GRWG/GDWG Web Meeting 2014-06-05
GRWG+GDWG Web Meeting on DCC: File formats + Plotting
- Sebastien Wagner (EUMETSAT) - Towards a GSICS DCC product & Plotting Tool Requirements
- Masaya Takahashi (JMA) Plotting Tool & netCDF formats for GEO-LEO VIS NRTC & RAC using DCC
- Dave Doelling (NASA) ATBD and Dataset to generate SBAFs
- All - Discussion
- Sebastien Wagner (EUMETSAT) - How to handle the seasonal cycle - looking forward - postponed until next meeting
Guest Chair: Dave Doelling (NASA)
CNES: Bertrand Fougnie
EUMETSAT: Sebastien Wagner, Tim Hewison, Pete Miu, Rob Roebeling, Bartolomeo Viticchie, Viju John
JMA: Masaya Takahashi, Arata Okuyama, Ryoko Yoshiro, Keita Hosaka
KMA: Minjin Choi
NASA: Dave Doelling
NOAA: Fangfang Yu, Manik Bali
Sebastien presented his DCC plotting presentation.
Dohyeong Kim from KMA has indicated by email last month that they will also follow the GSICS DCC calibration approach.
There was a discussion of the frequency of the netCDF calibration files. Pete pointed out that the frequency of the netCDF file updates is more of a user requirement. Each GRPC will determine for their domain the range of the moving window to perform the DCC PDF and to compute the temporal trend coefficients based on the frequency of the rolling window udpates.
Bertrand pointed out that uncertainty for daily analysis is larger than for longer time periods. Sebastien pointed out that rolling windows are used in order to facilitate shorter update times. Fangfang pointed out that the GEO domain will determine the rolling window temporal range.
The biggest challenge is the transfer of the Aqua-MODIS calibration through DCC. Ultimately, the optimum method may be DCC ray-matching, which work is still in progress at NASA. The other method of performing the same DCC analysis using Aqua-MODIS as the GPRC GEO will get more complicated as more analysis is performed such as deseasonalization, as Sebastien has pointed out.
The current thought is to provide blended calibration coefficients, when multiple calibration methods are available and to provide the temporal trend equation with coefficients.
The current DCC method will concentrate on the 0.65µm channel to focus on providing a demonstration product in the near term. However, the next generation of GEOs will have multiple visible channels. DCC works best from 0.3-1.0µm and will be validated, however, for wavelengths greater than 1µm ice particle dependencies are quite variable and its accuracy and stability are greatly diminished.
Tim suggested and there was a consensus that the either the degradation or the gain time series for the target instrument (as shown on Sebastiens slide 8) is the minimum plot to provide users (as soon as all the information is available in the file, both can be derived). It displays both the DCC rolling window updates and the line that identifies the temporal trend equation. Fangfang and Dave suggested that the gain plots have more user interest. Masayas slide 6 plot from Fangfang would be the plot when multiple methods become available for the blended plotting package. Most of other plots were considered to be either validation or QC plots for the individual GPRCs.
At the end of the meeting the visible calibration coefficients will be adjustments made to the pre-launch or nominal calibration coefficients. This then addresses the question of whether to provide radiance or reflectance coefficients. Whatever units the nominal calibration incorporates would then answer this question. The nominal coefficients are static and will be included in the global attributes. It would follow that the degradation of the nominal calibration would be the final plot.
Masaya presented the netCDF strawman.
There was considerable discussion on the file naming convention of the netCDF file. Pete did mention that the filename should be kept short and that metadata should be contained inside the file, not in the filename. Dave mentioned that the visible algorithm type should not be stated in the filename. It would be confusing to have a list of methods in the filename. Masaya suggested that the spectral range should be listed in the filename. Minjin suggested providing only one file for all of the solar reflected bands.
Pete reconfirmed that grouping of variables or attributes is difficult in the current netCDF structure and that dumping the variables and attributes will be dependent on the software and not the way the netCDF file was created. Dave mentioned that the variable associated with the visible calibration to be prominent in the file somehow, to distinguish from attributes or variables of the methods, which will overwhelm the netCDF file. As long as the variables names are clearly defined in the user document the user should find them in the netCDF file. Dave mentioned that domain range that is DCC specific, without specifying it is a DCC attribute may be confused for the GEO domain range to apply the calibration coefficients and suggests that DCC method variables attributes have a DCC somewhere in the variable name. Pete suggested this information should go into the long name in order follow CF standards. It was further agreed that the domain range in the netCDF file shall be specified as the geographical area over which the product is applicable, not the area over which it was derived.
There was a discussion on whether the blended method or all method coefficients should be provided. It was agreed that a README or short user oriented document should be provided to help the user decide the correct method for their application, such as for the ocean color community. It was proposed to have in the product the blended calibration correction (compulsory) and the calibration correction from each and every method. This point is still subject to discussion and maybe addressed by giving the flexibility to each GPRC to have the detailed results or not.
In order to provide the user with information regarding the blending of the calibration corrections, a table indicating the weight (expressed as percentages) of each and every method in the composite correction could be added to the product for each channel.
Pete mentioned the use of DOIs to document the method and version, the DOI can point to the ATBD. DOIs also provide traceability of the method.
Dave the presented the DCC SBAF and apologized to CMA for not providing the SBAFs for the FY2 satellites. He mentioned that NASA is developing a tool for the community. Tim mentioned an option for users to upload their own SRFs. The tool will use the GSICS SRF page
as the official SRF until otherwise notified. Another improvement will be to use the Thuillier solar incoming. Dave mentioned that the tool is foreseen to come live in about one month time.
We look forward to Sebastiens deseasonalization presentation at the next DCC web meeting.