Rivers export high loads of nitrogen from inland catchments to the marine environment (Smith et al., 2006). The ecological and economic problems caused by eutrophication of coastal and riverine ecosystems have been recognised years ago (Patsch and Radach, 1997; Artioli et al., 2008; Skogen et al., 2014). Decades of nutrient studies have revealed that rivers cycle rather than only transport nutrients (Garcia-Ruiz et al., 1998a; Seitzinger et al., 2002; Galloway et al., 2003). In agriculturally dominated areas, in-stream processes may for example retain up to 38 % of nitrate (NO3-) and 48 % of soluble reactive phosphate (SRP) inputs (Mortensen et al., 2016). The hyporheic zone, the subsurface region of streams and rivers that exchanges water, solutes and particles with the surface (Valett et al., 1993) and may mix streamwater during the transport through the sediments with underlying groundwater (Triska et al., 1989; Fleckenstein et al., 2010; Trauth et al., 2014), is one key compartment for in-stream nutrient cycling (Fischer et al., 2005; Zarnetske et al., 2011b; Basu et al., 2011; Stewart et al., 2011). For instance, denitrification, the anaerobic reduction of NO3- to gaseous N2 and in most river systems the dominant dissimilatory process which removes N out of the system (Laursen and Seitzinger, 2002; Bernot and Dodds, 2005; Lansdown et al., 2012; Kunz et al., 2017), often exclusively happens at “reactive sites” in the hyporheic zone (Duff and Triska, 1990; Rode et al., 2015). In addition to biological nutrient uptake, intermediate physical storage in the hyporheic zone disperses the propagation of pollutant and nutrient spikes which could be harmful for receiving water bodies (Runkel, 2007; Brookshire et al., 2009; Covino et al., 2010; Findlay et al., 2011). For those reasons, it is of interest to quantify the amount of nutrients actually reaching the reactive sites in the subsurface collateral to the processes they undergo there (Seitzinger et al., 2006; Zarnetske et al., 2012). Transport rates of water and nutrients from the surface to the subsurface could be attributed to water level, sediment properties and various other hydrological, biological, chemical and physical factors (Böhlke et al., 2009; Boano et al., 2014; Trauth et al., 2015). The complex interactions between these influencing factors and the temporal variability and local heterogeneity of hyporheic processes often cause high uncertainties in quantitative models. However, due to methodological restrictions, experimental investigations of nutrient dynamics in the hyporheic zone are rare and commonly exclusively of qualitative nature (Mulholland et al., 1997; Grant et al., 2014).

Attempts to quantify hyporheic nutrient processing rates have primarily been based on benthic chamber and incubation experiments (Findlay et al., 2011; Kessler et al., 2012). Those laboratory (mesocosm and flume) experiments can estimate the denitrification potential of the substrates, usually via denitrification enzyme assays. However, the realised denitrification rates depend equally on environmental and hydrological conditions rather than on substrate type or denitrification potential alone (Findlay et al., 2011). Thus, owing to the high variability and complexity of natural systems, hyporheic transport of nutrients cannot satisfactorily be mimicked in artificial set-ups (Cook et al., 2006; Grant et al., 2014).

Direct in-stream measurements of nutrient dynamics based on whole-stream tracer injections, mass balances (McKnight et al., 2004; Böhlke et al., 2009) and more recently high-resolution time series from automated sensors (Pellerin et al., 2009; Hensley et al., 2014; Rode et al., 2016a, b) can be used for determining general uptake rates on the reach scale, but do not allow to identify the reaction sites (hyporheic versus in channel or algal canopies) or specific local uptake processes (Ensign and Doyle, 2006; Ruehl et al., 2007). Further, in-stream measurements exclusively account for water which is re-infiltrating into the main stem after passage through the hyporheic zone. Under loosing conditions, where most of the nutrient-influx is flowing towards the groundwater, processing rates in the subsurface cannot be observed in the surface water. Likewise, if groundwater is contributing significantly to surface water chemistry, surface water mass balances do not characterise nutrient cycling in the hyporheic zone realistically (Trauth et al., 2014).

Conclusively, in situ assessments of hyporheic nutrient fluxes are indispensable. Hyporheic nutrient fluxes are commonly calculated from separate measurements of infiltration rates and pore-water concentrations (Kalbus et al., 2006; Ingendahl et al., 2009). The exchange of water between the surface and subsurface is traditionally derived from hydraulic head differences or tracer injections (Fleckenstein et al., 2010; USEP, 2013). Time series of high-resolution vertical temperature profiles have efficiently been used to derive vertical Darcy velocity (qy) (m d-1) in the streambed. While measurements of vertical Darcy velocities are a valuable asset, primarily horizontal fluxes are needed to assess hyporheic transport and residence time (Binley et al., 2013; Munz et al., 2016). Active heat-pulse tracing enables highly resolved in situ measurements of direction and velocity of hyporheic flow (Lewandowski et al., 2011; Angermann et al., 2012). These methods are profitable in shallow sediments (max. 15–20 cm) and rivers with fine sediments, but may not be implementable in streams with coarser sediments.

Pore-water solute concentrations are typically determined from analysis of grab samples extracted with drive points (Saenger and Zanke, 2009; USEP, 2013). Alternatively, dialysis cells – so-called peepers (Hesslein, 1976; Teasdale et al., 1995) – or gels (Krom et al., 1994), based on the diffusive equilibrium between the solute concentration in the pore water and the receiver solution in the peeper or the gel, have been used to measure small-scale solute distribution over highly resolved profiles on the millimetre to centimetre scale. These techniques provide valuable insights into the time-specific conditions at the target site. However, they delivers effectively a snap shot of the highly temporally variable hyporheic zone processes (Cooke and White, 1987) that may not be representative for the overall conditions in the system. Only repeated sampling at high frequencies and over longer time spans as conducted for example by Duff et al. (1998) can account for the short-term variability. Attempting to characterise larger areas with these methods is laborious and costly.

In sum, direct quantitative methods for hyporheic flux measurements have two major deficiencies. First, separate measurements of water flow velocities and pore-water concentrations are necessary to calculate mass fluxes. This is not only labour intensive but can, due to the high temporal and spatial variability of hyporheic flow, also lead to incorrect estimates of real flux rates (Kalbus et al., 2006). Second, most techniques for measuring pore-water concentrations exclusively capture the concentration at the sampling time, which may not reflect the overall conditions at the sampling location. New, affordable and efficient methods for the long-term measurement of nutrient fluxes through the hyporheic zone are therefore required to validate and improve models (Boyer et al., 2006; Wagenschein and Rode, 2008; Alexander et al., 2009) and to determine the site-specific extent of nutrient processing in the hyporheic zone (Fischer et al., 2009).

In groundwater studies, passive flux meters (PFMs) have successfully been used to quantify fluxes of contaminants (Hatfield et al., 2004; Annable et al., 2005; Verreydt et al., 2013) through screened groundwater monitoring wells, integrating time spans ranging from days to weeks. PFMs consist of a cylindrical, screened PVC casing filled with activated carbon (AC) as a porous sorbent. As long as the PFM is residing in the monitoring well, dissolved contaminants in the groundwater flowing passively through the meter are retained on the AC. Furthermore, the AC is preloaded with water-soluble resident tracers. The horizontal water flux through the screened media can then be determined from the displacement of these resident tracers, while simultaneously the contaminant flux is quantified based on the solute mass captured on the AC. First bench-scale and column experiments for PO4- have been conducted in the laboratory (Cho et al., 2007). As AC did not prove efficient in capturing nutrients, an anion-absorbing resin, a granular matrix originally manufactured for purification processes, was used as sorbent for PO4- instead. In theory, the PFM principle should be extendable to other nutrients or other environments. For example, development of sediment bed passive flux meters (SBPFMs) for the measurement of vertical contaminant seepage has been initiated (Layton, 2015). However, specific tests for NO3- as well as assessments under the complexities associated with hyporheic zone processes have not been conducted yet.

In this study we evaluate the applicability of PFMs for the measurement of horizontal nutrient fluxes in hyporheic zones, focusing on NO3-. We hypothesised that, while the principal concept of PFM can be maintained, several adaptations will still be necessary: most importantly, a suitable sorbent for NO3- as target nutrient is required. The market of anion-absorbing resins is large and offers a wide range of products with varying characteristics (Annable et al., 2005; Clark et al., 2005). Various criteria, like possible interference of resin compounds with the resident tracer analysis or pre-existing background nutrient loads on the resin, have to be considered. As experience on resin behaviour under field conditions is so far rare, we also expected unforeseen associated challenges, including for example biofouling of resin and/or nutrients. Additionally, a new deployment and retrieval procedure had to be developed. Existing groundwater PFMs have been installed into land-based wells. In hyporheic studies, underwater installation requires a technique which impedes contamination with surface water. Corrections for convergence and divergence of flowlines into or around the flux meter have been established in earlier studies (Klammler et al., 2007). However, accounting for an impermeable outer casing of a flux meter is much more complicated and requires additional factors which have to be determined experimentally for each specific application (Hatfield et al., 2004; Klammler et al., 2007; Annable et al., 2005). We therefore intended to deploy the PFM in a way that allows direct contact with the surrounding sediments and minimal manipulation of the natural flow pattern.

Considering these requirements, we developed a modification of the PFM for the application in the hyporheic zone (hyporheic passive flux meter, HPFM). Based on the results from laboratory analysis and a first field test in a nutrient-rich 3rd-order stream (Holtemme, Germany), we demonstrate prospects and remaining limitations for hyporheic nutrient studies with HPFMs.

The application of the HPFM for quantitative in situ measurement of horizontal NO3- and SRP fluxes through the hyporheic zone is novel. An earlier study on passive flux meter (SBPFM) in river beds (Layton, 2015) only assessed vertical flow of contaminants and is therefore not comparable to the application presented here. In the current work, adaptations were developed, tested and improved. Those include the choice of an appropriate resin, assessment of biofilm growth on the instruments and an approach that avoids challenges with contamination of the sorbent with nutrients. The results from the control HPFM showed that the uncertainty in measurement related to handling of the HPFM and processing of samples as conducted in this study is acceptable. Finally, the minimum and maximum deployment time will depend on the Darcy velocity and nutrient concentrations at a study site. Since the values derived from the control incorporate all the processing steps of HPFM and samples, they can be regarded as the method detection limit (MDL) (Greenberg et al., 1992). The MDL defines the lower limit for the use of HPFM in cases where nutrient fluxes are very low and deployment time cannot be extended. Based on the accumulated values detected in the control, the minimum deployment time can be estimated. In systems with high nutrient concentrations, usually the flow velocity qx will be the limiting factor. In our application the MDL for qx derived from the control was 8.4 cm for the complete deployment time (7 days). If the method inherited uncertainty should not be more than 5 % of the total measurement, the product of duration (in days) and velocity (in cm) should be at least 168 (20×8.4). As an example: if measured qx is around 200 cm d-1, 1 day (24 h) of deployment is sufficient. The lowest qx detected in our assessment was 21 cm d-1 (in HPFM AC4, see Fig. 3f), so that actually a deployment duration of 8 days would have been optimal. The same estimation can additionally be derived for expected nutrient fluxes. In systems with low nutrient concentrations, it would be preferable to start estimating the minimum deployment time based on the nutrient fluxes. We recommend that a control HPFM is incorporated in each field application of HPFM in order to determine the specific MDL. The upper limit is given by the loading capacity of the resin or complete displacement of all resident alcohol tracers.

The high-nutrient background on the AC required the separation of resin and AC in the HPFMs. We tested two different HPFM designs in this study, of which each inherits designated characteristics being more or less beneficial for different specifications. The first approach consists of pairs of two HPFMs where one is used to assess the water flux and the second to capture nutrients, and is preferable if a highly resolved depth profile is needed (a heterogeneous horizontal flux in the vertical direction). Since this approach assumes that local horizontal heterogeneity is negligible in the range of 20–30 cm, we recommend this type only for the use in uniform systems such as channelised river reaches. Even in those systems however, small-scale variability in streambed and sediment characteristics can cause spatially heterogeneous flow distributions (Lewandowski et al., 2011; Mendoza-Lera and Mutz, 2013). The second approach with alternating nutrient sorbents and water flux measuring segments is therefore preferable in most other cases as long as a high resolution over the vertical profile is not required. In general, several HPFMs should be grouped together in order to obtain representative results.

Further improvements of the HPFM for nutrient studies in the subsurface of rivers could be achieved by identifying a nutrient-free carrier for the tracers. First, because this would allow measuring nutrient and water flux at the same location within the device and thereby increase spatial resolution, and second, because in a mixed texture of nutrient absorber and tracer carrier the antibacterial nature of the activated carbon would suppress biofouling on the absorbent. We observed substantial biofilm growth on the resin in the laboratory and on the top 2 cm of the field-deployed HPFM R2. The results of the column experiments suggest that biofilm growth on the resin porous media did not affect its loading capacity. Further, biofilm growth was only visible on columns which were run beyond breakthrough, suggesting that considerable biofouling only started after the loading capacity of the tracer was exhausted. R2 detected higher NO3- fluxes in the top layer than the other HPFM. This could be due to contamination of the top layer of this HPFM with surface water (if the HPFM was not introduced sufficiently deep into the sediments). The further implication would be that this layer was exposed to much higher water and nutrient infiltration, so that the loading capacity was exhausted before the end of the experiment allowing biofilm accumulation. At the current state it is unclear to what extent the biofilm bound nutrients can be extracted by the procedure used here. Further experiments would also be needed to clarify under which conditions biofilm growth can occur and if bacterial uptake, transformation and release of nutrients influence the concentrations of nutrients inside the HPFM. HPFM segments on which biofilm is visible should be interpreted with caution. Finally, identifying a procedure or materials which completely inhibit biofouling will be an important step in the further development of HPFM.

In addition to instrumental adaptations we presented an installation procedure, which allows for smooth deployment with minimal disturbance of the system. Unlike typical well screen deployments where PFMs (Annable et al., 2005; Verreydt et al., 2013) or SBPFMs (Layton, 2015) have been inserted into a screened plastic or steel casing, our technique enabled the direct contact of the HPFM with the surrounding river sediments. Disturbing the natural structure of the sediment, potentially resulting in artificial flow paths, is intrinsic to all intrusive techniques, including HPFMs. Still, dispensing of a solid wall improves the integration of HPFMs in the natural system and prevents the generation of preferential flow paths along the wall of the device. Additionally, the HPFMs include a measurement time that is long relative to the duration of the installation, suggesting that the presented method causes lower disturbance compared to other intrusive measurements. While the installation of mini-drive points or heat pulse sensors in sediments coarser than sand may be difficult or even impossible and also proved unfeasible at our field site, installation of the HPFM with the presented technique was successful. The correction for convergence of flowlines into the device or divergence around it is relatively simple and already incorporated in the equation for the flux calculation. Heterogeneous permeability of the hyporheic zone around the HPFM does not distort the correction term as long as the permeability of the surrounding is coherently higher or coherently lower than the permeability of the HPFM matrix. Pre-measurements are therefore necessary for the selection of a suitable resin and tracer carrier. We believe that the presented approach and equations are applicable for a wide range of field conditions. However, for very coarse sediments, a protection of the HPFM with a solid screen might still be preferred. If fine particles are observed to bypass the mesh and enter the HPFM, a finer mesh should be chosen. We did not observe clogging of the mesh or intrusion of particles at our study, though in highly permeable systems with fine particle transport this might have to be considered.

A major advantage of the HPFM method is highlighted by the findings of the 7 day long field testing. In June, we found discrepancies between the average concentrations measured in the HPFM and the concentration found using the MLS. From our measurements it is not possible to prove that the HPFM results are correct and the MLS results biased. Nevertheless, the HPFM showed the expected decline in JN with depths, whereas the MLS pore-water concentrations were similar at all depths. This can be related to two reasons. First, we might have sampled surface water which bypassed along the wall of the MLS. The question would then be why that happened in June but not in October; second, we might have sampled the MLS at a time point when the hyporheic zone was inactive in respect to nutrient processing. Considering the high diurnal amplitudes in hyporheic oxygen concentration, we assumed that the discrepancy between HPFM and MLS arose from oscillations in hyporheic nutrient concentrations similar to the oxygen pattern. Microbial consumption of O2 in the sediments can, depending on nutrient concentration in surface water and transfer of these nutrients to the sediments, result in O2 depletion in the subsurface. Especially in nutrient-rich streams the related diurnal oscillations in O2 concentration favour night time denitrification in the hyporheic zone (Christensen et al., 1990; Laursen and Seitzinger, 2004; Harrison et al., 2005; O'Connor and Hondzo, 2008; Nimick et al., 2011). The redox conditions in the subsurface may also regulate the mobilisation/demobilisation of phosphate (Smith et al., 2011). The repeated manual sampling of pore water from MLS in October showed diurnal variations of SRP and NO3- in the subsurface of the testing reach, supporting the hypothesis that diurnal cycles in benthic metabolism caused temporal variations in hyporheic SRP and NO3- concentrations at our study site. As the majority of sampling is commonly conducted during daylight hours, night time conditions are under-represented in studies relying on single manual sampling events. Flux average concentrations can deviate by more than 50 % from estimates based on single event sampling, as was illustrated by comparison between our manual samples and the average pore-water concentrations calculated from the HPFM data.

Repeated pore-water sampling at high frequencies can be used to determine diurnal dynamics. However, continuing this over a longer time span is laborious, whereas if only few single time-specific snap shot samplings are conducted, the results may not realistically represent the overall conditions at the target site. Our comparison between MLS and HPFM reinforce the need for long-term recording of nutrient transport through the hyporheic zone. In general, most of our knowledge on hyporheic nutrient dynamics is based on measured surface water dynamics and models which project these dynamics on hyporheic processing. Theoretically, we could measure nutrient fluxes in the hyporheic zone and estimate whole-stream uptake rates from these measurements. However, the substantially higher effort to obtain subsurface data is not justified in most cases. As long as the overall in-stream retention is the focus, surface water monitoring will remain the method of choice. Innovative tracer experiments may even allow quantifying hyporheic exchange in streams. Haggerty et al. (2009) proposed a “smart” tracer approach, where the injected substance resazurin converts irreversibly to resofurin under metabolic activity. While a promising tool for detecting metabolic activity at the sediment–water interface in streams, first, uncertainties about sorption and transformation characteristics of these tracers remain (Lemke et al., 2014) and second, those methods give no evidence about nutrient transport to those reactive sites.

Thus, whenever the nutrient processing function of the hyporheic zone and its quantitative contribution to stream nutrient retention is of interest, for example in the evaluation of restoration measures including a rehabilitation of the river bed, direct measurements of hyporheic fluxes are indispensable. The HPFMs are a valuable approach that can be efficiently used to characterise and quantify nutrient dynamics in a sediment system. We consider that a combination of HPFM, MLS and concurrent measurements of pore water oxygen concentrations, as presented in this study, provide a practical set-up to interpret hyporheic nutrient dynamics.

Like solute concentrations and water flow patterns, the vertical extension of the hyporheic zone varies in time and space and between different rivers and reaches. Our set-up assessed exclusively the upper 50 cm of the hyporheic zone. We found continuously decreasing NO3- concentrations with depths, suggesting that this entire area (and potentially deeper) of the subsurface contained active sites for nitrate removal. While it was stated that denitrification is limited to the upper few cm of the hyporheic zone close to the sediment–water interface (Hill et al., 1998; Harvey et al., 2013), our results are in accordance to findings by Zarnetske et al. (2011b) and Kessler et al. (2012) who also report extended active hyporheic zones. Conducting collateral tracer tests, as suggested for example by Abbott et al. (2016), could deliver further evidence and characterise distinct flow paths. Nevertheless, since vertical water movement was overall downward and the lowest concentrations of NO3- were observed in the deepest segments of the HPFM, it is very likely that the hyporheic zone at our study site extends deeper than the 50 cm evaluated. The length of an HPFM can easily be increased, depending on the individual site conditions.

Considering the high spatial heterogeneity of the hyporheic zone, a larger number of HPFM would be needed to derive reliable and statistically supportable rates of hyporheic nutrient dynamics. The following example aims to display further possibilities of interpreting HPFM measurements. At our study site, the hyporheic removal potential RN of more than 50 % of infiltrating NO3- and 30 % of SRP suggests an active hyporheus. Evaluation of the effect of hyporheic removal activity on overall NO3- removal in the stream or the normalisation of hyporheic uptake to a benthic area requires a flow path length. In the presented example, this length refers to the horizontal vector of the distance the water travels in the subsurface before infiltrating the HPFM. The horizontal vector can be derived from the residence time of water and solutes in the hyporheic zone τHZ and the horizontal Darcy velocity qx. Assuming a downward flow direction, τHZ could be inferred from the vertical Darcy velocity qy as assessed from the temperature profiling and the hyporheic zone depths of 50 cm. Thereafter, τHZ conceptually corresponds to the time the water travels through the hyporheic zone before exiting to groundwater and sHZ to the horizontal vector of the flow paths. The nitrate uptake rate UNO3-HZ (mg NO3--N m-2 d-1) is then the difference between the theoretically transported NO3- mass MNO3-HZ theor, which is the product of QHZ and CNO3-SW and the measured mass flux MNO3-HZ real. During the testing phase UNO3-HZ was calculated as 693 mg NO3--N m-2 d-1. The same procedure yields a removal (uptake or adsorption) rate for SRP of UPO4-HZ=24 mg PO4- m-2 d-1. Calculating UNO3-HZ in the same way for each single depth assessed with the HPFM can deliver additional information about vertical gradients on nutrient processing rates and help to identify the most active depth in hyporheic zone. UNO3-HZi of a particular layer in the hyporheic zone can be derived by the differences in uptake rate between the regarded layer and the overlying layer. For instance the removal rates attributed to the different layers of HPFM L6 would be UNO3-HZ15=567 mg NO3--N m-2 d-1 in the shallow layer (0 to 15 cm depths), UNO3-HZ30=174 mg NO3--N m-2 d-1 in the layer from 15 to 30 cm depths and UNO3-HZ45=256 mg NO3--N m-2 d-1 in the deepest layer from 30 to 45 cm depths. From this example one could conclude that the shallowest sediments are the most efficient ones in terms of nitrate removal. While removal activity is first declining with depths it later increases again. This finding is consistent with the higher amplitudes of oxygen concentration in 45 cm depths compared to 25 cm depths, also suggesting higher biotic activity at the deepest layer. Potential reasons for this pattern could be decreasing NO3- penetration with depth (lower uptake at the middle layer than the shallowest one) which is in the deepest parts counter-balanced by increased residence time and stronger reducing conditions.

The role of the hyporheic zone as a hotspot for in-stream nutrient cycling is indisputable (Mulholland et al., 1997; Fellows et al., 2001; Fischer et al., 2005; Rode et al., 2015). Quantitative and qualitative knowledge about the influence of mass transfer on hyporheic nutrient removal is crucial to manage streams and river, especially in the light of increasing worldwide morphological alterations (Borchardt and Pusch, 2009), eutrophication (Ingendahl et al., 2009) and sediment loading (Hartwig and Borchardt, 2015). Despite decades of research on hyporheic nutrient cycling, robust quantitative data on nutrient fluxes through the hyporheic zone are limited, which is mainly due to methodological constraints in measuring nutrient concentrations and water flux in the subsurface of streams (O'Connor et al., 2010; Boano et al., 2014; Gonzalez-Pinzon et al., 2015). Passive flux meters have the potential to fill the gap in measured quantitative nutrient fluxes to the reactive sites in the sediments of rivers. To date, HPFMs are virtually the only method which can simultaneously capture nutrient and water flux through hyporheic zone within the same device and at the same spatial location. The field testing of several devices proved the general applicability of passive flux meters for quantifying NO3- and PO4- flux to reactive sites in the hyporheic zone. The hyporheic flux rates of nutrients and nitrate uptake rates measured in an agricultural 3rd-order stream were generally in agreement with rates reported in the literature. Our results clearly highlight the advantages of HPFM compared to commonly used methods (i.e. grab sampling of pore water and separate measurements of hyporheic exchange and Darcy velocities), first of all the capacity to integrate over longer time periods.

Quantifying nutrient flux to the potentially reactive sites in the hyporheic zone is an essential step to further improve our process-based knowledge on hyporheic nutrient cycling. In the future, long-term measurements of nutrient fluxes as obtained from HPFM can feed into and advance the transport part of nutrient cycling models.

We anticipate further improvement and increased use of passive flux meter approaches in order to advance conceptual models of nutrient cycling in the hyporheic zone. We demonstrated modifications which extended PFM application from groundwater to hyporheic zones. Current limitations related to the potential bias of results due to biofilm growth on sorbents require further analysis for the identification of more suitable sorbents. While we focused on nutrients, PFMs may also be used for a wide range of other substances like contaminants or trace elements.

Being labour efficient and attractive with respect to relatively low costs, numerous HPFMs can be efficiently used to cover larger areas and assess the degree of local heterogeneity. Further, neither advanced technology, maintenance, or power supply are needed which can be extremely advantageous for the use in remote areas or study sites without infrastructure.

All relevant data are either incorporated in figures/tables in the article or provided in the Supplement.