SPIE Digital Library Get updates from SPIE Newsroom
  • Newsroom Home
  • Astronomy
  • Biomedical Optics & Medical Imaging
  • Defense & Security
  • Electronic Imaging & Signal Processing
  • Illumination & Displays
  • Lasers & Sources
  • Micro/Nano Lithography
  • Nanotechnology
  • Optical Design & Engineering
  • Optoelectronics & Communications
  • Remote Sensing
  • Sensing & Measurement
  • Solar & Alternative Energy
  • Sign up for Newsroom E-Alerts
  • Information for:
SPIE Defense + Commercial Sensing 2017 | Register Today

OPIE 2017

OPIC 2017




Print PageEmail PageView PDF

Remote Sensing

Developing an airborne hyperspectral system to detect coastal and submarine springs

Optical characteristics of seawater are measured to identify freshwater inputs, enabling the location of valuable springs.
15 August 2006, SPIE Newsroom. DOI: 10.1117/2.1200607.0322

The water shortage in several regions of our planet has led to intensive research regarding the detection of underground water, such as discharges into the sea under some types of geological beds and other structures. Many of these springs, both submarine and coastal, are thought to discharge into Greece's seas.1 Their detection is of great importance, especially if they are found on or near islands where lack of water is a serious problem. Moreover, freshwater and coastal seawater interactions have received considerable attention during the past decade as the importance of coastal-zone inputs—such as those of nutrients and pollutants—has become increasingly recognized.

Submarine groundwater discharges (SGD) have only been studied in the last few years, as their importance as pathways of fresh water, nutrients, and other inorganic and organic substances into the sea has gained recognition. As a result, hyperspectral sensors have been designed for water applications. These include detectors of spectrally fine structures—materials like phytoplankton, sediment, and dissolved organic matter that are discernible by subtle spectral differences—in coastal and inland waters.

The research program Competitiveness is being undertaken by the Laboratory of Remote Sensing of the National Technical University of Athens, the Hellenic Center for Marine Research, ANCO S.A., and the Laboratory of Hydrology and Hydrogeology of the National Technical University of Athens. Together they are developing an operational system to detect coastal and submarine springs based on airborne hyperspectral data.2 For the project, a Cessna aircraft has been outfitted with a hyperspectral sensor (CASI II, 0.40–0.97μm, 244 bands), a thermal sensor (TABI, TIR 8–12μm), an inertial measurement unit (C-MIGITS III), an inertial base, a pre-processing computer, and flight control software.

Data provided by the system is intended to highlight changes in the optical seawater characteristics that are caused by the influence of freshwater inputs from coastal springs and SGD. These changes include differences in temperature, salinity, chlorophyll-a, and particulate matter. Thus, in-situ data regarding the spatial and temporal distribution of the aforementioned parameters was required for band selection during experiments with the airborne hyperspectral system, and for identification of the most important parameters to be studied through airborne hyperspectral data.

The study area, Agios Stefanos Bay, is situated at the northern sector of south Evvoikos Gulf in central Greece. It is an area well-known for strong tidal currents and freshwater discharges. Oceanographic measurements were conducted there during April 2005, June 2005, and February 2006 using standard field equipment (a portable SeaBird CTD with transmissometer and optical backscatter sensors, and a GER 1500 radiometer). Water samples were taken to assess chlorophyll-a and particulate-matter concentration (PMC), radiance and reflectance were measured, and the salinity of the uppermost centimeter of seawater was determined.

The data collected in-situ revealed that the inflow of surficial and submarine freshwater discharges substantially affected the horizontal distribution of the parameters measured. At inflow locations, temperature was lower, salinity was lower, PMC was higher, and chlorophyll-a concentration was higher.3 Freshwater flowing along the coast from small sub-aerial springs was distributed over the area as a thin surficial film (∼1cm), decreasing as the distance from the sources increased. By contrast, SGD appeared as permanent circular surface currents (3–5m diameter), that were visible clearly to the naked eye. They were marked by increased turbulence and rapid mixing, resulting in high PMC. Processing of ground radiometer data revealed 531–554μm, 667–690μm, and 732–751μm as the most appropriate wavelength intervals for selecting hyperspectral bands.

Based on the in-situ investigation, hyperspectral (96 bands, 3m resolution) and thermal (1m resolution) images were captured in June 2005 and February 2006 (see Figure 1).

Figure 1. Thermal infrared images (top) and temperature diagrams (bottom) are based on in-situ measurements for (a) June 2005 and (b) February 2006.

The hyperspectral images cover a broad spectrum range forming a stack of images in several different spectral channels known as the hyperspectral cube. In processing the images, we aim to identify which hyperspectral bands correspond to the various changes in optical seawater characteristics, representing salinity, turbidity, chlorophyll-a, and particulate matter changes. This selection process, and the resultant shrinkage of the hyperspectral cube, is of major importance for hyperspectral imagery processing.

Several different methods have been applied to minimize the dimensions of the hyperspectral cube, including principal component analysis (PCA), independent component analysis (ICA), and minimum noise fraction (MNF). However, each of these methods presents a major drawback, such as the production of statistically correlated imagery (PCA), the severe computational time loss when working with large hyperspectral cubes (ICA), and the excessive shrinkage of the hyperspectral cube causing loss of useful information (MNF).

Thus, a new method is under development, which combines wavelet analysis, ICA, and the Harsanyi-Farrand-Chang de-noising method, which compares corrleation and covariance eigenvalues to assess noise. The methodology will use the kurtosis criterion to discriminate between hyperspectral bands with Gaussian and non-Gaussian properties. Those with non-Gaussian properties will be subject to wavelet analysis and the Harsanyi-Farrand-Chang de-noising method.

Final results and evaluation of the system are expected at the end of 2006.

Vassilia Karathanassi
Laboratory of Remote Sensing, School of Rural & Surveying Engineering, National Technical University of Athens
Athens, Greece 
Since 1998, Vassilia Karathanassi has been a faculty member in the remote sensing laboratory at the National Technical University of Athens, first as a lecturer and currently as an assistant professor. She has authored many articles in international journals and various conference proceedings. Previously, she worked in the commercial sector as a researcher in image processing.