Optical and biogeochemical properties of diverse Belgian inland and coastal waters

. From 2017 to 2019, an extensive sampling campaign was conducted in Belgian inland and coastal waters, aimed at providing paired data of optical and biogeochemical properties to support research into optical monitoring of aquatic systems. The campaign was focused on inland waters, with sampling of four lakes and a coastal lagoon along the growth season, in addition to samples of opportunity of other four lakes. Campaigns also included the Scheldt estuary over a tidal cycle and two sampling campaigns in the Belgian coastal zone. Measured parameters include inherent optical properties (absorption, 5 scattering and beam attenuation coefﬁcients, near-forward volume scattering function, turbidity), apparent optical properties (Secchi disk depth, substrate and water-leaving Lambert-equivalent bi-hemispherical reﬂectance), and biogeochemical properties (suspended particulate matter, mineral fraction of particle mass, particle size distribution, pigment concentration, DNA metabarcoding, ﬂow microscopy counts, and bottom type classiﬁcation). The diversity of water bodies and environmental conditions covered a wide range of system states. The chlorophyll a concentration varied from 0.63 mg m − 3 to 382.72 mg m − 3 , 10 while the suspended particulate matter concentration varied from 1.02 g m − 3 to 791.19 g m − 3 , with mineral fraction varying from 0 to 0.95. Depending on system and season, phytoplankton assemblages were dominated by cyanobacteria, green algae (Mamiellophyceae, Pyramimonadophyceae) or diatoms. The dataset is available from Castagna et al. (2022)

2. The majority of the data concerns multispectral measurements, particularly at wavebands typical of ocean colour sensors; 3. The majority of the paired biogeochemical data includes only broad features (e.g., chlorophyll a concentration) deemed suitable for operational retrievals with multispectral instruments.
In order to fully exploit the potential of hyperspectral satellite missions, hyperspectral datasets paired with detailed composition information of aquatic systems are required (Dierssen et al., 2020). In addition, more extensive data is necessary to 25 develop and validate regional optical retrievals over complex optical systems such as lakes, lagoons, estuaries and rivers.
The data presented here were gathered and processed druring three projects funded by the Belgian Science Policy Office (BELSPO) and one project funded by the Research Foundation -Flanders (FWO). The PONDER project (BELSPO SR/00/325) focused on developing tools for spaceborne remote sensing of inland water systems using high spatial resolution The goal of this data report is to provide a detailed description and validation of the methods used during the research, present a summary of the observations and briefly discuss aspects of the data that might be relevant for potential users.

Methods and data description
Measurements are presented in three groups: (1) Inherent optical properties (IOPs) consisting of absorption, scattering and 40 beam attenuation coefficients, near-forward volume scattering function and turbidity; (2) Apparent optical properties (AOPs) consisting of Secchi disk depth, water-leaving and substrate Lambert-equivalent bi-hemispherical reflectance; and (3) biogeochemical properties consisting of suspended particulate matter concentration, mineral fraction of particle mass, particle size distribution, pigment concentration, DNA metabarcoding, flow imaging microscopy counts, and bottom type classification. Table 1 presents the description of relevant acronyms, symbols, constants and subscripts used in this study. The studied Secchi disk depth varied from 1 m to 6 m. Submerged macrophytes were confined to near-shore locations due to a steep basin slope.
The Donkmeer (51 • 02'23.4"N 3 • 58'47.2"E) is the second largest lake in Flanders, with a surface area of 0.86 km 2 and an average depth of 2 m. It experiences recurrent cyanobacterial blooms of Anabaena spp. and Planktothrix agardhii (Descy 75 et al., 2011). It is used for recreational activities and fishing. During the observation period (2017 to 2018) it experienced cyanobacterial blooms from summer to autumn, reaching a Chl a concentration of ≈400 mg m −3 and a Secchi disk depth of 0.2 m. Its northern and southern portions are connected through a narrow and shallow passage, with the northern portion experiencing a shorter period of cyanobacterial blooms due to management actions.
Other inland water systems included were the Scheldt estuary and the Leuven-Dijle canal. The Scheldt is a rain-fed lowland river, with an estuary environment subjected to tides from the mouth at Vlissingen (The Netherlands) to Ghent (Belgium,90 160 km upstream), where a system of locks prevent further propagation of the tide (Meire et al., 2005). It has a large economical importance as a transport waterway, connecting the harbours of Antwerp and Ghent to the North Sea. The Scheldt estuary was sampled in mid October 2019 at two locations near the city of Sint-Amands (51 • 03'18.0"N 4 • 11'59.6"E and 51 • 04'24.1"N, 4 • 11'24.0"E), with a time series including a full tidal cycle (tidal range of 6 m). During the observation period the Chl a concentration reached 55.7 mg m −3 at high tide, while the suspended particulate matter (SPM) concentration 95 reached 791.2 g m −3 . In the same campaign performed in the Scheldt, two samples were taken at the Leuven-Dijle canal, a highly transparent artificial waterway running parallel to the Dijle river, connecting Leuven to the Zenne-Dijle confluence, and ultimately to the Scheldt. The sampling position was close to the lock of Zennegat (51 • 03'46.7"N 4 • 25'50.0"E).
The stations sampled in the BCZ are part of the regular LifeWatch sampling campaigns (Mortelmans et al., 2019). The April and July 2018 campaigns were augmented to include spectroscopic measurements. The BCZ is a shallow part of the North 100 Sea (1 m to 40 m), experiencing high tidal fluctuations (average of 4 m) and strong tidal currents (1 m s −1 ). Those conditions, combined with limited freshwater discharge into the region, result in a well mixed-water column (van Beusekom and Diel-Christiansen, 1993). It experiences high turbidity (SPM concentration from 1 g m −3 to 200 g m −3 ) with large influence from particulate material imported through the Strait of Dover (Fettweis and Van den Eynde, 2003). The BCZ develops a turbidity maximum zone near Zeebbrugge, also influenced by the decreasing magnitude of the residual transport vectors from the East 105 border (Fettweis and Van den Eynde, 2003). The inflow of the Yser and Scheldt rivers influence the availability of nutrients, with an increase of nitrogen and phosphorus since the second half of the 20th century, followed by a de-eutrophication phase during which nitrogen and especially phosphorus decreased (Desmit et al., 2020). The phytoplankton seasonal dynamics are well described (e.g., Reid et al., 1990), with an early spring diatom bloom followed by a mixed bloom of the haptophyte Phaeocystis globosa and diatoms. A recent review of the phytoplankton seasonal dynamic was provided by Castagna et al. 110 (2021). The IOPs in this region were extensively studied by Astoreca et al. (2006Astoreca et al. ( , 2009Astoreca et al. ( , 2012.

Sampling
Sampling was performed from a diverse set of platforms. For the inland water campaigns, samples were taken from pontoons or an inflatable boat, depending on the system and date. The water was sampled just below the surface, taking care not to draw in materials floating at the surface (e.g., pollen, debris, etc). Field samples were stored in 5 L semi-transparent plastic carboys, 115 kept in the dark and cold during the transport to the laboratory, and processed within 4 to 6 hours from sampling. Sampling for the coastal campaigns was performed from the Research Vessel (RV) Simon Stevin (Flanders Marine Institute, VLIZ). For most marine samples, water was sampled just below the surface using Niskin bottles (General Oceanics, Inc.), attached to a rosette system. The exception were subsurface samples for flow imaging microscopy, taken with a bucket from the side of the ship. Filtrations were performed on board and water subsamples were kept in the dark and cold during transport to the laboratory for spectrophotometric measurements.
Macrophytes, sediment and biofilm were sampled in the Spuikom lagoon. Macrophytes were sampled during 2017 and 2018, by collecting floating specimens or recovering specimens from the bottom using a rake. Specimens were stored in transparent plastic bags containing water from the lagoon, and kept in the dark and cold during transport to the laboratory. Sediments and biofilm were sampled in April 2018 using polymethyl methacrylate (PMMA) tubes attached to a short corer. The cores were 125 retrieved with care not to disturb the surface of the sediment, sealed and transported to the laboratory in a vertical position. In July 2018, biofilm patches had detached from the bottom of the Spuikom and were sampled floating at the surface and stored in plastic bags for microscopic examination. The cores and macrophytes were stored in a climate room at 4 • C for up to 3 days until analysis.

130
Most of the IOP measurements were performed ex situ, with a benchtop spectrophotometer (Lambda 650S, PerkinElmer) equipped with a integrating sphere (150 mm internal diameter). The interior of the integrating sphere is made of highly reflective (nominal 99 %) sintered polytetrafluoroethylene (PTFE). The spectral IOP measurements made with the spectrophotometer include particle absorption (a p ), chemically decomposed into in vivo pigment absorption (a φ ) and depigmented particle absorption (a d ), chromophoric dissolved organic matter absorption (a g ), particle beam attenuation (c p ) and scattering (b p ) Napierian 135 coefficients. Methods followed recommendations from Pegau et al. (2002) andIOCCG (2018). Measurements covered the range from 250 nm to 850 nm in 1 nm steps, with a 2 nm integration slit and 0.24 s integration time (250 nm min −1 ). Data is provided in the range of 380 nm to 850 nm, and include a single pass of a smoothing function to reduce noise (rectangular filter, 10 nm window moving average).
Turbidity, defined as the side-scattering at 860 nm relative to Formazin standards (ISO 7027:1999;Dogliotti et al., 2015;140 Boss et al., 2009a), was measured in discrete samples with a portable turbidimeter (2100P ISO, HACH). Additional IOP data were measured from in situ instrumentation for a subset of water systems and stations. These include the diffraction peak (ψ < 15 • ) of the particle volume scattering function (VSF; β p (670)) and non-water beam attenuation at an acceptance angle of 0.018 • measured at 670 nm (c nw (670, 0.018); LISST, Sequoia Scientific).
2.3.1 Absorption coefficient from dissolved components (< 0.45 µm) 145 Absorption due to chromophoric dissolved organic matter (CDOM) was determined in the laboratory from fresh and undiluted filtered subsamples, using a 5 cm pathlength quartz cuvette. Subsamples were filtered with 0.45 µm polyamide (nylon) fiber syringe filters. Between samples, the syringe and cuvette were rinsed with deionized water. The filters were first rinsed by filtrating 15 mL of deionized water before use with samples. The first sample filtration volume was used for final rinse of the quartz cuvette and the volume discarded. The cuvette was then gently filled with the second filtration volume to avoid the 150 formation of bubbles, and allowed to rest for at least 3 min before measurements. The blank was determined with deionized water, using the same cuvette and at the same temperature (room temperature, ≈20 • C). Quality control evaluated the presence of absorption peaks at 676 nm as an indication of cell leakage and the offset from zero absorption at 750 nm as a indication of the presence of hydrosols.
The use of deionized water as blank for all measurements resulted in a mismatch of salinity for marine and brackish samples.

155
Dissolved salts change the complex refractive index of water (Quan and Fry, 1995;Röttgers et al., 2014b), affecting reflection and refraction interactions with the cuvette wall, water scattering and absorption. In the wavelength range reported here, the scattering effects and the absorption of salts are negligible, however the dissolved salts form what is known as salt-solvated water, mixed with pure water (Max and Chapados, 2001). Salt-solvated water can have higher or lower absorption than pure water, depending on the wavelength, with a magnitude defined by the salt type (Max and Chapados, 2001;Röttgers et al., 160 2014b). This difference is particularly important in the NIR range, due to the spectral shape of a g . The effect of the mismatch in salinity was investigated by comparing the effects of salinity on blank readings for a g determinations from deionized water and artificial seawater at a salinity of 35 (NaCl at 35 g kg −1 ). The results are shown in Fig. S1 (supplementary material) and the resulting difference spectrum was subtracted from a g of the Spuikom lagoon and the BCZ for an approximate correction.
Dissolved molecular oxygen (O 2 ) also contributes to absorption in the UV range (Jonaz and Fournier, 2007, and references 165 therein), however in the wavelength range reported here CDOM dominates the signal.
The Lambda 650S was not available in 2017 and a g was measured for a subset of samples using another spectrophotometer (UV-1601, Shimadzu). Sample preparation was the same, except that a g was only measured at selected wavelengths (380 nm, 400 nm, 433 nm, 550 nm and 750 nm). To retrieve the full spectrum for these samples, an hyperbolic model was fitted to the available wavelengths (Twardowski et al., 2004). The hyperbolic model fit was also applied to all measurements in the range 170 of 380 nm to 850 nm to provide a smooth, fitted version of a g and to calculate the CDOM absorption spectral hyperbolic slope coefficient, S hg : Most samples had negligible absorption in the NIR (< |0.1| m −1 nm −1 ), fluctuating around zero due to instrument noise or residual blank salinity difference effect. Four stations had positive or negative offsets that were larger than |0.1|, and the 175 negative offsets were attributed to the presence of bubbles in the blank of the day. All spectra were subtracted by the average a g between 750 nm and 850 nm. Five freshwater samples had a pronounced Chl a absorption peak at 676 nm, and since NIR signals were close to zero, those peaks likely result from cell leakage. The a g for those stations were flagged. 2.3.2 Absorption coefficient from particulate components (> 0.7 µm) Particle absorption was measured using the filterpad method, with particles concentrated onto a glass fiber filter (GF/F, effective mesh of 0.7 µm). To improve the homogeneity of particle deposition over the filtration area, two stacked filters were used and 185 samples with high abundance of particles were diluted in the filtration funnel with deionized water. Immediately after filtration, filters were transferred to PetriSlides (Merck), wrapped in aluminum foil and frozen in liquid nitrogen. Filters were then stored at -80 • C until analysis.
Before analysis, the filters were allowed to thaw to room temperature and kept hydrated with deionized water. To avoid dislocating large particles deposited on the filter fibers, hydration was performed by raising the filter, adding a droplet of water 190 to the PetriSlide base and gently lowering the filter onto the droplet, resulting in water spreading by capillarity. The "inside sphere" variant of the quantitative filterpad method was used (Stramski et al., 2015), using a center mount coated with PTFE.
Filters were read twice, with a 90 • rotation between reads to average small deviations from homogeneous deposition. This was especially important to account for large cyanobacteria colonies when these occurred in low densities (Fig. S2, supplementary material). Pigments were then oxidized by treating the filters with sodium hypochlorite (NaClO; Ferrari and Tassan, 1999), 195 using the same approach as for hydration. Filters were read after 15 min or complete oxidation, following the same orientations as of the pigmented readings. The pathlength amplification correction was taken from Stramski et al. (2015) as recommended in IOCCG (2018). The in vivo pigment absorption coefficient (a φ ) was calculated as the total particulate absorption coefficient (a p ) subtracted by the depigmented particle absorption coefficient (a d ). The chemical oxidation step was not performed for the Scheldt samples.

200
Sodium hypochlorite has an absorption peak at ≈300 nm, with residual absorption up to 500 nm. One method to remove the oxidant's signal is to rinse the filter with pure water before the determinations of a d and this method was used for a subset of samples to evaluate the impact of sodium hypochlorite and rinsing. We observed that some samples had an apparent loss of a d observed between determinations with sodium hypochlorite and after rinsing (Fig. 2). The magnitude of this effect varied across the lakes, with Spuikom, Donkmeer, Hazewinkel and Bocht suffering little to no loss between 550 nm and 850 nm, 205 while Dikkebus and Zillebeke showed large reductions in a d after rinsing. This effect might provide additional information on the nature of the particles in a given system, but needs to be further explored to understand its causes. A direct consequence for measurements of a d and a φ is that rinsing might underestimate a d .
Therefore we developed a statistical method to remove the oxidant's signal, instead of applying rinsing. We observed that when there was no loss of signal with rinsing, the minimum difference between a p and rinsed a d in the UV was at 305 nm, 210 with a d (305) typically between 80 % and 90 % of a p (305) ( Fig. 2A). Based on these observations, we fitted an exponential function to measured a d with sodium hypochlorite in the range of 550 nm to 850 nm, and included a point-estimate of rinsed a d at 305 nm as 0.8a p (305). The additional data point at 305 nm helps to set the curvature of the exponential model. Finally, an offset was included as necessary to match the a p absorption in the NIR end. The exponential model with an offset was:

215
where α d (m −1 nm −1 ) is a spectrally flat offset (Estapa et al., 2012) and S ed (nm −1 ) is the spectral exponential slope of a d .
The treatment with sodium hypochlorite can also introduce another artifact to the measurements. For a set of samples we observed a baseline offset between a p and a d in the NIR even before rinsing, propagating to increased baseline absorption of a φ . This likely results from sodium hypochlorite removing the adsorbed organic layer over particles (Binding et al., 2008), 300 400 500 600 700 800 A 300 400 500 600 700 800 with the magnitude of the effect proportional to the concentration of particles and organic matter. In our samples this effect was 220 larger in the maximum turbidity zone of the BCZ. This loss of absorbing material was not observed in a study by Röttgers et al. (2014a) including samples from a diverse set of environments, though the authors did not apply NaClO to the North Sea or Baltic Sea samples. The baseline effect was compensated for by adding the minimum offset necessary to match a p and a d in the 400 500 600 700 800 range of 800 nm to 850 nm without generating negative a φ values. This constant offset likely underestimates the contribution of organic absorption in the blue end due to the exponential spectral shape of this component (e.g., Cael and Boss, 2017).  ues higher than 0.565 m −1 nm −1 . The range of the S ed was between 0.0064 nm −1 and 0.0194 nm −1 . The fitted values of a d and S ed are presented in Fig. 3. Since particle absorption samples from the Scheldt were not depigmented but were dominated 230 by a d (Fig. 4A), an estimate of S ed is provided by fitting Eq. 2 directly to a p (Fig. 3B).
The integral-normalized a φ presented a diversity of pigment absorption peaks (Fig. 4B), showing spectral shapes associated with dominance of cyanobacteria, green and red lineages of algae. Considering that a φ was retrieved from fitted a d , an independent validation was performed against the total Chl a concentration (described later). In the blue end of the spectrum, we used Soret band of Chl a in vivo (435 nm; Fig. 5A). In the red end of the spectrum, we compared the a φ (676) against models fitted to global datasets (Nardelli and Twardowski, 2016, Fig. 5B). We note that our observations are similar to those reported by Hoepffner and Sathyendranath (1991), while the value of Nardelli and Twardowski (2016) are lower likely due a combination of their lower Chl a concentration range and their estimation of a φ (676) from the a nw (650, 676, 715) line height (Roesler and Barnard, 2013). The median spectral mass-specific in vivo pigment absorption coefficient at 676 nm, * a φ (676), was equal to 240 0.022 m 2 mg −1 nm −1 (Fig. 5C), similar to the value of 0.020 m 2 mg −1 nm −1 presented by Hoepffner and Sathyendranath (1991). The comparisons in the blue and red spectral regions we observe a linearity in log scale, with larger spread in the comparison a Gauss (435) likely due to a combination of the uncertainty in the estimate of the fitted a d and the variable pigment packaging effect (Morel and Bricaud, 1981;Latimer, 1983), both larger in the blue spectral range. those deployment conditions and that our systems were mostly turbid, we expect negligible influence of environmental light on the measurements. The instruments were used in a subset of stations. Calibrations with deionized water were performed monthly and the most recent calibration used for blank values. The VSF was retrieved from the raw LISST data file, using the 255 procedures described in Agrawal (2005) and the instrument calibration files.
Quality control was performed by flagging entries in which the single scattering transmittance (T b ) was lower than 30 %, to avoid artifacts produced by multiple scattering: where b t is the total scattering coefficient and l LISST is the pathlength of the LISST instrument. The non-water beam attenuation In addition to the c nw (670, 0.018) measured by the LISST instruments, the spectral particle beam attenuation (c p ) and scattering (b p ) coefficients were calculated from the c nw measured on fresh samples with the Lambda 650S spectrophotometer, by subtracting a g and a g +a p , respectively. To reduce the acceptance angle of the spectrophotometer (Boss et al., 2009b;Leymarie et al., 2010), a black barrier with a central circular aperture of 2.4 mm diameter was placed in the entry port of the integrating sphere. With a distance of 69.5 cm from the center of the sample cell to the integrating sphere (due to system of mirrors 270 extending the pathlength), the detector acceptance angle in water for this configuration is 0.074 • . The integration time of the instrument was increased to 0.4 s (150 nm min −1 ) to compensate for the reduced signal, but noise levels were still noticed throughout the spectra. On later samples, the reference beam was partially blocked with a 1 % transmittance filter to reduce the dynamic range imposed by the lack of a similar aperture in the reference beam.
As with the a g , c nw was measured in undiluted water subsamples, using 1 cm, 5 cm or 10 cm quartz cuvettes, depending  The additive error due to the contamination with the near-forward scattered signal within the acceptance angle can be accurately corrected for only when the VSF of each sample is known. However, an approximate correction is possible. Fig. 6B shows that the scattering signal in the range of 0.018 • to 0.074 • represents an average of 43.2 % of the particle scattering coefficient, with a stable relation across systems and seasons. Assuming that the angular shape of the VSF presents minor spectral variability and that the VSFs of different systems and seasons in our dataset are well represented by the subset for 295 which LISST data is available, it is possible to use the approximate correction: b p (λ, 0.074) = c p (λ, 0.074) − a p (λ), 300 c p (λ, 0.018) = a p (λ) + b p (λ, 0.018).
The LISST-equivalent (acceptance angle of 0.018 • ) version of the spectral b p and c p are also provided in conjunction with the measured values (acceptance angle of 0.074 • ), as recommended by IOCCG (2018).
Quality control included the flagging of c p spectra showing irregular behavior. Particle beam attenuation is a smooth spectral function (Roesler and Boss, 2003) due to anomalous diffraction that cause a complementary pattern between scattering and 305 absorption (Zaneveld and Kitchen, 1995). Disturbances can be present in the form of small peaks caused by pigment absorption or oscillations caused by anomalous dispersion when there is a large contribution of small particles to the beam attenuation, but their shape and position are characteristic. Large, abrupt and otherwise irregular spectral behavior likely arises from motion of large particles in the beam cross sectional area. Six samples from Belgian lakes were flagged by this procedure, all with presence of large cyanobacteria aggregates. All BCZ samples of April, taken during a major Phaeocystis bloom, were also 310 discarded due to similar effects of the large colonies. Beam attenuation remains elusive for standard bentchtop spectrophotometric analysis (sequential spectral scanning) under these conditions. As an example, cyanobacteria aggregates were observed floating on the surface of the cuvette at the end of the spectral run, even though the samples were mixed at the start of the measurement.
Another source of bias related to beam attenuation measurements in a spectrophotometer is that total internal reflection will 315 occur for a fraction of the scattered light, dependent on the incident angles on the water side of the quartz wall. Some fraction of the internally reflected light could scatter back into directions in the detector's acceptance angle and artificially decrease beam attenuation measurements (IOCCG, 2018). Our experimental procedure did not include a dark baffle inside the cuvette to reduce this potential effect. Finally, our measurements show an oscillation centered at ≈500 nm, regardless of the system, particle type or concentration. This suggests that this oscillation is an artifact of the measurement procedure, though we could 320 not identify its source. The spectral (LISST-equivalent) beam attenuation and particle scattering coefficients are presented in Fig. 7

Turbidity
Turbidity was quantified as formazin nephelometric units (FNU), a NIR side-scattering (90 • ) VSF magnitude relative to formazin standards. For the BCZ and Scheldt campaigns, turbidity was measured just after water sampling, while for the lakes 325 and lagoon the turbidity was measured after transportation to the laboratory. The transportation might result in changes in particle aggregation due to changes in turbulence. However, we found a single linear relation between turbidity and suspended particulate matter (described later) across all systems and seasons.

Apparent Optical Properties
The AOP measured in situ were the Secchi disk depth (d Secchi ) and the Lambert-equivalent water-leaving bi-hemispherical 330 reflectance (ρ L wl ). ρ L wl was measured with the above water or on-water protocols, depending on the system. The Lambert- equivalent bi-hemispherical reflectance (ρ L s ) measurements of substrate samples (sediments, macroalgae) were performed ex situ, underwater and in air, using natural illumination.
The "Lambert-equivalent" qualification indicates that the bi-directional reflectance distribution function (BRDF) of the targets is assumed to be well represented by the Lambert model, and the hemispherical-directional measurement is converted to 335 bi-hemispherical by scaling it with the cosine-weighted solid angle integral of an hemisphere, π sr. The water-leaving signal is not strictly Lambertian, however this approximation is commonly used for remote sensing purposes (cf. Frouin et al., 2019).

Secchi disk depth
A standard quadrant Secchi disk (black and white, 30 cm diameter) was used to measure transparency. The Secchi disk was deployed from the shady side of the sampling platform, recording the depths of disappearence and reappearance. The Secchi 340 disck depth (d Secchi ) was recorded as the average of the two depths. A tentative correction for the effect of sun zenith angle (described later) at the time of measurement (Lee et al., 2015) was applied following the formulation of Verschuur (1997).
This correction normalizes the Secchi disk depth measurements to the Sun in the zenith. Measured and corrected values are provided. A comparison between turbidity the inverse Secchi disk depth, corrected for the Sun zenith angle, shows a log-linear relation and is presented in Fig. S5 (supplementary material).

Lambert-equivalent water-leaving bi-hemispherical reflectance
Reflectance spectroscopy measurements were performed in Belgian lakes and the coastal lagoon using the "on-water" method, also known as the skylight-blocked approach (Lee et al., 2013(Lee et al., , 2019Ruddick et al., 2019b). Measurements were made with a hand held spectrometer (FieldSpec HH, Analytical Spectral Devices) equipped with a 7.5 • Field-of-View (FOV) foreoptics.
The instrument has bands with 3.6 nm full width at half maximum (FWHM) and a spectral sampling of 1.6 nm, covering the 350 range from 325 nm to 1075 nm. Water-leaving radiance (L wl ) was recorded at 0.5 m horizontal distance from the deployment platform, aligned with the Sun azimuth, and with the opening of the lens' cylindrical shield at 2.5 cm below the water surface.
The global downwelling plane irradiance just above the surface (E d (0 + )) was estimated from near-coincident measurements of the exitant radiance of a sintered PTFE reference target (nominal reflectivity of 12 %), held parallel to the surface Ruddick et al., 2019a). An example of the measurement approach is presented in Fig. S6 (supplementary material).

355
A total of 10 L wl spectra and 5 plaque exitant radiance spectra were averaged per station, with the measurement sequence completed within 2 min. The Lambert-equivalent water-leaving bi-hemispherical reflectance (ρ L wl ) was estimated according to: where is the estimated fractional shadowing error from the instrument and platform, θ is the nadir angle (0 • ) and ∆φ here refers to the platform-sensor system relative azimuth to the Sun (0 • ). The superscript "Meas" for L wl in Eq. 7 refers to the 360 measured water-leaving signal, biased due to shadowing of the water system. Measurements were resampled to a regular 1 nm interval before Eq. 7 was applied. The shadowing error was calculated for the range of IOPs observed in the lakes and lagoon and for the deployment from the inflatable boat and sensor as used in the campaigns, using a backward Monte Carlo radiative transfer code (unpublished). Formally, is a function of the deployment setup, Sun zenith angle, diffuse fraction of E d , a t and c t . However, the main IOP defining is a t (Fig. S7, supplementary material). The correction was applied using 365 field observations of the required parameters. For stations without measurements of IOPs, spectral c p was estimated from a multivariate linear regression against turbidity calculated over all available data and a g was estimated as the median a g of given water system. The diffuse fraction of E d was simulated from the Sun zenith angle for clear skies  and was set to 1 when field observations recorded clouds covering the direct Sun illumination.
The same spectrometer used in the lakes and lagoon campaigns was used in the Scheldt campaign, with the above water approach (Mobley, 1999;Ruddick et al., 2019b, a). For this measurement, ∆φ refers to the view direction relative to the Sun, set at 135 • . The water-system radiance (L ws ) was recorded at a nadir angle of 40 • , the sky radiance (L sky ) was measured at the nominal specular angle from L ws , 140 • , and E d estimated from the exitant radiance of the reference sintered PTFE plaque. ρ L wl was estimated according to:

375
whereρ f is the effective Fresnel reflectance.
For the BCZ campaigns of 2018, reflectance spectroscopy measurements were also made with the above water approach, using a set of three spectroradiometers (VIS-ARC RAMSES, TriOS) fixed on the bow of the RV Simon Stevin (Castagna et al., 2021). The effective Fresnel reflectance was estimated from wind speed and L sky according to Ruddick et al. (2006).
Further details of processing and quality control are described in Ruddick et al. (2006). The RAMSES instruments have 380 a typical bandwidth of 10 nm FWHM with spectral sampling every ≈3 nm. Radiance and irradiance measurements were resampled to a regular 2.5 nm interval before Eq. 8 was applied, and ρ L wl was linearly interpolated to 1 nm interval to match the other reflectance data. The above water measurements were not corrected for possible disturbances by the platform in the spectroscopic measurements (Shang et al., 2020) and were not corrected for the non-nadir viewing geometry (Gleason et al., 2012).

385
The available ρ L wl (0 • ) and ρ L wl (40 • , 135 • ) are presented in Fig. 8. As expected from the wide range of SPM concentrations, ρ L wl at 810 nm varied over three orders of magnitude, from 0.00036 to 0.10560. Similarly, the large diversity in terms of particle composition (e.g., mineral fraction, taxonomy, pigment) and relative contribution of a φ translate to a diversity of ρ L wl spectral shapes. A validation of ρ L wl and the shadowing correction for the on water approach is provided in Fig. 9, by estimating turbidity from ρ L wl (730) following the algorithm proposed by Nechad et al. (2010). The Nechad et al. (2010) algorithm was calibrated 390 with data between ≈1 g m −3 and ≈100 g m −3 and the comparison is restricted to that range, covering all systems with the exception of the Scheldt.

Lambert-equivalent bi-hemispherical reflectance of bottom substrates
Measurements of reflectance spectroscopy of the surface of the sediment cores were made with the portable spectrometer described previously (FieldSpec HH, Analytical Spectral Devices). The PMMA tubes were cut 5 cm above the sediment's 395 surface level, and measurements made in nadir view, with 2.5 cm of water above the sediment. Measurements were performed with the on-water approach as described previously, i.e. an extension to the shield of the foreoptics was submerged to 2.5 cm below the water surface. A circular NIST-traceable Munsel card was used as a submerged reference, gently placed over the sediment, receiving the same illumination as the sediment's surface. An example of the measurement setup is shown in Fig. S8 ( for microscopy examination, revealed an assemblage composed of representatives of the benthic diatom genera Pleurosigma,
Reflectance spectroscopy measurements of the most conspicuous macrophyte species occurring in the Spuikom (Cladophora glomerata, Ulva sp. and Sargassum muticum) were performed in air, with an hyperspectral camera (SOC710-VP, Surface Optics Coorporation). The macrophyte reflectance measurements were made in nadir view with natural illumination. The hyperspectral camera was set up at 1 m above the samples, using a f/2.8 aperture following the observation of lower spatial 410 variability in the visible to red-edge spectral region. The 12 % (nominal) reflectivity sintered PTFE plaque was used to estimate E d and a spectrally flat 5 % reflectivity sheet was used as background. Specimens were folded over a supporting petridish to avoid signal from the background and the reflectance averaged over circular areas to average over the three dimensional structures (Fig. S10, supplementary material). As the macroalgae were washed free of the sediment before the measurements, the determined ρ L s represent pure end members of substrate reflectance.

415
The average ρ L s spectra of macroalgae and the spectra of sediment samples are presented in Fig. 10. ρ L s spectra of macroalage reached values as high as 0.4 in the NIR range, while sediments, independent of biofilm thickness, had an average ρ L s in the NIR of ≈ 0.11. In the visible range, the spectral shape of Sargassum muticum seems indistinguishable from the sediments with thickest biofilms, as expected due to similar pigment composition of brown algae and diatoms. The similarity in pigment composition also explains the similarity between the average ρ L s of healthy specimens of the green algae Ulva sp. and Cladophora samples, the filters were rinsed with distilled water to remove salt from the filtration area and rim of the filter (Strickland and Parsons, 1968). The filters were then dried overnight at 60 • C and cooled to room temperature before mass determinations.
The pre-treatment of filters involved combustion for one hour at 450 • C to eliminate organic components, followed by washing to remove loose glass fibers and blank mass determinations. To calculate the mineral fraction (f m ) of the SPM, the filters were heated to 500 • C for one hour for thermal oxidation of organic matter and cooled to room temperature before new mass 430 determinations. For some filters, the last mass determinations were lower than the blank mass, indicating a combination of very low mineral fraction and loss of glass fibers during manipulation. Those samples had the mineral fraction set to zero. The observed range of SPM across all systems was from 1.02±0.09 g m −3 to 791.19±0.10 g m −3 and the range of mineral fraction was from 0±0.00 to 0.95±0.08. A comparison between SPM and turbidity is presented in Fig. 11.

Particle size distribution 435
The particle size distribution (PSD) was inverted from the LISST measurements of VSF, using the random shaped particle VSF kernels provided by the manufacturer. The LISST-100X with random shaped particle inversion provides the PSD between ppm). The PSDs converted to particle number concentration, using the volume of a sphere with the median diameter of each retrieved bin size range (Buonassissi and Dierssen, 2010), are shown in Fig. 12. Assessing differences in the PSD measured with the two instruments is under further investigation, but beyond the scope of this data report.

Pigment concentration
Pigment mass concentrations were determined using High Performance Liquid Chromatography (HPLC), following the method 445 of Van Heukelem and Thomas (2001). Cells were broken with sonication and the suspension was cleared by filtration through a 0.22 µm syringe filter. The HPLC was equipped with a reverse-phase column (Eclipse XDB C 8 ) and the detection was performed with spectral absorption (Agilent 1100 series, Diode Array Detector). Pigment standards were acquired from the Danish Hydrographic Institute (DHI) and quantified pigments are listed in Table 6. For the BCZ samples, the measurements are part of the regular marine LifeWatch BE sampling campaigns (Flanders Marine Institute, 2021a), described in Mortelmans were processed with the DADA2 algorithm (Callahan et al., 2016) to resolve amplicon single variants (ASVs). Probable contaminant sequences were removed using negative controls, following the method of Davis et al. (2018). Taxonomic assignment to the ASVs was based on the Protist Ribosomal Reference database (PR 2 version 4.12; Guillou et al., 2012). The raw molecular data can be found at the Sequence Read Archive (SRA) of the National Center for Biotechnology Information (NCBI) under the accession number PRJNA778668 (https://www.ncbi.nlm.nih.gov/sra/PRJNA778668).

465
The raw data was further processed to an aggregation level that is relevant for optical monitoring. The assigned taxonomy with PR 2 was updated to follow the taxonomy of the World Register of Marine Species (WoRMS Editorial Board, 2021), aggregated to species rank and filtered to remove non-pigmented organisms. Heterotrophic organisms were filtered at division rank, with the exception of exclusively heterotrophic dinoflagellates, which were filtered at the lowest rank possible based on reference sources (Hasle et al., 1997;WoRMS Editorial Board, 2021). The data was further annotated to indicate: (1)  TChl c = chlorophyll c1c2 + divinyl chlorophyll c3 (cf. Table 6). A value of 0.01 mg m −3 was added to represent in the log-log plot the samples with concentrations below the detection limit.

Flow imaging microscopy
As part of the regular LifeWatch BE samples in the BCZ, organisms in the range of 55 µm to 300 µm were counted and

Bottom cover
For systems in which the bottom can be visible from the surface, the bottom type was described. Bottom cover was classified 485 based on visual inspection from the surface or from images taken with a submersible camera (Fig. 15). For some stations the bottom cover was classified based on sampling of bottom material. The discrete classes were: (1) Sediment (potentially with biofilm); (2) Shells; (3) Cladophora; (4) Sargassum; (5) Brown algae (not specified); and (6) Heterogeneous. The last class was used when the bottom was covered by a complex mixture of other classes. This classification is qualitative in the sense that describes the major composition of the bottom at a given station, but does not provide fractional cover. Heterogeneous.

Ancillary parameters
In addition to the optical and biogeochemical parameters, a series of ancillary parameters were also determined: Time, position, local depth, sampling platform, Sun zenith angle, and visual descriptions of the sky and water system.
For the inland water campaigns, time and position of each station were recorded from a handheld Global Navigation Satellite System (GNSS) receiver (GPSmap 62s, Garmin), which is enabled to receive corrections from the European Geostationary 495 Navigation Overlay Service (EGNOS). The local depth at the time of sampling was estimated with a handheld single beam echosounder (Echotest II, Plastino) at the start of each station. The exception was one very shallow station (0.2 m), where a folding ruler was used. For marine stations, position, time and local depth were taken from the ship's navigation data, acquired with a differential GPS and a mounted single beam ecosounder (JFE 380-25, Japan Radio Co., Ltd.). Sun zenith angles were calculated with the HORIZONS system (Jet Propulsion Laboratory, NASA; https://ssd.jpl.nasa.gov/horizons/). All ancillary parameters, with the exception of the Sun zenith angle, were combined into a single metadata file.

Conclusions and recommendations
This study described in detail the first open dataset of paired optical and biogeochemical measurements in a diverse set of water systems located in Belgium. The wide range of observed conditions and the relative scarcity of similar open datasets in inland waters make this a relevant contribution to the community. Potential users of this dataset are encouraged to contact the authors 505 for further inquiries concerning the data and an updated status of studies in development using the dataset.
Although the raw, measured IOPs and AOPs are provided, their fitted or corrected versions are recommended for use. For example, the fitted a g removes noise and oscillation in the NIR due to salinity mismatch between blanks and brackish or marine samples. The fitted a d was found as the least biased way to remove the absorption signal of NaClO, providing the best estimate of a d and a φ . The b p , and consequentially c p , corrected to an acceptance angle of 0.018 • (equivalent to the LISST 510 instruments) provides a more accurate estimate of the scattering and beam attenuation. The d Secchi was corrected for the Sun zenith angle, normalizing all measurements to the Sun in the zenith, and ρ L wl measured with the on-water method were corrected for platform and instrument shadowing. The results of the fitting and corrections were evaluated through consistency checks between different data types and instruments as presented in the text.

pangaea.de/tok/c67200d99ea9bbbeadd9edec9690f937b5bacbff
Author contributions. AC, RD, and ID performed field campaigns in Belgian lakes. AC, HL, KR, MB, AD, and DD participated on field campaigns in the Spuikom, Scheldt and BCZ. LA joined the Sheldt and BCZ campaigns, JM joined BCZ campaigns and HD joined Spuikom campaigns. AC performed laboratory measurements of IOPs, field and laboratory measurements of AOP, processing of LISST data, measure-520 ments of turbidity, SPM and mineral fraction, classification of the bottom cover, curated the data and wrote the manuscript. MB performed reflectance spectroscopy measurements in the BCZ. ID and RD performed HPLC analysis of pigments. RB and AC performed analysis of DNA metabarcoding. LA performed flow microscopy counts. HD, KS and WV provided infrastructure, planning and methodological discussion. All authors reviewed the manuscript.
Competing interests. The authors declare that they have no conflict of interest.

525
floating biofilms, to Olivier De Clerck for identification of macroalgae, to Koen De Rycker for operating the short corer, to Dieter Vansteenwegen and André Cattrijsse for support during the BCZ and two extended campaigns in the Spuikom, and to the Café Zates for support during the Scheldt campaigns. We are appreciative of all researchers and steering committee members of the PONDER and HYPERMAQ projects for the interesting discussions along the development of this research. We acknowledge the R Core team and the authors of the R packages for developing and maintaining the free software used in this research. BCZ data provided as part of the Flemish contribution to and Mees, J.: Nutrient, pigment, suspended matter and turbidity measurements in the Belgian part of the North Sea, Scientific Data, 22,   Table 5. Summary of measured and derived biogeochemical parameters per water system.