Home Articles Radiometry cross-calibration of RazakSatT and SPOT 5 sensors

Radiometry cross-calibration of RazakSatT and SPOT 5 sensors

Shu Chuen, Woo
Department of Engineering
Astronautic Technology Sdn Bhd

This paper presents a method for radiometric cross-calibration of the solar reflective spectral bands of Medium-Sized Aperture Camera (MAC) on board of RazakSat™, a due to launch Malaysian second remote sensing satellite. SPOT 5 High Resolution Geometry (HRG) sensor has 3 spectral bands equivalents to RazakSat™ MAC Green, Red and Near Infrared region; while VEGETATION2 sensor’s blue band is equivalent to RazakSat™ blue band. RazakSat™ MAC has no on-board calibration assembly. This study attempts to calibrate MAC against the well calibrated SPOT 5. Radiative transfer model will be employed to convert SPOT 5 reflectance measured at the top-of-the atmosphere to MAC viewing geometry reflectance using bidirectional reflectance distribution function (BRDF) as measured on the ground. Consequently, bands calibration gain will be derived based on regression line of digital count against predicted MAC radiance. The value of gain derived here could be used as a temporal updated calibration gain for image MAC.

1. Introduction
RazakSat™ will be orbiting at Near Equatorial Orbit (NEqO) with nominal altitude of 685 km and 9 degrees inclination after its launch. The capability of high frequent passing time (14 per day) over Malaysia region will make earth observation easier than ever. MAC system is a medium-sized electro-optical push broom camera utilizing linear Charge-Coupled Devices (CCD) to produce images. The MAC system produces high-resolution images in one panchromatic and four multi-spectral bands with ground sampling distance of 2.5 m and 5.0m, respectively. At the nominal altitude of 685 km, MAC has a swath width of 20 km. The MAC system is designed for a three-axis stabilized platform, with tilting capability in the across and along-track directions to support stereoscopic and target specific imaging.

One of the key factors to obtain high accuracy optical remote sensing satellite image depends on radiometric calibration. There are 2 ways of conducting radiometric calibration; Pre-flight measurements in laboratories and On-board calibration. Accuracy of pre-flight sometimes dissatisfy due to difficult to match the harsh situation that will be encountered while in-orbit. Additionally, the instruments are subject to degradation after launch because of the aging of the optics or of the outgassing which occurs when the instrument leaves the atmosphere (Hagolle et al., 1999). On-board calibration assembly is equipped in many satellites to solve this problem. It is mandatory procedure of choosing a maintained sensor for cross calibration. The similarity of MAC and SPOT 5 sensors spectral bands are listed in Table 1.

Table 1: Comparison of bands MAC, HRG and VEGETATION2 (VGT2) Sensor Wavelength (um) Band Resolution (m) MAC 0.51-0.73 panchromatic 2.5 HRG 0.49-0.69 panchromatic 2.5 MAC 0.45-0.52 blue 5 VGT2 0.45-0.49 blue 1000 MAC 0.52-0.60 green 5 HRG 0.50-0.59 green 10 MAC 0.63-0.69 red 5 HRG 0.61-0.68 red 10 MAC 0.76-0.89 NIR 5 HRG 0.78-0.89 NIR 10

SPOT satellites have an inner lamp and an optical fiber system to observe the sun (Meygret et al., 1994). SPOT 5’s VEGETATION2 onboard calibration system consists of a lamp and associated optical devices, mounted on a carbon fiber bar which moves in front of the optics. About 100 detectors of each of the four cameras are simultaneously lighted. The scanning of the entire field of view, obtained by rotating the bar, ensures full coverage of the detector array. The complete calibration lasts about 3 min. and is performed once a month. The in-orbit behavior of the VEGETATION2 lamp is excellent and it is used to monitor the cameras sensitivity (Henry et al., 2000). VEGETATION2 and HRG acquire data exactly same time and there will be no difference in illumination and observation angle.

Cross calibration is one of the radiometric correction approaches to monitor post launch degradation and relative sensitivity of satellite sensor calibration. This approach has been doing over years for sensors without on-board calibration and sensors have significant calibration bias. Bidirectional Reflectance Distribution Function (BRDF) can be used to standardize reflectance observation with varying sun-view geometries to a common standard geometry (Leroy et al., 1994). Remote sensing BRDF has been widely used by previous research. In a past study, cross calibration between 2 different spatial resolution images, China’s FY-1D onboard MVIRS (1.1km) with MODIS (250m, 500m, 1000m) had archived accuracy ±5% (Liu et at, 2004). Cabot et al. (2000) used Africa desert sites to cross calibrate sensor AVHRR with respect to SPOT VEGETATION. Teillet et al. (1990) used White Sand as calibration site to update the calibration of the NOAA-9 and NOAA-10 AVHRR based on Landsat 5 and SPOT image data acquired on the similar day.

2.Vicarious calibration
2.1 Calibration Site

BRDF correction used in cross calibration will rely on data of in-situ collection. Thus, site selection is become vital. Earlier field data collection work for Landsat TM/ETM+ cross calibration had been done over the Railroad Valley Playa, Nevada and Niobrara, Nebraska (Teillet et al, 2001). Liu et al., (2001) had selected China desert site, DunHuang, as their in-situ observation place. The favorable site should be low aerosol loading, low humidity, low precipitation and small atmospheric and pseudo-invariant surface (Xiao et al., 2001). Desert site is thought to be suitable land cover in previous studies as it has weak BRDF effect (geometric and volume scattering) (Liu et al., 2004). The site in Malaysia will be determined on further finding in near future. Thanks to the orbit of RazakSat™, the imaging opportunity of RazakSat™ to acquire image with near coincident time to SPOT 5 is apparently higher than any satellite.

2.2 Field data collection
Top-of-atmosphere (TOA) reflectance of SPOT 5 must be converted to surface reflectance by using Radiative Transfer Model. The SPOT 5 surface reflectance here will be corrected to MAC viewing geometry using bidirectional reflectance distribution function measured on the ground. A narrow Field-of -View (FOV) spectroradiometer will be used to ease relocation of instrument (James et al., 2004). BRDF measurement requires a solar radiometer to measures at a particular solar zenith angle with respect to certain relative azimuth angle. Spectroradiometer will measures at certain viewing zenith angle with similar relative azimuth angle to solar radiometer. Both measurements are conducted simultaneously. Wu et al. (1995) had used kernel driven BRDF model to extrapolate BRDF beyond angle geometry collected. Bidirectional reflectance factor (BRF) is an alternative for easier derivation of BRDF, but a reference Lambertian panel, usually Spectralon, required during the measurement.

3. Methodology
Cross calibration must take into account the different in both viewing geometry and spectral bands between the both sensors (Teillet et al. 2001) if both image taken in different date. Reflectance will be used within whole process to remove cosine effect of different solar zenith angle and it also compensates different values of exo-atmospheric solar irradiance from spectral band differences (Teillet et al., 2001). BRDF angular measurement is based on the assumption that the observed site is homogeneous surface. Moreover, the selected site is sufficient to change the look angle observing different sections of the surface (Deering et al., 1987).

3.1 Viewing geometry correction
The BRDF describes the variations in reflectance with illumination and view geometry; it can be expressed as Equation 1 (Nicodemus et al, 1977)

… (Equation 1)

Where fr is the BRDF (sr-1), dL is the reflected radiation for an incident irradiance of intensity dE at wavelength ?. ? and ? are the zenith and relative azimuth angles respectively. The subscripts s and v denotes the angles in solar and view directions respectively.

However, previous study by Teillet et al. (2001) used tandem pair images which viewing geometry had only contributed small effects. This temporarily orbit placing provided a great opportunity to minimize viewing geometry-led illumination difference for Landsat TM/ETM cross calibration. The processing flow is depicted in figure 1. BRDF-predicted surface reflectance will be interpolated and applied to the whole image to get continuous surface reflectance spectrum. The meteorological parameters defined in radiative transfer model will be identical.

Figure 1: Processing flow of cross calibration The adjusted SPOT 5 surface reflectance is considered as predicted MAC surface reflectance. TOA reflectance is related to TOA radiance by Equation 2,

…. (Equation 2) Where Eoi is the exo-atmospheric solar irradiance in spectral band i (in W/m2μm) based on the Modtran-3 spectrum, ? is the solar zenith angle, and ds is the Earth-Sun distance is Astronomical units. Inverse of Equation 1 will get TOA radiance. The TOA radiance from inversed Equation1 is considered predicted MAC TOA radiance.

Graph will be plotted using MAC DN against predicted MAC TOA radiance to generate gain (in unit count per unit radiance) of each spectral band. The gain generated here can be used to validate MAC respontivity. User of MAC will get latest calibrated gain instead of pre-launch gain to compute TOA radiance.

3.2 Spectral band differences compensation
Spectral response of corresponding band differs for different sensor dependent on solar illumination, atmospheric transmittance and surface reflectance. A proper adjustment is required to compensate this behavior. A normalized spectral response factor will be generated for every band prior to BRDF correction in order to normalize the varying response profiles between RazakSat™ and SPOT 5. Figure 2 illustrates the examples of RazakSat™ spectral responses. The difference of band spectral response will be normalized by multiplying a factor called spectral band adjustment factor (Teillet et al., 2001). Band adjustment factor is calculated as

Figure 2: Spectral response profiles against wavelength

4. Expected observation
Cross calibration of Landsat TM with Landsat ETM+ will be used to produce the lifetime gain model of Landsat TM reflective bands (1-5, 7). Hence, this study also could be used to generate MAC calibration coefficients, by using linear regression over predicted radiances and the MAC image DN.

Liu et al., (2001) demonstrated considerably difference of gain value obtained from their study. Comparison between pre-launch calibration coefficients and cross-calibrate result will be investigated. The investigation result may show the sensor degradation over time. A comparison table (Table 2) will be produced to monitor MAC degradation,

5. Discussion
The accuracy of cross calibration of MAC using SPOT 5 is depending on 4 major aspects:

  • SPOT 5 uncertainties Cross calibration is solely sensor stability dependent. The biggest error introduced will from the reference sensor bias. However, validation could be done for few more sensors (e.g. Landsat ETM, MODIS) in order to get best result.
  • Radiative Transfer Model Radiative Transfer Model robustness in computation plays a major role in accuracy. Every Radiative Transfer Model has its limitation. Vermote et al., (1997) found 6S code cannot handle spherical atmosphere and as a result, it cannot be used for limb observation.
  • Image geometric correction Inaccurate geometric correction will introduces error in BRDF correction.
  • Miscellaneous bias (zero radiance DN) The capability of on board calibration will provide zero radiance reading by looking at deep space (e.g. AVHRR). However, RazakSat™ lacks of this configuration.

Despite these major errors present in this method, but cross calibrate still considered a good approach due to its high calibration accuracy archived by previous studies.

Field data collection discussed in section 2.2 will be occasionally unsuccessful to get it completed especially in tropical country. Loss of observation due to persistent cloud cover poses a challenge in collecting all necessary angular geometries. Lucht et al., (2000) employed a back-up algorithm for deriving in a situation of insufficient sampling. The back-up algorithm uses angular geometry available to adjust it to match the priori angular geometry observation made.

6. References

  • Cabot, F., Hagolle, O., Henry, P. (2000) “Relative and multitemporal calibration of AVHRR, SeaWiFS, and VEGETATION using POLDER characterization of desert sites” IEEE 2000 International Volume 5, Issue, 2000 Page(s):2188 – 2190 vol.5
  • Deering, D.W., Eck, T.F., (1987). “Atmospheric Optical Depth Effects on Angular Anisotropy of Plant Canopy Reflectance”. International Journal of Remote Sensing. 8:893-916.
  • Hagolle O., P. Goloub, P.-Y. Deschamps, H. Cosnefroy, X. Briottet, T. Bailleul, J.M.Nicolas, F. Parol, B. Lafrance, and M. Herman (1999), “Results of POLDER In-Flight Calibration,” IEEE Transactions on Geoscience and Remote Sensing, vol. 37, pp. 1550-1566, 1999.
  • Henry P. ,Meygret A. (2001) “CALIBRATION OF VEGETATION CAMERAS ON-BOARD SPOT4”. Proceedings of the VEGETATION 2000 conference, Belgirate-Italy, 3-6 April 2000 , Saint G. Ed, CNES – Toulouse & JRC – Ispra, pp 23-32
  • Leroy, M. , Roujean, J.L., (1994) “Sun and view angle corrections on reflectances derived from NOAA/AVHRR data”, IEEE Trans. Geosci. Remote Sensing, vol. 32, pp. 684-697.
  • Liu, J.J., Qiao Y,L., Liu Y, J (2004). “A new method for cross calibration of two satellite sensors” International Journal of Remote Sensing, 10, Vol 25, No.23 5267-5281.
  • Lucht, W., Schaaf, C.B., Strahler, A.H., (2000). “An algorithm for the retrieval of albedo from space using semiempirical BRDF models”. IEEE. Trans. Geosci. Remote Sens. 38:977-998.
  • Meygret, A., Hagolle, O., Henry, P., Dinguirard, M.,Hazane, P., Santer, R., and Deuz´e, J.L., (1994) “SPOT 3: First in-flight calibration results,” in Proceeding IGARSS’94, Pasadena, CA.
  • Nicodemus, F. E., Richmond, J. C., Hsia, J. J., Ginsberg, I. W. and Limperis, T., (1977), Geometrical considerations and nomenclature for reflectance. (Washington D.C.:National Bureau of Standards).
  • Privette J.L, T. F. Eck, and D. W. Deering (1997) “Estimating spectral albedo and nadir reflectance through inversion of simple BRDF models with AVHRR/MODIS-like data,” J. Geophys. Res., vol. 102, pp. 29 529-29 542
  • Shell, J.R., II; Salvaggio, C.; Schott, J.R. (2004). “A novel BRDF measurement technique with spatial resolution-dependent spectral variance”. IEEE International Volume 7, Issue , 20-24 Sept. 2004 Page(s): 4754 – 4757 vol.7 Strahler, A. H. and Muller, J., (1997), “MODIS BRDF/Albedo Product: Algorithm Theoretical Basis Document (Version 3.2)”. URL: https://modis.gsfc.nasa.gov/data/atbd/atbd_mod09.pdf
  • Teillet, P. M., Slater, P. N., Ding, Y., Santer, R. P., Jackson, R.D., and Moran, M. S., (1990), “Three methods for the absolute calibration of the NOAA AVHRR sensors in-flight”. Remote Sensing of Environment, 31, 105-120.
  • Teillet, P.M., J.L., Barker, B.L., Markham, R.R., Irish, G., Fedosejevs, J.C., Storey, (2001). “Radiometric cross-calibration of the Landsat-7 ETM+ and Landsat-5 TM sensors based on tandem data sets”. Remote Sensing of Environment, 78, 39-54.
  • Vermote, E., Tandre´ , D., Deuze´ , J. L., Herman, M., and Morcette, J. J., (1997), “Second simulation of the satellite signal in the solar spectrum: an overview”. IEEE Transactions on Geoscience and Remote Sensing, 35, 675-686.
  • Xiao, Q., Liu, J., Yu, H., and Zhang, H., (2001), “Analysis and evaluation of optical uniformity for Dunhuang calibration site by airborne spectrum survey date”. Proceedings of China Remote Sensing Sensors Radiometric Calibration (Beijing: Ocean Press), pp. 136-142.