GRWG VIS/NIR Sub-Group Web Meeting 2014-10-22
Review Priorities & Combining Methods
- Bertrand Fougnie (CNES) Combining VIS/NIR Methods
- Fangfang Yu (NOAA) An Integrated Calibration Method to Improve the GOES Visible Calibration Accuracy
- Arata Okuyama (JMA) Himawari-8 launch and its calibration approaches
- Dave Doelling (NASA) Reviewing Priorities for which VIS/NIR methods to develop into GSICS products
Guest Chair: Dave Doelling (NASA)
CNES: Bertrand Fougnie
EUMETSAT: Sebastien Wagner, Tim Hewison, Bartolomeo Viticchie, Viju John
JMA: Masaya Takahashi, Arata Okuyama, Keita Hosaka, Ryoko Yoshino
NOAA: Fred Wu, Fangfang Yu, Manik Bali
RAL: Dave Smith
Bertrand and Fangfang each presented their combining methods using multiple calibration approaches. Lots of very good discussion resulted. For example, it was suggested that ozone in the visible channel caused the DCC seasonal oscillations. Also, several suggestions were offered to Fangfang to predict the Sonoran desert outliers during ENSO events, including NDVI and MODIS weekly BRDF models.
Essentially the difference of Bertrands and Fangfang combination strategy is that Bertrand is more inclined to look for the best method for trending (other methods being used for confirmation) and apply it to all methods to examine the absolute calibration difference, whereas Fangfang uses all approaches simultaneously to remove outlier monthly gains, such as over deserts, which have their unique natural variability.
The goal of combining methods is to reduce the uncertainty of the combined calibration from each of the individual method uncertainties. Before attempting to combine methods the uncertainty of each method needs to be understood for each band and each configuration (i.e. sensor). Uncertainty budgets have to consider not only the theoretical uncertainty of each method, but also as much as possible the impact of radiometric artifacts coming from the sensor.
Since the monthly trending noise (type-A) is usually greater than the absolute calibration differences (type-B) between methods, the trend needs to be determined first, and deseasonalisation applied before finally deriving the absolute difference.
Bertrand pointed out that there is no overall combination strategy that works for every band, every sensor and method. Each band and method has its own limitations. Unexplainable differences between calibration methods maybe instrument related, such as sensor nonlinearity, say between DCC and dark targets, stray-light, polarization, and non-gray spectral response degradation. Spectral band degradation may explain drifts in the long-term trend as the satellite ages.
The combining of methods will be easier once a long enough time record of at least two years is known. Real-time calibration monitoring is not achievable with most methods, especially to determine the timing of an instrument anomaly that effects the calibration. For this reason, it may be that the GSICS Near-Real-Time and Re-Analysis Corrections for the VIS/NIR channels could be based on different methods (or combinations of methods).
Tim encouraged the group to consider whether the concept he developed to combine results from multiple references could be applied to combining multiple methods. (See previous web meeting
for details.) In this case, one method would be defined as the Primary
method, and results from other methods adjusted to be consistent with it, before merging the results together.
It was mentioned to follow up during the annual meeting about combining methods. However, given the priority of completing the DCC and lunar calibration, in order to have two methods combine, it will be the VIS/NIR sub-group's second priority at the annual meeting.
Himawari-8/AHI Calibration Update
Arata presented JMAs approach to calibrate their new Himawari-8 AHI imager. For the visible they plan on validating their onboard calibration using DCC, lunar, RTM, Ray-matching with VIIRS and GEO to GEO. JMA showed preliminary daily MTSAT-2/NPP-VIIRS ray-matched radiance pair regressions over clouds. Arata notes that the greatest challenge in VIIRS ray-matching is the VIIRS HDF file sizes. The new Himawari-8 reached geostationary orbit located at ~140 E.
Dave presented the survey results. The GEO satellite priority was consistent between GPRC and personal research priorities. The current GEOs were top priority, followed by future and historical. The consensus on calibration methods was to finish the DCC and lunar methods first, followed by SNO/ray-matching to either reference broadband or hyper-spectral imagers. For new methods, RTM, desert, and Rayleigh in no clear particular order had a greater priority than polar ice, sunglint or star calibration.
Dave thought that the number of respondents were few, but there was representation from most GRPC, and considered representative for GSICS. Fangfang mentioned that the survey was not clear that all GISCS members could vote, rather than one vote per GPRC. Sebastien mentioned that his GPRC is responsible to provide scope-CM calibration of historical GEOs. The DCC and lunar VIS calibration should be applicable to historical GEOs. The VIS/NIR group will discuss this at the next annual meeting.
Forthcoming Web Meetings
Sebastien is still on track to set up a lunar preparation workshop for the lunar workshop Dec. 1-4, 2014. Dave will set up a DCC BRDF/Seasonal variability web meeting and preparation for the annual meeting VIS/NIR web meeting in January. Tim would follow up and set up a joint UV and VIS/NIR hyper-spectral visible inter-calibration GSICS web meeting before the next annual meeting.