Abstract
The United States Naval Research Laboratory (NRL) has conducted hyperspectral remote sensing of the coastal ocean environment for nearly two decades, employing sensors that span the wavelength range from the near ultraviolet through the short wave infrared. Of these, four operate in the visible to near infrared (VNIR), including three used on airborne platforms: the Portable Hyperspectral Imager for Low Light Spectroscopy (PHILLS), a commercial CASI-1500, the UAV capable micro-Small Hyperspectral Imager for the Naval Environment (microSHINE), and one aboard the International Space Station (ISS)- the Hyperspectral Imager for the Coastal Ocean (HICO).
It is interesting to note that all four sensors have shown field radiances that are lower than expected at the shortest wavelength up to approximately 450-500nm, depending on the sensor. This necessitates the use of an empirical scaling at these wavelengths. There is also some evidence that hyperspectral imagers used by other groups have a similar issue. Suspected causes include atmospheric correction or calibration. This poster discusses sensor calibration and measurements taken at NRL at ground level (i.e. little atmosphere contribution) that seem to indicate that calibration does contribute to the problem of low radiances at the blue end of the spectrum.
The NRL sensors were calibrated using NIST traceable integrating spheres, one using only QTH lamps operating at about 3000K, and another blue-enhanced sphere equipped with a 300W Xe arc lamp in addition to QTH lamps. It should be noted that the first sphere was included in the first SIMBIOS Radiometric Intercomparison (SIMRIC) in 2001 and was shown to be within 2% agreement of the SeaWiFS Transfer Radiometer SXR-II at all wavelengths. In addition, the calibration of this sphere is checked using commercial Field Spec Pro and 3 spectrometers with current calibrations and it is consistently within a few percent agreement at all wavelengths. The calibration spheres themselves are not the problem.
It is more likely that the problem is related to the difference between the blue-deficient spectrum used to calibrate the sensors and the blue-rich spectra measured in the field. Almost all of these sensors are calibrated using an integrating sphere illuminated by quartz halogen QTH sources that operate at a color temperature of about 3000k. These sources have a very low output at the problematic blue wavelengths and a relatively much larger output at the red and infrared wavelengths. This is exactly opposite of what is measured in the field. In this case, the problem with calibration could be related to stray light or linearity issues in the sensor. This poster will discuss the investigations into resolving this issue. For instance, it appears that stray light may not be the issue as demonstrated during the calibration of HICO. At that time, HICO was placed in front of the QTH only sphere and the blue-enhanced sphere, which have very different spectral shapes. Both yielded almost the exact same calibration coefficients at all wavelengths and still the on-orbit radiances required an empirical correction in the blue.
Hyperspectral Imager Calibration in the Blue: Issues and Experiments
The United States Naval Research Laboratory (NRL) has conducted hyperspectral remote sensing of the coastal ocean environment for nearly two decades, employing sensors that span the wavelength range from the near ultraviolet through the short wave infrared. Of these, four operate in the visible to near infrared (VNIR), including three used on airborne platforms: the Portable Hyperspectral Imager for Low Light Spectroscopy (PHILLS), a commercial CASI-1500, the UAV capable micro-Small Hyperspectral Imager for the Naval Environment (microSHINE), and one aboard the International Space Station (ISS)- the Hyperspectral Imager for the Coastal Ocean (HICO).
It is interesting to note that all four sensors have shown field radiances that are lower than expected at the shortest wavelength up to approximately 450-500nm, depending on the sensor. This necessitates the use of an empirical scaling at these wavelengths. There is also some evidence that hyperspectral imagers used by other groups have a similar issue. Suspected causes include atmospheric correction or calibration. This poster discusses sensor calibration and measurements taken at NRL at ground level (i.e. little atmosphere contribution) that seem to indicate that calibration does contribute to the problem of low radiances at the blue end of the spectrum.
The NRL sensors were calibrated using NIST traceable integrating spheres, one using only QTH lamps operating at about 3000K, and another blue-enhanced sphere equipped with a 300W Xe arc lamp in addition to QTH lamps. It should be noted that the first sphere was included in the first SIMBIOS Radiometric Intercomparison (SIMRIC) in 2001 and was shown to be within 2% agreement of the SeaWiFS Transfer Radiometer SXR-II at all wavelengths. In addition, the calibration of this sphere is checked using commercial Field Spec Pro and 3 spectrometers with current calibrations and it is consistently within a few percent agreement at all wavelengths. The calibration spheres themselves are not the problem.
It is more likely that the problem is related to the difference between the blue-deficient spectrum used to calibrate the sensors and the blue-rich spectra measured in the field. Almost all of these sensors are calibrated using an integrating sphere illuminated by quartz halogen QTH sources that operate at a color temperature of about 3000k. These sources have a very low output at the problematic blue wavelengths and a relatively much larger output at the red and infrared wavelengths. This is exactly opposite of what is measured in the field. In this case, the problem with calibration could be related to stray light or linearity issues in the sensor. This poster will discuss the investigations into resolving this issue. For instance, it appears that stray light may not be the issue as demonstrated during the calibration of HICO. At that time, HICO was placed in front of the QTH only sphere and the blue-enhanced sphere, which have very different spectral shapes. Both yielded almost the exact same calibration coefficients at all wavelengths and still the on-orbit radiances required an empirical correction in the blue.