News
July 19, 2011:Coming Soon - GSICS Mail and News Subscriptions
In August 2011, GSICS plans to launch a mail and news subscription service through the cloud-based provider MailChimp. The service will provide information about GSICS inter-calibration activities, product beta-testing opportunities, product updates, workshops and meetings, and access to the GSICS Quarterly newsletter. We at the GCC plan to send you an e-mail in the coming weeks about how you can sign up for this new subscription service.
October 28, 2011: From GSICS Quarterly Vol. 5 No. 3, 2011, (PDF, 304 KB)
Cross–Comparison of GOME-2, AVHRR and AASTR Reflectance
Figure 1: AVHRR-3 Channel 1 spectral response function and GOME-2 spectral bands with red arrows showing the detector readout direction
The coincidence in space and time, and spectral coverage of the instruments on Metop-A provides an excellent opportunity to gain insight into their performance and the characteristics of their data. To assist in the calibration/validation of the GOME-2 spectrometer on Metop-A, comparisons with AVHRR-3 data have been performed to establish both the accuracy of the geo-location of the GOME-2 data and ultimately to examine long-term trends between the instruments’ calibration. The method can be extended to similar classes of instruments, e.g., inter-comparisons of GOME-1, SCIAMACHY, ATSR-2 and AATSR. An initial evaluation of consistency between GOME-2 and AATSR has already been carried out. This multi-mission framework will allow further analysis of the long-term consistency of each instrument dataset.
The GOME class of instruments (GOME-1, SCIAMACHY, GOME-2 on ERS-2, ENVISAT and Metop-A respectively) provide visible measurements collocated with ATSR-2, AATSR and AVHRR-3 respectively. Due to the different spectral and spatial resolutions, the spectrometer data must be spectrally averaged, using the appropriate radiometer filter function, while the radiometers must be spatially averaged with the appropriate spectrometer ground pixel footprint (40x320km, 30x120km, and 40x80km respectively).
For each ground pixel of the spectrometer, its reflectance is convolved with the Spectral Response Function (SRF) of each imager channel. For AVHRR Channel 1, a small part of the SRF is not covered by GOME-2 channel 4 (see Figure as illustration). The spectrometer reflectance is currently extrapolated at a constant value over this region, on the basis of an examination of spectra from multiple scenes, for which this seems a good approximation. Comparison of GOME-2 with AVHRR channel 2 is currently only considered indicative, due to partial spectral coverage and the presence of water vapour absorption lines across the channel.
The imager data in each band is then spatially averaged based on the latitude/longitude information provided for each instrument. Due to the sequential readout of the GOME-2 detectors, each detector pixel sees a slightly different ground scene and therefore a wavelength dependent correction to the geolocation information is required. Based on the corner values for each spectrometer ground pixel, plus some margin, the associated latitude/longitude data of the imager data is extracted.
Imager pixels within the FoV are identified, allowing for any specified misalignment correction and taking into account skew in off-nadir view, longitude varying with latitude, and earth curvature. Imager data are averaged and some scene variability statistics are stored for each spectrometer pixel (including cloud information where available).
Figure 2: GOME-2 spectrally convolved reflectance versus AVHRR channel 1 (0.632) reflectance – both on Metop-A. The red line shows the fit forced through [0, 0] i.e., no offset is assumed, only a potential gain error. Dash-dot line is 1:1
Assuming the geolocation of the higher spatial resolution imager is correct and that scatter is due to differences in scene viewed by the two instruments, a collocation shift can be derived which produces the minimum scatter. Applying this method to GOME-2 and AVHRR data allowed an error in the GOME-2 geolocation calculation to be isolated and corrected. The UTC time per scanner position was shifted by 375 ms due an incorrect processor setting and as a consequence all latitude/longitude and geolocation parameters were shifted by 2.5 to 5% of the along-track pixel size at nadir position, i.e., by 1 to 2 km along the orbit. This was fixed in the new processor version 4.5 installed on 9th September 2010.
After implementing the bug fix, GOME-2 data were compared to both AATSR and AVHRR data to examine differences in radiometric accuracy between the instruments. In the case of AATSR the 30 minute separation of Metop-A and ENVISAT adds noise to the comparisons due to scene variability. This effect can be reduced by processing several orbits to provide a stable estimate of the bias, although systematic scene differences also need to be considered.
Comparison of GOME-2 with the AATSR 670nm channel, which is believed to be very well calibrated, showed excellent agreement for the period 2007 to September 2009 based on a selection of one orbit with near-nadir coincidence per month over a period of two years. The mean difference in Sun-normalised reflectance was <1% with some seasonal variation (assumed to be due to systematic differences in scene and surface BRDF).
By contrast, comparison of GOME-2 data with collocated Metop-A/AVHRR channel 1 data for the same period indicates an offset of ~9% in AVHRR (see Figure 1). This is consistent with the offset of ~9% reported by Cao et al. 2008 between NOAA-16 & -17 AVHRR and MODIS 0.63µm data.
This work will be continued with a view to investigating long-term trends between the instruments and extending the method other instruments of a similar class.
References
Cao, C., X. Xiong, A. Wu, X. Wu, 2008. Assessing the consistency of AVHRR and MODIS L1b reflectance for generating Fundamental Climate Data Records, J. Geophys. Res., 113, doi: 10.1029/2007JD009363.
(by B. Latter [Rutherford Appleton Laboratory, U.K.], R. Lang and R. Munro [EUMETSAT])
