Introducing Life Cycle Assessment in Costs and Beneﬁts Analysis of Vegetation Management in Drainage Canals of Lowland Agricultural Landscapes

: Nitrate pollution remains an unsolved issue worldwide, causing serious e ﬀ ects on water quality and eutrophication of freshwater and brackish water environments. Its economic costs are still underestimated. To reduce nitrogen excess, constructed wetlands are usually recognized as a solution but, in recent years, interest has been raised in the role of ditches and canals in nitrogen removal. In this study, we investigated the environmental and economical sustainability of nitrogen removal capacity, using as a model study a lowland agricultural sub-basin of the Po River (Northern Italy), where the role of aquatic vegetation and related microbial processes on the mitigation of nitrate pollution has been extensively studied. Based on the Life Cycle Assessment (LCA) approach and costs and beneﬁts analysis (CBA), the e ﬀ ectiveness of two di ﬀ erent scenarios of vegetation management, which di ﬀ er for the timing of mowing, have been compared concerning the nitrogen removal via denitriﬁcation and other terms of environmental sustainability. The results highlighted that postponing the mowing to the end of the vegetative season would contribute to bu ﬀ ering up to 90% of the nitrogen load conveyed by the canal network during the irrigation period and would reduce by an order of magnitude the costs of eutrophication potential.

Multiple experimental pieces of evidence collected in the same canals indicated that denitrification occurring in the vegetated portions of the canal network accounts for the majority of in-stream N removal [27,28]. Subsequent upscale studies in the same hydraulic-regulated and simplified watershed have shown that canal management may deeply affect the balance between N sources and sinks and, thus, also determine the quality of waters delivered to the coastal areas [6,29].
Nitrate pollution has serious environmental effects and may also cause threats to human health [30]. There is evidence of the causal effects of NO 3 − in drinking water on methemoglobinemia and increased incidence of colon cancer [31]. On the other hand, several efforts have been made to estimate the economic losses and potential societal costs due to human welfare and from human-induced environmental impacts, such as eutrophication [32]. Such costs, which are not part of any prices paid or direct compensation, are defined as externalities and might cause market distortions, encouraging activities that induce private benefits even if they are costly to the environment and society [33]. Currently, there is a lack of large consensus on accounting for the present and future values of ecosystem functions and potential economic costs deriving from their loss, even though the most recognized methodology to evaluate environmental and human health costs was to calculate the willingness to pay (WTP) to prevent damages and protect the environment from the perspective of society as a whole [34]. This was recently applied to the N social costs and benefits analysis by Brink and van Grinsven [35] and Keeler et al. [36].
To deepen the analysis of the conservative management of aquatic vegetation, we have investigated its economical and environmental sustainability by the means of a costs and benefits analysis based on Life Cycle Assessment (LCA). LCA is an accepted method that was used for years to evaluate the potential environmental impacts of a product or a service concerning its function and throughout its whole life cycle [37], including wastewater treatment plants and different nutrient removal technologies [38][39][40]. LCA can quantify the impacts, determining the aspects which influence the environmental performance most, and identifying possible improvement potentials. To the best of the authors' knowledge, this is the first attempt to link the results of environmental and economic analysis to evaluate different options of vegetation management in the canal network. In the present study, LCA has been used to compare the effectiveness of two scenarios on N abatement in the canal network of the BVN basin. The potential of the canal network to mitigate NO 3 − pollution was quantified for two levels of vegetation maintenance, i.e., 5%, and 50% of the network total length. The 5% scenario corresponds to the present condition, while the more conservative 50% scenario is achieved only by postponing the mowing timing, from the middle of summer (present management) to the end of the vegetative season, in October. We hypothesize that this environmentally friendly choice, which may be sustainable also for the hydraulic security since it interests only half of the hydrological network length, may offer new management opportunities, reducing both the costs and environmental impacts and effectively contributing to mitigate NO 3 − loads.

Study Area
The BVN basin, accounting for~2600 km 2 (Figure 1a), overlaps the administrative territory of the Ferrara Province (Emilia-Romagna Region, Northeastern Italy) and is bordered by the Adriatic Sea and the embankments of the Po, Reno and Panaro rivers. The basin is a completely flat territory, highly fertile and well served by irrigation, with all features favoring intensive farming, making agriculture the dominant land use (>85% of the basin area).

Vegetation Management in the Canal Network
The Consorzio di Bonifica Pianura di Ferrara is one of the largest Land Reclamation Consortia in Italy for both contribution dimension and the extent of reclamation work (256,733 ha). It operates over a land district approximatively coinciding with the Ferrara provincial territory (263,500 ha), one of the most fertile zones of the Po Plain. Maintenance operations of the canal network (i.e., dredging, sediment removal, mechanical mowing, bank slope reinforcement, channel reshaping) are performed on a regular basis by the Consortium to maximize the conveyance capacity of the network, ensuring water drainage and water supply for irrigation. Similar to other irrigated watersheds [13,15], aquatic vegetation is routinely removed from the banks and the canal bottom using flail mowers, a standard The water used in the basin for irrigation is almost entirely derived from the Po river via mechanically controlled water diversion points and distributed to the arable lands through a capillary network of open-earth canals and ditches. The drainage of waters is artificially regulated by a complex system of main collectors converging towards several drainage plants that pump excess water out of the basin. The drainage system, managed by Ferrara Land Reclamation Consortium (Consorzio di Bonifica Pianura di Ferrara) and comprised of a network of 4208 km of canals (Figure 1b), 170 pumping stations and more than 13,000 hydraulic adjustment structures, is constantly operative to ensure hydraulic safety conditions for the Ferrara territory and water supply for agriculture during the irrigation period (from April to September).

Vegetation Management in the Canal Network
The Consorzio di Bonifica Pianura di Ferrara is one of the largest Land Reclamation Consortia in Italy for both contribution dimension and the extent of reclamation work (256,733 ha). It operates over a land district approximatively coinciding with the Ferrara provincial territory (263,500 ha), one of the most fertile zones of the Po Plain. Maintenance operations of the canal network (i.e., dredging, sediment removal, mechanical mowing, bank slope reinforcement, channel reshaping) are performed on a regular basis by the Consortium to maximize the conveyance capacity of the network, ensuring water drainage and water supply for irrigation. Similar to other irrigated watersheds [13,15], aquatic vegetation is routinely removed from the banks and the canal bottom using flail mowers, a standard procedure that has been in place in the BVN basin since the late 1980s and is carried out every year independently of a wet or dry summer season (Consorzio di Bonifica Pianura di Ferrara, personal communication). The mowing of in-stream emergent vegetation is performed once a year (in the middle of the summer, i.e., end of July, on 30% of the total canal network length) in canals with low flood risk, or twice a year (in June and October, on 37% of the total canal network length) in canals with high hydraulic risk. The remaining 33% is not subjected to mowing, since it accounts for canals with deep water column and high turbidity that prevent macrophyte development (Consorzio di Bonifica Pianura di Ferrara, personal communication).

Calculation of the N Removal Capacity of the Canal Network
The potential capacity of the canal network to remove N via denitrification was predicted for the irrigation period (i.e., from April to September) under two scenarios: (1) current condition, where vegetation is present throughout the vegetative season in 5% of the total network length (95% is considered completely unvegetated), and (2) conservative management, where vegetation is maintained throughout the vegetative season in 50% of the total network length by postponing the mowing operations to the end of the vegetative season. The end of the vegetative season corresponds to the end of the irrigation period when most of the canals switch from the primary function of irrigation to one of drainage. The mowing of in-stream vegetation is performed before the autumn period of intense rainfall, to maximize the discharge capacity of the drainage network.
The N removal capacity of the canal network under the two scenarios, i.e., actual and conservative, was estimated by employing a detailed upscale of extensive datasets of field measurements (i.e., water quality monitoring of the canal network, denitrification rates measured in vegetated and unvegetated canals) previously acquired in the study area [6,12,27,28]. Denitrification rates in the sediments of unvegetated canals (Dr UV ) were calculated by applying the Christensen model [41], proposed for NO 3 − -rich agricultural waterways and previously tested in bare sediments of shallow slow-flow aquatic ecosystems of the Po River Plain, including several selected unvegetated canals of the BVN basin [6,42,43]. A good correlation was obtained between modeled and experimental rates along the NO 3 − concentration range 1-7 mg N L −1 , overlapping the NO 3 − availability detected in NO 3 − -rich agricultural streams employed for the validation of the model developed by Christensen and collaborators [41]. The denitrification rates of NO 3 − diffusing from the water column to the anoxic sediments were calculated according to the following equation: where: The stations were located on the canal network of the Ferrara province and belonged to the official surface-water-monitoring network of the Emilia-Romagna Regional Agency for the Environmental Protection (ARPAE). The diffusion-reaction model by Christensen [41] was applied to all monthly ARPAE surveys for which measurements of water temperature, NO 3 − , and O 2 concentrations were concomitantly available, thus a dataset of daily denitrification rates was obtained for each month, from April to September ( Figure S1, Supplementary Materials). Denitrification rates in vegetated canal sediments (Dr V ), expressed per unit of canal length (kg N km −1 day −1 ), were calculated as a function of water NO 3 − availability (mg N L −1 ) by employing a predictive relationship previous developed on a large dataset of experimental measurements of denitrification rates acquired in several vegetated canals of the studied area [6]: The equation was applied to all monthly ARPAE surveys for which measurements NO 3 − concentration were available, thus a dataset of daily denitrification rates was obtained for each month, from April to September ( Figure S2, Supplementary Materials). Daily denitrification rates calculated for unvegetated and vegetated sediments were extended, being assumed constant, to the number of days per month (from April to September) and the unvegetated and vegetated canal surfaces, respectively, under the two different management scenarios. To determine the likely variation in the NO 3 − removal capacity of the entire canal network, the interquartile range (first and third quartiles as inferior and superior extremes) of each monthly dataset of denitrification rates was considered. Monthly contributions, calculated for unvegetated and vegetated canals, were finally summed to obtain the total N amount removed in each month and at the annual scale by the whole canal network under the two different management scenarios.

Scope Definition, Functional Unit and System Boundaries
The N removal capacity of the canal network, predicted under the two scenarios of vegetation management, was compared to the N load exported, during the vegetative season, by the canal network to the coastal lagoons and coastal waters, previously estimated by Castaldelli et al. [21]. This N load was obtained by combining measured discharge and N concentration dataset and represents the N output in the LCA scheme reported in Figure 2. The N input entering the canal network during the irrigation period was calculated by summing up the N load exported from the basin and the N removal capacity predicted under the current vegetation management. The scope of the LCA was to compare the environmental impacts of the current and the conservative vegetation managements on the N load conveyed by the same canal network and to provide potential environmental impacts for the quantification of costs and benefits. In the current management scenario, aquatic vegetation mowing is routinely carried as mentioned (Section 2.2), whereas in conservative management scenario, in 50% of the canal network vegetation, mowing is performed only once a year, at the end of the vegetative season, in October, while the remaining 50% is managed following the current practice, with one or two mowing interventions, according to the hydraulic needs.
The N removal capacity of the canal network, predicted under the two scenarios of vegetation management, was compared to the N load exported, during the vegetative season, by the canal network to the coastal lagoons and coastal waters, previously estimated by Castaldelli et al. [21]. This N load was obtained by combining measured discharge and N concentration dataset and represents the N output in the LCA scheme reported in Figure 2. The N input entering the canal network during the irrigation period was calculated by summing up the N load exported from the basin and the N removal capacity predicted under the current vegetation management. The scope of the LCA was to compare the environmental impacts of the current and the conservative vegetation managements on the N load conveyed by the same canal network and to provide potential environmental impacts for the quantification of costs and benefits. In the current management scenario, aquatic vegetation mowing is routinely carried as mentioned (Section 2.2), whereas in conservative management scenario, in 50% of the canal network vegetation, mowing is performed only once a year, at the end of the vegetative season, in October, while the remaining 50% is managed following the current practice, with one or two mowing interventions, according to the hydraulic needs.
In LCA studies, the functional unit is a key element because it is the reference to which the inputs and outputs can be related and enables comparison of two or more different systems, whereas the system boundaries are selected to determine which unit processes have to be included in the LCA study. For both management options, the functional unit selected was 1 km of canal network (average depth, during the vegetative season, which is overlapped with the irrigation one, 0.5 m; average width of the bottom 3 m) and the system boundaries covered mechanical mowing operations on canals, diesel and lubricant oil production and consumption, vehicle maintenance and emissions in air related to fuel combustion ( Figure 2). Vehicle production and transport from vehicle garages to canals are not included in the analysis, as well as fuel consumption for the other maintenance operations of canals (i.e., dredging, resectioning).  In LCA studies, the functional unit is a key element because it is the reference to which the inputs and outputs can be related and enables comparison of two or more different systems, whereas the system boundaries are selected to determine which unit processes have to be included in the LCA study. For both management options, the functional unit selected was 1 km of canal network (average depth, during the vegetative season, which is overlapped with the irrigation one, 0.5 m; average width of the bottom 3 m) and the system boundaries covered mechanical mowing operations on canals, diesel and lubricant oil production and consumption, vehicle maintenance and emissions in air related to fuel combustion ( Figure 2). Vehicle production and transport from vehicle garages to canals are not included in the analysis, as well as fuel consumption for the other maintenance operations of canals (i.e., dredging, resectioning).

Life Cycle Inventory (LCI) and Impact Calculations
The inventory data are reported in Table 1 and were collected from a dedicated questionnaire submitted to the Ferrara Land Reclamation Consortium in 2020. Data collected for the present study Water 2020, 12, 2236 7 of 15 covered the entire canals network of the hydrographic basin. For cleaning banks and canals, hydraulic mulching heads mounted on tractors and excavators with bucket-mowers are usually used, respectively. A type of trimmer equipped with a rotating cab and rotating mulching head was recently introduced for cutting vegetation in about 30% of the network length, reducing the number of forth and back passes necessary to complete the cutting, and consequently the amount of fuel consumed. After mowing operations, the residues (i.e., aboveground vegetation biomass) are usually left on the banks, so waste management has been neglected. All background data for fuel and engine oil productions, as well as for vehicle maintenance, were derived from Ecoinvent TM v3.6 (Ecoinvent Association, Zurich, Switzerland) [45] and Agribalyse TM v1.3 (Ademe, Angers, France) [46] databases. The ReCiPe Midpoint (H) v1.11 (PRé Consultants, Amersfoort, The Netherlands) method [47] and the open source package OpenLCA TM v1.8 (GreenDelta, Berlin, Germany) were used for the impact assessment and the overall LCA modeling, respectively. Emissions to air were calculated directly by the software based on input data and mainly derived from diesel combustion. Impact categories are Eutrophication potential (EP), Global Warming Potential (GWP), Photochemical Oxidant Formation Potential (POFP), terrestrial Acidification potential (AP), Particulate Matter Formation potential (PMFP), Human Toxicity Potential (HTP) and Marine Aquatic Eco Toxicity Potential (MAETP). In this study, allocation was not necessary because the sole N flux was considered.

Uncertainty Analysis
The life cycle inventory (LCI) data were evaluated according to the semi-quantitative "pedigree matrix" [48], where all input data are scored (1 to 5, where 1 is better) based on the data quality features of reliability (sampling methods and verification procedures), completeness (statistical representativeness of the datum and periods for data collection), temporal, geographic and a further technological correlation (for data used outside its proper context). In Ecoinvent TM , an uncertainty factor is assigned to each of the five data quality indicators [49] to calculate the total uncertainty of the result of each impact category, expressed as a 95% confidence interval. The pedigree matrix and the uncertainty factors based on the data quality ratings were used for the Monte Carlo simulation (1000 runs).

Costs and Benefits Analysis
The results of the LCA were used to calculate the costs related to the different capacity of the ecosystem, i.e., the canal network under study, to remove N and thus mitigate the eutrophication potential, and to the potential release of pollutants in environment, related to the mechanical operations, Water 2020, 12, 2236 8 of 15 in the two scenarios. To identify and analyze the annual costs for the environment, the damage costs concept was used [50] which takes into account possible outcomes to human health, linked to the release of pollutants. We assumed the definitions given by Mueller et al. [51] for ecological damage costs such as the loss of value of ecosystem services, and by Brink and van Grinsven [35] for human health costs, as the sum of market costs (i.e., medical treatments, productivity losses), non-market costs (i.e., individual's WTP to avoid the risk of pain and disease) and costs for life expectancy reduction caused by acute or prolonged inhalation of pollutants. The monetary values of environment and human health damages and the corresponding source used in this study are reported in Table 2. The potentiality of the canal network to remove N, quantified for the two levels of vegetation maintenance (5% and 50%), was translated into an avoided cost. We used the replacement cost method for the economic valuation of the N removal function of the canal network by quantifying the cost that would be avoided in obtaining an equivalent N removal capacity by employing constructed wetlands [54,55]. Cost data for N removal in constructed wetlands (expressed in term of € per kg of N removed) are scarce in Italy [56,57], so we adopted a literature range of costs (range 2-50 € kg −1 N, average value 7€ kg −1 N) of reducing N by means of surface flow constructed wetlands acquired in European agricultural landscapes with similar climate and anthropogenic pressures to the study area [35,58,59].

Results and Discussion
The mowing of aquatic vegetation in the canal network determines the loss of multiple interfaces in water (biofilms) and benthic compartments (oxic-anoxic microniches in the rhizosphere) representing active denitrification hotspots responsible for the canal depuration capacity against NO 3 − [3, 8,11].
The N removal performed by the canal network of the BVN basin under the current vegetation management was predicted in 385 ± 55 t N year −1 , of which >80% ascribed to unvegetated canal sediments ( Figure 3). Summing up this N removal capacity to the N output to the coastal zone (897 ± 210 t N year −1 ) previously estimated by [21], the N input of~1280 t N year −1 was quantified, representing the N load generated the BNV basin during the spring-summer months. The predicted N abatement in the canal network would be by about three times, increasing up to 1145 ± 177 t N year −1 in the case of postponing vegetation mowing to the end of the vegetative season in 50% of the network length ( Figure 3). This increased capacity, almost completely due to the interfaces created by the presence of aquatic vegetation, would contribute to buffering~90% of the N load conveyed by the canal network during the irrigation period.
The results of LCAs of current and conservative scenarios are presented in Table 3.
representing the N load generated the BNV basin during the spring-summer months. The predicted N abatement in the canal network would be by about three times, increasing up to 1145 ± 177 t N year −1 in the case of postponing vegetation mowing to the end of the vegetative season in 50% of the network length (Figure 3). This increased capacity, almost completely due to the interfaces created by the presence of aquatic vegetation, would contribute to buffering ~90% of the N load conveyed by the canal network during the irrigation period. The results of LCAs of current and conservative scenarios are presented in Table 3. Except for EP, which is affected by N loads, mostly as NO3 − , carried by canals, all other impact categories are caused by the emissions of tractors. EP strongly depends on the NO3 − loads transported by the canal network, and so the adoption of the conservative scenario, corresponds to a reduction of about 90% of the risk of eutrophication by the BVN basin in the terminal water bodies, i.e., the lagoons and coastal waters of the Po Delta. Thus, EP accounts for the environmental disturbance due to NO3 − pollution and the relative risk connected to phytoplankton and macroalgae blooms. These  pollution and the relative risk connected to phytoplankton and macroalgae blooms. These phenomena may lead to dissolved oxygen depletion and, ultimately, to long-lasting anoxia and dystrophy [23,60]. The other impact categories are mainly influenced by emissions to air caused by diesel combustion in tractors during vegetation mowing operations. In particular, nitrous oxides contribute to POFP, together with volatile organic compounds (NMVOC), both acting as precursors of the ground-level ozone layer, a harmful air pollutant being the main ingredient in smog, because of its effects on people and the environment [61]. Sulfur oxides contribute to AP, which measures the potential occurrence of atmospheric acidification, whereas human and environmental toxicity is due to traces of heavy metals (i.e., cadmium) in diesel-exhausted gases. GWP accounts for climate change deriving from the emissions of greenhouse gases in the air. GWP and EP values in the two scenarios are statistically different, unlike the other impact categories. In the study case, GWP is lower in the conservative scenario, because the new vegetation removal management in 50% of the canals network consists of a reduction in tractors' interventions and, consequently, lower fuel consumption. However, it has to be remarked that the effective benefit of the conservative strategy does not come from diminishing fuel consumption and emissions, but consists mainly in the EP reduction.
Multiplying unitary costs reported in Table 2 with the impact values as results of LCA in Table 3, the potential costs of damages on the environment and human health due to air emissions of fuel combustion by tractors, used for vegetation removal in the two scenarios were estimated ( Table 4). As a result, in the current scenario, each kilometer of canal theoretically charges society and the environment almost 52 EUR per year, slightly more than the 44 EUR of the conservative ones. For the entire canals network of the BVN basin, this means an overall potential extra-cost per year of about 206,000 EUR in the current scenario. These costs, which are comparable to those for the management of all structures of our societies, such as roads, rivers, and green areas, are still theoretical since they are not attributable to any public or private entity. Moreover, in our study case, the very low human density and the presence of a vegetation cover in the Spring-Summer season make the term almost irrelevant in the analysis of environmental sustainability, if compared to the same amount of emissions in an urban area. In the conservative scenario, potential costs due to air emissions of fuel combustion can be estimated at about 175,000 EUR per year, a value not markedly lower than in the previous case. On the other hand, between the two options of vegetation management (current and conservative) the main difference is not so much in terms of the number of mowing interventions or fuel consumption as it is in terms of the relevant benefits on the N balance in the canal network that could be achieved with the sole postponement of vegetation mowing to the end of the vegetative season.
The current denitrification capacity of the canal network was estimated at 385 t N year −1 , which can save 243 EUR of potential costs calculated on the base of the eutrophication potential, per kilometer of canal (Table 5), corresponding to about 950.000 EUR, calculated on the overall BVN basin. However, in the conservative scenario, the N abatement in canals has been estimated at 1145 t N year −1 , more than three times higher than in the current scenario, corresponding to 722 EUR per km of canals, or circa 3 million EUR on the entire basin of avoided costs for the eutrophication potential. This saving of more than 2 million EUR, passing from the current to a more conservative scenario, may be achieved by simply postponing the mowing period.
It is worth noting that the more effective NO 3 − removal is purely the net-benefit of an ecosystem, provided by the interlinked action of aquatic emergent vegetation and microbial communities, which can be achieved without any further investment in equipment or machinery from public or private bodies. The only term which cannot be accounted for now, because it needs dedicated research, is the increase in hydraulic risk in the case of extreme meteoric events, whose occurrence is increasing in the study area due to global climate change. However, in the studied basin, with very low population density and in the season of interest, i.e., spring and summer, this risk is reasonably low and likely manageable with low costs. In the literature, several experiences are reported regarding the strategies for mitigating NO 3 − pollution in agricultural landscapes, mostly focused on limiting the losses from the fields, such as limitations in the quantity and timing of fertilizers' distribution [7,62], or at the interface between the mainland and transitional waters, such as the construction of artificial wetlands [63]. None of them is focused on the potential costs that could be avoided through exploiting the natural removal capacity of the canal network. The only data available on costs of N removal are related to the construction of artificial wetlands. Moreover, the cost of building and maintaining constructed wetlands to reduce NO 3 − emissions to superficial waters is largely site-specific, making any kind of comparison particularly difficult. Based on average data acquired in European agricultural landscape, it has been estimated as, on average, 7 EUR kg −1 N [35,64,65], which would represent the cost for the society to achieve water depuration, to be added to the current cost for canal and ditch management in the basin. The denitrification capacity of the canal network predicted for the conservative vegetation scenario would equal an avoided cost of~8 million EUR per year (~2000 EUR per km of canal per year) obtaining an equivalent N abatement by employing constructed wetlands ( Figure 2). The decrease in EP in the conservative management scenario would bring a substantial reduction in NO 3 − loads, very likely leading to the achievement of an ecological good state in the coastal waters, fulfilling the requirements of the WFD. In monetary terms, this decrease in EP would equal a net saving of about~5 million with respect to the current scenario ( Figure 2), otherwise needed for the N abatement by employing artificial wetlands.

Conclusions
The results of this study highlighted that the implementation of conservative management practices of in-stream vegetation significantly improves the N removal capacity of the canal network, resulting in an increased net benefit for the society. This can be seen as one of the positive externalities, i.e., multiple environmental benefits for the society, that water-managing authorities provide through their operations in the canal network. It equals also to a net increase in sustainability in the management of an agricultural basin, which is quantitatively more important for the reduction in the eutrophication potential but also for the decrease in atmospheric emissions. We have to point out that the economic analysis still needs to be implemented with the evaluation of the works needed to maintain hydraulic security in the conservative scenario. However, the economic quantification done in this paper is the first step for including this paramount ecosystem function (i.e., N removal) in the overall management budget of these environments, in addition to the conventional hydraulic functions.