Toward a synergy between onorbit lunar observations Sophie
Toward a synergy between on-orbit lunar observations Sophie Lachérade CNES GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Lunar observations The SADE Lunar database at CNES: SENSOR Spectral range Nb of spectral bands Spatial resolution Acquisition Phase angle Number of Dates range measurements PHR-1 A Vis-Nir 4 2. 80 m 2012 -2013 [-115°; 115°] 166 PHR-1 B Vis-Nir 4 2. 80 m 2013 [-115°; 115°] 970 AQUA/MODIS Vis-Nir 7 250 -500 m 2002 -2013 [51°; 55°] 108 MSG 1 Vis-Nir-Swir 3 2500 m 2003 -2012 [-150°; 152°] 393 MSG 2 Vis-Nir-Swir 3 3000 m 2006 -2013 [-145; 145] 366 LANDSAT 8* Vis-Nir-Swir 8 30 m 2013 -7° and +8° 148 -> A lot of sensors with different characteristics. One common thing: there able to look at the Moon ! PHR: Pleiades High Resolution 2 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT *Results on-going !
Lunar observations – Spectral response PHR 1 A PHR 1 B MSG 1 MSG 2 MODIS LANDSAT 8 3 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Lunar calibration – But what kind of calibration ? The current lunar reference, internationally used, is ROLO. Different calibration methods could be perfomed, based on lunar acquisitions, all needed the ROLO model : - Stability monitoring - Inter-band calibration - Absolute calibration - Cross-calibration 4 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Multi-temporal calibration Goal: guaranty the stability of the sensor better than 1% PHR 1 B_B 0 All phases MSG 1_VIS 06 All phases -> Limitation of the ROLO model to take into account the phase angle. -> Can be bypassed by using a restricted phase angle (PHR: 40°, MODIS: 55°, LANDSAT 8: 7°) 5 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Multi-temporal calibration After the phase angle selection: MSG 1_VIS 06 Phase: -40°± 5° PHR 1 B_B 0 phase: ± 40° MODIS_555 phase: 55° Good accuracy of the method for one chosen phase ! 6 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Multi-temporal calibration After the phase angle selection and adjustment of the temporal range: PHR-1 A_B 2 phase: 40° MODIS_555 phase: 55° 2 years of PLEIADES versus 11 years of MODIS -> Seasonal cycles observed both on MODIS and PLEIADES with similar shape and level. Is it a residual effect of the lunar librations modelised by ROLO? 7 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Interband calibration Goal: estimation of a simulated irradiance in one band knowing the irradiance in another band the albedo of the Moon in these two bands. Accuracy to be achieved: better than the absolute one -> The accuracy of the method depends on the relative spectral accuracy of the albedo in the two bands Which ROLO to use ? Ref: The spectral irradiance of the Moon H. H. Kieffer and T. C. Stone The astronomical Journal, 129: 2881 -2901 2005 June Apollo correction: Fig 8 8 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Interband calibration – Phase dependence Calibration of the Blue band of PHR 1 B (B 0) from the other spectral bands of PHR 1 B All phase angles PHR 1 B [-70°; 70°] -> The phase dependence of ROLO has less influence on the interband calibration results than on the multi-temporal calibration results. 2% versus 5% for phase [-100°; 100°] This allows us to perform an accurate interband calibration whatever the phase ! -> Useful for sensors which cannot choose their acquisition phase angle 9 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Multi-temporal calibration Goal: guaranty the stability of the sensor better than 1% PHR 1 B_B 0 All phases MSG 1_VIS 06 All phases -> Limitation of the ROLO model to take into account the phase angle. -> Can be bypassed by using a restricted phase angle (PLEIADES: 40°, MODIS: 55°, LANDSAT 8: 7°) 10 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Interband calibration – Phase dependence Calibration of the red band of MSG 1 (VIS 06) from MSG 1 VIS 08 MSG 1 All phase angles MSG 1 [-70°; 70°] -> The conclusion is less obvious on the MSG dataset but the residual dispersion may be due to the integration step at high phase angles -> When limiting phase angles to [-70°; 70°], the conclusions are the same than for PHR : method with very few dispersion ! 11 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Interband calibration – Spectral dependence Interband calibration results for PHR 1 B (phase angles range: [-70°; 70°]) and MODIS: PHR 1 B AQUA/MODIS The results highlights a dispersion depending of the reference band which is used. This dispersion is up to 3% but the absolute accuracy of the results strongly depend of the spectral knowledge of the Moon albedo. -> Necessary to compare these results with other calibration methods to estimate the accuracy of the method: -> very good consistence of the results other methods 12 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Lunar calibration – But what kind of calibration ? The current lunar reference, internationally used, is ROLO. Different calibration methods could be perfomed, based on lunar acquisitions, all needed the ROLO model : - Temporal stability observation - Inter-band calibration - Absolute calibration - Cross-calibration 13 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Absolute calibration ROLO versus on-orbit measurements -> AQUA/MODIS, PHR 1 A and PHR 1 B show a very good agreement PHR-1 A and PHR-1 B absolute calibration based on ENVISAT/MERIS -> Uncertainty of the absolute calibration of the ROLO model up to 10% at 55° 14 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Cross-calibration results PHR-1 B versus AQUA/MODIS: Final accuracy: ± 3% - Retrieve calibration differences between MERIS and MODIS observed on desert sites (Lachérade et al. , IEEE, 2013) - Uncertainties for Moon cross-calibration due to the lack of correct spectral interpolation between PHR and MODIS (cross-calibration limited to a band-to-band cross-calibration) - Estimation of the interband accuracy based on Moon acquisitions ≈ 3% 15 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Cross-calibration results Where is the truth ? Need of a lunar reference in the SWIR (LANDSAT 8 – VIIRS ? ) MSG 2 versus AQUA/MODIS Very good agreement between cross-calibration results using the Moon (PHR 1 B and MODIS) and results obtained over desert sites. Dispersion of the interband calibration results using the SWIR band of MSG. 16 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
Lunar observations – Perfect sensor ? A perfect lunar reference sensor should be characterised by the following parameters: - its spatial resolution (PHR-LANDSAT 8) A image of the Moon with a lot of pixels will decrease the uncertainties when integrating its irradiances - the range of phases it has/will observe(d) (PHR-MSG) To be sure to take into account the phase effect, it is better to cross-calibrate sensors using the same phase of the Moon - its spectral resolution in term of band width and spectral range (SWIR range needed to constraint the spectrum) (LANDSAT 8 -VIIRS? ) The ideal sensor is characterised by a large number of spectral bands covering the full spectrum from 400 nm to 2500 nm - the accuracy of its absolute calibration (MODIS-LANDSAT 8 -VIIRS? ) No current sensor corresponds to this ideal one !!! It is necessary to take advantage of several of them to derive a perfect lunar reference 17 GSICS Annual Meeting – 24 -28 March 2014 - EUMETSAT
CONCLUSIONS To do list: - Improvement of the lunar cross-calibration method by implementing an interpolation function in the MUSCLE/SADE tools - Analyses of the LANDSAT 8 lunar dataset (VIS-NIR-SWIR bands) - Cross-calibration and interband calibration of all the sensors to better discriminate uncertainties linked to the lunar albedo and uncertainties linked to the methods - Improvement of the reference lunar model (based on available datasets): PLEIADES MSG MODIS / VIIRS ? LANDSAT 8 Phase angle Effect ROLO 18 Spectral range Inter-band calibration Absolute calibration Reference lunar Model
- Slides: 18