Our coastal ecosystems are extremely vulnerable and encounter pressures from natural hazards, climate change and coastal development leading to e.g. habitat loss, coastal erosion and pollution. Remote sensing can contribute to a sustainable coastal management and development by providing area wide information on i) the status and evolution of the coastal vegetation, coral reefs and ii) near real time and historic information on the water quality. These remotely sensed data will significantly contribute to our knowledge of the coastal zone and will complement more traditional field measurements and modelling efforts. In particular the recently launched sensors such as Sentinel-2 and Landsat-8, with a spatial resolution of 10 and 30 m respectively, allow for a factor 10 improvement in spatial detail. This means that remote sensing can now be used to monitor small scale coastal features such as harbors, ports, coral reefs and smaller patches of vegetation (e.g. mangroves) in a recurrent way (thanks to the high revisiting time of those satellites) and at no cost (thanks to the open data policy). New and high end image processing tools are being developed to process these new remote sensing data towards usable end products. These include tools for the detection of clouds, the removal of atmospheric absorption and scattering, the classification of the image in a number of pre-defined classes (e.g. vegetation species, coral associations,…) and the retrieval of bio-geophysical parameters (e.g. water quality parameter concentrations such as chlorophyll-a, suspended sediments; biomass estimation of aquatic and terrestrial vegetation; evaluation of structural indicators of vegetation (e.g. Leaf Area Index)). Here we present a complete end-to-end chain for the processing of Sentinel-2 and Landsat-8 data. This chain consists of all the necessary tools to convert raw Sentinel-2 and Landsat-8 imagery into ready-to-use maps and deliverables for end users. In detail, this comprises state-of-the-art tools and algorithms to i) automatically download and preprocess the imagery (including cloud detection and atmospheric correction) and ii) automatically convert these processed imagery into end products. Regarding the preprocessing tools, first the cloud detection and atmospheric correction module, OPERA, will be explained. The cloud detection algorithm builts further on the Automated Cloud-Cover Assessment (ACCA) algorithm(Irish, 2000). This algorithm uses a set of thresholds on different bands and band ratios. Here, the orginal ACCA is extended including tresholds on the ‘Coastal Aerosol’ and the ‘Cirrus’ band and provides improved cloud estimates particularly over water. The second preprocessing module OPERA is a scene-generic atmospheric correction, which means it can correct for atmospheric effects in scenes with land, coastal water and inland waters. OPERA is based on the atmospheric radiative transfer model Modtran5. It calculates the absorption and scattering in the atmosphere, the scattering effects at the air-water interface and neighbouring effects. Inputs for Modtran5, such as the aerosol optical thinkness, are estimated from the image itself. Applying OPERA on an uncorrected Sentinel-2 and Landsat-8 image removes all the unwanted effects and derives the ground reflectance for each pixel in the image. After atmospheric correction the resulting surface reflectances are further processed into end products providing information on the different aspects of the coastal ecosystem: water quality, coastal vegetation and coral reefs. To obtain information on water quality, the water pixels are processed to derive the concentrations of suspended sediment and chlorophyll-a. The suspended sediment can be retrieved because there is a clear relationship between its concentration and the reflectance of the pixel. In general, an increase in suspended sediment will result in an increase in water reflectance. This increase is particularly evident in the red, near infrared and short wave infrared spectral bands. The exact relationship is however different for each spectral band. For shorter wavelengths, like for instance the red band, the relationship will be linear for low concentrations (up to 50 mg L− 1), and then starts to saturate. This saturation will occur at higher concentrations for longer wavelength. This has led to the development of multiwavelength switching algorithms, where shorter wavelengths are used for low concentrations and longer wavelengths are used for higher concentrations. Short wave infrared bands can be used for the most extreme cases, where total suspended sediment reaches more than 500 mg L− 1. It should be noted that the suspended sediment concentrations refer to surface concentrations because of the limited penentration depth. Chlorophyll-a can be determined based on its specifc absorption characteristics in the red and NIR. The concentration of Chlorophyll-a can be accurately retrieved from airborne and satellite data in marine, lacustrine and transitional waters using site specific NIR-red-based reflectance ratios. The technique is based on a statistical relationship between the water quality variable of interest and the remote sensing signals. This approach is easily implemented with existing in situ data, and often produce reliable results for the areas and datasets from which they are derived. The concentration of Chlorophyll-a can then be used as an indicator of trophic status and to detect (harmful) algae blooms in coastal waters. Land pixels can be further classified according their land cover/land use. This allows to differentiate between several classes such as urban areas, bare soil, cropland and different vegetation types. Once classified into general broad classes, we further zoom in on those classes which are of particular interest for a substainable coastal management. One of those classes are for instance mangrove forests which are hugely valuable, both ecologically and economically, but are also highly threatened by clearing and rising sea levels. Here we present tools and methods to look at the mangrove extent, their species composition and the temporal changes in both. To classify the extent of mangroves at each time-step, all Landsat and Sentinel-2 channels are considered. A combination of an object-based approach to classify mangroves is applied. Using detailed knowledge on the biophysical properties of mangroves and changes over time, obtained from remote sensing, the histories of human-induced events and processes can be quantified.


Article metrics loading...

Loading full text...

Full text loading...

This is a required field
Please enter a valid email address
Approval was a Success
Invalid data
An Error Occurred
Approval was partially successful, following selected items could not be processed due to error