Methods Optical

Coastal environments are intrinsically dynamic and respond to a wide array of natural and anthropogenic drivers across a broad range of time steps.  In addition, coastal environments are under increasing pressure from land use intensification and climate change.  The development of the Australian Geoscience Data Cube has delivered an unprecedented capability to support environmental change monitoring applications through rapid processing and analysis of standardised Earth Observation (EO) time-series data in a High Performance Computing environment. Standardised long-term EO data records provide the capacity to monitor coastal changes processes and understand current changes from a historical perspective. The ability to visualise environmental changes in a spatio-temporal context provides the opportunity to assess whether the change phenomena are rapid / gradual onset, and/or episodic / cyclical in nature.  Understanding the spatio-temporal nature of the changes also enables the attribution of observed changes to the potential causes.

 Hovmöller diagrams, typically used to plot meteorological data, can be applied for visualising large datasets in a meaningful way. In this study, we apply Hovmöller plots to examine coastal change processes and estuarine dynamics, based on a time-series of Landsat based surface reflectance data over a 27-year period (1987-2014), within the Australian Geoscience Data Cube.  The Hovmöller plot in Figure 1 highlights the timing of a sea wall installation and associated land reclamation processes near Fremantle, Western Australia (see attachment).

Three coastal change processes are illustrated in this study: 

The opening, closing and migration of the mouth of the Glenelg River in Victoria; the Hovmöller plots show that the river mouth moves on an episodic basis and remains closed for periods of time.

The installation of a sea wall and subsequent land reclamation near Fremantle in Western Australia; the results illustrate rapid anthropogenic change in the coastal zone and highlight the timing of  the sea wall installation and land reclamation processes.

The migration of coastal dune fields north of Perth in Western Australia; the results show slow coastal change processes through the gradual northward migration of the dune field over multi-decadal time scales. 

The availability of standardised long-term Landsat data, in conjunction with new data becoming available from the Copernicus Sentinel-2 missions, point to the need for cross calibrated multi-sensor data, to enrich the global long-term EO record, in support of the detection and characterisation of coastal change phenomena.

Capability of ocean colour sensor to provide sea surface reflectance over sediment-dominated waters relies on the so-called Bright Pixel Atmospheric Correction (BPAC). The challenge is to decouple marine and atmospheric signal in the near-infrared (NIR), so that the classical atmospheric correction can properly operate as if over clear waters. Inspection of recent satellite data over the Amazon plume, Rio de la Plata river and smaller estuaries shows evidence that growing turbidity is still a limiting factor and generally yields to complete failure of the atmospheric correction.

The present work is a successive attempt to push the limits of the BPAC in the frame of MERIS 4th reprocessing. The core element is a coupled atmosphere-hydrological model defined by a set of free variables (sediment concentration and aerosol) and fixed constants (spectral shape of inherent optical properties). Information is restricted to NIR bands in order to avoid too large uncertainties in the model, due to lower absorption of pure seawater near the visible and larger absorption of non-modelled components; also we discard SWIR bands, not present on most of past and future sensors.

We first demonstrate that inversion is mathematically ambiguous at high turbidity for some band combinations, what explains the classical confusion between aerosol and sediment.  We then show that the problem can be solved unambiguously through a spectral fitting algorithm (χ2 minimization) on five NIR bands, taking into account input uncertainties. Finally we identify particulate absorption as a key modelling parameter over very turbid waters and manage to retrieve it through remote-sensing data in different estuaries of the world. Validation results on sediment backscattering and marine reflectance demonstrate the interest of such approach for MERIS 4th data reprocessing and OLCI, using same NIR bands.

Water quality and river inputs to oceans are major causes of concern for coastal managers. The Water Framework Directive, the Marine Directive Framework Strategy and the Grenelle of the Sea are many tools that set ambitious objectives in terms of water quality, especially concerning large estuaries. To guaranty the achievement of the established commitments, operational monitoring networks are multiplied over such complex, dynamic and fragile environments. Turbidity is a key parameter surveilled, for instance by continuous measurement networks located along riverbanks. This contribution details the performance and relevance of medium to high resolution Earth Observation satellites to complete field measurements and derive consistent time-series at space and time-scales adapted to water quality management in estuarine environments.

Within the framework of three independent projects (RIVERCOLOR /CNES ; HYMOSED / Seine Aval ; Take5 / CNES), optical oceanographic instruments have been deployed in order to develop and test various inversion algorithms. For validation purposes at a large scale, long-term turbidity records in the Seine, Loire and Gironde estuaries have been used. The tests performed show three worthwhile prospects regarding the potential of using space-derive observations in estuary monitoring and management programs.

First, we demonstrated that high resolution satellite data allow retrieving turbidity with a maximum accuracy of about 10%. Accuracy decreases down to about 50% for medium resolution. However, degraded accuracy still allows to appreciate turbidity gradient, and, in particular, maximum turbidity zones (MTZ) that have dominating impacts on sediment budget, biogeochemical processes and biological quality of the environment (oxygenation). Higher quality products are of interest to quantify sediment budgets and transports.

Second, although local to regional algorithms allow achieving the best performances in terms of SPM retrieval, we also demonstrated that most recent global algorithms can be selected for monitoring purposes. As a first step, the derived SPM products appear reasonably adequate for addressing the major upstream-downstream physicochemical gradients occurring in the estuaries.

Third, image time series derived from MODIS, LANDSAT and SPOT (Take5 simulation of Sentinel-2) databases were processed with the RIVERCOLOR processing chain to get frequently refreshed SPM map. The map dataset obtained in the Seine and the Gironde is currently used to control 3D hydro-sedimentary simulations. It is shown in the Gironde that, based on 12 years of MODIS-derived SPM maps (about 1200 figure cases), modeled SPM distributions exhibit systematical discrepancies from observed ones when considering specific tide conditions. This may help improving parameterization of the processes involved in the sediment vertical transport in the water column. Coupling high to medium resolution data allow improving knowledge for incoming sediment management plans. In particular archives seem appropriate to analyze inter-annual changes and main trends in sediment distribution and concentration patterns and gradients. Finally, high resolution maps of the turbidity are vital from a socio-economic point of view for instance to help managing dredging activities and also to help managing water quality and resources during low water level periods.

Consequently, the use of space imagery constitutes a major issue to improve the spatiotemporal dynamics knowledge of estuarine SPM. In addition, this study prefigures the interest of integrating Sentinel-2 data for inland water monitoring to complete Sentinel-3 observations useful in the inlets and plumes of estuaries. Finally it demonstrates the emergence of a Copernicus user service.

Micronekton organisms are both the prey of large ocean predators, and themselves also predators of eggs and larvae of many species from which most fishes. The micronekton biomass concentration is therefore a key explanatory variable that is usually missing in ecosystem models to understand individual behaviour and population dynamics of large oceanic predatorsthat are either targeted by fisheries (tuna, swordfish, marlin, etc.) or strictly controlled in by-catch (bluefin tuna, sharks), or fully protected (marine turtles, seabirds, marine mammals).

In that context, the objective of the OSMOSIS (Ocean ecoSystem Modelling based on Observations from Satellite and In-Situ data) project was to demonstrate the feasability and to prototype an integrated system going from the synergetic use of many different variables (Sea level, Earth Marine geoid, Sea Surface Salinity, Sea Surface Temperature, Ocean color) measured from space to the modeling of the distribution of micronekton organisms. This objective has been successfully reached, as demonstrated in this paper:

First, we present how data from CRYOSAT, GOCE, SMOS, ENVISAT, together with other non-ESA satellites and in-situ data, can be merged to provide the required key variables needed as input of the micronekton model: the 3D description of the ocean state (Temperature, Salinity, currents), the primary production and the euphotic depth.

Then we validate the micronekton density maps obtained in the Southern Indian and Pacific Ocean by forcing the model with the OSMOSIS input fields and show that the accuracy of the results is comparable, and locally even superior, to the values obtained using a full 3D numerical ocean circulation model with data assimilation from the Mercator-Ocean system.

This opens the door to a wide range of applications, from the simulation of the feeding and spawning habitats, to the animals movements based on the gradients of these habitats and ultimately the dynamics of the entire population from the larvae up to the oldest individuals.

To gain knowledge on the role of marine phytoplankton in the global marine ecosystem and biogeochemical cycles, information on the global distribution of major phytoplankton functional types (PFT) is essential. The Synergistic Retrieval of Phytoplankton Functional Types from Space from Hyper- and Multispectral Measurements project (SynSenPFT) aims to improve the retrieval of PFTs from space by exploring the synergistic use of low-spatial-hyper-spectral and high-spatial-multi-spectral satellite data. Three PFTs are investigated: diatoms, coccolithophores and cyanobacteria. The work involves the improvement/revision of existing PFT algorithms based on hyper- (PhytoDOAS, Bracher et al. 2009) and multi-spectral (OC-PFT, Hirata et al. 2011) datasets, development of synergistic PFT products combining the retrievals of these two algorithms and intercomparison of the synergistic PFT products with those derived from other methods (Ciotti and Bricaud 2006, Brewin et al. 2015). Further activities include the validation of the satellite PFT products against in situ data and a sensitivity study using radiative transfer calculations to determine the band-set requirements of multispectral instruments (i.e. Ocean and Land Colour Instrument - OLCI) to retrieve PFTs. The project started in December 2014 and the first results on the synergistic PFT products, validation and sensitivity studies are presented.