Mechanical removal of macrophytes in freshwater ecosystems: Implications for ecosystem structure and function

• Macrophyte removal affects ecosystem structure and functions. • Most removal studies were done in rivers and evaluated single ecosystem properties. • Modelling of removal on interrelated ecosystem properties with a Bayesian network. ⁎ Corresponding author at: Norwegian Institute for Wa E-mail address: Kirstine.thiemer@niva.no (K. Thiemer https://doi.org/10.1016/j.scitotenv.2021.146671 0048-9697/© 2021 Elsevier B.V. All rights reserved. a b s t r a c t a r t i c l e i n f o Article history: Received 14 January 2021 Received in revised form 2 March 2021 Accepted 17 March 2021 Available online 27 March 2021

Macrophytes are generally considered a nuisance when they interfere with human activities. To combat perceived nuisance, macrophytes are removed, and considerable resources are spent every year worldwide on this practice. Macrophyte removal can, however, have severe negative impacts on ecosystem structure and functioning and interfere with management goals of healthy freshwater ecosystems. Here, we reviewed the existing literature on mechanical macrophyte removal and summarised current information from 98 studies on shortand long-term consequences for ecosystem structure and functioning. In general, the majority of studies were conducted in rivers and streams and evaluated short-term effects of removal on single ecosystem properties. Moreover, most studies did not address the interrelationships between ecosystem properties and the underlying mechanisms. Contrasting effects of removal on ecosystem structure and function were found and these discrepancies were highly dependent on the context of each study, making meaningful quantitative comparisons across studies very difficult. We illustrated how a Bayesian network (BN) approach can be used to assess the implications of macrophyte removal on interrelated ecosystem properties across a wide range of environmental conditions. The BN approach could also help engage a conversation with stakeholders on the management of freshwater ecosystems.

Introduction
Mass development of aquatic macrophytes is a worldwide problem (Barrett, 1989;Hussner et al., 2017;Verhofstad et al., 2017) and considerable resources are spent every year on macrophyte removal (Hilt et al., 2006;Vereecken et al., 2006;Verhofstad and Bakker, 2019). Macrophytes are often negatively perceived as water weeds, notably during periods of mass development (nuisance growth) when very high densities of aquatic plants interfere with human activities. The removal of macrophytes is desired for the prevention of flooding of adjacent land (Boerema et al., 2014;Vereecken et al., 2006), prevention of clogging of hydropower plants (Dugdale et al., 2013), facilitation of irrigation (Armellina et al., 1996), disease control (Bicudo et al., 2007), trade and commerce (Güereña et al., 2015) together with recreational activities such as angling, swimming, boating and water skiing (Bickel and Closs, 2009;Verhofstad and Bakker, 2019).
Mass development generally results from a combination of multiple drivers promoting plant growth (light, temperature, nutrients) and minimising plant loss (lack of disturbance and biological control), when a species is present that is able to benefit from these conditions (Riis and Biggs, 2001). The controlling role of light and nutrient supply from diffusive or point sources is well established in shallow lakes with the theory of alternative stable states (Hilt et al., 2006;Verhofstad et al., 2017), and ecosystems recovering from nutrient enrichment have likewise been shown to exhibit excessive plant growth (Hilt et al., 2011). Increases in nutrient supply may also be indirect. Liming in oligotrophic systems, for example, can promote the degradation of sediment organic matter and boost CO 2 , NH 4 and PO 4 supply causing excessive plant growth (Roelofs et al., 1994). In river systems, water regulation (discharge and depth), nutrient supply (sewage effluents, fine sediment accumulation upstream of weirs, land use) and management of river bank (clearance of river banks leading to improved light availability) may boost the development of plant biomass across the channel (Chambers and Prepas, 1994), even in oligotrophic systems with perennial aquatic plants ( Moe et al., 2013;Rørslett, 1988). Degraded ecosystems may be sensitive to the introduction of new species able to invade for lack of biological control or because they have biological traits more suited to the modified environment (Hussner et al., 2017).
Solutions to combat perceived nuisance growth of macrophytes include mechanical removal (cutting, dredging), chemical treatment (herbicide, salt) or biological control (biocontrol agents such as herbivorous fish and insects or shading) (Hussner et al., 2017). However, these management practices are costly and generally only have short-term effects due to plant regrowth. They can therefore not be considered a sustainable solution. A more progressive management is using naturebased solutions to promote sustainable economic, societal, and environmental benefits (Boerema et al., 2014;Güereña et al., 2015).
The disadvantages macrophytes have for humans' conflict, at the same time, with the societal benefits that macrophytes provide (i.e. ecosystem services). The benefits of aquatic macrophytes are often overlooked by the public and might be underestimated in decision making by water managers. The ecosystem services provided by aquatic macrophytes include supporting (e.g., habitats for fish and macroinvertebrate), provisioning (e.g. food, fodder, fertiliser, biomass fuel), regulating (e.g. nutrient cycling, water purification, pest and disease control), and cultural services (e.g. aesthetic pleasure, inspiration for culture, art and design, recreation and tourism) (Boerema et al., 2014;García-Llorente et al., 2011).
These ecosystem services rely on the role that aquatic macrophytes play in ecosystem structure and function (Caraco et al., 2006;Engelhardt and Ritchie, 2001;Gurnell, 2014;Jeppesen et al., 1998). Many individual studies have quantified the effects of macrophyte removal on individual ecosystem properties, and time has come to synthesise this research and explore the implications at the level of the whole ecosystem level. Today, many countries prohibit chemical treatment and while biological control with nonnative species has been successful in different parts of the world (Hill and Coetzee, 2017), it introduces additional ecological uncertainties for native species (Hussner et al., 2017). Here, we focus our review on the effects of mechanical removal of aquatic plants (both submerged and free-floating), hereafter referred to as macrophyte removal, as is used worldwide in rivers and lakes. We distinguish short-term from long-term consequences on aquatic ecosystems. Short-term effects were defined as the necessary period for plant regrowth and ecosystem recovery, which may take weeks (Bal et al., 2017;Garner et al., 1996;Spencer et al., 2006) to years (Caffrey and Monahan, 2006;Painter, 1986;Rørslett and Johansen, 1996). Long-term consequences may emerge from repeated macrophyte removal (Baattrup- Pedersen et al., 2002). We also briefly discuss the complexity of assessing the consequences of aquatic plant removal on aquatic ecosystems, depending on the local context and removal methods used. For this, we used a Bayesian network (BN) approach as a first attempt to synthesise how macrophyte removal affects ecosystem structure in different freshwater ecosystems and how the current lack of a holistic approach may influence the conclusions derived from single organism studies. Finally, we identified research needs.

Publication search criteria
A systematic search was conducted to find relevant literature concerning mechanical removal of macrophytes (last search, 10.07.2020). Web of Science, PubMed and Google Scholar academic search engines were used to find the relevant scientific peer-reviewed papers using combinations of the following search terms in title and author keywords : ((fish* OR macrophyte* OR*macroinvertebrate*OR periphyton OR "aquatic weed*" OR "water weed" OR "aquatic plant*") AND (dredg*OR cut*OR mow* OR remov*)). Studies retrieved from the automatic search that clearly did not concern macrophyte removal were discarded. In addition, relevant articles from reference lists of papers and our own general knowledge were used to identify additional important literature. The initial search yielded 532 studies in total. From these studies, we selected all papers which met the criteria in Table 1 which gave a total of 98 papers of which 86 had an experimental setup. The other 13 papers were mainly review papers or papers on ecosystem services. Grey literature, in the form of reports and management plans were not included. However, conclusions from these were indirectly used in this review as several peer-reviewed papers used the local knowledge. Information on the effect of removal was extracted from each study which met the inclusion criteria: species removed, removal area, size of study and each ecosystem property measured.

General trends in publications
The 86 experimental papers covered studies on mechanical macrophyte removal in 25 countries, with the largest proportion of studies from America and Europe (Fig. 1). The majority of studies have been conducted in streams and rivers ( Fig. 2A) and evaluated the effects of single-event removal of submerged macrophytes (Fig. 2B). The effects of removal on macrophytes, fish, macroinvertebrates and hydraulic properties have been the most frequently studied ecosystem properties, whereas the consequences of removal on benthic algae, mussels and zooplankton have only been evaluated in very few studies (Fig. 2C).
Only nine studies have examined more than one ecosystem property (Fig. 2D). The consequences of macrophyte removal on separate ecosystem properties have mostly been documented through short-term studies with a mean range of 14 months including before and after sampling (Fig. 2E) and the effects of partial removal have been the most studied (Fig. 2F). An overview of the consequences of macrophyte removal for several ecosystem properties is summarised below from the data compiled in Appendix A.

Macrophytes
The influence of aquatic plant removal on the growth and survival of macrophytes causes long-term effects on community structure. Following removal, increased relative growth rates have been reported for several species such as Sparganium erectum (L.) (Bal et al., 2017), Myriophyllum spicatum (L.) (Crowell et al., 1994) and Lagarosiphon major (Ridl. Moss ex Wager) (Bickel and Closs, 2009) which may be a compensatory mechanism and a response to plant damage similar to herbivory (van Zuidam and Peeters, 2012). Increased growth rates may further be stimulated by improved light conditions and low selfshading post removal (Binzer et al., 2006). Despite increased growth rates, ten studies found reduced standing biomass by the end of the growth season (Armellina et al., 1996;Bal et al., 2017;Bal et al., 2006;Caffrey and Monahan, 2006;Crowell et al., 1994;Garbey et al., 2003;He et al., 2019;Schooler et al., 2007;Thiébaut et al., 2008). Removal was found to have more severe effects on survival of species with an apical meristem growth point, such as Potamogeton compressus (L.) and Potamogeton lucens (L.), with both being less tolerant to cutting (van Zuidam and Peeters, 2012) than species with basal meristem growth points, e.g. Sparganium emersum (Rehmann) (Baattrup- Pedersen et al., 2003) and free-floating macrophyte species such as Eichhornia crassipes (Mart.) (Spencer et al., 2006).  The effects of differences in tolerance to removal may be more pronounced for the composition of macrophyte communities in ecosystems where macrophyte removal is repeated annually. In lowland streams, annual macrophyte removals, over a period of 20 years, made the species composition of macrophyte more homogeneous and dominated by fast-growing species with basal meristem growth points, rhizomes and high dispersal capacities (e.g. Sparganium sp. and Elodea canadensis (Michx.)) (Baattrup- Pedersen et al., 2002). Shannondiversity and speciess richness have also been reported to decrease 19-66% and 16-40%, respectively (Baattrup-Pedersen et al., 2003; Baattrup- Pedersen et al., 2002;Baattrup-Pedersen and Riis, 1999;Best, 1994;Strien and Strucker, 1991). The altered species composition towards a community with a less complex structure, are likely to have knock-on effects on ecosystem structure and functioning, however these inter-relationships have not yet been revealed.

Macroinvertebrates
The consequences of macrophyte removal for macroinvertebrate abundance and diversity have been frequently studied in rivers (Appendix A, Table 1). In small rivers with submerged macrophytes, removal has been shown to instantly reduce abundance by 20-70% (Armitage et al., 1994;Dabkowski et al., 2016;Kaenel et al., 1998;Lusardi et al., 2018). The highest declines in abundance were found in taxa using macrophytes directly as substrate (i.e. Simuliidae and Chironomidae), whereas taxa living in or on the bottom sediments proved to be more resistant (Kaenel et al., 1998). Four studies did not find a significant effect on abundance, likely due to the late sampling of invertebrates post removal enabling a possible recovery (Armitage et al., 1994;Buczyński et al., 2016;Laughton et al., 2008;Ward-Campbell et al., 2017). Similar findings were reported from lakes with submerged macrophytes (Habib and Yousuf, 2014;Miliša et al., 2006). We did not find studies on invertebrate abundance in systems dominated by free-floating plants, although it is likely that the absolute number of invertebrates will be less affected as changes in surface for colonization are smaller and the sediment is usually not disturbed during removal of free-floating plants. Changes in overall macroinvertebrate diversity were less clear. Shannon-diversity was found either to decline significantly (23-44%) (Gray et al., 1999;Habib and Yousuf, 2014;Miliša et al., 2006), to stay unchanged (Bickel and Closs, 2009;Buczyński et al., 2016;Dabkowski et al., 2016) or to increase (14% increase) (Lusardi et al., 2018). These contrasting findings likely reflect the different contexts of the studies, such as differences in methods and time of sampling following the removal.
Reduced abundance and potential changes in diversity post removal may only be temporary. Invertebrate abundance has been reported to recover within 1-10 months depending on the time for plants to regrow and colonization from upstream or nearby areas (Habib and Yousuf, 2014;Kaenel et al., 1998;Monahan and Caffrey, 1996). Long-term studies on invertebrate community response are currently lacking.

Phyto-and zooplankton
Phyto-and zooplankton require a sufficient water residence time to develop (Reynolds, 2000) and thus the effects of macrophyte removal for phyto-and zooplankton abundance and diversity have mostly been studied in lakes rather than rivers (Appendix A). Removal of macrophytes in eutrophic waters may cause regime shifts: a study modelling combined effects of removal and high external nutrient loads found that removal of >30% of the submerged macrophytes in a lake with high nutrient input was sufficient to trigger a regime shift to an alternative stable state with higher phytoplankton biomass (Kuiper et al., 2017). Accordingly, an increase in phytoplankton biomass up to 83% and a shift towards communities dominated by fast growing cyanobacteria has been reported after removal of free-floating macrophytes (Bicudo et al., 2007;James et al., 2002;Zhu et al., 2019). The increase in phytoplankton biomass was explained by increased light availability due to the lack of macrophytes, and more resuspended sediment leading to increased availability of nutrients for phytoplankton growth. However, removal of submerged macrophytes has also been reported to lead to an initial decrease in phytoplankton biomass, followed by recovery after several weeks and possible exceedance of phytoplankton biomass in control sites (Alam et al., 1996;Morris et al., 2006;Wile, 1978). This was likely due to sediment disturbance during the removal leading to considerably increased turbidity, hence reducing light availability and impairing initial phytoplankton growth, while the lack of competition by macrophytes later boosted phytoplankton growth.
The effects of macrophyte removal on zooplankton have scarcely been studied (Appendix A). A decline in zooplankton abundance and a shift in community composition towards small zooplankton species has been reported post removal. These were suggested to be a result of increased fish predation on the larger Cladocera, and downstream displacement in rivers as macrophytes provide velocity refuge and are a food source for zooplankton (Garner et al., 1996;Mangas-Ramírez and Elías-Gutiérrez, 2004). Partial removal of macrophytes has been suggested to promote higher zooplankton abundance and diversity. In a study with removal of only free-floating macrophytes, zooplankton diversity increased from 11 to 40 species and was explained by higher habitat heterogeneity in partially cut areas where macrophytes with different life forms coexisted (Choi et al., 2014).

Fish
Macrophyte removal can be detrimental to fish populations, either directly when plants are harvested (Engel, 1990;Mikol, 1985) or indirectly through enhanced predation risk from larger fish (Unmuth et al., 1999), reduced food availability due to increased flow velocity (Garner et al., 1996) or deterioration of important spawning habitats (Lusardi et al., 2018;Swales, 1982). Reduced survival and abundance following macrophyte removal have been reported for fish fry and smaller fish in both rivers and lakes (Engel, 1990;Mikol, 1985;Mortensen, 1977). Mechanical harvesting of submerged macrophytes was also found to remove 2-25% of the standing juvenile population (Engel, 1990;Mikol, 1985). In eight studies, fish abundances were reduced by up to 60% after macrophyte removal (Greer et al., 2012) (Appendix A). In one study, a more severe outcome was found in a highly eutrophic lake, as no fish were recorded post-removal of macrophytes (Mangas-Ramírez and Elías-Gutiérrez, 2004). This was explained by oxygen depletion and increased ammonia concentrations following removal which were deemed lethal for fish. Conversely, in some cases, macrophyte removal has had no significant effect on fish abundance (Bickel and Closs, 2009;Laughton et al., 2008;Unmuth et al., 1999;Wile, 1978) and actually increased survival and growth of some fish age classes (Holmes et al., 2019;Olson et al., 1998;Unmuth et al., 1999;Unmuth et al., 1998). Increases in larger fish classes were suggested to be most profound when partially removing the dense vegetation of submerged macrophytes, thus allowing fish to spread out into formerly unoccupied areas, likely causing less cannibalism and competition (Unmuth et al., 1999). Hence, making general conclusions on the consequences of macrophyte removal on fish population structure are complex, as fish community structure will depend greatly on the local context, such as the species present in the system, their interaction, the trophic state, together with removal practice (partial or full removal).

Hydraulic properties
Macrophytes provide important protection to riverbanks and the lake littoral zone and stabilise the sediment by reducing flow velocities (Kaenel et al., 1998;Verschoren et al., 2017;Wilcock et al., 1999). In rivers, macrophyte removal generally enhanced discharge capacity, where flow velocities increased by 30-40% (Old et al., 2014;Wilcock et al., 1999), water level was lowered by up to 50% (Kaenel et al., 2000) and the Manning roughness coefficient was reduced by 25-73% (Bal and Meire, 2009;Old et al., 2014;Vereecken et al., 2006;Verschoren et al., 2017). The most profound effects on hydraulics were found when macrophytes were removed from larger areas . We found no studies describing the consequences of macrophyte removal on hydraulic functioning in lakes. Removal of submerged and free-floating macrophytes will likely increase shore wave exposure and resuspension of sediment to the water column, as suggested in comparative studies (Horppila and Nurminen, 2005;James et al., 2004).

Sediment transport
Hydraulic transport and retention capacity of dissolved and particulate material is tightly coupled to physical properties and will be affected by macrophyte removal ). An increase in suspended sediment concentration has been reported downstream of removal sites with highest maximum peaks during or shortly after (hours to days) removal (Greer et al., 2017;Rasmussen et al., 2021). Elevated suspended sediment concentrations which are detrimental to fish have been measured at stations several km downstream of a removal, lasting up to 77 days (Greer et al., 2017). Similarly, in lakes, turbidity increased during or shortly after removal with the use of mechanical shredding (Alam et al., 1996;James et al., 2002).

Nutrient cycling
Aquatic macrophyte removal is likely to impact on nutrient cycling and metabolism in freshwater ecosystems (Bernot et al., 2006;Levi et al., 2015;O'Brien et al., 2014). Aquatic plants can play a significant role in nutrient cycling in oligotrophic ecosystems, but despite several attempts (Ensign and Doyle, 2005;O'Brien et al., 2014) we found no studies that successfully quantified the impact of macrophyte removal on nitrate, ammonium and phosphate cycling rates. O'Brien et al.
(2014) recorded a marginal increase in water phosphate concentration, but not ammonium or nitrate concentrations after plant removal. The retention of nutrients by quatic plants (net uptake) is generally very small in nutrient rich rivers relative to fluxes (House et al., 2001).

Ecosystem metabolism
Submerged and emergent macrophytes can be major contributors to primary production in freshwater ecosystems, thus influencing ecosystem metabolism and diel variation in oxygen concentration (O'Brien et al., 2014). Gross primary production (GPP) was found to decrease by up to 70% after removal in streams with high biomass of submerged macrophytes (Kaenel et al., 2000;O'Brien et al., 2014). However, the reduction in GPP may only last a short time, as partial recovery of GPP can be caused by enhanced growth of filamentous algae, stimulated by higher nutrient concentrations and increased light availability post removal (Kaenel et al., 2000). We found no studies in ecosystems dominated by free-floating macrophytes, however GPP increases following removal is likely as better light conditions may stimulate growth of submerged macrophytes or phytoplankton depending on nutrient availability in the system. Ecosystem respiration (ER) was found either to decrease (Kaenel et al., 2000;Madsen et al., 1988) or to stay unchanged (Carpenter and Gasith, 1978;O'Brien et al., 2014). The discrepancy in ER responses may be caused by differences in removal practices. Studies finding lower ER also report elimination of organic sediment retained in plant beds and epiphytic heterotrophs following plant removal (Kaenel et al., 2000). The effects on ecosystem metabolism and oxygen balance following macrophyte removal may be different in different ecosystems and more research is needed to understand how these relationships may differ.
We did not find any studies on how removal may impact other metabolic pathways, notably those involving green-house gases (N 2 O, CH 4 and CO 2 ), such as denitrification, methanogenesis or methanotrophy. This said, rooted aquatic plants with large radial oxygen loss in the root system can increase the coupling of nitrification-denitrification sediment fluxes (Kreiling et al., 2011) and oxidation of methane into carbon dioxide (Ribaudo et al., 2017). Floating plants may considerably lower dissolved oxygen in the water column (where respiration largely exceeds aquatic photosynthesis) and increase denitrification (Tall et al., 2011). Denitrification may not otherwise be significantly altered (Pinardi et al., 2009;Tall et al., 2011), unless denitrification is limited by the availability of organic carbon in the sediment (generally higher in aquatic plant patches). The decomposition of aquatic plant dead tissue in the sediment is known to produce methane ebullition in anoxia, predictable from plant water content and stoichiometry (Grasset et al., 2019).

The complexity of evaluating consequences of macrophyte removal
The overall effects of macrophyte removal for ecosystem structure and function are complex and making generalisations isnot straightforward. The shifts in species abundances and composition, as well as trophic interactions following removal are poorly understood. The derived effects of macrophyte removal, including alterations in biochemical cycles and hydraulic conditions, may likely stimulate further changes in food-web structure. Moreover, current studies have very distinct contexts e.g. macrophyte species removed, removal method, ecosystem types, trophic states, time of removal, size of study and study design. Replication within each combination is infrequent or completely lacking (Appendix A). Due to the scarcity of studies on the consequences of macrophyte removal with regard to different ecosystem properties, performing an unbiased formal meta-analysis of previous work is unfeasible. Furthermore, the effect reported in the studies on single ecosystem properties does not necessarily reflect the direct influence of removal, as indirect effects, such as inter-relationships with other ecosystem properties are not considered and the underlying drivers for the potential change remain unaddressed. This suggests that a new approach to evaluating the consequences of macrophyte removal at the ecosystem level is needed.
7. Synthesising effects of macrophyte removal on ecosystem levelan example using a Bayesian network approach Our review of the existing literature showed that consequences of macrophyte removal have mainly been documented through short-term studies evaluating single ecosystem properties without considering the underlying mechanisms and interrelationships between ecosystem properties (Fig. 2D). Moreover, the results were highly dependent on the context of each study, making meaningful quantitative comparisons across studies very difficult. We therefore chose a Bayesian network (BN) approach to identify important consequences of macrophyte removal. These networks can be used to explore and understand the interrelationships between environmental factors and their influence on the response variable (end-point) of interest (Stewart-Koster et al., 2010), thus BNs can be helpful in management decisions of freshwater ecosystems with mass development of aquatic macrophytes. A Bayesian network (BN) is a model based on probabilities and consists of the main factors of a system (nodes) and their conditional dependencies illustrated by arrows connecting the nodes (Stewart-Koster et al., 2010) (Fig. 3). The network is quantified by conditional probability tables (CPTs) for each node and can consist of observed data or expert knowledge (Korb and Nicholson, 2004;Pollino et al., 2007). Here we use the BN approach as a first attempt to synthesise potential short-term effects of macrophyte removal from different freshwater ecosystems for a specified end-point. The structure of the network is important to guide the collection of measurements in specific case studies, so that node states and CPTs could be derived from measurements prior to and following removal events. We used the NETICA software v. 6.07 (Norsys, 2005) to construct the BN. The CPTs used in the network were based on general (qualitative) knowledge for illustrative purposes. Detailed information on the construction of the BN and the conditional probability tables can be retrieved in Supplementary Information 1.

Description of mechanical macrophyte removal network
One major short-term consequence of cutting aquatic plants is to increase the risk of phytoplankton bloom (Kuiper et al., 2017). We illustrate how BN can help us quantify this risk through an understanding of causal mechanisms. Phytoplankton growth is controlled by changes in resource supply (light and nutrient availability) and disturbance frequency (flow and trophic cascades) (Fig. 3A) (Bernes et al., 2015;Reynolds, 2000).
In the BN, the availability of resources is a function of three predictor variables: light, nutrient and bioturbation (benthic fish foraging). Light is a function of plant removal and ecosystem. Plant removal indicates the proportion of macrophyte removal (i.e. none, partial or full) and ecosystem represents either lake dominated by submerged or floating macrophytes or rivers dominated by submerged macrophytes. Nutrient loading represents the nutrient supply in the system and has three rates (low, moderate and high). Benthic fish foraging is an inverse function of the availability of epiphytic invertebrates and indicates the proportion of fish feeding on benthic invertebrates and thus a higher risk of bioturbation and associated nutrient release to the water column Fausch et al., 1997). The availability of epiphytic invertebrates is a function of plant removal and ecosystem.
The variable disturbance is a function of flow and zooplankton, which describes the hydrological disturbance (including water residence time) and the potential grazing pressure. Flow is a function of two predictor variables, plant removal and ecosystem and has three categories (low, moderate and high) representing hydrological disturbance conditions for zooplankton development. Zooplankton abundance is a function of flow and planktivorous fish. High zooplankton abundance results from hydrological stability and low predation pressure (i.e. low flow and low planktivorous fish). Planktivorous fish abundance preying on zooplankton is a function of piscivorous fish predation itself dependent on piscivorous fish presence and plant removal. Finally, plant removal indicates the proportion of macrophyte removal (none, partial or full) dependent on desired ecosystem services, e.g. full removal benefits recreational users, reduce the risk of flooding or eradiate specific target species that may be invasive to the area (Baattrup- Pedersen et al., 2003;Verhofstad and Bakker, 2019), partial removal can benefit fisheries (Bickel and Closs, 2009) and no removal can be beneficial for nutrient retention and birds (Klaassen and Nolet, 2007). In this BN each node in the model contains two to three states.

Evaluating short-term consequences of mechanical removal using a Bayesian network approach
In this hypothetical example, we used the BN model described above to synthesise and illustrate potential short-term effects of macrophyte removal on different ecosystem properties. The a priori assumption for the BN model is that the ecosystem of interest has a mass development of aquatic macrophytes conflicting with human interest, such as prevention of flooding and/or recreational activities. It is possible to specify conditions by setting the probability of a given state in the nodes and the nodes are then updated via the CPTs (Stewart-Koster et al., 2010). A given BN can therefore be adjusted to local conditions. Let's assume a lake is filled with submerged macrophytes, has high nutrient loading and hosts piscivorous fish, by setting the probabilities to 100% of the states in the respective nodes (ecosystem, nutrient loading and piscivorous fish) (Fig. 3A). What is the risk of phytoplankton bloom if macrophytes were partially removed? The probability of high phytoplankton abundance following partial removal of submerged plants is 59.4%, due to high resources (50%) and despite high disturbances (50%, zooplankton grazing and removal of plant protection). By only changing the management practice to full removal (plant removal; Full 100%) in the BN, the probability of high phytoplankton abundance now increases to 100% (Fig. 3B). More interesting are the effects on the trophic cascade in the two BNs. For BNs with partial and full plant removal, the probabilities for moderate Epiphytic invertebrates are 75% and 0%, high Planktivorous fish 50% and 0% and high zooplankton 0% and 100% respectively ( Fig. 3A-B). Thus, the choice of management practice can have very different implications for ecosystem structure. Again, we emphasise that the probabilities were obtained by expert knowledge and are used for illustrative purpose only. The states of the nodes and conditional probability tables in the BN should be based on values derived from the system under study for more realistic probabilities.
In addition, the BN can also be used to identify possible options for managing mass development of aquatic macrophytes. Assuming the same conditions as the previous example but allowing the variable nutrient loading, a goal for managers could be to reduce the risk of high phytoplankton abundance when removing the macrophytes in order to ensure recreational activities such as swimming. Further specifying the conditions by setting phytoplankton abundance to low and swimming to 100%, the BN suggests that the only option for this is to choose partial plant removal (Fig. 3C, plant removal; Partial 100%) and to reduce the nutrient loading to either low or medium (Fig. 3C, nutrient loading, Low 52.5%, Medium 27.5%).
These examples illustrate how BNs can be used to assess effects of mechanical macrophyte removal in a holistic way as the interrelationships between ecosystem properties are also considered and not only the direct effects on single ecosystem properties. The BN approach could help engage the stakeholders in conversation.

Research needs
In the future, mass development of aquatic macrophytes will likely increase in many freshwater ecosystems interfering with human activities and potentially resulting in more frequent removal (Hussner et al., 2017;Verhofstad et al., 2017). Currently, no studies have evaluated the effects of macrophyte removal on interrelated ecosystem properties for the whole ecosystem, thus a holistic evaluation of the consequences of macrophyte removal is lacking Appendix A. Considering the social and economic importance of freshwater ecosystems and knowing the important role of macrophytes, there is an urgent need for more research on macrophyte removal in order to understand the implications for whole ecosystem structure, functions and services. This will require large scale experiments covering different macrophyte species, ecosystem types and geographical gradients, where both parameters on ecosystem structure and functions are estimated. Consistent and comparable data can then be used to make general conclusions on consequences of macrophyte removal. This would enable management decisions to be based on balanced knowledge rather than just the prevailing negative perception of macrophytes. However, the long-term consequences of macrophyte removal on other ecosystem properties have received little attention and few studies exists, meaning that more research is needed to understand these long-term effects.

CRediT authorship contribution statement
KTH, BDE and SSC conceived the idea for the paper. KTH did the literature search and extracted data from relevant publications. KTH led the writing of the paper and made the figures. All authors participated in writing and critically read the final MS. The authors have no conflicts of interest to declare.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgement
This study was funded by the Research Council of Norway (297202/ E10), the German Federal Ministry of Education and Research (033WU005), the French Agence National de Recherche (N°ANR-18-IC4W-0004-06), the South African Water Research Commision (K5/ 2951), and the Fundação Araucária in Brazil (N°186/2019) for funding of MadMacs (Mass development of aquatic macrophytes -causes and consequences of macrophyte removal for ecosystem structure, function, and services) in the frame of the collaborative international consortium of the 2017 call of the Water Challenges for a Changing World Joint Programme Initiative (Water JPI). Additional funding was provided by Krypsivprosjektet på Sørlandet, NIVA and NMBU to support PhDstudent KTH in Norway. Finally, we would like to thank Joanna Kemp for linguistic corrections and two anonymous reviewers for improving the manuscript.