W.J. Rhea

Optical scattering and backscattering by organic and inorganic particulates in U.S. coastal waters

We present the results of a study of optical scattering and backscattering of particulates for three coastal sites that represent a wide range of optical properties that are found in U.S. near-shore waters. The 6000 scattering and backscattering spectra collected for this study can be well approximated by a power-law function of wavelength. The power-law exponent for particulate scattering changes dramatically from site to site (and within each site) compared with particulate backscattering where all the spectra, except possibly the very clearest waters, cluster around a single wavelength power-law exponent of -0.94. The particulate backscattering-to-scattering ratio (the backscattering ratio) displays a wide range in wavelength dependence. This result is not consistent with scattering models that describe the bulk composition of water as a uniform mix of homogeneous spherical particles with a Junge-like power-law distribution over all particle sizes. Simultaneous particulate organic matter (POM) and particulate inorganic matter (PIM) measurements are available for some of our optical measurements, and site-averaged POM and PIM mass-specific cross sections for scattering and backscattering can be derived. Cross sections for organic and inorganic material differ at each site, and the relative contribution of organic and inorganic material to scattering and backscattering depends differently at each site on the relative amount of material that is present.

Automatic classification of land cover on Smith Island, VA, using HyMAP imagery

Automatic land cover classification maps were developed from Airborne Hyperspectral Scanner (HyMAP) imagery acquired May 8, 2000 over Smith Island, VA, a barrier island in the Virginia Coast Reserve. Both unsupervised and supervised classification approaches were used to create these products to evaluate relative merits and to develop models that would be useful to natural resource managers at higher spatial resolution than has been available previously. Ground surveys made by us in late October and early December 2000 and again in May, August, and October 2001 and May 2002 provided ground truth data for 20 land cover types. Locations of pure land cover types recorded with global positioning system (GPS) data from these surveys were used to extract spectral end-members for training and testing supervised land cover classification models. Unsupervised exploratory models were also developed using spatial-spectral windows and projection pursuit (PP), a class of algorithms suitable for extracting multimodal views of the data. PP projections were clustered by ISODATA to produce an unsupervised classification. Supervised models, which relied on the GPS data, used only spectral inputs because for some categories in particular areas, labeled data consisted of isolated single-pixel waypoints. Both approaches to the classification problem produced consistent results for some categories such as Spartina alterniflora, although there were differences for other categories. Initial models for supervised classification based on 112 HyMAP spectra, labeled in ground surveys, obtained reasonably consistent results for many of the dominant categories, with a few exceptions.

A credit assignment approach to fusing classifiers of multiseason hyperspectral imagery

A credit assignment approach to decision-based classifier fusion is developed and applied to the problem of land-cover classification from multiseason airborne hyperspectral imagery. For each input sample, the new method uses a smoothed estimated reliability measure (SERM) in the output domain of the classifiers. SERM requires no additional training beyond that needed to optimize the constituent classifiers in the pool, and its generalization (test) accuracy exceeds that of a number of other extant methods for classifier fusion. Hyperspectral imagery from HyMAP and PROBE2 acquired at three points in the growing season over Smith Island, VA, a barrier island in the Nature Conservancys Virginia Coast Reserve, serves as the basis for comparing SERM with other approaches.

Manifold learning techniques for the analysis of hyperspectral ocean data

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.

A new data-driven approach to modeling coastal bathymetry from hyperspectral imagery using manifold coordinates

Recently a new approach to modeling nonlinear structure in hyperspectral imagery was introduced [Bachmann et al., 2005]. The new method is a data-driven approach which extracts a set of coordinates that directly parameterize nonlinearities present in hyperspectral imagery, both on land and in the water column. The motivation for such a parameterization and its applicability to coastal bathymetry is based on the physical expectation that in shallow waters in a region that is homogeneous in bottom type and dissolved constituents, the reflectance at any particular wavelength should decay exponentially as a function of depth. If the rate varies with wavelength, then the reflectance should best be described by a nonlinear sheet or manifold in spectral space. Other changes in the structure of the data manifold can be expected as inherent optical properties (IOP) and bottom type vary. The manifold coordinates can be used to extract information concerning the latter as well. In the present work, we compare a manifold coordinate based approach to extracting bathymetry with prior work [Maness et al., 2005] based on radiative transfer modeling; the latter defined a set of look-up tables produced by repeated execution of a radiative transfer software package known as EcoLight. Comparative results for the two approaches are presented for the same Portable Hyperspectral Imager for low-light spectroscopy (PHILLS) airborne hyperspectral scene, acquired over the Indian River Lagoon in Florida in July 2004 and described in [Maness et al., 2005].

Modeling Coastal Waters from Hyperspectral Imagery using Manifold Coordinates

In [1] [2], we introduced a direct data driven method of modeling nonlinear structure in hyperspectral imagery based on Isometric Mapping [15]. More recently, we have further improved the scaling of the approach [2], making it a practical method for large-scale hyperspectral scenes. The new method extracts a set of data manifold coordinates that directly parameterize nonlinearities present in hyperspectral imagery, both on land and in the water column. In the water column, this is particularly important because of the nonlinear, attenuating properties of the medium. In this paper, we model hyperspectral imagery acquired by the NRL PHILLS [5] at the Indian River Lagoon, Florida in July 2004. In our previous efforts [3] using a small subset of data derived from the surf zone outside of the lagoon, dominant manifold coordinates were shown to parameterize bathymetry directly with a high degree of correlation to a radiative transfer look-up table (LUT) approach. In the present work, we construct a full scene manifold coordinate representation and use this as the basis of a LUT for samples with known depths as determined by the SHOALS LIDAR. Sequestered test data presented to the manifold based LUT yield a mean estimated depth which differs from the LOAR retrieved depth by less than 0.44 m for depths between 0-10 m with a standard deviation less than 1.2 m.

A profiling optics and water return system for validation and calibration of ocean color imagery

We describe a Profiling Optical and Water Return (POWR) system that has been developed and used extensively at sea. The POWR system is a collection of oceanographic instruments used to measure the inherent optical properties (IOPs) of the upper 100m of the ocean while simultaneously collecting up to eight water samples at various depths for chemical and biological analysis. IOPs are local measurements that are directly related to the properties of the water at the depth sampled; hence it is critical that the water samples be taken at the same time and location as the IOPs. Used during three major experiments, the POWR system has proven valuable for relating IOPs to in-water constituents in support of ocean color remote sensing data product validation, optical model validation, and other interdisciplinary programs.

Airborne hyperspectral imaging of cyanobacteria accumulations in the Potomac River

High-resolution spectroscopy using the Portable Hyperspectral Imager for Low-Light Spectroscopy (PHILLS) was applied to the problem of detecting potentially harmful algae blooms in the coastal environment. Data were collected on two aircraft passes, 30-min apart, over the tidally influenced part of the Potomac River. Use of two wavelengths, 0.676 and 0.700 μm, permitted the detection of surface algae accumulations while avoiding the need for atmospheric corrections, which are problematic in Case-2 water. The analysis identified algal accumulations derived from frontal processes, and narrow, linearly coherent streaks, derived from Langmuir circulation. The streaks increased markedly in number between the two passes and formed a two-dimensional pattern across the river, consistent with the advection time of surface material into windrows. The effect of wind on the patches is primarily a local reorganization of the algal material into new streaks. Spectra from within the streaks compared to those from ambient water showed absorption characteristics consistent with the presence of cyanobacteria. This interpretation is reinforced by available in-situ data. This study illustrates the value of high spectral and temporal resolutions in observing the spatial distribution of the algae, in identifying dominant functional groups, and in understanding the response of the algae to physical forcing.

Automated land-cover models of barrier islands in the Virginia Coast Reserve derived from multi-season HyMAP and PROBE2 imagery

This paper addresses the problem of automated land-cover classification in a chain of barrier islands, collectively known as the Virginia Coast Reserve (VCR), an NSF Long Term Ecological Research site. Previously we have reported on the development of automatic land-cover models derived from a HyMAP scene acquired in May 2000 of Smith Island, Virginia, one of the VCR islands. The present work considers five of the six islands covered by additional hyper-spectral collections from another similar sensor, PROBE2. PROBE2 was flown over the VCR on August 22 and October 18, 2001. Between the two sensors, we now have three-season coverage of Smith Island, and two-season coverage of five other VCR islands: Myrtle, Ship Shoals, Wreck, Cobb, and Hog. Automatic classification Models have been derived, using both exploratory unsupervised classification models, and supervised models derived from labelled spectra. Results have been compared with in situ spectrometry measurements taken by us at the VCR and at another site in Great Bay, NJ. We also have acquired PROBE2 imagery in August 2001 at the Great Bay site and collected in-situ spectra there in late July and early August. These data serve as the beginning point of cross-region comparisons.

HYDICE data from Lake Tahoe: comparison to coincident AVIRIS and in-situ measurements

Coordinated flights of two calibrated airborne imaging spectrometers, HYDICE and AVIRIS, were conducted on June 22, 1995 over Lake Tahoe. As part of HYDICEs first operational mission, one objective was to test the system performance over the dark homogeneous target provided by the clear deep waters of the lake. The high altitude and clear atmosphere makes Lake Tahoe a simpler test target than near-shore marine environments, where large aerosols complicate atmospheric correction and sediment runoff and high chlorophyll levels make interpretation of he data difficult. Calibrated data from both runoff and high chlorophyll levels make interpretation of the data difficult. Calibrated data from both sensors was provided in physical units of radiance. The atmospheric radiative transfer code, MODTRAN was used to remove the path radiance between the ground and sensor and the skylight reflected from the water surface. The resulting water-leaving spectrometer, and with values calculated form in-water properties using the HYDROLIGHT radiative transfer code. The agreement of the water-leaving radiance for the HYDICE data, the ground-truth spectral measurements, and the results of the radiative transfer code are excellent for wavelengths greater than 0.45 micrometers . The AVIRIS flight took place more than an hour closer to noon, which makes the radiance measurements not directly comparable. Comparisons to radiative transfer output for this later time indicate that the AVIRIS data is strongly by sun glint. Because water-leaving radiance is dependent upon the characteristics of the water, it can be analyzed for some of those properties. Using the CZCS algorithm based on the water-leaving radiance at two wavelengths, the chlorophyll content of Lake Tahoe was computed from the HYDICE and ground-truth data. Resulting values are slightly higher than measurements made two weeks earlier from water samples, indicating a growth in the phytoplankton population which is very plausible given the intervening atmospheric conditions. The success in determining water-leaving radiance and interpreting it for pigment concentration are very positive results for this early HYDICE flight. The interpretations made so far do not make use of the full spectral content of the data, so much room for advancement remains.