of periodically resetting part of the habitat in a landscape to a (mostly) bare state in order to create a spatially dynamic mosaic of habitat patches in different stages of succession (Hinsch and Poethke, 2007). CRM in floodplain management, aimed at systematic removal of part of the floodplain forests, increased biodiversity in channels where natural reset of succession was absent (Baptist et al., 2004). Likewise, controlled fire management can help preserve grassland diversity through CRM (Richards et al., 1999). While good results have been shown, the inherent success of this management approach rests strongly on the assumption that successional stages are complementary to one another. The more unique stages are with respect to their community composition, the larger the gain to the regional species pool and thus overall landscape diversity. Conversely, when successional stages are highly similar, CRM will have little effect on landscape diversity.

Manmade water systems such as agricultural drainage ditches are a good example of anthropogenic ecosystems which are under continuous management to protect the hydrological drainage of agricultural land (Hill et al., 2016). However, their value for conservation of biodiversity is increasingly recognized as well (Armitage et al., 2003; Herzon and Helenius, 2008; Clarke, 2015). Reshaping ditch banks, removing vegetation and dredging organic sediment helps to maintain hydrological functioning, but may also support ecological function and diversity (Twisk et al., 2003). Conventional management of these waterways results in the constant resetting of succession (van Strien et al., 1991; Clarke, 2015; Hill et al., 2016), making an often unorganized form of CRM the norm in these systems. Formal CRM potentially offers an approach for this management to take place, while also realizing a biodiversity increase in comparison to existing management.

In human dominated landscapes, management primarily aims at maintaining provisioning services [e.g., water storage, food production (Power, 2010)]. For ditch ecosystems, the application of CRM has been suggested as a promising way to increase diversity at a landscape scale (Watson and Ormerod, 2004; Clarke, 2015; Hill et al., 2016) while maintaining associated services. Landscape level diversity (γ) in these ecosystems has been shown to be largely caused by differences in community composition between individual sites (Goldenberg Vilar et al., 2014; Whatley et al., 2014a), as indicated by a large β-diversity component, stressing the importance of landscape heterogeneity. Despite the claims on the potential usefulness of CRM (Clarke, 2015), so far no formal evaluation of its potential has been made, nor has it been widely adopted in practice in ditch management. A first step in the evaluation and adoption of such a management practice is to illustrate the importance of successional stage heterogeneity for landscape-wide diversity. Differences in community composition between sites may be caused by two inherently different underlying patterns, namely one of species richness difference and one of species replacement (Baselga, 2010; Legendre, 2014). Differences in community composition between sites (β-diversity) may be partitioned into a richness difference and species replacement component (Baselga, 2010; Podani and Schmera, 2011; Legendre, 2014). Species replacement refers to the simultaneous gain and loss of species along an ecological gradient (Legendre, 2014). A large replacement component is indicative for high levels of complementary between sites in terms of species composition. Alternatively, compositional differences among sites may also be generated merely by differences in species numbers (richness difference, sensu Legendre, 2014, also see Podani and Schmera, 2011). In contrast to the latter, it is the replacement component that will contribute to γ-diversity. Translating this to ditch networks, the degree to which successional stage diversity will be important in generating a high γ-diversity at the landscape scale will largely depend on the relative importance of the replacement component in the β-diversity among successional stages. For CRM to truly be useful for increasing landscape biodiversity, species replacement between different successional stages is required (i.e., complementarity) and not difference in richness alone.

In this study we performed 504 vegetation surveys in ditch reaches spread over 21 different polder landscapes in the Netherlands to study the importance of maintaining successional stage diversity on the landscape-wide species diversity of aquatic vegetation. We characterized the dissimilarity among successional stages and its underlying patterns to assess the complementarity of communities of different successional stages and their contribution to landscape biodiversity. In addition, we tested the potential effect of landscape-wide heterogeneity in successional stages on the landscape-wide diversity. This was done by calculating the biodiversity of simulated landscapes with different combinations of successional stages. Based on these results we evaluate the hypothesized merit of CRM on promoting landscape-wide diversity in ditch systems (Clarke, 2015) through proliferation of a diverse spatial-temporal mosaic of habitats.

Our diversity partitioning analyses show that β-diversity forms a very important component of the γ-diversity of macrophytes in ditch networks of polder systems. These results are in agreement with several other studies that have shown large differences in community composition among sites for a variety of organism groups in agricultural ditches, including macroinvertebrates, diatoms and helophytes and aquatic vegetation (Leng et al., 2010; van Zuidam and Peeters, 2013; Goldenberg Vilar et al., 2014; Clarke, 2015). New to our analyses is that more than 70% of this β-diversity seems to arise from differences in successional stages of ditch vegetation and that at least 40% of this β-diversity stems from complementarities in species composition between successional stages. Our results thus clearly demonstrate that heterogeneity in successional stages contribute strongly to the regional diversity of macrophytes in the ditch network of a landscape.

The importance of landscape heterogeneity for biodiversity has been shown in a variety of biomes such as boreal forests, grasslands and rivers (respectively, Niemala et al., 1996; Richards et al., 1999; Ward et al., 2002). Here we showed similar results for aquatic plants in agricultural drainage ditches within polder landscapes, illustrating the importance of maintaining landscape heterogeneity. The large contribution of β-diversity to γ-diversity observed in our study emphasizes the necessity for a focus of management on diversity at the landscape scale rather than at the scale of local ditch reaches. Given that differences among successional stages form an important part of the landscape heterogeneity influencing the β-diversity in our polder landscapes, the creation and maintenance of successional heterogeneity may be an important way through which management may be able to enhance and sustain landscape wide biodiversity. In the absence of any management or natural disturbances, heterogeneity among communities of different sites will gradually be reduced as succession progresses (Vandvik et al., 2005). The landscape would increasingly be dominated by a limited number of late successional stages. As the ‘No Management Scenario’ of our simulations illustrate, this is expected to result in a gradual decrease of β- and γ-diversity through time, as the early successional stages are lost from the landscape.

Currently, management in the agricultural ditch networks of Netherlands is almost entirely focused on the maintenance of their hydrological function and largely involves regular dredging, vegetation removal and bank reshaping (Twisk et al., 2003). Given that these measures involve a reset of succession, they contribute to the rejuvenation of vegetation (Wade, 1993) and may as such also contribute to the maintenance of successional stage heterogeneity and γ-diversity. However, the impact of these measures on the latter variables will also depend on the frequency and the scale of their application. The frequent and simultaneous application of these measures over large spatial areas will inevitably result in the disappearance of late successional stages. According to our diversity partitioning results this would be expected to result in a reduced β- and γ-diversity as well. However, such expectation is poorly supported by our simulation results which suggest that a predominance of young successional stages would be of little consequence to γ-diversity. Indeed, in these simulations, the progressive loss of later successional stages resulted in a minor reduction of α-diversity, but this was entirely compensated by an increase in β-diversity. Across the entire dataset, β-diversity of early successional stages was relatively high compared to late successional stages (Supplementary Figure S5). The high compositional variability among communities of early successional stages likely reflects a large impact of stochasticity during community assembly. Compared to late successional stages, interspecific interactions in young habitat patches are still weak and the composition of pioneering communities is likely more determined by coincidental dispersal and colonization events than by the outcome of competition (Tilman, 1994).

When interpreting the simulation results it is, however, important to note that diversity indices were calculated based on random draws from the entire dataset. Communities of ditch reaches were thus regarded to be freely interchangeable among polders. This implicitly assumes absence of any dispersal limitation and ignores historic trajectories of metacommunities in individual polders. This may have led to an underestimation of the negative effects of late successional stage loss. By removing late successional stages, their function as source of populations recolonizing the earlier stages may be compromised. The diversity of early successional stages in the absence of late successional stages may therefore have been overestimated in our simulations. In reality the reassembly of communities after the reset of succession may largely depend on colonization, which may take 3–4 years to recover full vegetation diversity (Milsom et al., 2004). Colonization by dispersing propagules from proximate source communities is also likely to be more influential than from more distant ones (Brederveld et al., 2011; Van Dijk et al., 2014). Furthermore, the remaining seed bank can also play an important role in community reestablishment (Sarneel et al., 2014; Van Leeuwen et al., 2014), although it is unknown to what extent dredging activities reduce local community resilience by removing large parts of this reservoir. If management resets succession in large parts of the landscape simultaneously, recolonization of empty sites may be severely impeded by dispersal limitation reducing overall β- and γ-diversity of the landscape. Although the contribution of early successional stages to γ-diversity may potentially be disproportionally high, the role of nearby mid and late successional stages as source communities of dispersal propagules should not be ignored in real plant metacommunities.

Conceptually, CRM is likely the type of management scheme that best guarantees the presence of a range of successional stages in the landscape. However, much depends on the specific parameters of management, namely the delay and frequency of visit by management (Hinsch and Poethke, 2007). The delay is defined as the time that each site is left to develop naturally before it is revisited (Morris, 2000). Frequency of visits by management to the landscape is defined as the proportion of sites being managed during each visit. More frequent visits lead to a lower proportion of sites being managed per visit, but creating a higher variation in successional age within the landscape (Hinsch and Poethke, 2007). A low delay (the duration of a few years) may entirely exclude some late successional stages from the landscape and therefore eliminate potential sources for recolonization of recently managed ditch reaches. In contrast, high delays (decades) may result in a predominance of late successional stages that may interfere with the hydrological function of the ditches and contribute less to γ-diversity. When a large number of ditch reaches are visited at once (low frequency), a landscape will likely consisting of large, homogenous blocks with relatively low successional stage diversity across the landscape. Low frequency management is typically found in the currently employed, large scale dredging operations where large swaths of the landscape are managed at the same time. We advocate that application of CRM management schemes with more frequent visits and possibly also higher delays would lead to more heterogeneous mosaic-like landscape patterns (see Niemala et al., 1996 for an example in boreal forests). Such mosaic may be further elaborated through a spatial planning that is focused on minimizing distance between recently managed and other reaches as to facilitate dispersal and recolonization (Watson et al., 2012; Van Dijk et al., 2014).

Although the potential advantages of CRM for biodiversity conservation in agricultural ditch networks are obvious, it is difficult to make generalizations about what should be the optimal CRM scheme. Ideally, CRM schemes are based on knowledge about the development rates of the different successional stages. Succession rates appear to be highly variable as the time needed for succession to turn an empty ditch into a carr may range between 10 and 27 years (Bakker et al., 1994). Succession rates in aquatic systems are highly context specific and depend on several factors such as sedimentation rates, flow velocity (Sousa, 1979), sediment characteristics, intensity of land use and associated nutrient loading (Carpenter, 1981). Obviously, more productive systems with high rates of sediment loading will require CRM schemes with shorter delays than less productive systems. Ideally, CRM schemes are adapted and improved based on information acquired through the monitoring of succession in parts of the ditch network that have been managed, a form of adaptive management (Haney and Power, 1996). Using social-benefit analyses (e.g., Fliervoet et al., 2013) the benefits to farmers (reduced management effort and cost), water quantity managers (maintained drainage capacity) and nature water quality organizations (increased ecological diversity) may be quantified in detail to arrive at an optimal management plan for a specific landscape.

It’s important to note that the ditch networks considered in our study have historically been associated with intensive land use. They are biotically impoverished in comparison to ditch networks in areas with less intensive agriculture (Leng et al., 2010). High nutrient loading is likely to enhance the intensity of interspecific competition especially in the later successional stages, leading to monotonous vegetation (van Zuidam and Peeters, 2013). This may have contributed to the relatively high degree of homogeneity of these stages in the landscape. Furthermore, conclusions of this study only pertain to the successional stages that were encountered in our field study. Current management practices in the Dutch agricultural polder landscape prevent ditches from containing successional stages that are older and more developed that the ones observed in our studies (e.g., alder carrs, floating fens). Our analyses and simulations do not take into account the potential contribution of such successional stages to β- and γ-diversity. Although such stages likely interfere with the drainage function of the ditch network, they would potentially have a unique and important contribution to regional diversity. Hence it would be interesting to also evaluate such stages in future work and consider how they can be maintained in the landscape.

The same case can be made for other groups of organisms that make their home in the ditch systems. Despite the important structural and functional role of macrophytes in the aquatic habitat, conservation management in ditch networks should also aim at supporting biodiversity of other organism groups. Extrapolation of our conclusions regarding the contribution of successional stages to macrophyte biodiversity should therefore be done with prudence. Although the diversity of macrophytes may be an important determinant for the diversity of some associated groups (Whatley et al., 2014b), such association is not unequivocally strong for all aquatic groups. Producers are shown to be more strongly driven by patterns of replacement than higher trophic groups (Soininen et al., 2018). Therefore, studies on the impact of succession on the β- and γ-diversity of those other groups and how these can be adequately managed is vital. Our data driven approach may be easily applied to other species groups to assess the potential efficacy of a CRM scheme. Herein lays the power of our approach, all it requires is data on species composition of different sites in a landscape and a conceptual notion of successional trajectories. This type of data is generally gathered by nature- and water managers in their standard monitoring schemes. These data may serve as an ideal first step in evaluating the inherent requirements (differences in, and replacement of species between successional stages) for useful application of CRM, before extensive management activities are carried out that may be hit-or-miss. Further development of a specific management scheme could benefit from specific deterministic models (see Perona et al., 2009 for CRM in floodplains and Larocque et al., 2013 for an overview of similar models in forests).

We have shown the importance of succession as a driving force in agricultural drainage ditches, non-natural aquatic habitats that nonetheless can contribute to the biodiversity of the landscape (Davies et al., 2008; Verdonschot et al., 2011). Promoting successional heterogeneity is possible through cyclic rejuvenation of parts of the landscape through management. As management of ditch systems is needed to preserve hydrological functioning (Herzon and Helenius, 2008), the change to a system-specific, spatially explicit management scheme employing the principles of CRM can be both viable and cost-effective. This combination of drainage and ecological function fits well within the context of reconciliation ecology (Rosenzweig, 2003), allowing for coexistence of relatively high levels of biodiversity in a seemingly unhospitable landscape of anthropogenic agricultural activity.

ST, MV, and SD conceived the conceptual idea for the study. ST and SD designed the study and ST carried out the field work. ST carried out the data analysis with help from SD and MV. The first draft of the manuscript was written by ST, MV, EB, and SD with all authors contributing substantially to revisions.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The handling Editor is currently co-organizing a Research Topic with one of the authors EB, and confirms the absence of any other collaboration.