Ocean Color 1

Shelf-seas and coastal zones provide essential goods and services to humankind, such as fisheries, aquaculture, tourism and climate regulation. The understanding and management of these ecosystems can be enhanced by merging marine ecosystem models and satellite data in “biogeochemical reanalysis”. Such reanalysis uses data assimilation algorithms to correct long-term simulations of marine biogeochemistry, by using decadal time series of ocean colour data, such as chlorophyll concentration. Though applications of long-term reanalysis have been carried out with open ocean models, their effectiveness has not been tested in shelf-sea ecosystems yet.

In this work we performed the first decadal reanalysis of the biogeochemistry of the North West European shelf, with a full evaluation of its skill and value; this reanalysis can be used to describe the variability of the ecosystem and to support marine policy. We assimilated Ocean Colour from ESA’s Climate Change Initiative (ESA CCI_OC), into a state-of-the-art marine ecosystem model of the shelf sea, in a reanalysis spanning the period 1998-2009. The assimilation method is based on the localized Ensemble Kalman filter. By computing the percentiles of the ensemble distribution, we were able to provide the confidence level of the reanalysis estimates. Crucially, we exploited per-pixel error characterization of the ESA CCI_OC data to define objectively the parameters of the assimilation algorithms.

Assimilation of CCI_OC improved the model predictions of the assimilated product in 70% of the North East Atlantic region. The reanalysis showed skill in matching a large dataset of in situ biogeochemical variables, including Essential Climate Variables (ECVs) that cannot be observed from space, such as nutrient concentrations. Spearman rank correlations were significant and higher than 0.7 for physical-chemical variables (temperature, salinity, oxygen), ~0.6 for chlorophyll and nutrients (phosphate, nitrate, silicate), and significant, though lower in value, for partial pressure of dissolved carbon dioxide (ρ=0.4; p<0.01). The reanalysis captured the size of the pH and ammonia data, but not their variability. The potential value of the reanalysis for assessing ecosystem status and variability has been exemplified in two case studies. The first assessed that between 350,000-400,000 km² of shelf bottom areas were oxygen deficient - depending on the confidence level applied in the assessment - potentially threatening bottom fishes and benthos. The second application confirmed that the shelf is a net sink of atmospheric carbon dioxide, but the flux can vary between 36-46 Tg C yr^-1 if a 90% confidence level is applied to the estimates.

This novel reanalysis of the North East Atlantic biogeochemistry, with full error quantification, enhances the understanding and informs the management of the European shelf ecosystem, in relation to eutrophication, fishery, and the air-sea exchange of a crucial greenhouse gas. This work is contributing to the Project “Now Maps” of the Copernicus Marine Environment Monitoring Service.

The chlorophyll-a concentration (Chla) is an Essential Climate Variable, and the study of its variability at global scale requires a succession of satellite ocean colour missions to cover a period adapted to climate research. This study aims at documenting the challenges faced in the construction of a multi-mission Chla data record suitable for climate research and trend detection, as well as practical solutions.

Inter-mission differences and drift in sensor outputs can introduce artefacts affecting trend evaluations as well as our capacity to meet GCOS stability requirements. The impact of these issues on studies of time series is addressed through a sensitivity analysis that compares slopes of linear regression obtained for varying levels of inter-mission bias and drift with respect to a 15-year reference series. The relationship, constructed for a representative set of ocean provinces, between bias and the level of significance associated with the comparison of slopes shows that a bias on the order of ±5-6% generally induces a trend slope that is significantly different from the reference case, while a threshold on bias values not exceeding 2% largely alleviates this effect. Moreover, the study suggests that a drift larger than 2% per decade on the Chla series can result in misleading conclusions from a trend analysis. All results have a clear regional dependence that needs to be taken into account in bias-correction and merging efforts. Low-Chla regions, such as the oligotrophic subtropical gyres, appear particularly sensitive to perturbations and require still higher levels of consistency and stability.

These results serve to set the stage for a trend analysis performed on the available ocean colour Chla time series, either based on a single mission such as SeaWiFS, MERIS and MODIS, or resulting from the merging of these data. Particular attention is given to the data record generated by ESA’s Climate Change Initiative (CCI) project.  The solution adopted by CCI to cope with inter-mission biases is based on the construction of bias maps associated with the remote sensing reflectance spectra expressed on a common set of wavelengths through a band-shifting scheme. The trends obtained from the various products are presented and compared through statistical tests and contingency matrices. The various products show a coherent picture for the main trend patterns, such as diagonal features of positive trends in the South Pacific or negative trends in the North Atlantic over the period 1998-2012. Moreover, over 1998-2007, significant trends of opposite signs are virtually absent if the CCI and SeaWiFS records are compared. However differences in features, amplitudes and significance are seen as well. As far as merged products are concerned, these differences are analysed as a function of the biases existing between single-mission products.

According to remote-sensing and modelling estimates, phytoplankton in the oceans are responsible for the production of some 50 Gigatonnes of carbon globally and per annum, which is approximately half of the net primary production on Earth.  The estimation of phytoplankton biomass (in units of chlorophyll concentration) through remote sensing of ocean-colour methods was first demonstrated with the Coastal Zone Colour Scanner (CZCS), over 3 decades ago,  and has since been improved significantly in the lifetime of successive sensors such as SeaWiFS, MODIS-A and MERIS.  The fields of chlorophyll concentration can be combined with models of photosynthesis that are based on available irradiance, to compute primary production at the global scale.  However, the primary production models require additional data on photosynthesis-irradiance parameters that is not detectable by remote sensing methods and consequently these photophysiological parameters remain difficult to assign on a global scale.  Both culture and field experiments have shown that these parameters are dependent upon environmental variables such as temperature, light and nutrient availability, and also on the phytoplankton community structure.  The ‘Marine Primary Production: Model Parameters from Space’ (MAPPS) project is part of the Support To Science Elements (STSE) initiative and focusses on creating the best estimate of phytoplankton photosynthesis parameters at the global scale, using remote-sensing data.  The foundation of this project has been the compilation of a unique global database of measurements of photosynthesis-irradiance parameters from 36 biogeochemical provinces (Longhurst 2007). In this project, the existing methods for the remote estimation of photosynthesis-irradiance parameters are compared and their performance is analysed. A new methodology, built on a mechanistic understanding of the principal factors governing algal photophysiology, is shown to perform better than the pre-existing methods at the global scale.  The new methodology is then combined with input data from multiple remote-sensing data sources (chlorophyll, sea-surface temperature, photosynthetically available radiation) to derive global fields of the photosynthesis-irradiance parameters.  These fields are then used in a spectrally-resolved primary-production model at the global scale.  

Many coastal seas and inland waters have optically complex (Case-2) water properties which cause ambiguous interpretations of the remote sensing signal and therefore cause high uncertainties of ocean colour (OC) products. In the framework of the ESA-funded “Case-2 Extreme” (SEOM-C2X) project, retrieval methods for ocean colour products in extreme Case-2 waters are tested, implemented, and validated with reference to the capabilities of the Sentinel-3/OLCI and SLSTR and the Envisat/MERIS instruments. Based on hyperspectral radiative transfer simulation, in situ, and Earth Observation data specific features of extreme scattering and absorbing waters are reviewed. A sub-classification scheme of Case-2 waters for various levels of turbidity as function of various optical properties and water constituents is introduced. The important challenges and opportunities for OC retrievals and for atmospheric correction in extreme Case-2 waters are summarized and compared to Case-1 (open ocean) waters. These opportunities include synergy of the Sentinel-3/OLCI and SLSTR sensors. For example, the SLSTR SWIR bands at 1610 and 2250 nm have significant potential for improving the atmospheric correction of extremely turbid waters. Furthermore, the new OLCI bands at 400 and 1020 nm have great potential for improving OC in extreme Case-2 waters. 

Ocean primary production is sensitive to changes in solar energy reaching the sea surface. If the photosynthetic response of phytoplankton is represented by a non-spectral model, it is possible to find an analytical solution to describe the sensitivity. The problem is more complex if a spectral representation is used, but it can be treated using numerical integration. We have analysed both cases using data from the northwest Atlantic Ocean. We find that the relative change in daily, watercolumn production for a given change in surface irradiance depends only on a function of the dimensionless irradiance (irradiance normalized to the photoadaptation parameter).  In the spectral case, it is important to keep separate the diffuse and direct components of sunlight in the radiative transfer calculations to account for the differences in their zenith angular distribution: the two components are added together before the computation of primary production. It is recommended that the Sentinel-3 products include spectrally-resolved daily Photosynthetically Available Radiation (PAR) to facilitate computation of marine primary production at daily to annual, and regional to global scales.