Calibration Status of the Aatsr Reflectance Channels

The radiometric gains of the AATSR visible to shortwave-infrared channels are calibrated using an on-board solar diffuser. The on-board calibration is supported by vicarious techniques such as comparisons over stable desert targets to establish relative biases in the radiometric calibrations of AATSR and other optical sensors and also to determine the long-term drifts. The authors present the results of comparisons over the Dome-C site in Antarctica, analysis of dark water measurements and multi-platform inter-comparisons. The data and the lessons learned from the calibration monitoring are being fed into the Sentinel-3 calibration planning. 1. AATSR Visible Calibration History Only 3 updates to AATSR visible calibration tables have been introduced into the L1b products since the launch of ENVISAT in 2002. • 14-December-2004 1.6m non-linearity correction introduced (PO-TN-RAL-AT-0540) • 29-November-2005 – Exponential Drift Correction is applied to VC1 files (PO-TN-RAL-AT-0542) • 18-December-2006 – Thin Film Drift Correction is introduced to VC1 files (PO-TN-RAL-AT-0552) • NO corrections for bias (e.g. AATSR vs. MERIS) have been introduced. Identifying which corrections have been applied to a specific product is obtained from the name of the VC1 and GC1 files used (DSD 30 and DSD 31 in L1B product). The GC1 filename determines whether or not the 1.6μm nonlinearity correction has been applied, so if GC1 file ATS_GC1_AXVIEC20020123_073430_20020101_000 000_20200101_000000 was used the nonlinearity correction is NOT applied, otherwise the correction has been applied and the user does not need to take any further action. The VC1 file generation date indicates which drift correction has been applied as follows • Date before 29-Nov-2005 13:20:26 then no correction applied • Date between 29-Nov-2005 13:20:26 and 18-Dec2006 20:14:15 then exponential drift correction is applied • Date after 18-Dec-2006 20:14:15 then thin film correction is applied The ESA Data Processing Centre has implemented the corrections when the updates were introduced. So for the 1st reprocessing, 1.6um channel non-linearity correction depends on when data was reprocessed in relation to updates to GC1 file i.e. if reprocessed after 14th December 2004 then L1B products should have nonlinearity correction incorporated. The drift correction is dependent on VC1 files used. For the 2nd Reprocessing all products should have 1.6um nonlinearity correction applied and the drift correction will be as for the 1 reprocessing as the same VC1 files are used. 2. Long-Term Trend Analysis Desert and Ice Targets have been used extensively for calibration and monitoring of a range of optical sensors including AVHRR, ATSR-2, GOES, POLDER and Vegetation. The principal assumptions are • Uniform reflectance over large area • Long term-radiometric stability of the calibration sites ensures long-term stability of the top-of-the atmosphere (TOA) albedo (and of seasonal variations, if any) or reflectance over large spatially uniform areas. • High surface reflectance to maximise the signal-to-noise and minimise atmospheric effects on the radiation measured by the satellite The long term calibration drift is established by comparing against reference measurements i.e. D(t) = R(t)/Rref either by using a stable reference sensor such as MERIS or by using a reference BRDF derived from early measurements, model or ground measurements. The test sites used in the analysis are primarily those identified by Cosnefroy and co-workers in 1996 [1], which are now well established reference targets. _________________________________________ Proc. of the '2nd MERIS / (A)ATSR User Workshop', Frascati, Italy 22–26 September 2008 (ESA SP-666, November 2008) Using the BRDF for the sites we can show that calibration systems of ATSR-2 and MERIS exhibit good long term stability. The results for AATSR, show excellent long term stability for the 1.6μm channel, but long term drift in the 0.87μm, 0.67μm and 0.56μm channels, see Fig 1. The latter result is confirmed by the trend where MERIS Figure 1: Time series of AATSR reflectances normalised to the reference BRDF for all desert sites (red), Greenland (blue) and Dome-C (cyan). Figure 2: Time series of AATSR reflectances normalised to MERIS refelectances for all desert sites (red), Greenland (blue) and Dome-C (cyan). Note that the measured biases obtained from early intercomparisons have been removed. For the early phase of the mission the calibration trend followed an exponential decay function as predicted from the experience of ATSR-2, AVHRR etc. such that D(t) = exp(-kt) as shown in fig 3. Figure 3: 0.56μm calibration drift for the first two years of operation showing the exponential drift model. After 2005, the data shows that the exponential decay model was incorrect in the case of the AATSR visible channels. The observed long term trends suggest that the drift is caused by a thin-flim interference effect (Etalon) of the form D(t) = 1+Asin2(2πnxt/λ) (where n is the refractive index and x is the thickness) as shown in fig 4. Figure 4: 0.56μm calibration drift for the first two years of operation showing the thin film drift model. Although the ‘thin film’ model provides a reasonable representation of the observed drift, analysis of most recent data shows that the model underestimates the trend when using the current parameters. Also, the model does not reliably predict the future evolution of the long term drift. Therefore, instead of attempting to fit a parametric function to the data, it is proposed to use a smoothed average of the measurements. The smoothing function given by ( ) ) ( 1 2 / 2 / t D N t D width i