Evaluation of the CDOM Absorption Coefficient in the Arctic Seas by Sentinel-3 OLCI Data

The main goal of our work is the revealing of problematic issues related to estimates of the absorption coefficient of colored organic matter in the northern seas from data of the Ocean and Land Color Instrument (OLCI) on the Sentinel-3 satellites. In particular, a comparison of the OLCI standard error estimates ADG443_NN_err., relating to the measurement and retrieval of the geophysical products, with the uncertainties in the real situation of the northern seas, where the natural conditions are extremely unfavorable (first of all, frequent cloudiness, low Sun heights). We conducted a comprehensive multi-sensor study of the uncertainties using various approaches, first at all, directly comparing the data from satellite (OLCI Sentinel-3 and four other ocean color sensors) and field measurements in five sea expeditions 2016-2019, by using the different processing algorithms. Our analysis has shown that the real uncertainties of the final product are significantly higher than the calculated errors of the ADG443_NN_err., which is 100% and ~10%. The main reason for that is the unsatisfactory atmospheric correction. We present the results of the analysis of the different effecting factors (satellite sensors, processing algorithms, use of the other parameters), and formulate the tasks of future work.


Introduction
The content of our study is an analysis of problematic issues related to estimates of the absorption coefficient of colored organic matter in the northern seas from data of the Ocean and Land Color Instrument (OLCI) on the Sentinel-3 satellites. Three factors determine the importance of the problem: firstly, a subject of our research -the CDOM absorption coefficient, which is usually the main part of the seawater absorption of solar radiation in the visible range; secondly, the area of our study -the Arctic Basin, the region where global warming is most pronounced; third, OLCI -a new generation of ocean color scanners designed to replace well-known devices such as MERIS and MODIS.
Our possibilities for carrying out such a study are based on a large number of field measurements carried out in the northern seas. in expeditions by the Shirshov Institute of Oceanology Russian Academy of Science (SIO RAS) RV Akademik Mstislav Keldysh in 2016 -2019.
Absorption of solar radiation in the surface layer is one of the factors affecting an increase in water temperature and a decrease in the ice cover area in the Arctic seas, where the global warming is most pronounced (see, for example, [1,2]). The absorption coefficient determines the seawater ability to absorb solar radiation, and the colored organic matter is one of the main absorbing components of seawater (also called yellow substance or gelbstoff).
Direct determinations of the concentration of CDOM are quite difficult; therefore, its content is usually estimated through the spectral absorption coefficient ag(λ), which can be obtained with acceptable accuracy using optical methods, both contact and remote [3,4]. The contribution of the Table 1. In total, we collected 341 stations with in situ measurements of the absorption spectra required for the validation of satellite data. However, due to frequent cloudiness, many of these stations could not be compared with satellite data. We focused on the Arctic seas, so for our analysis, we used data only for stations located north of the Arctic Circle; their location is shown in Figure 1. The selection of stations for various types of research is described in Section 3. cloudiness impedes the acquisition of data from simultaneous shipborne and satellite measurements; therefore, the estimation of errors in satellite algorithms is a rather difficult problem for the Arctic seas. Strong horizontal heterogeneity and temporal variability make the error estimation process difficult. Therefore, when using satellite data in the visible range, it becomes necessary to average them over several days or weeks. Figure 2 shows maps of distributions of the number of days with data for each bin (left) and cloud coverage maps, in accordance with the OLCI_CLOUD flags, averaged over the periods of the voyages. On the cloud coverage maps according to the OLCI_CLOUD flag, the proportion of cloudy days for the averaging period is the greater, the higher the gradation value.

Field Measurements
At all selected stations, the absorption coefficients of seawater and its components was determined; at some stations, a floating or a deck spectroradiometer measured the spectral values of the upwelling radiance just beneath the sea surface (a floating device) or the above sea surface (a deck instrument); two irradiance meters BIC and RAMSES (both or one of them) measured the spectral underwater downward and upward irradiances; a transparency meter PUM-A measured the seawater beam attenuation coefficient.

Measurements of the Spectral Absorption Parameters
Spectral seawater absorption coefficients were measured using a portable spectrophotometer in the ICAM (Integrated Cavity Absorption Meter) configuration. ICAM was developed at the Department of Biophysics, Biological Faculty, Moscow State University, Moscow, Russia [24]. Three measured absorption spectra (seawater, filtrate, and empty sphere) for each 330 ml water sample allow us, thanks to using ICAM technology, to eliminate the influence of light scattering.
Knowledge of the effective path length of photons, taking into account their multiple scattering, is necessary to determine the absolute values of the absorption coefficient. Calculations by the Monte Carlo method were carried out, which revealed the absence of dependence of absorption on the scattering properties of seawater. The scattering coefficient in the calculations varied from 0 to 5 m -1 [25].
The CDOM spectra ag(λ) was determined as the difference between the absorption spectral values of the filtrate and pure water [26]. A more detailed description of the measurement and calculation procedure is given in [25,27].

Floating and deck spectroradiometers
For measurements of the spectral upwelling radiance just beneath the sea surface, we used a PRO-1 floating spectroradiometer, developed at the SIO RAS [28]. It also measures the spectral surface irradiance, which makes it possible to calculate the spectral remote sensing reflectance Rrs. The measurements are carried out at drift stations, and the device floats off the ship at the distance of 0-50 m to avoid the ship's effects due to shading or reflecting. Depending on weather conditions, 20-50 spectral scans are averaged during processing using a specially developed algorithm. The subsurface radiance reflectance ρ(λ) is calculated as ρ(λ) = πLu(λ)/Ed(λ), where Lu(λ) and Ed(λ) are the upwelling radiance and the downwelling irradiance just beneath the sea surface. For continuous monitoring of changes in surface irradiance during measurements, a special DNO sensor is used, installed on an open deck in a place free from shading. The measurements were carried out on the AMK 68th, 72nd, and 76th voyages.
On the AMK-65 and AMK-71 cruises, we used a deck spectroradiometer. It was developed at the Marine Hydrophysical Institute (MHI) [29] to measure the spectral remote sensing reflectance in the range of 390-720 nm with a resolution of 5 nm. This deck spectroradiometer uses a prism monochromator. An optical receiving system is installed on its entrance slit, and the two-beam measurement principle is used with a cosine collector for measuring the sea surface irradiance and a radiance channel lens mounted on opposite ends of the tube. The tube itself is attached in front of the entrance slit of the monochromator.
The ratio of the water radiance and the surface irradiance is measured; the calibration was performed against the reference screen. The measurement is taken from the ship at a 45° zenith angle and the azimuthal angles where the sun's glints were minimal. After measuring the sea surface radiance reflectance, the spectrophotometer measures the radiance reflectance the cuvette with absorbing walls and bottom, filled with water. If the measurements are carried out under the same conditions, the use of a cuvette makes it possible to take into account the reflected component. The averaging several sequential measured values allow us to eliminate random errors.

Measurements of the diffuse attenuation coefficient Kd
To calculate the diffuse attenuation coefficient, we used data on underwater spectral irradiance Ed(z, λ), measured by two irradiance meter BIC and RAMSES.
A spectral underwater irradiation meter RAMSES ACC-VIS (TRIOS, Germany) has 190 spectral channels in the spectral range of 320-950 nm (spectral resolution is 3 nm) and accuracy of ±5%. To conduct measurements in the sounding mode, we used a special protective frame, and the data was transmitted to the ship laboratory by a cable. TRIOS provided calibration of the instrument to the National Institute of Standard and Technology (NIST) standards.

of 32
The other irradiance meter, E-meter (BIC 2100), has two measuring modules for the surface and underwater irradiance; both simultaneously measure the irradiance in four spectral bands 443, 490, 555, and 625 nm. The specialists from SIO RAS modernized the device to perform measurements in an autonomous regime, without a cable. We calibrated the instrument at the SIO RAS before each expedition, according to the working standard of the spectral irradiance density.
The measured Ed(z, λ) values allow us to compute the diffuse attenuation coefficients Kd(λ) for underwater irradiance, assuming the exponential law Ed(z) = Ed(0 -) exp[-Kd z], where wavelength λ is omitted for simplicity. In this work, we used Kd(λ) values for the upper uniform layer whose thickness is varied from 10 to 30 m.
The maximum depth Z*, which is available for satellite observations, is assumed equal to the depth for which the contribution of the radiance coming from depths greater than Z* is less than 10% of the total radiance. Then Z* = 1.15/Kd. According to our field measurements, the Kd(490) varied from 0.03 to 0.34 m -1 for the North Atlantic and the Barents Sea and within 0.06-0.36 m -1 for the Kara Sea. Hence the informative satellite depth Z* was within 3.2-18.5 m in the Kara Sea and 3.4-38 m in the Norwegian and Barents Seas. Usually, the thickness of the layer, where the Kd calculations were carried more than the Z* value. Only in the clear waters of the Barents Sea (stations 5557, 5560, AMK-68), the Z* value was more than the depth of the layer for field measurements.

The beam attenuation coefficient measurements
These measurements are used for comparison with the backscattering coefficient values bbp calculated from Rrs values measured with floating/deck spectroradiometers and satellite color scanners (see Section 3.3). The measurements were carried out using a PUM-A transparency meter in the laboratory mode -on water samples or as part of a ship's flow system.

PUM-A technical characteristics:
Attenuation coefficient measurement range 0.050 -1.0 m -1 ; Measurement error of the attenuation coefficient -0.005 m -1 ; Wavelength of attenuation measurement -530 nm; Half-width of the spectral range -25 nm; The amount of internal non-volatile memory -2 GB; Autonomy (time of continuous work)-5 hours; Volume of the measuring cuvette PUM-A -270 mL.
For control of the stability of measurements, the cuvette was filled with ultrapure type 1 water prepared at the Institute of Oceanology of the Russian Academy of Sciences, following ASTM, CAP, and NCCLS standards on a "Simplicity 185" apparatus by "Millipore." The measured values of the attenuation coefficient were c = 0.057 m -1 . The control of the PUM is also carried by the inter-calibration of the PUM-A device with a submersible PUM transparency meter; the beam attenuation coefficient values of the water sample from the bottle, measured by PUM-A in a cuvette, were compared with the PUM data on the same horizon. The difference did not exceed 0.01 m -1 .

Satellite data and software
We ordered the data of the first and second levels of the OLCI-A and OLCI-B sensors at https://archive.eumetsat.int/usc/, and the data MODIS-Aqua, VIIRS NPP and VIIRS NOAA -on the website https://oceancolor.gsfc.nasa.gov/.
Level 2 satellite data were processed using software developed by the Ocean Optics Laboratory of Shirshov Institute of Oceanology [30], designed to batch process large datasets of ocean color scanners. The set of programs includes the modules for calculating new products and recording results in HDF and/or NetCDF format, averaging data, exporting maps in a given projection, matching in situ and satellite observations, etc. The essential feature of this software is the dynamic loading of libraries that can be developed and implemented by users without having to recompile programs. For the convenience of using programs designed for working with data in the NASA format developed by the Ocean Biology Processing Group (OBPG) at NASA's Goddard Space Flight Center, we prepared a converter transforming OLCI data to the above-mentioned SeaWiFS-like format and applied this software for converting the OLCI level 2 data. The following standard products were used: OLCI water-leaving reflectance (16 bands), absorption coefficient, Total Suspended Matter (TSM) concentration, and the diffuse attenuation coefficient Kd. When processing, the following data flags were used: OLCI_LAND, OLCI_CLOUD, LCI_CLOUD_AMBIGUOUS, OLCI_CLOUD_MARGIN. For maps building, we binned the data on a 3 by 3 km grid.
The software [30] allows us to process the Level 2 data from MODIS-Aqua, VIIRS SNP, and VIIRS NOAA sensors. These data include the remote sensing reflectance Rrs as well as the diffuse attenuation coefficients Kd(490). Using the values of the Rrs, the absorption coefficients of yellow substance ag and the particle backscattering coefficients [4] were calculated for the field stations where a time difference between satellite and field measurements was no more 24 hours. Since the algorithms for these products were developed for MODIS color scanners, we interpolated the VIIRS Rrs to MODIS wavelengths. When calculating the parameters, the pixels flagged with LAND, CLDICE, SEAICE or STRAYLIGHT were used.
The software Sentinel Application Platform (SNAP), developed by Brockmann Consult, allows us to work with the OLCI Level 1 data and derive the water-leaving radiance from the TOA (Top Of Atmosphere) radiances. The neural network processor C2RCC (Case 2 Regional Coast Colour), developed by Doerffer and Schiller [31], provides us the opportunity to derive the seawater inherent optical properties, including the seawater absorption of detritus and yellow substance as well as the particle backscattering coefficient.

Algorithms
For calculating the IOP from remote sensing data, primarily for calculating the absorption coefficients, and also the backscattering coefficient bb and the TSM concentration, we used the algorithms listed below.

OLCI NN
The standard product ADG443_NN (colored Detrital and Dissolved Material absorption coefficient) is obtained in OLCI Level 2 files, as a result of the operation of IMT (Inverse Modeling Technique) neural networks. The spectral remote sensing reflectance Rrs is obtained to the input after standard or neural network atmospheric correction.
The SNAP program, developed by ESA (The European Space Agency), as a result of the operation of the C2RCC neural network processor, can also provide other IOPs data. GIOP A model of the optical properties of seawater -GIOP (Generalized Inherent Optical Property) was presented in [33,34]. The algorithm is based on the application of various analytical models, parameterized for the surface layer in different seasons. The algorithm uses Rrs (λ) in the visible range and can derive the absorption coefficients of the colored dissolved organic matter (ag(λ)) and of phytoplankton pigments (aph(λ)), as well as the particle backscattering coefficient bbp(λ). Based on the obtained data on the remote sensing reflectance Rrs (λ), the model performs an inversion to find the optimal set of the bio-optical characteristics of the water column, which minimizes the difference between the simulated and the measured values. The model provides for the possibility of adjusting input parameters (such as chlorophyll concentration, the spectral slope for the yellow substance absorption).
The Quasi-Analytical Algorithm (QAA) This algorithm was evolved by Lee et al. [22] to determine the backscattering and absorption coefficients from the spectral remote-sensing reflectance Rrs(λ), as a result of analysis of several empirical, analytical and semi-analytical models, in particular, Gordon et al. [35], Lee et al. [36]. The spectral bb(λ) is presented by the generally utilized expression [37,38].
Along with the bb(λ), the QAA algorithm can derive the seawater absorption coefficient and make estimates of the values of the detrit-gelbstoff and the phytoplankton pigment absorption 9 of 32 coefficients. The European Space Agency (ESA) Ocean Colour (OC) Climate Change Initiative (CCI) selected the QAA algorithm to derive bbp for creating a long-term series of satellite ocean color products.
Regional semi-analitycal algorithm (RSA) The regional semi-analytical algorithm was developed to estimate the yellow substance absorption coefficient ag, particle backscattering coefficient bbp, and diffuse attenuation coefficient Kd(λ) at four spectral channels 443, 490, 555, and 625 nm in the Kara and White Seas using MODIS satellite data [23]. It uses the Rrs data for the wavelength range  488 nm; the above-surface remote-sensing reflectances Rrs(412) and Rrs(443) are not used, as the probability of atmospheric correction errors for them is very high. The inverse problem of finding seawater absorption a(λ) and backscattering bb(λ) is solved using low-parametric models in which these seawater coefficients are represented as a superposition of the contributions of the main components. The absorption coefficient a(λ) is defined as the sum of the absorption of pure sea water, yellow substance or CDOM, and the phytoplankton pigments, while bb(λ) is defined as the superposition of the backscattering by pure sea water and the suspended particles. The contribution of chlorophyll is accounted for by using a regional algorithm for the Kara Sea [39]. An iterative approach was used to improve the accuracy when estimating the slope of the absorption spectrum. As a result of solving the inverse problem, two parameters are defined: ag(443) and bbp(555), these parameters also allow calculating the Kd(λ) value by using a Gordon's formula [40]. The development and validation of the RSA algorithm was performed with the shipboard measurement data, which contain no errors from atmospheric correction; the algorithm was also tested on satellite data and showed acceptable results [23]. In the regional algorithm [39], the chlorophyll concentration is calculated with the formula ln(Chl) = −6.64 ln[Rrs(531)/Rrs(547)] -0.265. This formula was derived based on the satellite data for Rrs and in situ measurements of the chlorophyll concentrations in the Kara Sea in 2007, 2011 and 2013-2015. In the present work, the RSA algorithm was modified to the OLCI spectral channels and now it uses the Rrs data for the wavelength range greater or equal 490 nm.
Algorithm for determining bbp and TSM.
The algorithm [41] was developed for cases where the semi-analytical algorithm cannot work properly due to large errors in atmospheric correction; this situation is typical in the Barents and White Seas. The simplified algorithm uses only two wavelengths SeaWiFS 510 and 555 nm, where the errors in atmospheric correction are much less than at 412, 443 and 490 nm. To obtain the particle backscattering coefficient bbp = bbp(555), the diffuse attenuation coefficient Kd(555) and the parameter X(555) are calculated, where X(555) = bb(555) / [a(555) + bb(555)], a(555) and bb (555) are the seawater absorption and backscattering coefficients. The value of X (555) is determined through the value of the normalized water-leaving radiances Lwn(555), and Kd(555) through the ratio Lwn (510) / Lwn (555). The value [a (555) + bb (555)] can be obtained from Kd(555) using the standard formula [40]. The backscatter bbp is calculated as the difference between the bb of seawater and the known backscatter bbw of pure water.
The TSM concentration from satellite data is calculated using a regional algorithm obtained on the basis of ship measurements on cruises 13 and 14 of the R/V Akademik Sergey Vavilov (August-September 1998). The data on quasi-simultaneous measurements of the upwelling radiance Lwn (floating spectroradiometer) and TSM were used. First, the bbp values were obtained, and then the regression equation TSM vs. bbp was derived: TSM = 73.5 bbp +0.016, where bbp is measured in m -1 and TSM is in mg/L. The average error in determining the concentration of suspended matter is about 30%.

Results
This section presents the results of comparing the absorption values of the colored organic matter ADG_443, available on the OLCI website as a standard product ADG_443_NN of the 2nd level, with the data of direct measurements of the absorption coefficient by the ICAM device (section Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 9 August 2020 doi:10.20944/preprints202008.0229.v1 2.2). For control, we also calculated the ADG_443 values using the SNAP program C2RCC processor [32].
Errors in estimating bio-optical parameters from satellite data are due to various factors, in particular, errors in an atmospheric correction, connected with bad observation conditions and errors of the algorithms used. In the Arctic seas, significant losses in satellite data occur due to cloudiness, a frequent phenomenon in this region. It was cloudiness that did not allow obtaining a full amount of data for the above comparison, even thorough a large number of field measurements was obtained in five voyages of the R/V «Akademik Mstislav Keldysh» in 2016-2019, It is worth to note a principal difference between the field data and satellite measurements: the former were obtained at a point, the latter for a sufficiently large area (the size of such an area (pixel) for OLCI data with a full spatial resolution of WFR is 0.3 km).
In our work, we try to compensate for the lack of a sufficient amount of data to obtain reliable statistical estimates by the complexity of the comparison. For that, we used data from other field measurements to confirm the results obtained. These data included the remote sensing reflectance Rrs(λ) measured by floating or deck spectroradiometers, the spectral diffuse attenuation coefficient of the underwater irradiance Kd(λ), the beam attenuation coefficient c, as well as data from other satellite color scanners, results of calculations by various algorithms (see Section 2.2-2.3).

Validation and analysis of the ADG_443 values derived from OLCI data
The absorption coefficient of the colored organic matter ADG_NN (starting now ag_olci_std) is available as a standard product of the 2nd level of the OLCI instrument on the Sentinel 3A, B satellites (see Section 2.2). Results of comparing ag_olci_std with the data of direct measurements of the absorption ICAM (ag _icam) are shown in Figure 3.  Figure 1 (map)); Fig. 3.b (89 points) excludes river mouths, as well as stations with large zenith angles of the sun (more than 75 0 ) and other factors that impede obtaining quality data (see Section 3.2).
Even after excluding the above stations, as we see in Figure 3.b, the scatter of points is enormous, especially for the Kara Sea, where the difference between the OLCI_ADG_443_standard values for the same ag value according to ICAM data can be more than 30 times. The main reason for such discrepancy is the errors of atmospheric correction since the results for the absorption values calculated from the Rrs data measured by the floating (deck) spectroradiometer are incomparably better (see below). An analysis of atmospheric correction errors given in Section 3.2 allows us to select 15 stations with acceptable atmospheric correction. Station 6240 in the Kara Sea was excluded due to the strong spatial variability around this station (see Section 3.2), and station 5704 as located south the polar front. Below, we have 13 selected stations, however, for some of them, several passes of Sentinel 3A, B satellites are acceptable, and the total number of OLCI data used was 19 (10 for the Barents Sea, 9 for the Kara Sea). Figure 4 shows the results of comparing the absorption values calculated from the Rrs measurements by the floating (deck) spectroradiometer using different algorithms with the values measured by the ICAM device at the 13 aforecited stations. The values of the determination coefficients R 2 turned out to be large enough for all the algorithms used: 0.656 for RSA, 0.922 for QAA, 0.960 for GIOP.
The calculation of the absorption coefficients from satellite data was also carried out for four other satellite sensors -MODIS-Aqua, MODIS-Terra; VIIRS SNPP, and VIIRS NOAA-20) using the above algorithms. Along with the standard ADG_443_NN values, the ag_OLCI_C2RCC values were calculated using the L1 level OLCI data by the C2RCC processor of the SNAP program [32]. All algorithms, except for neural networks, gave negative values in the calculations for individual flyby; that caused the difference in the number of data pairs (Table 2) The atmospheric correction errors are described in more detail in Section 3.2. Figure 5 shows the regressions of the calculated values of ag_olci_std and ag _OLCI_C2RCC versus the measured values of ag_icam; the regression parameters are presented in Table 3.
As can be seen from the Table 3, for the Barents Sea OLCI standard product, a quite acceptable correlation is obtained; the relationship for the Kara Sea is insignificant. For the ag _OLCI_C2RCC values, a significant relationship is observed for both seas, and this relationship is even closer for the Kara Sea than for the Barents Sea.
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 9 August 2020 doi:10.20944/preprints202008.0229.v1    Table 4. Unfortunately, for the Kara Sea, the correction equation for ag_OLCI_standard is impossible due to the insignificance of the correlation. Still, it is possible for the absorption coefficient values calculated with the C2RCC processor.
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 9 August 2020 doi:10.20944/preprints202008.0229.v1 Applying the C2RCC processor, as can be seen from Table 4, the coefficient of determination for the Kara Sea turns out to be even higher than for the Barents Sea, and the regression errors are also quite acceptable. As noted above, the statistical support of the obtained equations is insufficient, and it is too early to recommend them for practical use. In Section 3.4, we will test the obtained equations with the diffuse attenuation coefficient values.
At the end of this section, we will give an example of correcting the ag spatial distributions using the derived equation for the Barents Sea. Figure 6 shows the ADG_443 spatial distributions in the Barents and partly in the Norwegian Seas built with standard ag_OLCI_standard and corrected ag_corr data for AMK-65 and AMK-68 cruises. Adjusted distributions look more plausible; a vast white area between 30 and 50 deg. d. in Figure 6B due to a lack of satellite data due to cloudiness.

Results of atmospheric correction
As shown in the previous section, the quality of the atmospheric correction is critical for obtaining reliable absorption coefficient values. Unfortunately, in the conditions of the Arctic seas, ensuring an acceptable quality of atmospheric correction is a most challenging task due to extremely unfavorable natural conditions: it is often cloudy, which leads to a significant loss of data; low angles of the Sun. In this section, we estimate the atmospheric correction errors as a function of various factors, using Rrs as a reference to compare the values of the spectral radiance reflectance directly above the sea surface.
We selected 189 data pairs based on the requirement that a time difference between the in situ measurement and the satellite overpass is not exceeding 24 hours. Figure 7 shows the dependencies of the standard error RMSE of Rrs on the difference between the time of the in situ measurement and the satellite overpass ( Figure 7A) and on the solar zenith angle ( Figure 7B).
Both figures show a large scatter of points. As seen, a weak trend is observed in both cases -the errors increase with increasing the time difference between the field and satellite measurements and the solar zenith angle. Still, one can see the RMSE values of about 0.001 even though with a time difference of more than 20 hours and the solar zenith angles more 70°. Conversely, the errors exceeding 0.004 can be detected with a time difference of fewer than 5 hours and angles less than 50°. Next, we will consider the influence of various factors in more detail, and we will start with the errors of atmospheric correction associated not with the atmosphere and observation conditions, but with the optical characteristics and processes occurring in the surface water layer.
It is worth noting that in the Kara and Laptev Seas, the strong influence of river runoff leads to a decrease in the water-leaving radiation values in the short-wavelength part of the spectrum due to absorption by dissolved organic matter, which increases the errors in the determination of the absorption coefficient ADG_443. As an example, Figure 8 shows all the Rrs spectra for station 6240, performed in the Kara Sea on July 14, 2019. For this station, we collected the most significant amount of satellite data -23 files. The colors show the data of the various sensors; black line -the result of direct determinations; gray lines -the spectra with the RMSE > 0.001.  The spatial distribution of the beam attenuation coefficient c measured in the Kara Sea on the day of station 6240 (July 14, 2019) using the flow-through measuring complex is shown in Figure 10. The values of c(532) exceeding 1 m -1 correspond to the waters of the surface desalinated layer formed in the summer season by the Ob and Yenisei river runoff [43]. These waters are characterized by lower salinity values and a high content of colored dissolved organic matter (CDOM) compared to waters outside the influence of river runoff [42]. Station 6240 is located in the area where waters with different optical characteristics come into contact, which causes the observed differences in the values of the absorption coefficient, calculated using different algorithms from different data.
Comparison of the total amount of satellite data for the stations, where the shipboard reflectance measurements were carried out, with the amount of data remaining after applying the criterion for standard error values RMSE < 0.001 for all used ocean color scanners is presented in Table 5. It should be noted that the results of the OLCI AAC better match to shipboard data (13 spectra out of 27 remained) than the results of BAC correction (3 spectra out of 27 remained).  Figure 11 shows the RMSE values for all 27 OLCI spectra obtained using the alternative atmospheric correction, depending on the solar and viewing zenith angles, as well as the difference in the azimuthal angles. Flags are marked with symbols. There is a little amount of data without flags -only six spectra. That indicates unfavorable conditions for satellite observations in the studied seas. The most frequent flag is the 'cloud risk' (12 points), the 'sun glint risk' flag is observed at four locations, all other flags at five. The errors increase at the solar azimuthal angles more 75, and the viewing angles greater than 15. It should be noted that the risk of cloudiness or glint does not always lead to errors.

In situ measurements in the Barents Sea
In this section, we consider the measured data from the AMK-65 and AMK-68 cruises. The particle backscattering coefficient bbp values were not directly measured in our field studies. The bbp values were calculated using the Rrs values, measured by a floating (AMK-65) and a deck (AMK-68) spectroradiometers; the calculations were made for a wavelength of 555 nm using the different algorithms described in 2.2.2: the simplified regression algorithm (RSA) [23]; a quasi-analytical algorithm QAA [22], and the model of the optical properties of seawater -GIOP (Generalized Inherent Optical Property) [33,34].
The bbp (555) data for the selected stations were compared with the attenuation coefficient c, m -1 data for the subsurface layer (5 m). Direct determinations obtain the c (m-1) data with the PUM-A device (Section 2). Figure 12 shows the results of the comparison of the bbp(555) values using different algorithms and c (m -1 ) measured in situ.

Figure 12
Comparison of the backscattering coefficient of particles bbp (555) with attenuation coefficient c (m-1) data for the Barents Sea. Red circles -SRS algorithm, blue -QAA algorithm, green -GIOP algorithm. Black line is a correlation based on SRA, black dotted line -correlation based on QAA data, dash-dotted line -based on GIOP data. Table 6 presents the statistical parameters of the regression of the backscattering coefficient of particles bbp (555) through the values of the attenuation coefficient c (m -1 ) for different algorithms.  Figure 12 and Table 6 show a good agreement between the bbp (555) and the c (m -1 ) values: the coefficient of determination R 2 = 0.95, R 2 = 0.94 for the QAA and GIOP algorithms respectively, and R 2 = 0.96 for RSA. Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 9 August 2020 doi:10.20944/preprints202008.0229.v1

Comparison of the particle backscattering coefficient values derived from field and OLCI satellite data
Stations were selected in the Barents and the Kara Seas at the AMK-65, 68, 72, and 76 cruises for comparing the bbp data derived from field and OLCI satellite data. The stations included the data from a floating or a deck spectroradiometer, satellite data from OLCI, MODIS and VIIRS, and absorption data measured by ICAM. Stations 6240 and 6229_2 of the AMK-76 cruise were removed from the consideration (see 3.1, 3.2). For OLCI data, the calculation was performed using the Rrs values available as the results of the BAS (standard) and AAS (NN) atmospheric correction. In the last case, the Rrs data of the 2nd level were derived from the TOA Level 1 radiance data, using the C2RCC processor of the SNAP program [32]. For the calculations in the Kara Sea, the regional semi-analytical RSA was also used. Figure 13 shows the regression lines for the Barents (a), the Kara and Laptev seas (b). Table 7 presents the parameters for different regions, algorithms, and satellites.
In the Kara Sea, a correlation with R 2 > 0.3 is observed only in the case of the C2RSS processor, the QAA algorithm. That may be due to the strong influence of river flow in this region. A better correlation is observed for stations in the Barents Sea. For example, for standard OLCI level 2 data using the GIOP algorithm, R 2 = 0.92. Figure 13. Comparison of bbp (555) in situ and satellite calculations for the Barents Sea (a), the Kara Sea and the Laptev Sea (b). Red circles -RSA algorithm of the 2nd level data, red crosses -RSA_C2RSS algorithm; blue circles -QAA algorithm for data of the 2nd level, blue crosses -QAA_C2RSS algorithm; green circles -GIOP algorithm for data of the 2nd level, green crosses -GIOP _C2RSS algorithm; yellow circles -SRA. Straight lines -direct correlation -match the colors of the second-level algorithms, dashed lines -for C2RSS algorithms, black straight -for RSA  Figure 14 shows the results of comparison between the values of ag(443) and Kd(443) measured in situ, separately for the Barents and Norwegian Seas (AMK 68 - Figure 14A) and the Kara Sea (AMK 72 and 76 - Figure 14.B) The ag coefficient was measured using ICAM, Kd was calculated from the data of the underwater irradiance measured by two instruments (BIC and Ramses) in the upper homogeneous layer. If measurements for the station were made with both instruments, then the Kd coefficient was determined as the average between these measurements.  Table 8 presents the parameters of the regression between the values ag(443) and Kd(443). It can be seen that a reasonably good correlation is seen for the Kara Sea, with data from the cruises AMK 72 and 76: the coefficient of determination R 2 = 0.73. For the Barents Sea, we also get a high correlation (R 2 = 0.72), but the regression equations in these two seas differ significantly. At the same time, the stations with coccolithophore blooms (see Figure 14A) fit well into the obtained regression. For the waters of the Norwegian Sea, there is no stable relationship between ag(443) and Kd(443): R 2 = 0.21, but the resulting regression equation does not differ much from that obtained for the Barents Sea. The results obtained indicate the consistency of measurements of ag and Kd.

Consistency between the data derived from field measurements
For checking the consistency of the field measurements, we calculated Kd using a simple formula: Kd = (1+0.005 0) a + 3.47 bb, where a and bb -the seawater absorption and backscattering coefficients, 0 -the solar zenith angle [44]. This formula provides reasonable accuracy for waters with coccolithophore bloom and no bloom [45].
We use the spectral values of the absorption coefficient a(λ), measured by ICAM, and the backscattering coefficient determined as the sum of the scattering by the particles and pure water, the particles backscattering bbp was calculated from the in situ measured spectra Rrs(λ) using the algorithm [41].
The correspondence of the Kd values calculated by formula [44] and based on in situ measurements of underwater irradiance for 18 stations in the Barents Sea (AMK 68) is shown for two wavelengths 443 and 490 nm in Figure 15. It is seen an excellent agreement of the Kd estimates obtained for both spectral channels: in both cases, the coefficient of determination is R2 = 0.98; The RMSE equal to 0.028 and 0.035 m-1 for 443 and 490 nm, respectively. The stations with coccolithophore blooms do not worsen the obtained correspondence. That proves a good agreement between the data of our field measurements of the absorption coefficient a(λ), remote sensing reflectance spectra Rrs (λ), and the diffuse attenuation coefficient Kd (λ).   Table 9 shows the parameters of the correspondence between the Kd(490) values calculated from OLCI and field data if we use satellite data for the nearest pixel. As seen, from Figure 16 and Table 9, a good agreement between satellite and field data is observed for the Barents Sea (AMK 68): coefficient of determination R 2 = 0.76, a standard error RMSE = 0.05 m -1 , the relative error RE = 20%. But these calculations were carried out only for 4 stations. The stations with coccolithophore bloom did not show much difference from other stations in this region.
In the Kara Sea (AMK 72 and 76), in general, the satellite values of Kd (490) are approximately two times lower than Kd(490) obtained from in situ measurements: the ratio of the mean values is 0.54, RE = 45%, but for stations in clear waters with Kd(490)  0.1 m -1 satellite data errors do not exceed 15%. N, the number of pairs; R 2 , the coefficient of determination; <Y>/<X>, the ratio of mean Kd(490) values; RMSE, the root mean squared error, m -1 ; RE, the relative error Figure 17 shows the correspondence between the values of diffuse attenuation coefficient Kd(490) calculated from field measurements and satellite data (MODIS and VIIRS). As in the case of OLCI, the Kd(490) values are obtained from level 2 files, so they are the standard product of these scanners. The Kd(490) values in the nearest pixel are used. Since the time of overpass of the considered satellites differs from each other, the set of presented data for each scanner is different. Unlike OLCI, for MODIS and VIIRS scanners, we managed to obtain data for station 5580 in the Barents Sea with an intense coccolithophore bloom observed (directly measured concentration of the coccolithophore plated cells was approximately 5×10 6 cell/L, the detached coccoliths 1.  Table 10 shows the parameters of the correspondence between the Kd(490) values from satellite (MODIS and VIIRS) and field data if satellite data are used for the nearest pixel. The calculated parameters are almost the same for both scanners and depend little on the region: the regression equations are practically the same, and the errors also. The satellite estimates of Kd(490) reasonably coincide with field data in clear waters, where Kd(490) < 0.2 m -1 , and give underestimates in more turbid waters, which are the result of the influence of intense coccolithophore bloom in the Barents Sea, and the river runoff in the Kara Sea. Table 10. Correspondence parameters between the Kd(490) values obtained from field measurements (X) and from satellite data (Y). Satellite data are taken for the nearest pixel. All parameters are the same as in Table 9. If we compare the Kd(490) estimates from the OLCI data with the MODIS and VIIRS estimates, then for all three scanners a good agreement with the field data is obtained in clear waters (Kd (490) < 0.2 m-1), and in more turbid waters they underestimate the Kd (490) value by about a factor of 2.

Discussion
Specialists in the various scientific and engineering directions, using satellite data, need to know the uncertainties in the end product, not only the errors in measurement and retrieval of the geophysical parameter used. The goal of our study is to assess the real possibilities of evaluating the absorption of seawater and colored organic matter using OLCI satellite scanner data in the northern waters -the Barents Sea (partly, the Norwegian Sea) and the Kara sea ( partly, the Laptev Sea). Our possibility to reach the above goal is based on the availability of the appropriate data from field measurements in five expedition voyages of the R/V Akademik Mstislav Keldysh (SIO RAS) in the period 2016 -2019.
As we mentioned in the Introduction, the conditions for obtaining satellite data in the Arctic Basin are extremely unfavorable, and in such situations, the real uncertainties may differ significantly from the errors of the instrumental measurements and the parameter retrieval by the developed algorithms. An essential advantage of the OLCI data products is that it provides the user not only the data files with the error estimates but the annotations data files comprising the classification, Water Quality, and Science Flags (WQSF). The WQSF file provides information about the validity and suspicious quality; Level-1B products contain atmospheric and meteorological data, solar and satellite angles.
We executed our study by applying the comprehensive approach based on the analysis of multi-sensor and multi-variable data from satellite and in situ measurements. Our analysis included satellite data from OLCI Sentinel-3A and 3B, MODIS Aqua and MODIS-Terra, VIIRS SNPP, and VIIRS NOAA-20 from 2016 to 2019, and the field data on the spectral absorption coefficients of seawater and the colored organic matter, the spectral remote sensing reflectance above and beneath sea surface, the spectral diffused attenuation coefficient, the beam attenuation coefficient. For calculating the inherent optical properties (IOP) from remote sensing data, primarily for calculating the absorption coefficients, the backscattering coefficient, and TSM concentration, we used the different algorithms.
In particular, we compared the OLCI ADG443_NN values obtained as a standard Level 2 product and derived by the C2RCC neural network processor of the SNAP program, developed by the European Space Agency (ESA) [ [31,32]. We also performed such a comparison for the spectral The bad observation conditions are the main factor determining a significant loss of data entirely or in their quality. In section 2.1, Figure 2 shows the distributions of the number of days with data for each bin and cloud coverage following the OLCI_CLOUD flags for AMK-65, 68, and 76 expeditions, averaged over the periods of study. There are the bins with no pixel with data over the whole expedition period; the numbers of the day with at least one pixel per bin in the most of the Barents Sea did not exceed 6. The cloud coverage maps, according to the OLCI_CLOUD flag, characterize the cloud coverage distributions over the study period.
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 9 August 2020 doi:10.20944/preprints202008.0229.v1 Figure 18C shows, if we take the value of 0,02 m-1 as an absolute error for the values a_olci_std less than 0.1 m -1 , and 20% as a relative error for the values more than 0.1 m-1, we must reject more than a half the total number pairs. The main reason for the appearance of bad points is unsatisfactory atmospheric correction. In section 3.2, we consider this problem in detail.As a parameter for analysis of the atmospheric correction errors, we used the values of the remote sensing reflectance Rrs. We selected 189 pairs of the Rrs() spectra for which a time difference between the in situ measurement and the satellite overpass did not exceed 24 hours. For each of the selected pairs, we calculated the standard deviation for the spectral difference between the spectra. We built the dependences of the RMSE values on the time difference and from the solar zenith angle. Analysis of these dependencies allows us to present the reasonable limitations relating to the further use of the satellite Rrs spectra: the RMSE  0.001. This value is not a strict requirement; at present, we consider it as a reference point.
For analysis of the effectiveness of the retrieval OLCI ADG_NN, we selected 15 stations with acceptable atmospheric correction. Two of them were rejected due to other reasons, and we got 13 selected stations. For some of them, several Sentinel 3A, B passes were acceptable, and the total number of OLCI data used was 19 (10 for the Barents Sea, 9 for the Kara Sea). The main results of our work are formulated in Conclusions.

Conclusions
We conducted a comprehensive multi-sensor study of the uncertainties in satellite OLCI data on the color dissolved organic matter absorption under specific conditions of the Arctic Seas. For that, we used various approaches, comparing data from satellite (Sentinel-3 and four others) and field measurements in five sea expeditions 2016-2019, as well as different the data processing algorithms. The most significant, in our opinion, the results of the performed work, are presented below.

•
First of all, we should note the significant losses in satellite data occur due to cloudiness, a frequent phenomenon in the Arctic region. That did not allow obtaining a full amount of data for our analysis, using a large number of in situ data measured in five sea expeditions. We managed to select, from the satellite ADG_443 spatial distributions for our cruises, only 151 pairs for the comparison. We shortened this set to 89 pairs after rejecting the Ob and Yenisei estuaries in the Kara Sea as well as stations with large zenith angles of the sun and other factors that impede obtaining quality data.

•
The comparison of the OLCI ADG443_NN standard products with the directly measured by the ICAM spectrophotometer values of the yellow substance absorption coefficient ag has shown that the real uncertainties of the final product quite different from the calculated errors the ADG443_NN_err. The former can exceed 100%, while the latter even for a single OLCI file, ~10 %. The main reason for that is the unsatisfactory atmospheric correction.

•
An analysis of the atmospheric correction errors allowed us to select 19 pairs for further analysis (10 for the Barents Sea, 9 for the Kara Sea). For control, we also calculated the ADG_443 values using the L1 level OLCI data by the C2RCC processor of the SNAP program. We have derived the absorption coefficients, not only from OLCI data, but from other satellite sensors -MODIS-Aqua, MODIS-Terra; VIIRS SNPP, and VIIRS NOAA-20). We also used the different algorithms: QAA (Quasi-Analytical Algorithm) [22], SIORAS RSA (Regional Semi-analytical Algorithm) [23], GIOP (Generalized Inherent Optical Property) [33,34].
For the OLCI standard product, a quite acceptable correlation was obtained only for the Barents Sea; for the Kara Sea it was insignificant. For the ag _OLCI_C2RCC values, a significant relationship was observed for both seas.

•
We have derived the correcting equations ag_corr vs. ag_olci_std and ag_corr vs. ag_C2RCC, allowing us to obtain the satellite absorption coefficient values closer corresponding to the measured ones. For the Barents Sea, we obtained reasonable results for the regression ag_corr vs. ag_olci_std: the coefficient of determination R 2 = 0.670, the regression error sregr = 0.030 m -1 , the Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 9 August 2020 doi:10.20944/preprints202008.0229.v1 relative error (variation coefficient) RE = 55%. For the Kara Sea, there is not the acting equation for ag_OLCI_standard due to the insignificance of the correlation. Still, it is possible for the absorption coefficient values calculated with the C2RCC processor: R 2 = 0.624, sregr= 0.045 m -1 , RE = 30%. Of course, the statistical support of the derived equations is insufficient, and we present them not for practical use.

•
We have estimated the satellite atmospheric correction errors depending on the various factors, using the spectral radiance reflectance Rrs directly measured above the sea surface as a reference. The standard error RMSE increases weakly with increasing the time difference between the in situ measurement and the satellite overpass as well as on the solar zenith angle. Still, the RMSE values can be about 0.001 even though with a time difference of more than 20 hours and the solar zenith angles more 70°. Conversely, the errors exceeding 0.004 can be detected with a time difference of fewer than 5 hours and angles less than 50°. The other factors that affected RMSE are the azimuth angles difference and the situation in the surface water layer.

•
In our field studies, we did not directly measure the particle backscattering coefficient bbp. We calculated the bbp values using the Rrs values, measured by a floating (AMK-65) and a deck (AMK-68) spectroradiometers, and applied the different algorithms (the simplified regression algorithm (SRA) [41]), QAA [22], GIOP [33,34]. For the Barents Sea, we checked the consistency of the bbp(555) data for the selected stations with values of the attenuation coefficient c, m -1 directly measured in the subsurface layer (5 m) by the PUM-device. For all algorithms, we found a good agreement between the bbp(555) and the c(m -1 ) values: the coefficient of determination R 2 was equal to 0.95 (N=31) for the QAA, 0.94 (N=29) for GIOP, and 0.96 (N=31) for SRA.

•
We have also got the results of a comparison between the bbp values derived from field Rrs() data measured by a floating or a deck spectroradiometer and from OLCI satellite data. In the last case, the Rrs() data were obtained both as the results of the BAS (standard) and AAS (NN) atmospheric correction. The best results we obtained for the Barents Sea using the OLCI Rrs_standard values and the QAA and GIOP algorithms: N=11, R 2 = 0.87 and 0.92, RMSE = 0.0072 m -1 and 0.0059 m -1 , RE = 28 and 26%. For the Kara Sea, we have got the only result with the GIOP using the OLCI Rrs_C2RCC values: N=9, R 2 = 0.33, RMSE = 0.0002 m -1 , RE = 56%.
• comparison of its values with the values of the diffuse attenuation coefficient Kd(490) calculated from the data of field measurements. We found a good agreement between satellite and field data for the Barents Sea (AMK 68): coefficient of determination R 2 = 0.76, a standard error RMSE = 0.05 m-1, the relative error RE = 20% (but these calculations were carried out only for 4 stations). The stations with coccolithophore bloom did not show much difference from other stations in this region.
In the Kara Sea (AMK 72 and 76), in general, the satellite values of Kd (490) were approximately two times lower than Kd(490) obtained from in situ measurements: the ratio of the mean values was 0.54, RE = 45%, but for stations in clear waters with Kd(490)  0.1 m -1 satellite data errors did not exceed 15%.

•
For checking the consistency of the field measurements, we calculated Kd using a simple formula: Kd = (1+0.005 0) a + 3.47 bb, where a and bb -the seawater absorption and backscattering coefficients, 0 -the solar zenith angle [44]. This formula provides reasonable accuracy for waters with coccolithophore bloom and no bloom [45]. We used the spectral values of the absorption coefficient a(λ), measured by ICAM, and the backscattering coefficient determined as the sum of the scattering by the particles and pure water, the particles backscattering bbp was calculated from the in situ measured spectra Rrs(λ) using the algorithm [41].
We have found an excellent agreement of the Kd estimates obtained for spectral channels of 443 and 490 nm for 18 stations in the Barents Sea (AMK 68): for both the wavelengths, R 2 = 0.98, the RMSE equals 0.028 and 0.035 m -1 for 443 and 490 nm, respectively. The stations with coccolithophore blooms do not worsen the obtained correspondence. This proves a good agreement between the data of our field measurements of the absorption coefficient a(λ), remote sensing reflectance spectra Rrs(λ), and the diffuse attenuation coefficient Kd(λ).

•
Because the absorption coefficient usually make a main contribution in the diffuse attenuation coefficient, it was interesting to check the correlation between these parameters. The ag coefficient was measured using ICAM, Kd was calculated from the data of the underwater irradiance measured by the underwater spectroradiometers. For the Barents (N=20) and the Kara Sea, we have got the close values of R 2 equal to 0.72 and 0.73, but the regression equations in these two seas differ significantly. At the same time, the stations with coccolithophore blooms in the Barents Sea fit well into the obtained regression.
In our opinion, the most interesting and practically relevant results obtained in our work are related to the evaluation of the uncertainties in the retrieval of the absorption coefficient of the colored organic matter from satellite data in the real conditions of the Arctic seas. Some of the results should raise questions for specialists from the OLCI team. Why are the values of the absorption coefficient ADG_443_NN obtained as a standard product different from those calculated with the C2RCC processor, although, by the idea, both calculations should result in the same values? Our assessment showed that the apparent reason for such a difference is the unmatching results of atmospheric correction. Our data shows that the results from the C2RCC processor are better consistent with the data from field measurements. If this is the case, then perhaps it is possible to make appropriate changes in the calculation of the standard product?
A lack of appropriate satellite data for our validation studies was an unpleasant surprise for us. Having a large amount of field data, we have not expected that such a problem could arise. Now we see as the main task of our future work, the development of the statistically reliable method for assessment of the uncertainties in the retrieval of the bio-optical parameters from satellite data under the real conditions in the Arctic basin. We realize that such a problem requires the broader coverage of satellite data collecting combined with the field measurements, and it can be solved only by joint efforts of international cooperation. Acknowledgments: In this section you can acknowledge any support given which is not covered by the author contribution or funding sections. This may include administrative and technical support, or donations in kind (e.g., materials used for experiments).

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results".