DSCOVR Options

[scalbers]

Here is our "DSCOVR Transcendance" poster from AGU, highlighting Blueturn and the Simulated Weather Imagery. The authors included Michael Boccara, Jay Herman, and Zoltan Toth.

https://agu.confex.com/agu/fm17/meetingapp.cgi/Paper/232523

https://agu.confex.com/agu/fm17/mediafile/H...anscendance.pdf

[scalbers]

I've recently made some fixes to the handling of Rayleigh scattering in making simulated Earth images. This makes the simulated blue sky (over ocean areas) somewhat darker than previously. As a result less color adjustment to the DSCOVR web images is need to make a match. Hopefully this is now a more realistic color and contrast.



The left is simulated (ray-traced) from global 3-D weather and land surface data, right is a DSCOVR/EPIC image at the same time. In general my simulations come up a little short on the brightest cloud reflectance values. This can be partially addressed by adjusting the anisotropic reflectance factor for the bright cloud tops.

[scalbers]

Also for comparison, here is an image constructed from the DSCOVR calibrated counts data, converted to reflectance and then using my color processing algorithms to calculate the RGB values. The blue color looks somewhat brighter than in the DSCOVR web page images. This should be close to a true color / contrast astronaut view. The brightness is set to minimize clipping of the brightest white clouds. There's really an interesting variety of whiteness to the clouds with the deepest most opaque ones near the center being the brightest. Some additional processing would be needed to register the individual narrowband image locations better, considering the Earth's rotation.

[scalbers]

With some further adjustments I get this comparison that is hopefully a bit closer, including a slightly darker blue over the oceans. The land is now brighter relative to the scattered light in the atmosphere.

The ocean/sky blue is brighter in the image one post above for perhaps multiple reasons. One way to characterize this gap is to note that the simulated 551nm reflectance in the darkest ocean areas is about 4.4% compared with 5.0% observed. The blue channel and color saturation are also higher in the above post. The same dark ocean areas are observed to have ~12.4% reflectance at 443nm and the simulation has a lower value. This gap can be bridged by considering reflected sunlight from beneath the water surface, and also checking the reflected skylight from the surface.

[scalbers]

With this refinement to increase the reflected light from beneath the water surface we can see the comparison below.



The simulated image on the left is a closer match to the observed color processed version in post #153. The image on the right is from the DSCOVR website with an empirical color adjustment.

[scalbers]

To help with rendering the Earth, a guideline I've come across is that the chromaticity of the blue sky (due to Rayleigh scattering) should be about x=0.23 and y=0.23. This helps quantify the relative values of RGB in the clear sky portion of Earth images. This is assuming that the standard CIE color matching functions are correct in describing how this color is perceived.

A relation that I hadn't before known is that this "Rayleigh Blue" color is about the same color a star would have with an infinite (or very high) blackbody temperature. Thus on an imaginary planet orbiting a very early type O star a water cloud there would be about the same color as a haze free blue sky on Earth, instead of white.

[scalbers]

DSCOVR has been in safe mode since June 27th: https://spacenews.com/dscovr-spacecraft-in-safe-mode/

I wonder if an archive of L1B files is stil available in NetCDF or HDF format?

[scalbers]

A fix is being worked on: https://spacenews.com/software-fix-planned-...restore-dscovr/