Throughout the world, agriculture, fossil-fuel combustion, and other human activities are leading to increasing levels of reactive nitrogen in the environment (Galloway et al., 2003; Gruber and Galloway, 2008). As a result, human health and ecosystem functions increasingly suffer from the effects of eutrophication and hypoxia, both at regional and global scales (Diaz and Rosenberg, 2008; Conley et al., 2009, 2011; Villnäs et al., 2013). The Baltic Sea is a semi-enclosed sea surrounded by highly industrialized riparian states and is particularly sensitive to anthropogenic impacts and natural fluctuations (Matthäus, 1995).

High reactive nitrogen levels cause increased biomass production (Falkowski et al., 1998) and thus increased sedimentation of organic material. Subsequent remineralization leads to bottom water oxygen deficiency, increased sulfate reduction and ultimately to the production of hydrogen sulfide (H₂S), which is highly toxic to most organisms (Gray et al., 2002). In the Baltic Sea, as deep water renewal is hampered by the strong pycnocline and submarine sills, total sulfide (∑S²⁻ = H₂S + HS⁻ + S²⁻) accumulates in the water column and can reach concentrations of up to 200 μM. These stagnation periods can last for many years and are only interrupted by inflow events from the North Sea under certain meteorological conditions (Schinke and Matthäus, 1998).

Nitrate (NO3-), as the major component of reactive nitrogen as well as H₂S, are therefore of great importance for the ecological state of the Baltic Sea and thus are monitored by the coastal countries, which have adopted the Helsinki Convention for the Protection of the Marine Environment of the Baltic Sea in order to abate eutrophication by reducing nutrient loads from point and non-point sources (Backer et al., 2010). Many methods for the detection of NO3- (Moorcroft et al., 2001) and H₂S (Lawrence et al., 2000) have been developed, but in most marine laboratories the determination of both compounds is usually carried out via simple colorimetric measurements (Grasshoff et al., 2009). The frequency and the spatial coverage of these measurements are currently limited by the cost of the laboratory analysis and personnel. Furthermore, the sampling process and preparation of the mostly toxic reagents is time consuming and involves multiple sources of error. However, with the help of in situ sensors, it is possible to reduce the working time as well as contamination and altering of the sample. In addition, in situ sensors have a small size, low weight and relatively low energy consumption, and therefore are suitable for being deployed on moorings (Collins et al., 2013; Sakamoto et al., 2017), floats (Johnson et al., 2013, 2017; D'Ortenzio et al., 2014; Pasqueron de Fommervault et al., 2015), towed vehicles (Pidcock et al., 2010) or autonomous underwater vehicles (Johnson and Needoba, 2008). In situ sensors are thus helping to bridge the gap between large-scale but non-specific remote observations of satellites limited to the surface waters and highly sophisticated and specific laboratory analysis from discrete water samples (Zielinski et al., 2009).

In general, there are three different types of in situ sensors that can be used for NO3- measurements: (1) wet chemical analyzers, (2) optical UV sensors, and (3) ion selective electrodes (ISEs). Each of these methods has its own advantages and disadvantages. Wet chemical analyzers such as the WIZ probe (Vuillemin and Sanfilippo, 2010) or Lab-on-chip devices (Beaton et al., 2012) are using chemical reagents and thus produce high maintenance toxic waste. Furthermore, operating lifetime of those analyzers is strongly dependent on reagent stability. On the other side, they can be calibrated in situ and thus have small uncertainty of measurement. However, the preparation of stable standard solutions for H₂S and NO3- measurements is difficult as the shelf life of these standards is limited. By contrast, optical UV sensors such as Nitratax (Hach Lange GmbH, Germany), ISUS/SUNA (Johnson and Coletti, 2002; Johnson et al., 2006; MacIntyre et al., 2009) (Satlantic, Canada), or ProPS/OPUS (Zielinski et al., 2007; Prien et al., 2009) (TriOS GmbH, Germany) do not require any chemical reagents but are influenced by optical/ionic interferences including bromide (Br⁻), chloride (Cl⁻), turbidity and water color. ISEs such as NISE (Hach Lange GmbH, Germany) and EXO (YSI Inc., USA) are not influenced by turbidity and water color, but deteriorate quite rapidly when used under environmental conditions since the sensor molecules in the membrane change or are leached out. Therefore, they are often subject to significant drift and require regular re-calibration (Moorcroft et al., 2001; Le Goff et al., 2002). UV sensor technology is therefore used most frequently and has been increasingly used for NO3- measurements in a variety of environments such as rivers (Pellerin et al., 2009; Wade et al., 2012; Feng et al., 2013; Aubert and Breuer, 2016; Chappell et al., 2017; Kunz et al., 2017), groundwaters (Huebsch et al., 2015), coastal waters (Zielinski et al., 2011; Frank et al., 2014), open ocean waters (Omand and Mahadevan, 1966; Johnson, 2010; Ascani et al., 2013) and hydrothermal vents (Le Bris et al., 2000).

The aims of this study were to validate and optimize an optical in situ method for the precise and accurate simultaneous measurement of NO3- and HS⁻ in brackish waters of the Baltic Sea. We used the OPUS UV spectral sensor (TriOS, Oldenburg, Germany), which was deployed for the first time in coastal waters, allowing improved observations of both compounds. Near-real time measurements and online sensor control was realized by using a deep-sea telemetry system integrated into the conductivity-temperature-depth (CTD) rosette sampler. Furthermore, detailed protocols have been developed for both sensor calibration and data evaluation. Finally, results were confirmed by comparison with conventional measurements made by an independent laboratory based method.

During a monitoring cruise in the Baltic Sea in February 2018, in situ measurements were carried out with the OPUS. The OPUS was mounted on the CTD rosette sampler and could be controlled online via a deep-sea telemetry system. The hydrographic data of the measured stations (TF200, TF286, TF245) and the results of the MLR are shown in Figure 3. In addition, summarized results of laboratory and sensor measurements can be found in Table 1. The CTD data (Figures 3A,D,G) indicated a stratified water column with cold, low-saline surface water and warm, saline deep water. O₂, temperature and salinity values were fairly uniform in the surface layer at all three stations due to the winter mixing (TF200: 357 μM, 4.6°C, 7.5; TF286: 358 μM, 4.5°C, 7.3; TF245: 359 μM, 4.1°C, 7.3). Below 50 m depth, the salinity and the temperature increased and reached nearly constant values in the bottom water (TF200: 6.8°C, 17.3; TF286: 6.8°C, 12.8; TF245: 5.6°C, 10.3). In contrast, the O₂ concentrations sharply decreased from 50 m depth so that no O₂ was present in the bottom water of all three stations. In addition, intrusions of O₂-rich water could be located at station TF200 and TF286 at 60 m and 125 m depth, respectively. When O₂ was not present, ∑S²⁻ was determined using the methylenblue method (MBM) as well as the OPUS sensor (Figures 3B,E,H). Concentrations close to the bottom obtained by MBM were 4.3 μM (TF200), 14.6 μM (TF286) and 33.6 μM (TF245). Concentrations of ∑S²⁻ obtained by the OPUS sensor at about the same depths were 5.6 μM (TF200), 16.9 μM (TF286), and 34.3 μM (TF245). Furthermore, NO3- was measured using a segmented flow analyzer (SFA) and the OPUS sensor (Figures 3B,E,H). At the surface, NO3- concentrations were consistently high and slowly increasing northwards. Concentrations above 50 m depth obtained by SFA were in average 2.4 μM ± 0.1 μM (TF200), 3.6 μM ± 0.04 μM (TF245), and 3.9 μM ± 0.02 μM (TF286). Concentrations of NO3- obtained by the OPUS sensor above 50 m depth were in average 2.5 μM ± 0.2 μM (TF200), 3.3 μM ± 0.2 μM (TF245), and 5.9 μM ± 0.2 μM (TF286). NO3- maxima were observed at station TF200 and station TF286 at 60 m and 125 m depth, respectively, and coincided with above-mentioned O₂ intrusions.

Additional parameters (Br⁻, turbidity, and poly_4) are shown in Figures 3C,F,I. Br⁻ was measured with the OPUS sensor but was also calculated from salinity using CTD data whereby conservative behavior was assumed. Turbidity (reported in nephelometric turbidity units or NTU) was measured in parallel with temperature and salinity and was multiplied by 50 for plotting in the figures. In the upper water column turbidity values were rather low. High turbidity was observed especially in the euxinic deep water where O₂ was absent. Furthermore, poly_4 values resulting from the MLR are also shown Figure 3 and are presented as “percentage of sample absorption at a wavelength of 210 nm” in order to show how much of the sample absorbance could not be allocated to any of the calibration spectra included in the calculation group. Poly_4 values were about 24% (TF200), 25% (TF245), and 17% (TF286) in the surface water and 4% (TF200), 8% (TF245), and 4% (TF286) in the bottom water.

OPUS and laboratory results were measured at different depths and thus a direct correlation of the concentration data would be inappropriate. Instead, averaged NO3- concentrations above a depth of 50 m, as presented in the results section were compared for each station. Resulting deviations for NO3- were 4% (TF200), 8% (TF245), and 51% (TF286). In the case of ∑S²⁻ the bottom water concentrations between bottle data and OPUS data were compared. Deviations found here were 30% (TF200), 2% (TF245), and 14% (TF286). As mentioned above, dissociation of H₂S is pH-dependent and the calibration solution used in this study was buffered to a constant pH of 7.3. Therefore, the reported deviations could be explained by variations in the in situ pH of the study area. However, considering that the highest deviations observed for NO3- (51%) and ∑S²⁻ (30%), correspond to only 2 and 1.3 μM, respectively, it can be stated that NO3- and ∑S²⁻ measurements carried out at the Baltic Sea using the OPUS sensor are in satisfactory agreement with those obtained from reference methods. Although, the findings of our study are limited by the small number of measured stations, presented results show that both NO3- and ∑S²⁻, as important environmental parameters, can be determined in situ, simultaneously and without reagents during classical CTD work, providing near real time readings with sufficient accuracy without time and cost intensive traditional laboratory analysis. That the deconvolution of the sample spectra was successful is also shown by the MLR results of the other parameters that were included in the calculation groups. For instance, OPUS Br⁻ concentrations and theoretical Br⁻ concentrations correspond very well with each other (Figures 3C,F,I). Maximum deviations were found in the bottom water and were only 6% (TF200), 3% (TF245), and 6% (TF286). It should be noted here, that pressure effects on bromide absorbance were neglected in our study because we have worked in shallow waters, but should be taken into account at higher pressures (Sakamoto et al., 2017).

But although NO3-, Br⁻, and ∑S²⁻ concentration data of the OPUS sensor fit very closely with the laboratory and theoretical data and all features of the concentration distributions are reproduced some questions still remain. Figure 3B, for example, shows a sample absorption spectrum from station TF245 at 6 m depth and the results of the deconvolution using the calibration spectra during the MLR. The sum of the NO3-, Br⁻, HCO3-, Cl⁻, and poly_4 spectrum would give the sample spectrum. The polynomial coefficients provided by the MLR with regard to the light scattering term (poly_4) seem to be remarkable high. This is especially true for the upper water column as can be seen from Figure 3, where up to 25% of the sample absorption is allocated to the poly_4 spectrum.

On the one side, we assume that light scattering actually could be increased by suspended particulate matter (SPM), as certain colloidal fractions can have scattering coefficients significantly greater than pure sea water (Stramski and Woźniak, 2005). On the other hand, we think that absorption fractions of CDOM were added to the poly_4 term during the regression, since we did not add such a calibration spectrum to the calculation groups. This assumption is underpinned by the striking inverse correlation between poly_4 and salinity (Figure 3), as it has also been found for CDOM in other coastal waters (Fichot and Benner, 2011). Although CDOM should be the most relevant parameter at this point, one aspect of future research thus could be to investigate the impact of both the particulate and the dissolved components of organic material on in situ UV measurements. Since the optical properties of CDOM have been found to be empirically related to DOC concentration (Fichot and Benner, 2011) further research could also help to get a better understanding of DOC cycling. This could be of considerable interest, since dissolved organic material (DOM) and DOC, as a major component of the DOM pool, can enhance eutrophication and pollution of natural waters, for instance, by the mobilization of phosphorous and heavy metals (Fest et al., 2008; Ros et al., 2010). Since the aim of the current work was to detect NO3- and ∑S²⁻, correction for this offset as performed by other workers (Sakamoto et al., 2009; Zielinski et al., 2011) was not carried out.

Although, the findings of our study are clearly limited by the small number of measured stations, the OPUS sensor in combination with deep-sea telemetry allowed in situ determination of NO3- and ∑S²⁻ in the Baltic Sea, enhancing monitoring data set. This approach increases the simplicity of the determination of both parameters and the speed of data acquisition. Concentration gradients as well as maxima and minima can be determined more accurately as the spatial resolution of the data can be higher than using traditional bottle sampling. The OPUS is capable of resolving NO3- variability on a scale of 6 m at a CTD lowering speed of 0.3 m/s (OPUS sampling rate: 0.05 Hz). At lower profiling speed the vertical resolution could be even higher. A future cooperation with the manufacturer TriOS (Oldenburg, Germany) could help to further improve the sampling rate. Meanwhile, we have received a new firmware allowing OPUS measurements at a sampling rate of 0.1 Hz which results in 3 m vertical resolution. A relevant step to resolve e.g., steep gradients in the water column.

Finally, it can be concluded that there is a clear need of innovative and cost-effective monitoring (Zielinski et al., 2009; Borja and Elliott, 2013) and using sensors such as OPUS is a smart way of gathering relevant information quickly. Modern oceanographic tools like the instrument described here should be more widely used. This is especially important to identify anthropogenic impact (e.g., in the form of sporadic events) and their consequences for the marine ecosystems. The affected processes can have a wide range of temporal and spatial scales, requiring frequent and distributed observations. Only then can the feedbacks be unraveled and the ecosystem status adequately assessed.

DM, LR, MP, and RP conceived and designed the experiments. LR, MP, and DM performed the experiments. DM and LR compiled and processed all the field and laboratory data. DS-B, RP, and JW contributed materials, reagents, analysis tools. DM wrote the paper, with edits and contributions from all co-authors. RP, JW, and DS-B directed the study and its scope.