Ambio 2020, 49:1820–1837 https://doi.org/10.1007/s13280-020-01345-5 ENVIRONMENTAL EFFECTS OF A GREEN BIO-ECONOMY Efﬁciency of mitigation measures targeting nutrient losses from agricultural drainage systems: A review Mette Vodder Carstensen , Fatemeh Hashemi , Carl Christian Hoffmann, Dominik Zak, Joachim Audet, Brian Kronvang Received: 18 December 2019 / Revised: 5 April 2020 / Accepted: 5 May 2020 / Published online: 3 June 2020 Abstract Diffusive losses of nitrogen and phosphorus resulting in water quality problems and ecosystem degra- from agricultural areas have detrimental effects on dation worldwide (Kronvang et al. 2005; Diaz and freshwater and marine ecosystems. Mitigation measures Rosenberg 2008; Steffen et al. 2015). The intensiﬁcation treating drainage water before it enters streams hold a high and expansion of agriculture during the past decades have potential for reducing nitrogen and phosphorus losses from led to a drastic increase in nutrient loss from agricultural agricultural areas. To achieve a better understanding of the areas, as well as changes in land use. Wet landscapes have opportunities and challenges characterising current and been systematically drained to enable anthropogenic new drainage mitigation measures in oceanic and activities such as food production (Skaggs and van Schil- continental climates, we reviewed the nitrate and total fgaarde 1999). However, in addition to water, drainage phosphorus removal efﬁciency of: (i) free water surface systems also transport nutrients rapidly to surface waters, constructed wetlands, (ii) denitrifying bioreactors, (iii) thereby lowering the natural retention capacity of catch- controlled drainage, (iv) saturated buffer zones and ments. Thus, engineered ecotechnologies designed to (v) integrated buffer zones. Our data analysis showed that intercept and reduce nitrogen (N) and phosphorus the load of nitrate was substantially reduced by all ﬁve (P) losses from agricultural drainage systems have emerged drainage mitigation measures, while they mainly acted as over the last decades with the aim to improve water quality sinks of total phosphorus, but occasionally, also as sources. (Mitsch and Jørgensen 1989). Substantial changes in land The various factors inﬂuencing performance, such as use can also be expected in the future when addressing design, runoff characteristics and hydrology, differed in energy and food security such as transformation of the the studies, resulting in large variation in the reported society to a bio-economy (Marttila et al. 2020; Rakovic removal efﬁciencies. et al. 2020). Water quality and quantity are key elements in such a transformation, thus the development and imple- Keywords Agricultural drainage systems mentation of drainage mitigation provide valuable oppor- Catchment management Meta-analysis tunities for innovation in future bio-economies. Besides Mitigation measures Nutrient reduction Water quality reducing nutrient losses to surface water, these measures can be designed to provide multiple ecosystem services, such as water storage and biomass production, as well as INTRODUCTION recycling of nutrients. Drainage mitigation measures reduce the transport of N The high intensive agricultural production dominating from drainage systems primarily by enhancing denitriﬁca- parts of the world, such as Western Europe and North tion (O’Geen et al. 2010), i.e. the process by which nitrate America, is one of the main causes of eutrophication dissolved in water is converted to atmospheric nitrogen (Knowles 1982). Denitriﬁcation requires anoxic conditions, electron donors and availability of organic carbon. If these Electronic supplementary material The online version of this requirements are met, the rate of the denitriﬁcation is article (https://doi.org/10.1007/s13280-020-01345-5) contains sup- mainly controlled by temperature and the hydraulic plementary material, which is available to authorized users. The Author(s) 2020 www.kva.se/en Ambio 2020, 49:1820–1837 1821 retention time (HRT), which is inversely proportional to and sedimentation, while the shallow vegetation berms the water ﬂow rate (Kadlec and Knight 1996; Hoffmann supply organic carbon. Furthermore, FWS can capture et al. 2019). The water ﬂow from subsurface drainage surface runoff if located downhill. Free water surface systems is driven by precipitation and snowmelt and, thus, constructed wetlands are mostly established in areas with varies greatly on a temporal as well as a spatial scale low permeable soils, and if not, they are often sealed with (Skaggs and van Schilfgaarde 1999). This challenges the non-permeable layers such as clay membranes to prevent performance of drainage mitigation measures in some parts seepage to the groundwater. Construction of wetlands for of the world, for instance the Nordic countries, where high diffusive pollution control began in the late 1980s with the loading rates of nitrate often occur during autumn to early aim to create simple systems mimicking the processes spring when the water temperature and denitriﬁcation rates occurring in natural wetlands (Mitsch and Jørgensen 1989; are low. Therefore, we were particularly interested in Fleischer et al. 1994). Multiple types of FWS exist (Mitsch investigating the nitrate removal efﬁciency of drainage et al. 2001), although in this review, we focused only on mitigation measures treating drainage water in climate the subset of FWS designed to treat drainage water before zones, where high loading rates of nitrate often occur when it reaches streams. conditions for denitriﬁcation is suboptimal. In addition to nitrate removal, drainage mitigation measures have shown Denitrifying bioreactors (DBR) potential for retention of P as increased HRT allows set- tling of suspended material such as sediment and particu- In DBR, the drainage water is routed horizontally or ver- late P (PP). Yet, the anoxic conditions established by these tically through a basin ﬁlled with carbon-rich ﬁlter sub- mitigation measures might lead to net P release, depending strate (e.g. different types of wood chips mixed with gravel, on local hydrological and geochemical conditions (O’Geen soil or other materials) before it reaches the outlet (Blowes et al. 2010). et al. 1994) (Fig. 1). The substrate of the DBR can either be In this review, we focused on ﬁve types of mitigation in direct contact with air (David et al. 2016; Carstensen measures treating drainage water before it enters streams. et al. 2019b) or sealed off by a layer of soil on top of the These were the commonly applied free water surface ﬂow reactor (de Haan et al. 2010). Similar to FWS, the base of constructed wetlands (FWS), denitrifying bioreactors the DBR are sealed with non-permeable membranes to (DBR) and controlled drainage (CD) and the two emergent avoid seepage if establish on water-permeable soils. Den- technologies saturated buffer zones (SBZ) and integrated itrifying bioreactors are also known as subsurface ﬂow buffer zones (IBZ) (Fig. 1). To obtain a better under- constructed wetlands, denitrifying beds or bio-ﬁlters. The standing of the opportunities and challenges of current and ﬁrst pilot study with DBR, established in Canada in 1994, was inspired by wastewater treatment plants (Blowes et al. new drainage mitigation measures targeting the transport of nutrients from agricultural areas in oceanic and continental 1994). However, in contrast to wastewater treatment plants, climates, we examined nitrate and total P (TP) removal DBR was solely designed to promote anoxic conditions, efﬁciencies at 82 drainage sites established between 1991 and carbon-rich ﬁlter material was added to fuse and 2018 in eleven countries. Thus, this review compiles denitriﬁcation. the available evidence on nitrate and TP removal efﬁcien- cies from both pilot and full-scale ﬁeld studies on drainage Controlled drainage (CD) mitigation measures to provide a synthesis of the existing body of peer-reviewed literature. Controlled drainage is a groundwater management tech- nique, where the in-ﬁeld groundwater level is elevated using a water control structure to restrict the water ﬂow MATERIALS AND METHODS from the drain outlet (Gilliam et al. 1979) (Fig. 1). Thus, CD alters the hydrological cycle of the ﬁeld, which, Overview of the included types of drainage depending on location and season, increases some or all of mitigation measures the following ﬂow components: root zone water storage, seepage (shallow, deep), surface runoff, plant uptake and Free water surface constructed wetlands (FWS) evaporation (Skaggs et al. 2012). Experiments with CD were initiated in the late 1970s in the USA to investigate In FWS, drainage water typically passes one or more deep the potential for enhancing in-ﬁeld denitriﬁcation (Wil- basins or channels and shallow vegetated zones (berms) lardson et al. 1970), and CD were also practiced in the before reaching the outlet and eventually the stream (Ko- former German Democratic Republic to cope with summer vacic et al. 2000) (Fig. 1). The deep zones reduce the water droughts, though the technique disappeared with the fall of ﬂow and thus increase HRT and promote denitriﬁcation the wall (Heinrich 2012). The Author(s) 2020 www.kva.se/en 1822 Ambio 2020, 49:1820–1837 Fig. 1 Conceptual scheme of the ﬁve drainage mitigation measures Saturated buffer zones (SBZ) 2019)(Fig. 1). The inﬁltrating water saturates the riparian soil and creates anoxic conditions, though in In a SBZ, drainage water and riparian soil are recon- order for denitriﬁcation to occur, the soil carbon content nected by a buried, lateral perforated distribution pipe must be sufﬁcient. This novel technique was recently running parallel to the stream, which redirect the drai- developed and tested in the USA (Jaynes and Isenhart nage water into the riparian zone (Jaynes and Isenhart 2014). The Author(s) 2020 www.kva.se/en Ambio 2020, 49:1820–1837 1823 Integrated buffer zones (IBZ) which meant that recalculation of removal efﬁciencies were necessary in some studies. In IBZ, the drainage water is ﬁrst retained in a pond designed to capture particles and increase the HRT and to Meta-analysis buffer surface runoff (Zak et al. 2018) (Fig. 1). After the pond, the water inﬁltrates a vegetated shallow zone where The average nitrate and TP removal efﬁciencies of miti- the top soil has been removed. In this inﬁltration zone, gation measures treating agricultural drainage water were anoxic conditions develop and carbon is added from the quantiﬁed using meta-analysis. Prior to the analysis, the vegetation via root exudates or leached plant litter. Inte- assumption of normality was tested visually (Q–Q plot, grated buffer zones were recently developed and tested in histogram) and by the Shapiro–Wilk test, and where the Northwestern Europe with the aim to improve the nutrient assumptions were not fulﬁlled this is mentioned in the reduction capacity of traditional riparian buffer zones result section. Meta-analysis was only conducted for a bypassed by drainage pipes, while promoting multi-func- mitigation measure if sufﬁcient data were available, i.e. tionality, such as biodiversity and biomass production (Zak data from more than two sites originating from different et al. 2019). studies. The meta-analysis was performed in R software 3.6.1 (R Core Team 2019) using the R package ‘meta’ Literature search and inclusion criteria (Schwarzer 2019). The effect size of each study was expressed as the raw removal efﬁciency and was calculated To ﬁnd relevant studies for our review, a search of pub- as follows: lished studies was conducted via ISI Web of Science for Load Load in out Removal efficiencyðÞ % ¼ 100 1900–2019 employing four different search strings, which Load in are described in the Supplementary Material (Table S1). -1 where Load is the loading to the system in kg year and The relevant studies was selected considering the following in -1 criteria: Load the loss from the system in kg year ; for CD sites out -1 -1 the unit is kg ha year . – The inlet water had to originate from drainage systems Each effect size was weighted, and a higher weight was transporting water from agricultural ﬁelds, and must given to studies with small standard error (SE) and large not be mixed with water from other sources such as sample size, as these were regarded as more precise. The streams. summary effect was calculated based on the effect sizes – Based on the Ko¨ppen-Geiger climate classiﬁcation and their weight, using a random effect model, which allow system, the sites had to be located in oceanic (Cfb, Cfc) the true mean to vary between studies, as the selected or continental (Dfa, Dfb, Dfc, Dfd, Dsc) climates studies differed in design, materials and methods. To (Fig. 2), where the conditions for denitriﬁcation are account for this variability, the weighting factor assigned to often suboptimal. Thus, climate zones with dry winters each effect size incorporated both the within-study vari- (letter w) were excluded. 2 2 ance (r ) and the between-study variance (T ). The – The study had to be a ﬁeld study with sites exposed to DerSimonian and Laird (DL) method was applied to esti- ambient temperature and with a surface area larger than mate T , and the Hartung-Knapp method was used to adjust 10 m . the conﬁdence intervals (CI), producing more conservative – The study had to include a mass balance for either results, as recommended by Borenstein (2009), when nitrate—N, total phosphorus (TP) or total suspended dealing with a low number of studies (K \ 20). To evalu- solids (TSS) for at least one drainage season, whose ate whether the use of the overall summary effect was length depended on the climate region. appropriate, the degree of consistency of the effect sizes was assessed using forest plot, funnel plot and multiple If two studies were conducted at the same study site within overlapping monitoring periods, the study with the statistical measures. The observed variation (Q) was tested longest time series was selected. Not all extracted data to investigate if the true effect varied between studies and if application of the random effect model was appropriate could be separated into years or seasons, implying that standard deviation (r) for nitrate removal was not available (Borenstein 2009). The excess variation over the observed variation (I ) gave an indication of what proportion of the for nine sites and for TP removal for one site; still, these sites were included in the calculation of the arithmetic variation was real, and reﬂected the extent of overlapping CIs. However, care must be taken, as in the case of an I mean (Table S2). Absolute removal was reported in various -2 -1 -3 units (e.g. g m ,kgha ,gm ), and we therefore close to zero, it can either be ascribed to that all variance is due to sampling error within the studies, though it can also identiﬁed and used the most commonly reported unit, be caused by very imprecise studies with substantial The Author(s) 2020 www.kva.se/en 1824 Ambio 2020, 49:1820–1837 −160 −140 −120 −100 −80 −60 −40 −20 0 20 40 60 80 100 120 140 160 180 80 80 70 70 60 60 50 50 40 40 Country #sites 30 30 USA 34 Canada 12 20 20 Denmark 7 10 10 Sweden 7 UK 6 0 0 Norway 4 −10 The Netherlands 4 −20 New Zealand 3 −20 Finland 2 −30 −30 Switzerland 2 France 1 −40 −40 −50 −50 −60 −60 −70 −70 −80 −80 −90 −160 −140 −120 −100 −80 −60 −40 −20 0 20 40 60 80 100 120 140 160 180 Dfa Dfb Dfc Dfd Dsc Cfb Cfc Fig. 2 World map showing the climate regions included in the review and the number of study sites per country difference between effect sizes. Thus, large I values can data when conducting a review. In this study, the risk of either indicate the possible existence of different subgroups bias tool developed by Higgins et al. (2011) was used as a or that the analysis contain highly precise studies with very guideline, although it was originally developed based on small differences between the effect sizes. The estimate of evidence from randomised trials within the ﬁeld of meta- the absolute variance, T , was used as an indication of epidemiology. However, it has earlier been modiﬁed and dispersion, and it was compared with r . The standard used for environmental studies (Bilotta et al. 2014), such as deviation of the effect size (T) was also reported. In the wetlands (Land et al. 2016). In our study, the risk of bias funnel plot, the removal efﬁciencies were plotted against assessment included two steps (Fig. 3), where the ﬁrst step the SE, thus asymmetry or other shapes in the funnel plot was an evaluation of the water balance monitoring strat- might indicate bias related to publication bias, hetero- egy (1.A in Fig. 3). The water balance is especially of geneity or sampling error. Funnel plots were only inspected importance when quantifying the removal efﬁciency, as if the analysis contained more than ten studies, as recom- any errors here will propagate into the nutrient balance. To mend by Borenstein (2009). For each effect size and assess the monitoring strategy of the water balance, the summary effect, a 95% CI was reported. Additionally, a most important ﬂow paths were given a percentage, and 95% prediction interval (PI) was calculated for each sum- aggregated into an overall score. Thus, monitoring of mary effect, yielding the interval where 95% of future inﬂow and outﬂow accounted for 30%, groundwater for studies will fall (Borenstein 2009). To further explore 20%, surface runoff for 10% and precipitation and evapo- heterogeneity and the robustness of the summary effect, a ration for 5% each. A percentage of 100% implied that all meta-analysis was performed on two subsets of data for important ﬂow paths were monitored or otherwise each drainage mitigation measure if data sufﬁced. The ﬁrst accounted for. In the second step, the monitoring frequency subset of data contained only sites from the low risk of bias of ﬂow (2.B in Fig. 3) and the spatial and temporal fre- category (‘‘Risk of bias assessment’’), while the other data quency of nutrient sampling (2.C, 2.D) were assessed. set only contained sites where the within-study sample size Finally, the selection of control and impact sites was (N) was larger than two. evaluated (2.E in Fig. 3); however, this was only relevant for studies on CD, as these were the only studies with true Risk of bias assessment spatial replication. In the remaining studies, the inlet served as control and the outlet as impact. If all ﬁve attributes It is important to consider the extent of systematic errors were fulﬁlled, the site was considered as having low risk of resulting from different factors such as a poor study design bias; otherwise, it was considered having moderate to high or issues related to the collection, analysis and reporting of risk of bias. The Author(s) 2020 www.kva.se/en Ambio 2020, 49:1820–1837 1825 STEP 1 STEP 2 B Inflow/outflow was monitored continuously A Flow path Weight (at least hourly). Inflow 30 Low risk of bias Yes to all C Nutrients were assessed at the most important flow Outflow 30 paths (inflow, outflow, if influencial: surface run off Score ≥ 90 Groundwater 20 or groundwater). Surface run off 10 D The sampling of nutrients was evenly distributed Precipitation 5 during the entire run off season (at least two Moderate to high samples per month) or was flow proportional. Evaporation 5 No to one or more risk of bias E The control and impact sites were more or less similar. Score < 90 Fig. 3 Overview of the method for risk of bias assessment Table 1 Results from the process of ﬁnding and selecting relevant studies for free water surface constructed wetlands (FWS), denitrifying bioreactor (DBR), controlled drainage (CD) and saturated (SBZ) and integrated buffer zones (IBZ). WB: water balance Drainage Result of search After screening Passing inclusion Study sites Study years * WB score Studies with low mitigation measure title and abstract criteria replicates risk of bias of total sites (%) FWS 7550 173 17 33 109 85 55 DBR 7550 173 9 19 54 83 21 CD 213 100 14 25 93 80 20 SBZ 187 24 1 6 19 83 17 IBZ 176 13 1 1 2 100 100 Table 2 Size of each type of drainage mitigation measure, catchment area, DMMCAR (ratio of facility area and catchment area), age and HLR (hydraulic loading rate to the facility area) for free water surface constructed wetlands (FWS), denitrifying bioreactors (DBR), controlled drainage (CD) and saturated (SBZ) and integrated buffer zones (IBZ). SD: standard deviation. For CD, age refers to study length Size Catchment area DMMCAR Age HLR mean ± SD Mean ± SD Range (m ) Mean ± SD Range Mean ± SD Range Mean ± SD Range (m) 2 2 (m ) (ha) (m ) (%) (%) (year) (year) FWS 5486 ± 9377 20–51 000 65.1 ± 220.2 0.8–971.0 1.8 ± 2.1 0.03–7.06 5 ± 5 1–20 20 ± 22 DBR 71 ± 46 15–128 10.9 ± 6.7 0.8–20.2 0.1 ± 0.1 0.04–0.38 4 ± 2 2–10 685 ± 647 CD 10 572 ± 29 590 1005–149 000 2.1 ± 4.3 0.1–14.9 100 4 ± 1 1–5 0.2 ± 0.1 SBZ 4229 ± 2802 460–7392 14.1 ± 14.5 3.4–40.5 7 ± 7.2 0.65–15.73 4 ± 1 2–6 6 ± 6 IBZ 250 250 15.0 0.2 1 1 99 RESULTS having ‘moderate to high’ risk of bias (Table 1). The ratio of the drainage mitigation measure surface area to the Descriptive characteristics contributing catchment area (DMMCAR) was largest for SBZ (7%) and FWS (2%), while DBR (0.1%) and IBZ The initial search yielded 8126 studies in total, and after (0.2%) had the lowest ratios (Table 2). For CD, the evaluating the inclusion criteria, we had a master bibliog- DMMCAR was technically 100% if assuming that the raphy of 42 articles containing 84 sites distributed across groundwater level was elevated within the entire con- eleven countries (Table 1 and Table S2). According to our tributing catchment area, however, this can be a misleading risk of bias assessment, the risk of bias was low in 35% of term, as the control system only occupied very little of the the studies. Insufﬁcient monitoring of the water balance ﬁeld (app. one m per regulation well). The hydraulic was the main reason that many studies were categorised as loading rate (HLR) to the systems differed substantially, as The Author(s) 2020 www.kva.se/en 1826 Ambio 2020, 49:1820–1837 expected, being highest for DBR and lowest for CD due to DBR of 40% within a range from 6 to 79% (CI: 24 to 55%, the difference in size. Treatment of drainage water is a PI: - 9 to 89%) (Fig. 4). The funnel plot revealed asym- relatively new concept, as illustrated by that the oldest metry of data, where studies with low efﬁciency tended to facilities were two 20-year-old FWS and the second oldest have lower SE (Fig. 5). The heterogeneity analysis showed a 10-year-old DBR. The youngest and least studied mea- that the I was high (99%), as some of the studies were very sure was IBZ. precise, but showed different removal efﬁciency. Average 2 2 T (436%) was much higher than r (169%). The subset Free water surface constructed wetlands (FWS) analysis of data with either low risk of bias or sampling periods longer than two years/drainage seasons reported The weighted average obtained by meta-analysis showed lower removal efﬁciency (35%), and CI and PI were that FWS signiﬁcantly reduced nitrate loading by 41% slightly narrower for studies with N [ 2 (Table 4). Similar 2 2 within a range from - 8 to 63% (Fig. 4). The CI varied to FWS, studies with N [ 2 had lower T and higher r . from 29 to 51%, while the PI was rather broad, varying The arithmetic mean efﬁciency was 44% (CI: 35 to 53%), from 5 to 76%. The funnel plot did not indicate major while the absolute nitrate removal per DBR volume -3 -1 biases, as the studies were more or less evenly scattered amounted, on average, to 715 g N m year (CI: 292 to -3 -1 -3 (Fig. 5). However, the heterogeneity of the selected sites 760 g N m year ), ranging from 66 to 2033 g N m 2 2 -1 was rather high (I = 96%), and T (260%) was higher than year (Table 4). This corresponded to an area-based 2 -2 -1 r (70%). The subset analysis of data with either low risk nitrate reduction of 594 g N m year (CI: 333 to -2 -1 of bias or sampling periods longer than two years/drainage 855 g N m year ). seasons showed the average removal ranged between 40 Only two studies included TP balances for the full and 44%, and CI and PI were slightly more narrow than for drainage season, preventing meta-analysis. These two the full dataset (Table 4). Studies with N [ 2 had lower T , studies were somewhat contradictory in that one found 2 -2 -1 whereas r was slightly higher, which lowered the release of TP (- 208% or - 30 g P m year ) and the -2 -1 heterogeneity. According to the arithmetic mean, the other net removal (28% or 6 g P m year ) (Table 4). removal efﬁciency was 41% (CI: 29 to 51%) (Table 3). The absolute nitrate removal per FWS area amounted to 60 g N Controlled drainage (CD) -2 -1 -2 -1 m year (CI: 29 to 91 g N m year ). According to the meta-analysis the average TP removal The meta-analysis showed that CD signiﬁcantly reduced efﬁciency of FWS was 33%, ranging from - 103 to 68% the annual nitrate loading by, on average, 50% within a (CI: 19 to 47%, PI: - 2 to 69%) (Fig. 6). The removal range from 19 to 82% (Fig. 4) (CI: 41 to 59%, PI: 19 to efﬁciencies did not follow a normal distribution; the data 81%). However, both CI and PI should be interpreted with were skewed to the left due to net release of TP from care as the effect sizes did not follow a normal distribution. multiple sites. The funnel plot showed an asymmetrical The funnel plot displayed a more or less even scatter of 2 2 scatter of sites, as sites with TP release had much higher SE sites (Fig. 5). Heterogeneity was high (I = 79%), yet T (Fig. 5). As expected, the heterogeneity was rather high, was only slightly higher than r . The removal efﬁciency of 2 2 and T (226%) was much lower than r (838%). The studies including sampling periods longer than two years/ subset data analysis for TP removal reported a slightly drainage seasons was more or less similar to the result of higher removal (35%) than the initial data; however, r the full data analysis (Table 4). However, the subset anal- was still very high as the included studies reported both ysis pointed to the possible occurrence of two groups; one removal and release of TP (Table 4). The data were further with a removal efﬁciency \ 44% and one with a removal investigated by separating sinks and sources, showing that efﬁciency [ 61%. The mean (arithmetic) nitrate removal four sites exhibited a net release of TP (- 49%, CI: - 18 to efﬁciency was 48% (CI: 40 to 56%). The absolute nitrate -2 -1 - 83%) and eleven sites acted as sinks (38%, CI: 27 to removal amounted to 1.20 g N m year (1.16 to -2 -1 -1 49%). The arithmetic mean TP removal efﬁciency was 1.24 g N m year ), corresponding to 12 kg N ha -1 -1 -1 18% (CI: - 4 to 46%) with an average absolute removal year (CI: 8 to 16 kg N ha year ). The relative nitrate -2 -1 -2 -1 of 0.68 g P m year (- 1.16 to 2.52 g P m year ) reduction correlated well with the relative reduction of (Table 3). The removal efﬁciency of TSS was 41% (CI: 28 drainage ﬂow (R = 0.80 (Pearson), p \ 0.0001, K = 19), to 54%) when calculated as the arithmetic mean. and exclusion of studies with sub-irrigation, a practice implying an additional water supply, improved this corre- Denitrifying bioreactors (DBR) lation (R = 0.88, p \ 0.0001, K = 10) (Fig. 7). The average loss of TP via drainage water was reduced The weighted average calculated by meta-analysis showed by 34% (CI: 10 to 58%, PI: - 23 to 92%) according to the a signiﬁcant reduction of the annual nitrate loading by meta-analysis (Fig. 6). The removal efﬁciencies did not The Author(s) 2020 www.kva.se/en Ambio 2020, 49:1820–1837 1827 FWS ID N Weight RRE (%) & 95%−CI Haan et al. (2010) FWS20 2 6% Groh et al. (2015) FWS13 2 9% Groh et al. (2015) FWS14 2 9% Tanner and Sukias (2011) FWS10 2 8% Tournebize et al. (2014) FWS22 8 4% Tanner and Sukias (2011) FWS12 4 5% Kovacic et al. (2000) FWS14 3 9% Fink and Mitsch (2004) FWS05 2 7% Kovacic et al. (2000) FWS13 3 7% Kovacic et al. (2006) FWS17 2 8% Kovacic et al. (2000) FWS15 3 9% Kovacic et al. (2006) FWS16 2 6% Tanner and Sukias (2011) FWS11 5 4% Koskiaho et al. (2003) FWS19 2 9% Summary effect (t 8, p<0.0001) Prediction interval 2 2 Heterogeneity: I = 96%, = 260, p < 0.01 DBR ID N Weight RRE (%) & 95%−CI Haan et al. (2010) SSF04 2 9% Christianson et al. (2012) SSF22 2 10% Christianson et al. (2012) SSF21 3 8% Haan et al. (2010) SSF03 2 4% Carstensen et al. (2019) SSF08 5 9% Carstensen et al. (2019) SSF07 5 9% Christianson et al. (2012) SSF19 7 9% David et al. (2016) SSF02 3 3% Søvik et al. (2008) SSF18 3 7% Søvik et al. (2008) SSF16 3 10% Christianson et al. (2012) SSF20 2 10% Søvik et al. (2008) SSF15 2 10% Summary effect (t 6, p<0.0001) Prediction interval 2 2 Heterogeneity: I = 99%, = 436, p < 0 CD ID N Weight RRE (%) & 95%−CI Lalonde et al. (1996) CDS8D 2 4% Wesström and Messing (2007) CD20 8 7% Lalonde et al. (1996) CDS7D 2 6% Woli et al. (2010) CD16 3 5% Drury et al. (2009) CD06B 4 6% Drury et al. (2009) CD08B 4 4% Schott et al. (2017) CD19 5 6% Drury et al. (1996) CD02A 2 8% Drury et al. (2009) CD05B 4 6% Drury et al. (1996) CD04A 2 4% Drury et al. (1996) CD01A 2 7% Drury et al. (2014) CD11B 5 6% Drury et al. (1996) CD03A 2 5% Drury et al. (2014) CD10B 5 5% Wesström et al. (2014) CD21 8 3% Drury et al. (2009) CD07B 4 5% Jaynes et al. (2012) CD18 4 7% Carstensen et al. (2018) CD22 4 4% Wiliams et al. (2015) CD17 4 4% Summary effect (t 12, p<0.0001) Prediction interval 2 2 Heterogeneity: I = 79%, = 201, p < 0.01 −50 0 50 100 Fig. 4 Forest plots showing effect sizes (RRE) and 95% conﬁdence intervals (CI) of relative nitrate–N removal and summary effect with CI and prediction interval and heterogeneity analysis for free water surface constructed wetlands (FWS), denitrifying bioreactors (DBR) and controlled drainage (CD). N within-study sample size. ID represents FWS and DBR study sites; for CD the letter is unique for the research facilities The Author(s) 2020 www.kva.se/en 1828 Ambio 2020, 49:1820–1837 FWS Nitrate FWS Total phosphorus A B Reinhart et al. (2005) 0 0 Kovacic et al. (2006) Groh et al. (2015) Kovacic et al. (2000) Kynkäänniemi et al. (2013) Mendes et al. (2018) Kovacic et al. (2000) 2 10 Kovacic et al. (2006) Mendes et al. (2018) Mendes et al. (2018) Groh et al. (2015) Kovacic et al. (2000) 4 20 Fink and Mitsch (2004) Koskiaho et al. (2003) Kovacic et al. (2000) Kovacic et al. (2006) 6 30 Tanner and Sukias (2011) Kovacic et al. (2000) 8 Kovacic et al. (2000) 40 Tournebize et al. (2014) Tanner and Sukias (2011) Fink and Mitsch (2004) 10 50 Tanner and Sukias (2011) Tanner and Sukias (2011) Kovacic et al. (2006) 12 60 Haan et al. (2010) Tanner and Sukias (2011) Tanner and Sukias (2011) 14 70 –20 0 20406080 −100 −50 0 50 100 150 C CD Nitrate D DBR Nitrate Søvik et al. (2008) 0 0 Christianson et al. (2012) Søvik et al. (2008) Christianson et al. (2012) Drury et al. (1996) Carstensen et al. (2019) Jaynes et al. (2012) Carstensen et al. (2019) Wesström and Messing (2007) Haan et al. (2010) Christianson et al. (2012) Drury et al. (1996) Lalonde et al. (1996) Christianson et al. (2012) Drury et al. (2009) Drury et al. (2009) Drury et al. (2014) Schott et al. (2017) Søvik et al. (2008) 10 15 Drury et al. (1996) Woli et al. (2010) Drury et al. (2009) Drury et al. (2014) Drury et al. (2009) Wiliams et al. (2015) Lalonde et al. (1996) Carstensen et al. (2018) Drury et al. (1996) Haan et al. (2010) David et al. (2016) Wesström et al. (2014) 20 30 0 20406080 100 −40 −20 0 20406080 Mean removal (%) Fig. 5 Funnel plots of free water surface constructed wetlands (FWS), denitrifying bioreactors (DBR), controlled drainage (CD) and saturated (SBZ) and integrated buffer zones (IBZ) for data sets containing results on nitrate–N or total phosphorus follow a normal distribution as the data were slightly until now, only one study containing multiple sites has skewed towards the right. According to the statistical been published for each practice (Table 4). The annual 2 2 analysis, heterogeneity was moderate (I = 65%), while T arithmetic mean removal efﬁciency was 75% (CI: 35 to (406%) and r (403%) were more or less identical, sug- 53%) of the nitrate loaded into the SBZ. However, between gesting that the studies were similar enough to justify 6 and 77% of the water bypassed the SBZ; thus, taking all combination. The arithmetic mean was 29% (CI 10 to nitrate leaving the ﬁeld into account, the average nitrate 48%) (Table 4). The average absolute TP retention removal efﬁciency was 37% (CI: 17 to 57%) and varied -2 -1 -2 amounted to 0.03 g P m year (0.01 to 0.05 g P m from 8 to 84%. The absolute nitrate removal per SBZ area -1 -1 -1 -1 -2 -1 -2 -1 year ) or 0.30 kg P ha year (0.10 to 0.49 kg P ha was 23 g N m year (CI: 9 to 37 g N m year ). -1 year ) (Table 4). The relative reduction of TP loading There were no available data on TP balances for SBZ in the correlated well with the reduction of drainage ﬂow articles selected for this review. For IBZ, the annual nitrate (R = 0.87 (Pearson), p \ 0.01, K = 6) (Fig. 7). removal efﬁciency, calculated as the arithmetic mean, was 26% (CI: 20 to 32%) (Table 4). The absolute nitrate -2 -1 Saturated and integrated buffer zones (SBZ removal per IBZ area was 140 g N m year (71 to -2 -1 and IBZ) 209 g N m year ). The removal efﬁciency of TP was 48% (CI: 40 to 56%), while the absolute TP removal per -2 -1 -2 Removal efﬁciencies could not be aggregated using meta- IBZ area was 2.4 g P m year (CI: 1.4 to 3.5 g P m -1 analysis for the emergent technologies, SBZ and IBZ, as, year ). The Author(s) 2020 www.kva.se/en Standard error (%) Ambio 2020, 49:1820–1837 1829 Table 3 Relative and absolute removal of nitrate–N, total phosphorus (TP) and total suspended solids (TSS) based on raw data and the meta- analysis for free water surface constructed wetlands (FWS), denitrifying bioreactors (DBR), controlled drainage (CD) and saturated (SBZ) and am meta integrated buffer zones (IBZ). K is the number of study sites included when calculating the arithmetic mean, and K is the number of study sites included in the meta-analysis am% meta meta am Drainage mitigation K K Removal Mean ± SD Removal Mean ± SD K Removal Mean ± SD -2 -1 measure (%) (%) (g m year ) Nitrate–N FWS 18 14 40 ± 17 41 ± 21 21 60 ± 69 DBR 19 12 44 ± 21 40 ± 27 2 594 ± 481 CD 20 19 48 ± 18 50 ± 20 6 1 ± 1 SBZ 6 68 ± 39 13 23 ± 18 SBZ 637 ± 25 IBZ 2 26 ± 4 19 140 ± 50 TP FWS 16 15 18 ± 46 33 ± 28 8 0.68 ± 4.19 CD 7 7 29 ± 26 34 ± 32 2 0.03 ± 0.03 DBR 3 -50 ± 136 3 - 5.79 ± 20.96 IBZ 2 48 ± 6 20 2.44 ± 0.76 SS FWS 6 41 ± 16 6 1555 ± 936 Includes the water and nitrate–N bypassing the SBZ DISCUSSION with decreasing HLR (Vymazal 2017; Hoffmann et al. 2019) (Fig. S2), though temperature is at least as important. Removal efﬁciency and uncertainty of drainage The design of mitigation measures is commonly guided by mitigation measures DMMCAR, as a rough estimate of HLR and temperature; for instance, in New Zealand a guideline predicts that a Removal efﬁciency was quantiﬁed in both absolute and DMMCAR of 5% will yield an approximate nitrate relative values in our review. However, care should be reduction of 50 ± 15% (Tanner et al. 2010), while in taken when comparing values from different sites, as the Denmark a ratio around 1–1.5% is recommended for FWS absolute removal efﬁciency depended heavily on the to ensure a HRT of minimum 24 h during winter (Land- nutrient loading to the system (Fig. S1). The loading rate of brugsstyrelsen 2019). However, the optimal DMMCAR is nutrients are highly site speciﬁc, as it is determined by the site-speciﬁc and depends on hydrological and geochemical concentration of nutrients in the water and by HLR, which conditions, e.g. similar DMMCARs can have very different is highly variable from site to site. For example, for DBR, temperatures and HLRs (Fig. S3). the speciﬁc loading rate of nitrate per DBR area differed The quantiﬁcation of nutrient loading and removal is -2 substantially between sites (221 to 11,533 g N m DBR somewhat uncertain as it relies on a black-box approach -1 year ). Furthermore, the HLR varies from year to year, (i.e. input–output). This implies that the estimates depend although this variation can be accounted for to some extent especially on the frequency of nutrient sampling and the by monitoring over multiple years. In this review, it was water ﬂow monitoring strategy. Estimates of TP retention demonstrated by that study sites monitored for multiple might be more uncertain than those of nitrate as TP con- years (N [ 2) had higher r , and thus incorporated more centrations in tile drainage water tend to change quickly variation. Absolute removal was reported relative to miti- over time, especially at high ﬂow, which can be difﬁcult to gation measures surface area in our review, however, capture (Johannesson et al. 2017), whereas nitrate con- another possibility would be to report absolute removal per centrations tend to change more gradually. Johannesson catchment area, however, the estimate of catchment areas et al. (2017) tested the importance of ﬂow monitoring are often very uncertain, adding more uncertainty to the strategy and found that TP retention was underestimated removal estimate. The HLR also inﬂuence relative when based solely on outlet ﬂow measurements rather than removal, where the removal efﬁciency tends to increase on both inlet and outlet ﬂow measurements. The Author(s) 2020 www.kva.se/en 1830 Ambio 2020, 49:1820–1837 FWS ID N Weight RRE (%) & 95%−CI Kovacic et al. (2006) FWS16 2 8.6% Fink and Mitsch (2004) FWS05 2 3.7% Mendes et al. (2018) FWS03 3 10.6% Mendes et al. (2018) FWS02 3 7.9% Kovacic et al. (2006) FWS17 2 12.3% Mendes et al. (2018) FWS04 3 8.4% Kynkäänniemi et al. (2013) FWS21 2 10.5% Kovacic et al. (2000) FWS14 3 2.5% Reinhart et al. (2005) FWS06 2 13.2% Kovacic et al. (2000) FWS13 3 10.0% Koskiaho et al. (2003) FWS19 2 5.4% Kovacic et al. (2000) FWS15 3 4.0% Tanner and Sukias (2011) FWS10 2 0.8% Tanner and Sukias (2011) FWS12 4 1.3% Tanner and Sukias (2011) FWS11 5 0.7% Summary effect (t 3.51, p<0.05) Prediction interval 2 2 Heterogeneity: I = 82%, = 226, p < 0.01 CD ID N Weight RRE (%) & 95%−CI Wesström and Messing (2007) CD20 8 20% Wesström et al. (2014) CD21 8 12% Tan and Zhang (2011) CD09C 5 22% Zhang et al. (2015) CD12B 4 16% Carstensen et al. (2018) CD22 4 8% Zhang et al. (2015) CD13B 4 9% Zhang et al. (2015) CD14B 4 13% Summary effect (t 5.10, p<0.001) Prediction interval 2 2 Heterogeneity: I = 65%, = 406, p < 0.01 −100 −50 0 50 100 Fig. 6 Forest plots showing effect sizes (RRE) and 95% conﬁdence intervals (CI) of relative total phosphorus (TP) removal and summary effect with CI and prediction interval and heterogeneity analysis for free water surface constructed wetlands (FWS) and controlled drainage (CD). N within-study sample size. ID represents unique sites for FWS and DBR; for CD the letter is unique for the research facilities Free water surface ﬂow constructed wetlands (FWS) and urban stormwater. Our review of FWS showed that they did not always remove TP, as four out of 15 FWS sites The results showed that FWS signiﬁcantly reduced the acted as a source of P. This net release of P might be due to nitrate loss from drainage systems. However, as expected, mobilisation of dissolved reactive P (DRP) from the sedi- the efﬁciency varied considerably since the included ment or the size of the FWS being too small to adequately studies differed in design (e.g. HLR, aspect ratio, size, decelerate the ﬂow (Kovacic et al. 2000; Tanner and Sukias carbon availability), age monitoring schemes and run off 2011). The studies reporting a net release of TP had a very characteristics, factors that all affected the removal efﬁ- high within-study variance and they were therefore given ciency. At one site, nitrate release was reported, which was less weight in the meta-analysis, with the consequence that most likely due to the lack of monitoring of one of the the removal efﬁciency was higher than the arithmetic inlets (Koskiaho et al. 2003), which emphasises the mean. Both the relative and the absolute removal efﬁciency importance of the monitoring scheme. The removal efﬁ- were lower compared with Land et al. (2016), probably ciency found in this review was slightly higher than that of because the average TP loading was much higher in the an earlier review, which reported a removal of 37% (CI: 29 studies included in their review, where also FWS estab- to 44%) (Land et al. 2016). Compared with Land et al. lished in streams were represented. In our review, most (2016), the average absolute removal was much lower in studies on FWS had low risk of bias, although, often only -2 -1 our review (181 ± 251 g N m year ), which was not the inlet or the outlet was monitored, which were com- surprising, as their review included a broad range of cre- pensated for in the studies by adjusting the unmonitored ated and restored wetlands treating both agricultural runoff, ﬂow component with precipitation, evaporation or riverine water, secondary and tertiary domestic wastewater The Author(s) 2020 www.kva.se/en Ambio 2020, 49:1820–1837 1831 Table 4 Results from meta-analysis of all data and data from sites with more than two years or drainage seasons (N[ 2) and data from sites with low risk of bias (ROB) for free water surface constructed wetland (FWS), denitrifying bioreactors (DBR) and controlled drainage (CD). k: 2 2 within-study sites, SE: standard error, CI: conﬁdence interval, PI: prediction interval, T : between-study variance, r : within-study variance, I : proportion of unexplained variance 2 2 2 Data analysed k SE tp\ Range CI PI T r I Q test (p \) FWS Nitrate–N All 14 41 8 0.0001 - 8 to 63 29–51 5 to 76 260 70 96 0.0001 FWS Nitrate–N N [ 2 6 40 11 0.0001 22 to 54 31–49 19 to 60 41 79 61 0.02 FWS Nitrate–N Low ROB 12 44 14 0.0001 22 to 58 37–51 14 to 74 175 66 96 0.0001 FWS TP All 15 33 5 0.0002 - 103 to 68 19–47 - 2 to 69 226 838 82 0.0001 FWS TP N [ 2 8 35 2 0.06 - 102 to 49 20–49 - 29 to 81 373 1024 67 0.037 FWS TP Low ROB 14 35 5 0.0002 - 102 to 68 20–50 - 1 to 71 231 874 83 0.0001 DBR Nitrate–N All 12 40 6 0.0001 6 to 79 24–55 - 9 to 89 436 169 99 0.001 DBR Nitrate–N N [ 2 6 35 7 0.0007 18 to 45 23–47 12 to 82 267 209 88 0.0001 DBR Nitrate–N Low ROB 4 35 5 0.02 13 to 45 10–59 - 58 to 127 408 41 94 0.0001 CD Nitrate–N All 19 50 12 0.0001 19 to 82 41–59 19 to 81 201 119 79 0.0001 CD Nitrate–N N [ 2 13 49 9 0.0001 19 to 81 36–59 3 to 92 383 128 82 0.01 100 100 A B 80 80 60 60 40 40 20 20 0 0 0 20406080 100 0 20406080 100 Water reduction (%) Fig. 7 Percentage A nitrate and B total phosphorus removal vs. percentage reduction of drainage outﬂow at the outlet of ﬁelds with controlled drainage groundwater (if not lined with a non-permeable most likely a conservative estimate since many of the sites membrane). with suboptimal design were given a relatively high weight due to low SE. Many of the DBR sites were assessed to Denitrifying bioreactors (DBR) have a moderate to high risk of bias, as ﬂow was often only measured at either the inlet or the outlet, however, due to Our meta-analysis showed that DBR signiﬁcantly reduced their small size, the uncertainty caused by this might be the nitrate loss from drainage systems to surface water. The lower for DBR than for e.g. FWS. removal efﬁciencies generally displayed high variations, which reﬂected the differences (e.g. design, age) between Controlled drainage (CD) the studied sites, not least regarding nitrate loading rates. Many of the study sites were experimental facilities or pilot According to our results, CD signiﬁcantly reduced the studies, implying that they were established to investigate loading of nitrate at the drain outlet. However, hetero- and identify factors inﬂuencing performance. For example, geneity was relatively high and the efﬁciencies displayed among the studies included in our review, the low removal high dispersion around the mean. This was expected, efﬁciency could be ascribed to short-circuiting within the though, as the efﬁciency of CD is especially inﬂuenced by system (Christianson et al. 2012a), inadequate sizing, i.e. drain spacing and management, which differed between too short HRT (David et al. 2016), and scarce monitoring sites (Ross et al. 2016). For example, the target elevation of (Søvik and Mørkved 2008). Accordingly, the average the water table differed considerably between sites, from removal efﬁciency derived from the meta-analysis was 15 to 76 cm below the soil surface. The removal efﬁciency The Author(s) 2020 www.kva.se/en Nitrate reduction (%) TP reduction (%) 1832 Ambio 2020, 49:1820–1837 found in our review aligned very well with that from an efﬁciency of SBZ (Jaynes and Isenhart 2019) as higher earlier review of 48 ± 12% by Ross et al. (2016). The removal efﬁciency has been found at sites with established nitrate reduction was mainly regulated by the reduction of perennial vegetation. This might be due to addition of more the ﬂow at the drain outlet, which has also been stressed in labile carbon to the soil to support denitriﬁcation or to earlier studies (Skaggs et al. 2012; Ross et al. 2016). enhanced immobilisation of microbial N by the more Although many studies stated that CD was implemented to developed rhizospheres (Jaynes and Isenhart 2019). The increase denitriﬁcation, higher denitriﬁcation rates or lower removal efﬁciency of SBZ is difﬁcult to quantify as the nitrate concentrations in drain water were seldom reported outlet of the SBZ is the riparian soil where N and P con- despite denitriﬁcation measurements (Woli et al. 2010; centrations can only be measured with piezometers, and Carstensen et al. 2019a). This lack of denitriﬁcation was dilution by groundwater through ﬂow can occur. Another probably due to insufﬁcient amounts of soil organic carbon, concern is whether or not the piezometer measurements temperature limitation or absence of anoxic zones in the can be considered representative for the whole area. In our soil. Higher efﬁciencies could potentially be obtained if the review, IBZ had the lowest average removal efﬁciency of water level was elevated even closer to the surface where the mitigation measures, which probably can be ascribed to the organic C content is higher, but this could increase the that the two IBZs were experimental test facilities with too surface runoff (Rozemeijer et al. 2015) and/or harm the low DMMCAR and the vegetation was not fully developed crop yield. The redirected water is either stored in the root (Zak et al. 2018). A recent technical report on IBZ showed zone or directed to alternative ﬂow paths. If the excess that the removal efﬁciency of two full-scale facilities water moves towards the stream without passing conditions established in Denmark was 53–55%, which was even a suitable for denitriﬁcation, there will be no removal of conservative estimate (van’t Veen et al. 2019). The overall nitrate and thus no effect of CD. In contrast, if the water reduction of nitrate to the receiving water might be even passes deeper zones with reduced conditions or conditions higher than reported, as after passing the IBZ, the water favourable for denitriﬁcation, the nitrate will most likely be inﬁltrates the riparian zone between the IBZ and the stream removed. Higher removal efﬁciency of CD could be gained where nitrate can be further removed by denitriﬁcation or if CD was combined with, for example DBR, treating the vegetation. Thus, more studies on SBZ and IBZ are needed part of the water still leaving via the drainage system (Woli to critically assess their nutrient removal efﬁciency and the et al. 2010). A concern regarding the implementation of uncertainty related to the monitoring of the outlet. CD has been that the saturation of the root zone might cause desorption of redox-sensitive P, but none of the Applicability in the farmed landscape studies on CD reported TP or DRP release. However, in The ﬁve drainage mitigation measures can seamlessly be three studies the CI crossed the zero line, indicating that TP removal was not signiﬁcant, which was supported by the integrated into landscapes with existing drainage systems, PI. The retention efﬁciency determined in our study was but to optimise performance and cost efﬁciency their considerably lower compared with Ross et al. (2016), who individual applicability to the landscape must be evaluated reported a TP retention of 55 ± 15%. Almost all sites with carefully. Each measure varies in size and capacity to CD were categorised as having moderate to high risk of intercept water, where the size relative to the catchment bias, as the majority of the studies only quantiﬁed the area decreases in the order of FWS [ SBZ [ IBZ [ DBR. reduction in ﬂow and nutrients at the drainage outlet. Only Especially the size of the contributing catchment, slope and few attempted to quantify nitrate or P budgets for all ﬂow soil type determine how and where the measures can be paths leading nutrients to the surface water (Sunohara et al. implemented (Fig. 8). Flat landscapes (slope \ 1%) are 2014). suitable for implementation of CD as a single control structure will affect a large area; however, as the technol- Saturated and integrated buffer zones (SBZ and IBZ) ogy advances it might soon be possible also to implement CD in sloping landscapes. In gently sloping terrains, FWS, Two novel technologies, SBZ and IBZ, were included in DBR and SBZ ﬁt as a hydraulic gradient is needed to move our review to demonstrate the recent development in this the water through the systems. The hydraulic gradient research area. Until now, SBZ have mainly been investi- should preferably be minimum 2–3% for FWS and DBR, gated in USA and with variable results (Jaynes and Isenhart while for SBZ the slope of the landscape should be around 2019). Low performance of SBZ has been linked to 2–8% (Tomer et al. 2017). In addition, in sloping land- selection of unideal sites containing permeable soil layers scapes, IBZ are suitable as a hydraulic gradient of mini- or sites where a low fraction of water was diverted to the mum 4% is required to move water through the pond and SBZ, which is controlled by the length of the distribution the inﬁltration zone (Fig. 8). In sloping areas, surface pipe. Vegetation has also been argued to inﬂuence the The Author(s) 2020 www.kva.se/en Ambio 2020, 49:1820–1837 1833 Fig. 8 Conceptual diagram of potential locations of free water surface constructed wetlands (FWS), denitrifying bioreactors (DBR), controlled drainage (CD) and saturated (SBZ) and integrated buffer zones (IBZ) on mineral soils in a small catchment runoff is also more likely to occur, which can be inter- (Carstensen et al. 2019b). For instance, DRP release and cepted by the IBZ (Zak et al. 2019). methane emission have been reported from facilities Besides suitability to the landscape, implementation experiencing nitrate limitation (Robertson and Merkley strategies are often guided by cost efﬁciency. Cost efﬁ- 2009; Shih et al. 2011), while other processes need further ciency, including capital and operational cost of the drai- investigation (e.g. nitrous oxide emission, loss of dissolved nage systems, has been calculated earlier (Christianson organic carbon). Permanent removal and recycling of P et al. 2012b; Jaynes and Isenhart 2019). However, the cost require plant harvesting or sediment removal; another of preliminary examinations such as geological and soil course of action may be to combine mitigation measures investigations has often not been included despite that it with a P ﬁlter (Canga et al. 2016; Christianson et al. 2017). can constitute a substantial part of the budget, and is The possibilities of optimising ecosystem services and therefore important to consider when selecting mitigation synergies with the surrounding landscapes where drainage measure. Cost efﬁciency is inherently country speciﬁc mitigation measures are applied are manifold (Goeller et al. since, as besides local costs, such as land acquisition, it 2016) e.g. biodiversity, water storage, phytoremediation depends on national regulation and implementation (Williams 2002) or provision of biomass (Zak et al. 2019). strategies. For example, in Denmark, FWS can only be Current examples of multiple ecosystem service provi- implemented at a certain location if the catchment area is sioning are, the combination of CD, sub-irrigation and larger than 20 ha and if it removes more than 300 kg N ha reservoirs, which according to Satchithanantham et al. -1 -1 wetland year , and other requirements such as to soil (2014), can reduce the peak ﬂow in spring and delay short- clay content ([ 12%) also prevail. term water-related stress on crops in periods with less precipitation. In addition, sub-irrigation can increase crop Current advances in ecosystem service provisioning yields (Wesstro¨m and Messing 2007; Jaynes 2012). According to our review, CD was combined with sub-ir- The selection, implementation and design of drainage rigation at 14 out of 25 sites, while FWS were combined mitigation measures should ideally maximise the supply of with a sedimentation pond at 6 of the 33 sites. A sedi- ecosystem services and minimise undesirable by-products. mentation pond is a simple supplement, which can Thus, the management and design of mitigation measures increased the water storage capacity and give access to should not solely focus on nutrient reduction, but also take irrigation water and nutrients for recycling. Yet, the into consideration potential negative by-products, as some potential of mitigation measures for increasing the climate of these can be minimised by location or design resilience of agricultural areas by retaining and storing The Author(s) 2020 www.kva.se/en 1834 Ambio 2020, 49:1820–1837 more water in the landscape, thereby buffering hydrologi- In addition to considering the local geographical and cal peak events, needs to be investigated at catchment climatic conditions for selection and application of drai- scale. Due to the potentials for adaptation and synergies nage mitigation measures, integration with future changes with the surrounding landscape, these systems are inno- in climate and land use must be considered. Climate vative opportunities in future bio-economies, as the mea- change is predicted to cause more intense and frequent sures can reduce nutrient losses, while providing multiple precipitation events and prolonged summer droughts in the ecosystem services e.g. nutrient reuse, biomass production, investigated climate regions (Christensen et al. 2013). The biodiversity, etc, if designed accordingly. envisaged increase in temperature might improve the per- formance of the drainage mitigation measures, even though the intense precipitation events will challenge their PERSPECTIVE: OPPORTUNITIES hydraulic capacities and, thereby, their performance, AND CHALLENGES FOR IMPLEMENTATION potentially changing the need for mitigation measures at OF MITIGATION MEASURES AT CATCHMENT catchment scale. Human modiﬁcations of land use, land SCALE and water management induced by, for instance, a green shift to a new bio-economy (Marttila et al. 2020) might Effective implementation of drainage mitigation measures entail further expansion and intensiﬁcation of land uses requires a holistic approach encompassing both ecosystem such as agriculture and forestry, which will increase the services and potential negative by-products, while simul- demand for drainage and thereby the need for implemen- taneously maintaining a catchment scale perspective tation of drainage mitigation measures to reduce the (Hewett et al. 2020). This require a catchment scale nutrient losses. understanding of ﬂow paths, taking into consideration all Acknowledgement This paper is a contribution from the Nordic important transport paths inﬂuencing the quality of ground- Centre of Excellence BIOWATER, which is funded by Nordforsk and surface water (Goeller et al. 2016). Consequently, under Project Number 82263. We thank Tinna Christensen for detailed information on local nutrient ﬂow pathways and graphical design and Anne Mette Poulsen for language assistance. drainage systems is highly needed. It should also be Open Access This article is licensed under a Creative Commons emphasised that the mitigation measures discussed in this Attribution 4.0 International License, which permits use, sharing, review only target drainage water, while other mitigation adaptation, distribution and reproduction in any medium or format, as measures, such as cover crops, target the water before it long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate leaves the root zone (Beckwith et al. 1998) or restored if changes were made. The images or other third party material in this wetlands that target water further downstream (Audet et al. article are included in the article’s Creative Commons licence, unless 2014). Consequently, it is essential that the drainage miti- indicated otherwise in a credit line to the material. If material is not gation measures should complement and not compensate included in the article’s Creative Commons licence and your intended for farm management practices producing high pesticide, N use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright or P leaching that inﬂuences other ﬂow paths such as holder. To view a copy of this licence, visit http://creativecommons. groundwater or surface runoff. Choosing the most appro- org/licenses/by/4.0/. priate and avoiding incompatible mitigation measures require collaboration between the different actors in the catchment to align the interests of all stakeholders (Hashemi and Kronvang 2020). To guide this decision process, we propose a further development of the sustain- REFERENCES ability index developed by Fenton et al. (2014), where Audet, J., L. Martinsen, B. Hasler, H. De Jonge, E. Karydi, N.B. weighting factors are assigned to relevant parameters. This Ovesen, and B. Kronvang. 2014. Comparison of sampling index, serving as a tool for stakeholder involvement, could methodologies for nutrient monitoring in streams: Uncertainties, be expanded with more ecosystem services and cost costs and implications for mitigation. Hydrology and Earth effectiveness adapted to local conditions. Furthermore, System Sciences 18: 4721–4731. Beckwith, C.P., J. Cooper, K.A. Smith, and M.A. Shepherd. 1998. application of a combination of mitigation measures may Nitrate leaching loss following application of organic manures to be more cost efﬁcient than introducing only one option. In sandy soils in arable cropping. Soil Use and Management 14: correspondence with this, a study by Hashemi and Kron- 123–130. vang (2020) found that it may be more cost effective to use Bilotta, G.S., A.M. Milner, and I.L. Boyd. 2014. Quality assessment tools for evidence from environmental science. Environmental a combination of targeted mitigation measures rather than a Evidence 3: 14. single option for reduction of the nitrate loading to aquatic Blowes, D.W., W.D. Robertson, C.J. Ptacek, and C. Merkley. 1994. ecosystems. Removal of agricultural nitrate from tile-drainage efﬂuent water The Author(s) 2020 www.kva.se/en Ambio 2020, 49:1820–1837 1835 using in-line bioreactors. Journal of Contaminant Hydrology 15: Gilliam, J.W., R.W. Skaggs, and S.B. Weed. 1979. Drainage control 207–221. to diminish nitrate loss from agricultural ﬁelds. Journal of Borenstein, M. 2009. Introduction to meta-analysis. Oxford: Wiley- Environmental Quality. Blackwell. Goeller, B., C. Febria, J. Harding, and M. Ar. 2016. Thinking beyond Canga, E., G.J. Heckrath, and C. Kjaergaard. 2016. Agricultural the bioreactor box: Incorporating stream ecology into edge-of- drainage ﬁlters. II. Phosphorus retention and release at different ﬁeld nitrate management. Journal of Environmental Quality 45: ﬂow rates. Water, Air, & Soil Pollution 227: 276. 151. Carstensen, M.V., C.D. Børgesen, N.B. Ovesen, J.R. Poulsen, S.K. Groh, T.A., L.E. Gentry, and M.B. David. 2015. Nitrogen removal Hvid, and B. Kronvang. 2019a. Controlled drainage as a targeted and greenhouse gas emissions from constructed wetlands mitigation measure for nitrogen and phosphorus. Journal of receiving tile drainage water. Journal of Environmental Quality Environmental Quality 48: 677–685. 44: 1001–1010. Carstensen, M.V., S.E. Larsen, C. Kjærgaard, and C.C. Hoffmann. Hashemi, F., and B. Kronvang. 2020. Multi-functional beneﬁts from 2019b. Reducing adverse side effects by seasonally lowering targeted set-aside land in a Danish catchment. Ambio. (This nitrate removal in subsurface ﬂow constructed wetlands. Journal issue). https://doi.org/10.1007/s13280-020-01375-z. of Environmental Management 240: 190–197. Hassanpour, B., S. Giri, W.T. Pluer, T.S. Steenhuis, and L.D. Christensen, J.H., K. Krishna Kumar, S.-I.A.E. Aldrian, M.D.C. Geohring. 2017. Seasonal performance of denitrifying bioreac- Cavalcanti, W. Dong, P. Goswami, J.K.K. Hall, A. Kitoh, et al. tors in the Northeastern United States: Field trials. Journal of 2013. Climate phenomena and their relevance for future Environmental Management 202: 242–253. regional climate change. Cambridge, New York: Cambridge Heinrich, C. 2012 BalticCOMPASS. BalticCOMPASS Work Package University Press. 4 coordination meeting, 8–9 March. Christianson, L., A. Bhandari, M. Helmers, K. Kult, T. Sutphin, and Hewett, C.J.M., M.E. Wilkinson, J. Jonczyk, and P.F. Quinn. 2020. R. Wolf. 2012a. Performance evaluation of four ﬁeld-scale Catchment systems engineering: An holistic approach to catch- agricultural drainage denitriﬁcation bioreactors in Iowa. Trans- ment management. WIREs Water 7: e1417. actions of the Asabe 55: 2163–2174. Higgins, J.P. T., D.G. Altman, P.C. Gøtzsche, P. Ju¨ni, D. Moher, A.D. Christianson, L.E., A. Bhandari, and M.J. Helmers. 2012b. A Oxman, J. Savovic, K.F. Schulz, et al. 2011. The Cochrane practice-oriented review of woodchip bioreactors for subsurface Collaboration’s tool for assessing risk of bias in randomised agricultural drainage. Applied Engineering in Agriculture 28: trials. British Medical Journal. 861–874. Hoagland, C.R., L.E. Gentry, M.B. David, and D.A. Kovacic. 2001. Christianson, L.E., C. Lepine, P.L. Sibrell, C. Penn, and S.T. Plant nutrient uptake and biomass accumulation in a constructed Summerfelt. 2017. Denitrifying woodchip bioreactor and phos- wetland. Journal of Freshwater Ecology 16: 527–540. phorus ﬁlter pairing to minimize pollution swapping. Water Hoffmann, C.C., S.E. Larsen, and C. Kjaergaard. 2019. Nitrogen Research 121: 129–139. removal in woodchip-based bioﬁlters of variable designs treating David, M.B., L.E. Gentry, R.A. Cooke, and S.M. Herbstritt. 2016. agricultural drainage discharges. Journal of Environmental Temperature and substrate control woodchip bioreactor perfor- Quality 48: 1881–1889. mance in reducing tile nitrate loads in East-Central Illinois. Jaynes, D.B. 2012. Changes in yield and nitrate losses from using Journal of Environmental Quality 45: 822–829. water management in central Iowa, United States. Journal of Soil De Haan, J., J.R. Van Der Schoot, H. Verstegen, and O. Clevering. and Water Conservation 67: 485–494. 2010. Removal of nitrogen leaching from vegetable crops in Jaynes, D.B., and T.M. Isenhart. 2014. Reconnecting tile drainage to constructed wetlands. International Society for Horticultural riparian buffer hydrology for enhanced nitrate removal. Journal Science (ISHS), Leuven, pp. 139–144. of Environmental Quality 43: 631–638. Diaz, R.J., and R. Rosenberg. 2008. Spreading dead zones and Jaynes, D.B., and T.M. Isenhart. 2019. Performance of saturated consequences for marine ecosystems. Science 321: 926–929. riparian buffers in Iowa, USA. Journal of Environmental Quality Drury, C.F., C.S. Tan, O. Gaynor, and T.O. Welacky. 1996. Inﬂuence 48: 289–296. of controlled drainage-subirrigation on surface and tile drainage Johannesson, K.M., K.S. Tonderski, P.M. Ehde, and S.E.B. Weisner. nitrate loss. Journal of Environmental Quality 25: 317. 2017. Temporal phosphorus dynamics affecting retention esti- Drury, C.F., C.S. Tan, W.D. Reynolds, T.W. Welacky, T.O. Oloya, mates in agricultural constructed wetlands. Ecological Engi- and J.D. Gaynor. 2009. Managing tile drainage, subirrigation, neering 103: 436–445. and nitrogen fertilization to enhance crop yields and reduce Kadlec, R.H., and R.L. Knight. 1996. Treatment wetlands. London: nitrate loss. Journal of Environmental Quality 38: 1193. Taylor & Francis Group. Drury, C.F., C.S. Tan, T.W. Welacky, W.D. Reynolds, T.Q. Zhang, Knowles, R. 1982. Denitriﬁcation. Microbiological Reviews 46: T.O. Oloya, N.B. Mclaughlin, and J.D. Gaynor. 2014. Reducing 43–70. nitrate loss in tile drainage water with cover crops and water- Koskiaho, J., P. Ekholm, M. Raty, J. Riihimaki, and M. Puustinen. table management systems. Journal of Environmental Quality 2003. Retaining agricultural nutrients in constructed wetlands: 43: 587. experiences under boreal conditions. Ecological Engineering 20: Fenton, O., M.G. Healy, F. Brennan, M.M.R. Jahangir, G.J. Lanigan, 89–103. K.G. Richards, S.F. Thornton, and T.G. Ibrahim. 2014. Perme- Kovacic, D.A., M.B. David, L.E. Gentry, K.M. Starks, and R.A. able reactive interceptors: blocking diffuse nutrient and green- Cooke. 2000. Effectiveness of constructed wetlands in reducing house gases losses in key areas of the farming landscape. The nitrogen and phosphorus export from agricultural tile drainage. Journal of Agricultural Science 152: 71–81. Journal of Environmental Quality 29: 1262–1274. Fink, D.F., and W.J. Mitsch. 2004. Seasonal and storm event nutrient Kronvang, B., E. Jeppesen, D.J. Conley, M. Søndergaard, S.E. Larsen, removal by a created wetland in an agricultural watershed. N.B. Ovesen, and J. Carstensen. 2005. Nutrient pressures and Ecological Engineering 23: 313–325. ecological responses to nutrient loading reductions in Danish Fleischer, S., A. Gustafson, A. Joelsson, J. Pansar, and L. Stibe. 1994. streams, lakes and coastal waters. Journal of Hydrology 304: Nitrogen removal in created ponds. Ambio 23: 349–357. 274–288. Kynkaanniemi, P., B. Ulen, G. Torstensson, and K.S. Tonderski. 2013. Phosphorus retention in a newly constructed wetland The Author(s) 2020 www.kva.se/en 1836 Ambio 2020, 49:1820–1837 receiving agricultural tile drainage water. Journal of Environ- Robertson, W.D., and L.C. Merkley. 2009. In-stream bioreactor for mental Quality 42: 596–605. agricultural nitrate treatment. Journal of Environmental Quality Lalonde, V., C.A. Madramootoo, L. Trenholm, and R.S. Broughton. 38: 230–237. 1996. Effects of controlled drainage on nitrate concentrations in Ross, J.A., M.E. Herbert, S.P. Sowa, J.R. Frankenberger, K.W. King, subsurface drain discharge. Agricultural Water Management 29: S.F. Christopher, J.L. Tank, J.G. Arnold, et al. 2016. A synthesis 187–199. and comparative evaluation of factors inﬂuencing the effective- Land, M., J.T.A. Verhoeven, W.J. Mitsch, W. Grannli, A. Grimvall, ness of drainage water management. Agricultural Water Man- C.C. Hoffmann, and K.S. Tonderski. 2016. How effective are agement 178: 366–376. created or restored freshwater wetlands for nitrogen and Rozemeijer, J.C., A. Visser, W. Borren, M. Winegram, Y. Van Der phosphorus removal? A systematic review. Environmental Velde, J. Klein, and H.P. Broers. 2015. High frequency Evidence monitoring of water ﬂuxes and nutrient loads to assess the Landbrugsstyrelsen. 2019. Miniva˚domra˚de-ordningen 2019. Etabler- effects of controlled drainage on water storage and nutrient ing af a˚bne miniva˚domra˚der og miniva˚domra˚der med ﬁlter transport. Hydrology and Earth System Sciences 20: 347–358. matrice. Satchithanantham, S., R.S. Ranjan, and P. Bullock. 2014. Protecting Marttila, H., A. Lepisto¨, A. Tolvanen, M. Bechmann, K. Kyllmar, A. water quality using controlled drainage as an agricultural BMP Juutinen, H. Wenng, E. Skarbøvik, et al. 2020. Potential impacts for potato production. Transactions of the Asabe 57: 815–826. of a future Nordic bioeconomy on surface water quality. Ambio. Schott, L., A. Lagzdins, A. Daigh, K. Craft, C. Pederson, G. (This issue). https://doi.org/10.1007/s13280-020-01355-3. Brenneman, and M. Helmers. 2017. Drainage water management Mendes, L.R.D., K. Tonderski, B.V. Iversen, and C. Kjaergaard. effects over ﬁve years on water tables, drainage, and yields in 2018. Phosphorus retention in surface-ﬂow constructed wetlands southeast Iowa. Journal of Soil and Water Conservation 72: 251. targeting agricultural drainage water. Ecological Engineering Schwarzer, G. 2019. General package for meta-analysis. 120: 94–103. Shih, R., W.D. Robertson, S.L. Schiff, and D.L. Rudolph. 2011. Miller, P.S., J.K. Mitchell, R.A. Cooke, and B.A. Engel. 2002. A Nitrate controls methyl mercury production in a streambed wetland to improve agricultural subsurface drainage water bioreactor. Journal of Environmental Quality 40: 1586. quality. Transactions of the Asae 45: 1305–1317. Skaggs, R. W., N. R. Fausey, and R. O. Evans. 2012. Drainage water Mitsch, W., and S.E. Jørgensen. 1989. Ecological engineering: An management. Journal of Soil and Water Conservation 67. introduction to ecotechnology. New York: Springer. Skaggs, R.W., and J. Van Schilfgaarde. 1999. Agricultural drainage. Mitsch, W.J., J.W. Day, J.W. Gilliam, P.M. Groffman, D.L. Hey, Madison, WI: American Society of Agronomy, Crop Science G.W. Randall, and N. Wang. 2001. Reducing nitrogen loading to Society of America, Soil Science Society of America. the Gulf of Mexico from the Mississippi River Basin: Strategies Søvik, A.K., and P.T. Mørkved. 2008. Use of stable nitrogen isotope to counter a persistent ecological problem: Ecotechnology—the fractionation to estimate denitriﬁcation in small constructed use of natural ecosystems to solve environmental problems— wetlands treating agricultural runoff. Science of the Total should be a part of efforts to shrink the zone of hypoxia in the Environment 392: 157–165. Gulf of Mexico. BioScience 51: 373–388. Steffen, W., K. Richardson, J. Rockstro¨m, S.E. Cornell, I. Fetzer, Ng, H.Y.F., C.S. Tan, C.F. Drury, and J.D. Gaynor. 2002. Controlled E.M. Bennett, R. Biggs, S.R. Carpenter, et al. 2015. Planetary drainage and subirrigation inﬂuences tile nitrate loss and corn boundaries: guiding human development on a changing planet. yields in a sandy loam soil in Southwestern Ontario. Agriculture, Science 347: 1259855. Ecosystems & Environment 90: 81–88. Sunohara, M.D., E. Craiovan, E. Topp, N. Gottschall, C.F. Drury, and O’Geen, A.T., R. Budd, J. Gan, J.J. Maynard, S.J. Parikh, and R.A. D.R. Lapen. 2014. Comprehensive nitrogen budgets for con- Dahlgren. 2010. Mitigating nonpoint source pollution in agri- trolled tile drainage ﬁelds in Eastern Ontario, Canada. Journal of culture with constructed and restored wetlands. Advances in Environmental Quality 43: 617. Agronomy 108: 1. Tan, C.S., and T.Q. Zhang. 2011. Surface runoff and sub-surface Ockenden, M.C., C. Deasy, J.N. Quinton, A.P. Bailey, B. Surridge, drainage phosphorus losses under regular free drainage and and C. Stoate. 2012. Evaluation of ﬁeld wetlands for mitigation controlled drainage with sub-irrigation systems in southern of diffuse pollution from agriculture: Sediment retention, cost Ontario. Canadian Journal of Soil Science 91: 349–359. and effectiveness. Environmental Science & Policy 24: 110–119. Tanner, C.C., and J.P.S. Sukias. 2011. Multiyear nutrient removal Poe, A.C., M.F. Piehler, S.P. Thompson, and H.W. Paerl. 2003. performance of three constructed wetlands intercepting tile drain Denitriﬁcation in a constructed wetland receiving agricultural ﬂows from grazed pastures. Journal of Environmental Quality runoff. Wetlands 23: 817–826. 40: 620–633. R Core Team. 2019. R: A language and environment for statistical Tanner, C. C., J. P. S. Sukias, and C. R. Yates. 2010. New Zealand computing. Vienna: R Core Team. guidelines for constructed wetland treatment of tile drainage. Rakovic, J., M.N. Futter, K. Kyllmar, K. Rankinen, M.I. Stutter, J. Tomer, M.D., D.B. Jaynes, S.A. Porter, D.E. James, and T.M. Vermaat, and D. Collentine. 2020. Nordic Bioeconomy Path- Isenhart. 2017. Identifying riparian zones best suited to instal- ways: Future narratives for assessment of water-related ecosys- lation of saturated buffers: A preliminary multi-watershed tem services in agricultural and forest management. Ambio. assessment. In Precision conservation: geospatial techniques (This issue). https://doi.org/10.1007/s13280-020-01389-7. for agricultural and natural resources conservation, ed. J. Del- Reinhardt, M., R. Gachter, B. Wehrli, and B. Muller. 2005. gado, G. Sassenrath, and T. Mueller, 83–93. Madison, WI: Phosphorus retention in small constructed wetlands treating American Society of Agronomy and Crop Science Society of agricultural drainage water. Journal of Environmental Quality America Inc. 34: 1251–1259. Tonderski, K.S., B. Arheimer, and C.B. Pers. 2005. Modeling the Reinhardt, M., B. Muller, R. Gachter, and B. Wehrli. 2006. Nitrogen impact of potential wetlands on phosphorus retention in a removal in a small constructed wetland: An isotope mass balance Swedish catchment. Ambio 34: 544–551. approach. Environmental Science and Technology 40: Tournebize, J., C. Chaumont, C. Fesneau, A. Guenne, B. Vincent, J. 3313–3319. Garnier, and U. Mander. 2015. Long-term nitrate removal in a buffering pond-reservoir system receiving water from an agri- cultural drained catchment. Ecological Engineering 80: 32–45. The Author(s) 2020 www.kva.se/en Ambio 2020, 49:1820–1837 1837 Van’t Veen, S. G. W., B. Kronvang, D. Zak, N. Ovesen, and H. Address: Department of Bioscience, Aarhus University, Vejlsøvej 25, Jensen. 2019. Intelligente bufferzoner - Notat fra DCE. Nationalt 8600 Silkeborg, Denmark. Center for Miljø og Energi. e-mail: email@example.com Vymazal, J. 2017. The use of constructed wetlands for nitrogen removal from agricultural drainage: A review. Scientia Agricul- Fatemeh Hashemi is Post Doc at the Department of Bioscience of turae Bohemica 48: 82–91. Aarhus University. Her research activities focus on both land use and Wesstro¨m, I., A. Joel, and I. Messing. 2014. Controlled drainage and climate change effects on water and Nitrate transport at catchment subirrigation: A water management option to reduce non-point scale using numerical modelling. source pollution from agricultural land. Agriculture, Ecosystems Address: Department of Bioscience, Aarhus University, Vejlsøvej 25, & Environment 198: 74–82. 8600 Silkeborg, Denmark. Wesstro¨m, I., and I. Messing. 2007. Effects of controlled drainage on e-mail: firstname.lastname@example.org N and P losses and N dynamics in a loamy sand with spring crops. Agricultural Water Management 87: 229–240. Carl Christian Hoffmann is a senior researcher at the Department of Willardson, L.S., B. Grass, G.L. Dickey, and J.W. Bailey. 1970. Drain Bioscience of Aarhus University and his research interest is mitiga- installation for nitrate reduction. Environment Science Technol- tion of nutrient losses from agriculture with focus on restored and ogy 31: 2229–2236. constructed wetlands. Williams, J.B. 2002. Phytoremediation in wetland ecosystems: Address: Department of Bioscience, Aarhus University, Vejlsøvej 25, progress, problems, and potential. Critical Reviews in Plant 8600 Silkeborg, Denmark. Sciences 21: 607–635. e-mail: email@example.com Williams, M.R., K.W. King, and N.R. Fausey. 2015. Drainage water management effects on tile discharge and water quality. Dominik Zak is a senior researcher at the Department of Bioscience Agricultural Water Management 148: 43–51. of Aarhus University and his research interest is wetland biogeo- Woli, K.P., M.B. David, R.A. Cooke, G.F. Mcisaac, and C.A. chemistry with focus on peatland restoration. Mitchell. 2010. Nitrogen balance in and export from agricultural Address: Department of Bioscience, Aarhus University, Vejlsøvej 25, ﬁelds associated with controlled drainage systems and denitri- 8600 Silkeborg, Denmark. fying bioreactors. Ecological Engineering 36: 1558–1566. e-mail: firstname.lastname@example.org Zak, D., B. Kronvang, M.V. Carstensen, C.C. Hoffmann, A. Joachim Audet is a researcher at the Department of Bioscience of Kjeldgaard, S.E. Larsen, J. Audet, S. Egemose, et al. 2018. Aarhus University and his research interests are nutrient cycling and Nitrogen and phosphorus removal from agricultural runoff in greenhouse gas emissions from freshwater ecosystems. integrated buffer zones. Environmental Science and Technology 52: 6508–6517. Address: Department of Bioscience, Aarhus University, Vejlsøvej 25, Zak, D., M. Stutter, H.S. Jensen, S. Egemose, M.V. Carstensen, J. 8600 Silkeborg, Denmark. Audet, J.A. Strand, P. Feuerbach, C.C. Hoffmann, et al. 2019. An e-mail: email@example.com assessment of the multifunctionality of integrated buffer zones in Northwestern Europe. Journal of Environmental Quality 48: Brian Kronvang is professor at the Department of Bioscience of 362–375. Aarhus University. His research activities focus on catchment science Zhang, T.Q., C.S. Tan, Z.M. Zheng, T.W. Welacky, and W.D. and management – mainly focusing on ﬂuxes and sinks of matter Reynolds. 2015. Impacts of soil conditioners and water (sediment, organic matter, nutrients, pesticides, heavy metals) from table management on phosphorus loss in tile drainage from a land to coastal areas. clay loam soil. Journal of Environmental Quality 44: 572–584. Address: Department of Bioscience, Aarhus University, Vejlsøvej 25, 8600 Silkeborg, Denmark. e-mail: firstname.lastname@example.org Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afﬁliations. AUTHOR BIOGRAPHIES Mette Vodder Carstensen (&) is PhD fellow at the Department of Bioscience of Aarhus University. Her research activities focus on mitigation of nutrient loss from agricultural drainage systems using engineered systems. The Author(s) 2020 www.kva.se/en
"Ambio" – Springer Journals
Published: Nov 3, 2020
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera