Optical remote sensing

Definition of Passive optical remote sensing: Passive optical sensors detect natural energy (radiation) that is emitted or reflected by the object or scene being observed. Reflected sunlight is the source of radiation measured by passive optical sensors. [1]. This is the common definition for Passive optical remote sensing, other definitions can be discussed in the article

Remote sensing of coastal zones using satellite or airborne sensors became a powerfull monitoring instrument, because large scale and synoptically environmental information are essential for the understanding and management of the system. Optical multi- or hyperspectral sensors enable the detection of in-water properties such as suspended matter, phytoplankton concentrations, benthic surface and vegetation properties and bathymetry in optically shallow water areas (see “Applications”).

See also

...

Principles

Optical remote sensing using passive satellite or airborne sensors is the spatial resolved detection and utilization of sunlight, which has been transmitted trough the atmosphere and which is reflected from the earth surface or the water body backward to the sensor. The sunlight spectrum is modified on its way from the sun though the atmosphere, the sea surface and the water body. Matter in atmosphere, water and at the boundary layers is absorbing, scattering and reflecting the light in a very specific way and in dependency on wavelength. As result, the light is carrying spectral information about the composition of matter.

Sensors

Multi- or hyperspectral satellite- or airborne sensors are detecting this spectral information with a distinct accuracy in terms of temporal, spatial, spectral and radiometric resolution that is different and characteristic for each sensor.

Data processing

Different remote sensing data processing methods are evaluating the spectral information in order to retrieve the content information with regard to space and time.

Physics based retrieval algorighms

Physics based retrieval algorithms can be applied generally world wide and very flexible under all those conditions covered by the implemented optical models. Once set up, they do not rely on manual adaptations in order to generate products and frequently are completely independent on inputs from ground truth measurements.

Empirical algorithms

Nevertheless, the integration of the sometimes very multi-layered complex natural conditions in the physics based models is not always useful, although the remote sensing imageries clearly reflect effects of demanding properties such as species composition. This frequently is of importance to exploit also the potential of sensors, which are strictly speaking not perfect suitable for the independent detection of a environmental property. Here, also empirical algorithms are powerfull to exploit remote sensing data. But, empirical approaches usually rely on accompanied ground truth measurements and are typically not transferable to different types of aquatic systems.

Generic processing systems

Generic processing systems cover a wide range of applications and frequently can be applied to new sites, applications and sensors due to a systematic modular approach and easy adaptations for sensor and site specific properties.

Geo-rectification

Geo-rectification procedures for imaging data are applied in order to connect the spatial resolved pixel values with geographical coordinates and to deliver geo-coded maps, which can be implemented into Geographical Information Systems GIS. The trigonometric principles and procedures for operational geo-referencing are essentially developed and operational so far. Therefore, sub-pixel accuracies can be achieved for the condition that sufficient navigation data are available. However, in practice the attainable spatial accuracy of operational geo-coded products depends on the accuracy of the satellites or airborne navigation or metadata. Therefore, for high-precision tasks many satellite or airborne imageries have to be spatially refined using manual geo-rectification approaches or automatic matching algorithms.

Resctrictions

Restrictions apply, wherever the object specific signals are masked by others or interfered in ambiguous way with regard to both the sensor resolution and the retrieval methods. Therefore, only few restrictions can be stated in general, but many restrictions can be determined only with regard to the specific site to be observed, the sensors used and algorithms applied.

Examples for general restrictions

General restrictions are the masking of clouds for optical signals reflected from the earth surface or geometric recording conditions that effect strong sun glitter contributions to the signal from the water surface.

Examples of case dependent restrictions

a) Intermediate sun glitter conditions can be treated with only if the radiometric, spectral (and in dependency on the algorithm also spatial) resolution of the sensor is sufficient and the processing approach supports such a correction or consideration of this effect (> example).

b) The spatial resolution of the sensor in most cases must better than the spatial heterogeneity of the target areas to be mapped. Subpixel classification of surfaces is only possible for a very restricted number of spectral endmembers. At least, every contribution of members not implemented into the model will create errors.

c) The spectral behaviour of optical classes must significant differ to each other with respect to the spectral resolution of the sensor. E.g. Chlorophyll and Gelbstoff can not be estimated independently, if the sensor spectral and radiometrical resolution is unsufficient and/or the aquatic system specific absorption spectra are to simular for an independend quantitative estimation.