The Naval Research Laboratory (NRL) has been developing the Coastal Water Spectral Toolkit (CWST) to estimate
water depth, bottom type and water column constituents such as chlorophyll, suspended sediments and chromophoric
dissolved organic matter from hyperspectral imagery. The CWST uses a look-up table approach, comparing remote
sensing reflectance spectra observed in an image to a database of modeled spectra for pre-determined water column
constituents, depth and bottom type. Recently the CWST was modified to process multi-spectral WorldView-2 imagery.
Generally imagery processed through the CWST has been collected under optimal sun and viewing conditions so as to
minimize surface effects such as specular reflection. As such, in our standard atmospheric correction process we do not
include a specular reflection correction. In June 2010 a series of 7 WorldView-2 images was collected within 2 minutes
over Moreton Bay, Australia. The images clearly contain varying amounts of surface specular reflection. Each of the 7
images was processed through the CWST using identical processing to evaluate the impact of ignoring specular
reflection on coverage and consistency of bottom types retrieved.
The Hyperspectral Imager for the Coastal Ocean (HICO) is a hyperspectral sensor which was launched to the
International Space Station in September 2009. The Naval Research Laboratory (NRL) has been developing the Coastal
Water Signatures Toolkit (CWST) to estimate water depth, bottom type and water column constituents such as
chlorophyll, suspended sediments and chromophoric dissolved organic matter from hyperspectral imagery. The CWST
uses a look-up table approach, comparing remote sensing reflectance spectra observed in an image to a database of
modeled spectra for pre-determined water column constituents, depth and bottom type. In order to successfully use this
approach, the remote sensing reflectances must be accurate which implies accurately correcting for the atmospheric
contribution to the HICO top of the atmosphere radiances. One tool the NRL is using to atmospherically correct
HICO imagery is Correction of Coastal Ocean Atmospheres (COCOA), which is based on Tafkaa 6S. One of the user
input parameters to COCOA is aerosol optical depth or aerosol visibility, which can vary rapidly over short distances in
coastal waters. Changes to the aerosol thickness results in changes to the magnitude of the remote sensing reflectances.
As such, the CWST retrievals for water constituents, depth and bottom type can be expected to vary in like fashion. This
work is an illustration of the variability in CWST retrievals due to inaccurate aerosol thickness estimation during
atmospheric correction of HICO images.
In this paper, we investigate the use of nonlinear structure to derive the physical characteristics of coastal data. In particular, we show how the physics of shallow water coastal regions lead to well defined nonlinear structures (manifolds) in the corresponding hyperspectral data. The exact form of this structure is determined by both the Inherent Optical Properties of the water column as well as the boundary conditions (bottom reflectance, depth). This structure is then used to develop efficient algorithms for searching large 'lookup tables' of precalculated spectra with known physical characteristics, which are used for estimating the various physical parameters (bathymetry, bottom type, etc.) of the scene.
We assess our methods with data collected by the NRL PHILLS sensor at the Indian River Lagoon (IRL) in Florida. The IRL is a well-studied and characterized body of water that contains a number of different water and bottom types at various shallow (generally less than 8 meters, except in the shipping channel where depths can be as much as 18 m) depths. We show in particular that the search algorithm is able to produce valid results in a short amount of time, and compare our results with an IRL LIDAR bathymetry survey from early 2004.
This paper demonstrates the characterization of the water properties, bathymetry, and bottom type of the Indian River Lagoon (IRL) on the eastern coast of Florida using hyperspectral imagery. Images of this region were collected from an aircraft in July 2004 using the Portable Hyperspectral Imager for Low Light Spectroscopy (PHILLS). PHILLS is a Visible Near InfraRed (VNIR) spectrometer that was operated at an altitude of 3000 m providing 4 m resolution with 128 bands from 400 to 1000 nm. The IRL is a well studied water body that receives fresh water drainage from the Florida Everglades and also tidal driven flushing of ocean water through several outlets in the barrier islands. Ground truth measurements of the bathymetry of IRL were acquired from recent sonar and LIDAR bathymetry maps as well as water quality studies concurrent to the hyperspectral data collections. From these measurements, bottom types are known to include sea grass, various algae, and a gray mud with water depths less than 6 m over most of the lagoon. Suspended sediments are significant (~35 mg/m3) with chlorophyll levels less than 10 mg/m3 while the absorption due to Colored Dissolved Organic Matter (CDOM) is less than 1 m-1 at 440 nm. Hyperspectral data were atmospherically corrected using an NRL software package called Tafkaa and then subjected to a Look-Up Table (LUT) approach which matches hyperspectral data to calculated spectra with known values for bathymetry, suspended sediments, chlorophyll, CDOM, and bottom type.
A useful technique in hyperspectral data analysis is dimensionality reduction, which replaces the original high dimensional data with low dimensional representations. Usually this is done with linear techniques such as linear mixing or principal components (PCA). While often useful, there is no a priori reason for believing that the data is actually linear.
Lately there has been renewed interest in modeling high dimensional data using nonlinear techniques such as manifold learning (ML). In ML, the data is assumed to lie on a low dimensional, possibly curved surface (or manifold). The goal is to discover this manifold and therefore find the best low dimensional representation of the data.
Recently, researchers at the Naval Research Lab have begun to model hyperspectral data using ML. We continue this work by applying ML techniques to hyperspectral ocean water data. We focus on water since there are underlying physical reasons for believing that the data lies on a certain type of nonlinear manifold. In particular, ocean data is influenced by three factors: the water parameters, the bottom type, and the depth. For fixed water and bottom types, the spectra that arise by varying the depth will lie on a nonlinear, one dimensional manifold (i.e. a curve). Generally, water scenes will contain a number of different water and bottom types, each combination of which leads to a distinct curve. In this way, the scene may be modeled as a union of one dimensional curves. In this paper, we investigate the use of manifold learning techniques to separate the various curves, thus partitioning the scene into homogeneous areas. We also discuss ways in which these techniques may be able to derive various scene characteristics such as bathymetry.
This study focuses on Coastal land cover classification from airborne hyperspectral at two sites. Our primary study area, is a chain of barrier islands, collectively known as the Virginia Coast Reserve (VCR); the second site is located in and around Barnegat Bay, NJ. At the Barnegat Bay site, hyperspectral imagery was acquired by PHILLS during a two week campaign in late July and early August. The present work examines land-cover models for PHILLS imagery subsets acquired on August 2, 2001. At the VCR site, we
have acquired an extensive time-series of PROBE2 imagery over six of the barrier islands, as well as one HyMAP scene. Multi-season models
have been developed that take advantage of seasonal differences in land-cover to improve classification accuracy. Automatic classification experiments consider roughly 20-25 categories
of land-cover at the two different sites. Categories include a variety of wetland plant species (brackish and freshwater), beach, dune, and upland plant species and plant communities. We also examine in detail detectability and accuracy of mapping invasive plant species such as Phragmites australis, which pose a particular challenge to natural resource managers.
We present an overview of the Naval EarthMap Observer (NEMO) spacecraft and then focus on the processing of NEMO data both on-board the spacecraft and on the ground. The NEMO spacecraft provides for Joint Naval needs and demonstrates the use of hyperspectral imagery for the characterization of the littoral environment and for littoral ocean model development. NEMO is being funded jointly by the U.S. government and commercial partners. The Coastal Ocean Imaging Spectrometer (COIS) is the primary instrument on the NEMO and covers the spectral range from 400 to 2500 nm at 10-nm resolution with either 30 or 60 m work GSD. The hyperspectral data is processed on-board the NEMO using NRL's Optical Real-time Automated Spectral Identification System (ORASIS) algorithm that provides for real time analysis, feature extraction and greater than 10:1 data compression. The high compression factor allows for ground coverage of greater than 106 km2/day. Calibration of the sensor is done with a combination of moon imaging, using an onboard light source and vicarious calibration using a number of earth sites being monitored for that purpose. The data will be atmospherically corrected using ATREM. Algorithms will also be available to determine water clarity, bathymetry and bottom type.