Pre-flight radiometric calibration
The process of characterising a sensor before it is launched or used in-flight is a complex one, and involves the use of highly specialised instruments and access to traceable radiometric standards. Several organisations in the UK have such facilities, including:
Various ways to check the laboratory calibration under conditions similar to those in which the sensor is operated have been developed. These range from dedicated calibration lamps incorporated into the sensor to external devices such as the mobile integrating sphere shown in the figure below.
The stability of integrating spheres needs considering as these are widely used to transfer radiance calibrations (Guenther, 1987). Goetz et al. (1998) describe a ‘round robin' set of measurements using an ASD FieldSpec FR spectroradiometer to measure the spectral radiance of eight integrating spheres operated at several NASA centres, Los Alamos National Laboratory and the University of Arizona. The spectroradiometer used was assessed as having a precision of better than ± 2% at wavelengths greater than 500 nm, rising to 6% at 350 nm. The study found significant variability between the various integrating spheres, but overall this was within ± 5% for wavelengths greater than 500 nm. The authors identified the absorption of water vapour by the barium sulphate coating of several of the larger spheres as being a particular problem and recommended that they be kept flushed with dry nitrogen. The effect of sphere calibrations being interpolated from measurements made at a limited number of wavelengths was also noted. If uncorrected these could introduce spectral features from the material used to coat the surface of the integrating sphere into the resultant spectra.
In most applications to date, the accuracy of such integrating spheres has not been a critical issue, as applications and users have been satisfied with error levels of around 5 %. However, as quantitative Earth observation matures and demand from users stimulates the need for improvement this will become more of a limiting factor. The objective of an integrating sphere is to homogenise radiation into a spatially uniform (or at least Lambertian) source, but this is rarely achieved in practice, and spatial non-uniformities of several percent are typical from such sources (Knee, 1999). Without some form of independent and spectrally discriminating monitoring device, the output radiance will rapidly lose its traceability to primary sources. It is also important to note that when using integrating spheres in close proximity to any instrument, they are a very large source of stray light and highly susceptible to inter-reflections.
Recognising these issues, the National Physical Laboratory (NPL), the UK national standards laboratory, has designed a transportable integrating sphere specifically to calibrate remote sensing systems in the field and this has been used with satellite and aircraft sensors as well as field spectroradiometers. It is known as the NPL Transfer Standard and Absolute Radiance Source (TSARS) (Pegrum et al., 2004) and is calibrated at NPL by direct comparison with the primary black body source (Woolliams et al., 2002). It has been specifically designed to be spatially uniform (<0.5% over its exit port of up to 100 mm diameter) and encompasses a set of filter radiometers to monitor its calibration. The overall error budget for the NPL TSARS gives an uncertainty between 450–500 nm between 0.70% and 0.99% at the 95% confidence level. Within the spectral region of 500–700 nm, the 95% confidence uncertainty is better still, between 0.56% - 0.74%. Above 700 nm the uncertainties are between 0.61% - 1.18%. These uncertainties were improved before the TSARS was deployed during the NCAVEO 2006 field campaign and used to calibrate six spectroradiometers from ten different research groups participating in the experiment. A similar portable integrating sphere is described by Brown & Johnson (2003). Here, NASA worked with the National Institute of Standards and Technology (NIST) to develop a portable integrating sphere source, called the NIST Portable Radiance Source (NPR), aimed at complementing existing detector-based measurement strategies and to enhance the capabilities of the EOS calibration programme. The NPR is calibrated for spectral radiance over the range 400nm to 2400nm at the primary US facility for irradiance and radiance calibrations.
Since the 1980's National Metrology Institutes (NMI) have worked towards basing all radiometric scales, including spectral radiance and irradiance, on traceability to a primary standard detector, the cryogenic radiometer (Fox, 1996). This instrument defines a scale of spectral responsivity (a monochromatic characteristic of a detector) and then calibrates filter radiometers (detector-spectral filter combination) to allow measurements of real world polychromatic sources e.g lamps, Sun, Earth-reflected solar irradiance (radiance). This is carried out by spectrally tuning a beam of monochromatic radiation (calibrated with reference to a standard detector) over the entrance aperture of a filter radiometer. Alternatively, and more commonly, a few standard filter radiometers can be used to measure the thermodynamic temperature of a black body source from which spectral radiance and irradiance can then be derived. The most commonly used transfer lamps are tungsten halogen 1000 W FEL type. It should be noted that evidence recently obtained from a large study of such lamps undertaken by NPL as part of an international comparison of spectral irradiance suggests that one in three lamps is subject to a change in its output on transfer from NMI to the user (Woolliams et al., 2006). This is particularly worrying since many users only have one standard and in some cases this is used to establish secondary standards provided to others.
A transportable radiance standard based on a 1000 W calibrated quartz-halogen lamp and Spectralon reference panel has been developed for laboratory and hangar-based validation of the radiometric calibration of the AVIRIS imaging spectrometer, and found to have an estimated accuracy of around 1% in the visible region, reducing to 6.5% at 2,500 nm (Chrien et al., 2000). Such a transportable radiance standard could also be used to check the radiometric calibration of spectroradiometers in the field. As the science of remote sensing matures and models become more sophisticated, the demand for accurate measurements of spectral radiance (as opposed to reflectance) in the field will increase. Furthermore, advances in metrology mean that it is now more accurate to calibrate and measure in terms of spectral radiance than reflectance. Taken together, these two developments can be expected to steer the development of new methodologies to those which provide increased rigour, demonstrable traceability for individual measurements, and improved intercomparability between research groups (Fox, 2004).
Further information on specific sensors
Landsat Multispectral Scanner
Markham, B. L. and Barker, J. L., 1983. Spectral characterization of the Landsat-4 MSS sensors. Photogrammetric Engineering and Remote Sensing 49, 811-33.
Markham, B. L. and Barker, J. L., 1985. Spectral characterization of the LANDSAT Thematic Mapper sensors. International Journal of Remote Sensing 6, 697-716.
Thome, K. J., Markham, B., Barker, J., Slater, P. N. and Biggar, S., 1997. Radiometric calibration of Landsat. Photogrammetric Engineering and Remote Sensing 63, 853-858.