This study was designed to examine the spatial and temporal capabilities of ESA’s OLCI on the Sentinel-3A platform to monitor the water quality parameters: chlorophyll-a, turbidity, and CDOM in the Baltic Sea when compared to more traditional monitoring techniques such as monitoring by ship or mooring. The measurement frequency of OLCI/Sentinel-3A data is also compared to the frequency of MERIS/ENVISAT data (from 2008 and 2010).
OLCI S3A full resolution level 2 data from 2017 and 2018 of the NW Baltic proper’s coastal region was processed to remove pixels of low data quality using predetermined flags. The number of valid scenes per year was determined and the level 2 products: chlorophyll-a (chl_NN), turbidity (derived from TSM_NN), and CDOM (using ADG_443_NN as a proxy) were calculated by averaging the values of 9 pixels surrounding, and including, a central sampling site. These measurements were then compared to paired in situ measurements, taken on the same day +/- 3 hours of satellite overpass, to examine the correspondence and variability between measurements. Measurements from a WetLabs WQM mooring were also examined and compared to in situ measurements. Pearson’s correlation was used to determine covariance between OLCI S3A’s measurements and in situ measurements and mean normalized bias, root mean square error, and mean absolute percentage difference were used to evaluate the performance of OLCI S3A and the optical mooring compared to in situ measurements.
OLCI S3A produced a higher number of valid observations per month than its predecessor MERIS for both stations: B1 and BY31. It also produced more valid observations per month at stations B1 and BY31 than ship-based monitoring teams. OLCI S3A’s current method of processing underestimated chlorophyll-a concentrations (MNB = -7%, RMSE = 40%, APD = 49%, r = 0.48, p < .00001, N = 156) especially if chlorophyll-a concentrations were measured during peak production periods. The optical mooring showed a much higher correlation but more relative error and bias (MNB = -39%, RMSE = 43%, APD = 39%, r = 0.94, p < 0.00001, N = 12) also underestimating chlorophyll-a concentrations. OLCI S3A’s current ocean color processing method drastically overestimates turbidity (MNB = 189%, RMSE = 1011%, APD = 214%, r = 0.55, p = .000097, N = 45) whereas the optical mooring showed good agreement with in situ measurements and less variability (MNB = 21%, RMSE = 26%, APD = 21%, r = 0.69, p = 0.0132, N = 17). Lastly, OLCI S3A was strongly correlated to in situ CDOM values (MNB = -5%, RMSE = 37%, APD = 51%, r = 0.82, p < 0.00001, N = 36).
Overall, OLCI shows improved retrieval of chl-a at values below 10 mgl-1 as well as improved CDOM retrieval than MERIS (underestimation of about 40% vs. about 60-75%). Turbidity is highly overestimated, but can be corrected either using in situ data for calibration, or by applying the regional TSM-specific scatter before converting to turbidity.