Basic Principles

Sunlight is a spectrum of radiation from infrared, through visible, to ultraviolet. It is filtered by the atmosphere and ocean, some wavelengths being absorbed and scattered, and some reflected back through the ocean and atmosphere to the satellite sensor. The sensor is multispectral, it measures reflected sunlight at specific wavelengths, or "bands", typically RGB, which are combined to create a "true colour" image, but also at wavelengths that can be processed to provide a range of Earth Observation data: such as the amount of sediment or plankton chlorophyll in the water, or on land the amount and health of vegetation or classification of land use.

For satellite-derived bathymetry, the essential principle is that there is a relationship between the amount & type of reflected energy detected by the sensor and the depth of water. This relationship is limited to 20-30 metres, the depth of light penetration in water, needs to consider the bottom type, since sand reflects more than, for instance, seagrass, and can be impossible in areas with high sediment load or algal blooms. But, if you can see the bottom from space it is possible to calculate the sea depth.

Algorithms | How it Works

Satellite-based bathymetry uses a processing chain comprising distinct algorithms:

• spatial registration
• surface & cloud masking
• de-glint
• atmospheric correction
• radiometric normalisation
• depth invariant indices
• benthic classification
• physics based model inversion
• satellite derived bathymetry
• nautical charting

Spatial Registration

Using multiple images and/or images from differing sensors requires that image sequences are spatially aligned.

Land, cloud & white cap mask

For many of the processing steps it is advantageous to exclude pixels of land, clouds, cloud shadows and breaking waves (white caps).

Sun Glint Removal

De-glint is a pre-processing algorithm aimed at removing the direct reflection of the sun from the air-water interface or ‘sun glint’.

Atmospheric Correction

Subsequent to glint removal the next step it atmospheric correction i.e. the process of removing the effects of the atmosphere on the reflectance values of the image taken by the satellite sensor. Gases and aerosols in the atmosphere absorb and scatter radiation both when solar radiance is passing through the atmosphere and reflectance passing back out. The Sentinel-2 Sen2Cor SNAP module configured for maritime aerosols was used.

Atmospheric correction is difficult because 80-90% of the sensor signal in the blue-green is contributed by the atmosphere and the atmospheric path signal varies significantly. Over coastal zone images it is particularly problematic as a common assumption, that the open ocean is fully absorbent in NIR is invalid in turbid water and, in addition, coastal zone aerosols have contributions from both continental and maritime sources.

It is true to state that improvements in coastal zone atmospheric correction models would provide the most significant reduction of uncertainty in satellite derived bathymetry.

Depth Invariant Indices

Calculation of depth invariant indices is a water column correction step that is frequently a useful pre-processing step for benthic classification. Attenuation of the reflectance is approximately inverse exponential with water depth, therefore the transform, \[X_{i} = \ln\left(R_{i}{-R_{i}^{deep}}\right)\]approximately linearises the effect of depth on reflectance. \(R_{i}\) is the pixel reflectance in band \(i\), and \(R_{i}^{deep}\), is the deep water reflectance in that band.

Habitat Mapping

Habitat classification is a complementary output available from satellite data bathymetric surveys. Due to the submerged nature of the marine habitats, classification requires pre-processing techniques, common to satellite derived bathymetry, for the removal of surface glint and the effects of variable depth e.g. water attenuation effects.

Benthic habitat classes used for high spatial resolution bottom type classification.

Physics Based Inversion

The main recent development in satellite derived bathymetry algorithms is that of physics-based model inversion approaches. These methods work by establishing a forward model for the spectral water-leaving remote sensing reflectance, \(R_{rs}(\lambda)\), based on the depth, optical properties of the water and bottom type. The majority of methods are based on a set of equations derived by Lee et al (1998; 1999) which can be represented by a function, \[R_{rs}(\lambda) \approx f\left({P,G,X,H,R(\lambda),\lambda}\right)\]Where \(P\),\(G\),\(X\) represent the amount of phytoplankton, dissolved organic matter (CDOM) and particulate backscatter, \(H\) is the depth and \(R(\lambda)\) is the bottom type reflectance – which can be taken from a spectral library or linear mixtures thereof..

IHO Nautical Charting

The ultimate objective in satellite derived bathymetric surveys is that the depth estimates, sub-surface features and, importantly, the error estimates, are sufficiently accurate and acceptable for maritime navigation and are used to create nautical charts that conform to the strict standards of the International Hydrological Organisation (IHO).

Do you need to use costly commercial VHR satellite images?

Actually, not really necessary. In comparison testing with two solutions: "good-enough" using 10m Sentinel-2 images and "best-achievable" using 2m VHR images it was discovered that the "good-enough" method yielded comparable results to the "best-achievable" method with a cost saving of €20 per square kilometre. Now we concentrate on using Sentinel-2 images and only resort to VHR for making comparative studies.

What happens if my survey site is usually cloudy or the water turbid?

One of the significant innovations that we have introduced is instead of using single images we use a stack of images. Previous research projects required a perfect cloudless image that would be the best condition possible to produce SDB. ARGANS takes a series of SDB models and uses a statistical approach, based on the exceptional signal / noise ratio of Sentinel-2 images to combine these into one result and therefore producing the ‘perfect model’ from a large database. By using multiple satellite images, errors within the image can be removed. This includes sediment plumes, effects of turbid water and human artefacts such as boats.