Propagation of sensor noise in oceanic hyperspectral remote sensing.

In previous works, the authors have shown via numerical simulation that sensor noise, even assuming otherwise perfect knowledge of the environment, can cause large scale variations in the retrieval of concentrations of biophysical parameters in a water body, and also investigated methods for using statistical measures (such as the Mahalanobis distance) to help mitigate these issues. In this work, we derive explicit formulas that can be used to estimate how uncertainty in the sensor radiance is propagated to uncertainty in the remote sensing reflectanceRrs(λ), without the need for simulations. In particular, the formulas show that the variation in Rrs(λ)is affected by not only the noise characteristics of the sensor, but also by the conditions (atmospheric parameters, viewing angles, altitude) under which the data is collected. We include validation results for the formulas over a wide range of atmospheric conditions, and show by example how the collection conditions can affect the uncertainty in Rrs(λ).

Airborne system for multispectral, multiangle polarimetric imaging.

In this paper, we describe the design, fabrication, calibration, and deployment of an airborne multispectral polarimetric imager. The motivation for the development of this instrument was to explore its ability to provide information about water constituents, such as particle size and type. The instrument is based on four 16 MP cameras and uses wire grid polarizers (aligned at 0°, 45°, 90°, and 135°) to provide the separation of the polarization states. A five-position filter wheel provides for four narrow-band spectral filters (435, 550, 625, and 750 nm) and one blocked position for dark-level measurements. When flown, the instrument is mounted on a programmable stage that provides control of the view angles. View angles that range to ±65° from the nadir have been used. Data processing provides a measure of the polarimetric signature as a function of both the view zenith and view azimuth angles. As a validation of our initial results, we compare our measurements, over water, with the output of a Monte Carlo code, both of which show neutral points off the principle plane. The locations of the calculated and measured neutral points are compared. The random error level in the measured degree of linear polarization (8% at 435) is shown to be better than 0.25%.

Expected improvements in the quantitative remote sensing of optically complex waters with the use of an optically fast hyperspectral spectrometer-a modeling study.

Using simulated data, we investigated the effect of noise in a spaceborne hyperspectral sensor on the accuracy of the atmospheric correction of at-sensor radiances and the consequent uncertainties in retrieved water quality parameters. Specifically, we investigated the improvement expected as the F-number of the sensor is changed from 3.5, which is the smallest among existing operational spaceborne hyperspectral sensors, to 1.0, which is foreseeable in the near future. With the change in F-number, the uncertainties in the atmospherically corrected reflectance decreased by more than 90% across the visible-near-infrared spectrum, the number of pixels with negative reflectance (caused by over-correction) decreased to almost one-third, and the uncertainties in the retrieved water quality parameters decreased by more than 50% and up to 92%. The analysis was based on the sensor model of the Hyperspectral Imager for the Coastal Ocean (HICO) but using a 30-m spatial resolution instead of HICO's 96 m. Atmospheric correction was performed using Tafkaa. Water quality parameters were retrieved using a numerical method and a semi-analytical algorithm. The results emphasize the effect of sensor noise on water quality parameter retrieval and the need for sensors with high Signal-to-Noise Ratio for quantitative remote sensing of optically complex waters.

Improving the retrieval of water inherent optical properties in noisy hyperspectral data through statistical modeling.

The use of the Mahalanobis distance in a lookup table approach to retrieval of in-water Inherent Optical Properties (IOPs) led to significant improvements in the accuracy of the retrieved IOPs, as high as 50% in some cases, with an average improvement of 20% over a wide range of case II waters. Previous studies have shown that inherent noise in hyperspectral data can cause significant errors in the retrieved IOPs. For LUT-based retrievals that rely on spectrum matching, the particular metric used for spectral comparisons has a significant effect on the accuracy of the results, especially in the presence of noise in the data. In this study, we have compared the Euclidean distance and the Mahalanobis distance as metrics for spectral comparison. In addition to providing justification for the preference of the Mahalanobis Distance over the Euclidean Distance, we have also included a statistical description of noisy hyperspectral data.

Impact of signal-to-noise ratio in a hyperspectral sensor on the accuracy of biophysical parameter estimation in case II waters.

Errors in the estimated constituent concentrations in optically complex waters due solely to sensor noise in a spaceborne hyperspectral sensor can be as high as 80%. The goal of this work is to elucidate the effect of signal-to-noise ratio (SNR) on the accuracy of retrieved constituent concentrations. Large variations in the magnitude and spectral shape of the reflectances from coastal waters complicate the impact of SNR on the accuracy of estimation. Due to the low reflectance of water, the actual SNR encountered for a water target is usually quite lower than the prescribed SNR. The low SNR can be a significant source of error in the estimated constituent concentrations. Simulated and measured at-surface reflectances were used in this study. A radiative transfer code, Tafkaa, was used to propagate the at-surface reflectances up and down through the atmosphere. A sensor noise model based on that of the spaceborne hyperspectral sensor HICO was applied to the at-sensor radiances. Concentrations of chlorophyll-a, colored dissolved organic matter, and total suspended solids were estimated using an optimized error minimization approach and a few semi-analytical algorithms. Improving the SNR by reasonably modifying the sensor design can reduce estimation uncertainties by 10% or more.

Vicarious calibrations of HICO data acquired from the International Space Station.

The Hyperspectral Imager for the Coastal Ocean (HICO) presently onboard the International Space Station (ISS) is an imaging spectrometer designed for remote sensing of coastal waters. The instrument is not equipped with any onboard spectral and radiometric calibration devices. Here we describe vicarious calibration techniques that have been used in converting the HICO raw digital numbers to calibrated radiances. The spectral calibration is based on matching atmospheric water vapor and oxygen absorption bands and extraterrestrial solar lines. The radiometric calibration is based on comparisons between HICO and the EOS/MODIS data measured over homogeneous desert areas and on spectral reflectance properties of coral reefs and water clouds. Improvements to the present vicarious calibration techniques are possible as we gain more in-depth understanding of the HICO laboratory calibration data and the ISS HICO data in the future.

Interpretation of absorption bands in airborne hyperspectral radiance data.

It is demonstrated that hyperspectral imagery can be used, without atmospheric correction, to determine the presence of accessory phytoplankton pigments in coastal waters using derivative techniques. However, care must be taken not to confuse other absorptions for those caused by the presence of pigments. Atmospheric correction, usually the first step to making products from hyperspectral data, may not completely remove Fraunhofer lines and atmospheric absorption bands and these absorptions may interfere with identification of phytoplankton accessory pigments. Furthermore, the ability to resolve absorption bands depends on the spectral resolution of the spectrometer, which for a fixed spectral range also determines the number of observed bands. Based on this information, a study was undertaken to determine under what circumstances a hyperspectral sensor may determine the presence of pigments. As part of the study a hyperspectral imager was used to take high spectral resolution data over two different water masses. In order to avoid the problems associated with atmospheric correction this data was analyzed as radiance data without atmospheric correction. Here, the purpose was to identify spectral regions that might be diagnostic for photosynthetic pigments. Two well proven techniques were used to aid in absorption band recognition, the continuum removal of the spectra and the fourth derivative. The findings in this study suggest that interpretation of absorption bands in remote sensing data, whether atmospherically corrected or not, have to be carefully reviewed when they are interpreted in terms of photosynthetic pigments.

Interpretation of hyperspectral remote-sensing imagery by spectrum matching and look-up tables.

A spectrum-matching and look-up-table (LUT) methodology has been developed and evaluated to extract environmental information from remotely sensed hyperspectral imagery. The LUT methodology works as follows. First, a database of remote-sensing reflectance (Rrs) spectra corresponding to various water depths, bottom reflectance spectra, and water-column inherent optical properties (IOPs) is constructed using a special version of the HydroLight radiative transfer numerical model. Second, the measured Rrs spectrum for a particular image pixel is compared with each spectrum in the database, and the closest match to the image spectrum is found using a least-squares minimization. The environmental conditions in nature are then assumed to be the same as the input conditions that generated the closest matching HydroLight-generated database spectrum. The LUT methodology has been evaluated by application to an Ocean Portable Hyperspectral Imaging Low-Light Spectrometer image acquired near Lee Stocking Island, Bahamas, on 17 May 2000. The LUT-retrieved bottom depths were on average within 5% and 0.5 m of independently obtained acoustic depths. The LUT-retrieved bottom classification was in qualitative agreement with diver and video spot classification of bottom types, and the LUT-retrieved IOPs were consistent with IOPs measured at nearby times and locations.