Performance of a Woodchip Bioreactor for the Treatment of Nitrate-Laden Agricultural Drainage Water in Northeastern Germany

: Reactive barriers, such as denitrifying bioreactors, have been identiﬁed as a clean-up option for nutrient-laden agriculture runo ﬀ . Here we tested a 20 m long, 3.75 m wide and 2.2 m deep woodchip bioreactor receiving tile drainage water from a 5.2 ha ﬁeld site, aiming at testing the hydraulic functioning of a dual-inlet system and quantifying its impact on nutrient loads (nitrogen, reactive phosphorus, organic carbon) in a region with a drainage season taking place in the hydrological winter (November to April). The hydraulic conditions in the dual-inlet bioreactor system developed di ﬀ erently than expected; asymmetric ﬂow rates led to long average hydraulic retention times and a highly dispersed residence time distribution, which was revealed by a bromide tracer test. With a nitrate load reduction of 51 to 90% over three drainage seasons, the woodchip bioreactor proved at the same time to be very e ﬀ ective under the winter conditions of northeastern Germany. The bioreactor turned from an orthophosphate source in the ﬁrst year of operation into an orthophosphate sink in the second and third year, which was not expected because of anoxic conditions (favorable for denitriﬁcation) prevailing within the woodchips. Besides an e ﬃ cient nutrient retention, the woodchip bioreactor contributed to the total organic carbon load of receiving waters, which impairs the overall positive role of bioreactors within intensively agriculturally used landscapes. We consider this promising low-maintenance biotechnology particularly suitable for single drainage pipes with high discharge and high nitrate concentrations.


Introduction
The nitrogen load of agricultural tile drainage water can contribute to eutrophication of surface waters, particularly in naturally nitrogen-limited aquatic ecosystems [1,2]. The tile drainage water represents an accelerated runoff component from agricultural fields, which can transport high amounts of nitrate (NO 3 − ). The total export of nitrogen from field sites via the tile drainage pathway is a function of both the fertilization regime and the specific hydrological situation of a given year. More recently, it could be shown that discharge intensity overrides fertilization effects [3]. Even in years with little or no fertilizer application, nitrogen export rates might be high because of long-term nitrogen accumulation in the soil and high-volume fluxes. In order to overcome the dependence of nutrient export from agricultural land from the hydrological regime, constructed wetlands for the treatment of nitrate-laden tile drainage water have been proposed [4]. In case of limited available space, a denitrifying bioreactor, a rather technical variant, which falls into the category of reactive barriers, may be an option. It consists of an artificial ditch filled with organic substrates (e.g., woodchips) through which the drainage water is passed before it discharges to the surface water. The anoxic and carbon-enriched conditions in the bioreactor favor the occurrence of the microbially mediated denitrification converting nitrate to the non-hazardous dinitrogen gas (N 2 ). Due to the environmentally problematic intermediate products nitrite (NO 2 − ), nitric oxide (NO) and nitrous oxide (N 2 O), the completion of the denitrification process is an important factor for the bioreactor's positive performance, which furthermore depends on its impact on other water quality-relevant substances. The main concern lies on phosphorus and (organic) carbon, for which increased concentrations have been observed in the outflow of bioreactors [5,6] and which can also lead to eutrophication and oxygen deficiency in surface waters. The nitrate removal efficiency depends on different factors such as hydraulic retention time (HRT), nitrate and dissolved oxygen (DO) load and temperature [7][8][9]. The HRT depends on the hydraulic performance of the bioreactor, which can be evaluated with the help of tracer tests [10,11]. The field study presented here aimed at testing the operability of a woodchip bioreactor under northeastern German climatic conditions where discharge is limited to the winter season. The specific objectives were to evaluate (i) the hydraulic performance and possible hydraulic anomalies due to two inlets and (ii) the integrative impact of the system on nutrient loads (nitrogen (N), reactive phosphorus (o-PO 4 -P), organic carbon (OC)) heading towards the surface water. Three years of operation provided the database that was analyzed to examine the objectives.

Materials and Methods
The woodchip bioreactor was installed at the edge of a systematically tile-drained field site ( Figure 1a) at the brook "Saaler Bach" in Wiepkenhagen (Germany) in fall 2017. Two adjacent drainage pipes with a catchment of 2.9 ha (inlet 1) and 2.3 ha (inlet 2) were selected to load the facility with the help of diversion shafts (Figure 1c). The woodchip bioreactor was sized according to [12], envisaging an inflow of 12.5 m 3 d −1 for each inlet (total inflow 25 m 3 d −1 ) and a minimum theoretical (also termed "nominal") hydraulic retention time t n of 0.5 days, which is defined as with V sat as the saturated bioreactor volume, n as the porosity of the woodchips (generally estimated to be 0.7 [13]) and Q as the flow rate in the bioreactor. The trapezoidal bioreactor, filled with woodchips with a particle size of the main fraction of 5. The nitrate removal efficiency NRE (%) for a complete drainage season and the daily nitrate removal rate NRR (g m −3 d −1 ) were calculated from the NO 3 − -N loads in the inflow (NO 3 N in ) and outflow (NO 3 N out ): Further water quality parameters are monitored weekly, such as total organic carbon (TOC), total bound nitrogen (TN b ), i.e., all organic and inorganic nitrogen compounds except N 2 , and orthophosphate (PO 4 3− ), with a TOC/TN b -analyzer and a photometer. Organically bound nitrogen (N org ) is determined as the difference between TN b and (NO 3 concentrations turned out to be negligible in the system. Furthermore, on-site air temperature, as well as water temperature and redox potential at the bottom of the woodchip bioreactor, is constantly monitored (15 min time interval). Redox potential values are corrected to the standard hydrogen potential E 0 .
To get a deeper insight into the hydraulics of the system and the residence time of NO 3 − , a tracer test with potassium bromide (KBr) was carried out in the drainage season 2019/2020 during quasi-stationary low-flow conditions (Q ≈ 4.5 m 3 d −1 ). For this purpose, 1121 g KBr (=750 g Br − ) was dissolved in 5 L water and poured in inlet 1 of the woodchip bioreactor (Figure 1c). At the outlet, water samples were collected every 4 h for days 1-4, every 6 h for days 5-9, every 12 h for days 10-14 and every 24-72 h for days 15-40 and analyzed for Br − with ion chromatography to monitor the complete residence time distribution (RTD). The measured concentrations were reduced by the natural Br − background concentration of 0.305 mg L −1 . With the help of the mean tracer residence time t mean (d), the accuracy of the previously calculated t n can be verified [10]: with C i as the Br − -concentration in the i-th sample, ∆t i as the respective sampling interval and t i as the appearance time. If the complete saturated bioreactor volume is hydraulically active (no stagnant zones) and no retardation of bromide occurs, t mean should be equal to t n [11]. The water level during the tracer test was~1.55 m, resulting in a V sat of 61 m 3 for the complete bioreactor. Since the tracer was injected in inlet 1 and the outlet is located in the middle of the bioreactor, approximately half of the bioreactor (30.5 m 3 ) was expected to be hydraulically involved in the tracer transport. Several parameters have been proposed to evaluate the hydraulic performance of the bioreactor such as the hydraulic efficiency λ, the short-circuiting indicator S and the Morrill Dispersion Index MDI [10]: with t p as the peak appearance time (d) and t 10 , t 16 , t 50 and t 90 as the appearance time (d) of 10, 16, 50 and 90% of the cumulative tracer mass. Generally, hydraulic conditions similar to plug-flow (i.e., without considerable preferential flow or dispersion) are considered well-suited for the functioning of the bioreactor. Thus, parameter values of λ > 0.75, S ≈ 1 and MDI < 2 are classified as ideal [10,11].

Hydraulic Conditions in the Woodchip Bioreactor
Out of the three monitored drainage seasons, the seasons 2017/2018 and 2019/2020 are considered typical with generated run-off from November to April, while the drainage season 2018/2019 with run-off from February to April represents a special case due to the extreme dryness in the year 2018 ( Figure 2). The hydraulic conditions in the woodchip bioreactor differed from the envisaged flow conditions, i.e., symmetric inflow from two sides to the outlet in the middle of the bioreactor. The inlet 2 did not contribute to the flow rate due to an unknown hydraulic connection to the neighboring drainage system, which was not visible on the available drainage system map. This led to the discharge of the run-off of inlet 2 towards the neighboring drainage system, where thus more drainage water was discharged than previously observed. The outflow at the outlet consisted almost exclusively of inflow via inlet 1, which facilitated the calculation of the nitrate balance of system but led ultimately to an oversizing of the bioreactor. To make sure a woodchip bioreactor works hydraulically as planned, it is recommended to choose a single inlet with high inflow and high nitrate concentrations to load the bioreactor. If multiple inlets are envisaged, the distance to the neighboring drainage pipe should be as high as possible to reduce the risk of an unknown subsurface hydraulic connection. The woodchip bioreactor causes back pressure in the drainage system, which can lead to changes in the flow pattern. Furthermore, control elements such as gate valves on critical spots could be included to allow subsequent and easy adjustments to the hydraulic system. The RTD (Figure 3) of the tracer test resulted in a t mean of 6.75 days, indicating that more volume of the bioreactor was involved in the tracer flow from inlet 1 than expected, since an active flow volume of half of the bioreactor would have led to a t n of 4.75 days. Here, we assumed t n = t mean resulting in a saturated volume of 43.4 m 3 corresponding to 71% of the saturated bioreactor volume, which was, in turn, used as best estimate for the normalization of the RTD (Figure 3) and t n calculation (Table 1), accepting inaccuracies regarding the effective porosity of the woodchips. The tracer recovery at the outlet was 61%, which is low even for a field study and suggests, to some degree, an interference of the northern inlet and the connected drainage system with the tracer test.  Under the monitored low-flow conditions, the minimum residence time for Br − (thus, for NO 3 − ) was~24 h, which is even longer than the commonly recommended t n for sizing a woodchip bioreactor [12]. The latest increased bromide concentrations in the outlet were observed after~30 days, which shows that the residence time of NO 3 − can vary enormously. The shape of the RTD is very similar to previous tracer studies [10,11] with a steep increase at the beginning, a subsequent peak at~0.5 PV and a less steep decrease with tailing. The wide range of the RTD clearly shows that the bioreactor did not act as a plug-flow reactor (MDI >> 2). As stated for constructed wetlands by [14], a woodchip bioreactor will never show ideal plug-flow because the pore network of the woodchip will always cause a considerable hydrodynamic dispersion of the solute front, as can be observed in soils [15]. The plug-flow oriented target range of the performance parameters was derived for reactor types such as disinfection or sedimentation beds [16], which do not have sediment filling and require the absolute fulfilment of t n to function properly. It does not seem realistic to apply the same standards to a woodchip bioreactor and thus a poor hydraulic performance may not necessarily imply a poor nitrate removal performance. For the evaluation of the hydraulic performance, the parameters, which purely target at preferential flow (e.g., S or λ), should be preferred compared to parameters, which also target at dispersion (e.g., MDI). In this case, the resulting S-value of 0.56 indicates moderate short-circuiting in the third drainage season; it is likely that some pore water regions in the woodchips are more mobile than others (caused for example by local compaction or bio-clogging) and this might also change further over time as the woodchips are being decomposed, which in turn leads to a change of the pore network and the saturated hydraulic conductivity [17]. In this particular case, the minimum residence time due to short-circuiting was still sufficiently long for the nitrate removal, but for bioreactors with shorter t n , the replacement of the woodchips after several years might be more necessary due to hydraulic malfunctions than due to carbon-supply limitations [17].

Impact of the Woodchip Bioreactor on Nutrient Loads (N, o-PO 4 -P, OC)
With values from 51 to 90% (Table 1), the nitrate removal efficiency of the woodchip bioreactor was very high despite average water temperatures only slightly higher than the minimum of~2 • C and far from the optimum (25-35 • C) for denitrification [9]. This finding is consistent with previous studies that woodchip bioreactors still remove some nitrate at low temperatures [18,19]. Lower average air temperatures show that temperature amplitudes were mitigated in the bioreactor due to an insulation effect caused by the depth of the bioreactor and the covering geotextile, so that temporary sub-zero air temperatures did not bring nitrate removal to a standstill. Extra heat might also be generated by the anaerobic decomposition processes as it was observed in systems such as waste landfills [20]. The positive average redox potential at inlet 1 indicates the continuous inflow of high energy-providing electron acceptors, such as O 2 and NO 3 − , and is in the redox range, in which denitrification generally occurs (~+231 mV [9]). The negative average redox potential at inlet 2 is clearly below the redox range of denitrification and reflects the inlet's hydraulic malfunctioning, i.e., no inflow of nitrate-laden drainage water. It is likely that the high average HRT (t n ≥ 3.1 days) compensated the effect of low temperatures on the nitrate removal efficiency. The average NRRs (0.75-2.31 g m −3 d −1 ) are in the lower range of previously observed removal rates [5,7], because the NRR is normalized to V sat (see Equation (3)), which in this case was sized for a higher flow rate than ultimately achieved. Yet no statement about the temporal development of the nitrate removal rate is possible, since the lower NRR in 2019/2020 is related to lower NO 3 − concentrations in the inflow ( Table 1). The high HRT did not prevent the release of NO 2 − in the first drainage season, which indicates an incomplete denitrification. In the first month of operation, NO 2 − -N concentrations up to 10.8 mg L −1 were observed in the outflow and dropped afterwards to <0.005 mg L −1 ; in the following drainage seasons NO 2 − -N concentrations were also increased in the first month, but never exceeded 0.6 mg L −1 and were of minor importance for the overall N balance (see Figure 4b for drainage season 2019/2020). This coincides with remaining NO 3 − -N concentrations in the outflow in the first month, which did not reoccur during the rest of the drainage season. It seems that in this initial stage with high inflow and high NO 3 − -N concentrations (Figure 4a), denitrification was incomplete. During the remaining months, the denitrification was complete and only N org was released from the woodchip bioreactor. However, compared to TOC, the release potential of the bioreactor for N org is low due to the high C/N ratio (>200 [17]) of the woodchips.
In the drainage season 2019/2020, the N org -load in the outflow was only increased by 4% compared to the inflow, while for TOC an increase of 45.6% was observed. The decomposition of the woodchips leads to the release of organic and inorganic decomposition products [5,6]. During the first drainage season after installation of the woodchip bioreactor TOC and PO 4 3− emissions were very high, but in case of TOC the release was lessened in the following drainage seasons (Table 1). A possible reason is the decrease in easily and quickly degradable organic compounds within the woodchip matrix. For the PO 4 3− the concentrations were considerably increased (up to 4 mg L −1 compared to up to a maximum of 0.28 mg L −1 in the inflow) right after putting the bioreactor into service, but afterwards the woodchip bioreactor acted as a sink for PO 4 3− -possibly due to consumption of PO 4 3− by microorganisms [9]. The substance release in following drainage seasons is expected to be similar to drainage season 2019/2020 ( Figure 5), which would corroborate the overall positive environmental effect of the system and support the suggestion by [21] to use weathered woodchips as filling material to prevent the strong substance release in the first drainage season.

Conclusions
In this study, a high nitrate removal efficiency was obtained with a woodchip bioreactor for agricultural drainage water in northeastern Germany indicating the usability of this measure type in regions with drainage seasons in the hydrologic wintertime. Although temperatures did range near the minimum temperature required for denitrification, it seems that a longer hydraulic retention time can compensate low temperature effects. Hydraulic retention times were longer than envisaged due to a hydraulic malfunction of the second inlet, which is why more precaution needs to be taken regarding subsurface flow conditions in the drainage pipe system for multiple-inlet bioreactors than in the case of a single-inlet bioreactor. The residence time distribution resulting from a tracer test showed a dispersed shape with a minimum residence time still sufficient for a complete nitrate removal, although short-circuiting parameters indicate moderate preferential flow in the 3-year-old woodchips. Negative environmental impact due to nitrite (NO 2 − ), total organic carbon (TOC) and orthophosphate (PO 4 3− ) release was observed in the first month after installation of the woodchip bioreactor. In the following drainage seasons, the bioreactor turned into a sink for orthophosphate and the TOC and nitrite release faded.