MATERIAL AND METHODS Study site and sampling

Measurements were conducted at the Long Term Ecosystem Research Catchment (LTER) near the town of Pleine–Fougères, about 20 km from Rennes, France. The 2.5 km stream studied here drained a small agricultural sub-catchment of the LTER, known locally as Vilqué. It is surrounded by mostly agricultural land including dairy and low intensity crops. Fourteen 1L samples for radon and nitrate analysis were taken from the headwater areas to its confluence with l'Hermitage at intervals between 100 m and 500 m. Radon was measured in the field using a RAD7 radon in air detector, with each sample being purged for 5 minutes and then counted for 1 h. Nitrate was measured by ion chromatography.

The groundwater flux, hyporheic depth and mean residence times in the HZ was calculated using the FINIFLUX model (Frei and Gilfedder, 2015). This is a 1D finite element mass-balance model that includes terms such as radon loss to the atmosphere, radioactive decay, groundwater discharge and the hyporheic parameters. The hyporheic parameters and groundwater discharge are fitting variables in the mass-balance. The newest version of FINIFLUX uses three different residence time distributions (RTDs, exponential, gamma and power law distributions) to represent storage in the HZ.

The nitrate loss is based on the UPERflux mass-balance equations for the HZ (Pittroff et al., 2017). In its most basic form the mass-balance encompasses the concentration of nitrate in the stream, water flux through the sediment from the stream, RTD and the kinetics for nitrate reduction in the sediments. It does not include the groundwater nitrate flux.

𝑇_𝑁𝑂3= 𝑞_ℎ∆𝑥(𝐶_𝑖𝑛− 𝐶_𝑜𝑢𝑡)

(Eq. 1)

With TNO3is the nitrate loss, qh the water flux though the hyporheic sediments, x is the reach length,

and Cin and Cout are the concentrations of nitrate entering and leaving the HZ respectively. Cin is

measured in stream water, while Cout is calculated by convolving Cin, reaction kinetics (first order, or

Monod equation) and the RTD:

𝐶_𝑜𝑢𝑡(𝜏) = ∫ 𝐶₀∞ 𝑖𝑛𝑒−𝜆𝑧𝑅𝑇𝐷𝑑𝑧

(Eq. 2)

The new model has been extended E.q 2 to include a factor a which describes the RT of surface water in the oxic part of the HZ. This ‘lag’ essentially cuts away a part of the RTD, ranging from 0 to a hours, and simulates explicitly the exposure time of nitrate to conditions suitable for reduction. We have varied a from 0 to 100 h to capture a range of possible values, but based on kinetics of oxygen reduction

a is likely to range between 2-6 hours.

RESULTS AND DISCUSSION

The radon results show clearly that there is a hot-spot for groundwater discharge to the Vilqué stream

at around 500 m from the first sampling location, with activities increasing from 200 Bq m-3 _{to 15 kBq}

m-3_{. This was also observed in the field as spring entering the stream from the bank. The FINIFLUX}

model fitted the measured data very well (r=0.9), with a total groundwater discharge of 430 m3 _d-1

over 2.5 km. The mean residence time varied depending on the RTD, but in a rather narrow range from 2.3 to 2.48 hours. Over the length of the river measured nitrate concentrations decreased from 44 mg/l to 32 mg/l, due to a combination of dilution by groundwater and reduction in the HZ. The loss of nitrate attributed to reduction in the HZ using UPERflux with a=0, exponential RTD and first order

kinetics was 4.8 kg d-1_{. Implementing an a with a value of 2 h reduced loss to 2 kg d}-1 _{due to the}

inhibitory effect of the oxic zone. To systematically test the influence of RT and a, we varied these parameters between published values for 1-3rd order streams. We found that the loss of nitrate is highly sensitive to both the RT and a, with only a small combination of values (RT 1-20 h, a<2) leading to significant nitrate loss (Figure 1). It is also clear that the nitrate loss is inversely related to residence time. Thus to maximise nitrate mass loss in stream networks a combination of short residence times and low values of a are needed. This work shows that it is vital to consider the exposure of chemicals such as nitrate to conditions conducive to reaction in the HZ, rather than only relying on the residence time in the subsurface.

Figure 1: Nitrate mass loss in the hyporheic zone as a function of lag time (portion of RTD allocated to the oxic zone) and mean residence time.

REFERENCES

Frei, S., Gilfedder, B.S., 2015. Technical Note: FINIFLUX an implicit Finite Element model for

quantification of groundwater fluxes and hyporheic exchange in streams and rivers using Radon. Water Resources Research 51, 6776-6786.

Gomez-Velez, J.D., Harvey, J.W., Cardenas, M.B., Kiel, B., 2015. Denitrification in the Mississippi River network controlled by flow through river bedforms. Nature Geoscience 8, 941–945;

doi:910.1038/ngeo2567.

Kiel, B., Cardenas, M.B., 2014. Lateral hyporheic exchange throughout the Mississippi River network. Nature Geoscience, 413–417; doi:410.1038/ngeo2157.

Pittroff, M., Frei, S., Gilfedder, B.S., 2017. Quantifying nitrate and oxygen reduction rates in the

hyporheic zone using 222_{Rn to upscale biogeochemical turnover in rivers Water Resources Research}

SEARCHING FOR THE CONTROL MECHANISMS OF NITROGEN REMOVAL IN A MEDITERRANEAN