Re: ccplot profile vs. CALIOP browse image profiles

[Peter K. <pet...@wa...>]

On 04/03/2015 01:30 PM, Andrew Prata wrote:
> Hi,
> 
> Thank you for creating such a wonderful piece of software and for
> keeping all of the code open source!
> 
> I’ve been trying to generate CALIOP profiles that best match the NASA
> browse image accessible
> here: http://www-calipso.larc.nasa.gov/products/lidar/browse_images/production/
> 
> ccplot has come the closest! However, as with my own script, there is a
> lot of “noise” from ~ 0–8.2 km, that is not apparent in the NASA browse
> images.
> 
> Is there a way to overcome this?  Have the NASA browse images had some
> sort of smoothing or masking applied?  If so, is there any documentation
> on this?
> 
> I’ve attached some figures to demonstrate the problem.  While ccplot
> does a good job representing large profiles, there is generally too much
> noise to identify cloud/aerosol signals located within small latitude
> ranges and at low altitudes (0–8.2 km).
> 
> Kind Regards,
> 
> Andrew
> 
> 
> --
> Andrew Prata
> 
> PhD Candidate
> School of Earth, Atmosphere & Environment
> 9 Rainforest Walk
> Monash University, VIC 3800
> 
> Mobile: +61 414 873 225
> Office Number: 224

Hi Andrew,

At the moment ccplot performs nearest-neighbor interpolation, which
means every value displayed in the plot is an actual value from the
data. The NASA browse images likely use some sort of averaging, but I'm
not sure what the exact algorithm is.

The ccplot API (http://ccplot.org/api/) provides a function
ccplot.algorithms.interp2d_12 which does averaging when zooming out, and
nearest-neighbor interpolation when zooming in. In my experience, the
results of this interpolation are closer to the NASA produced images. I
would like ccplot to switch to the new algorithm in future. Since you
asked, I have done the modification in the latest revision, available
from github:

https://github.com/peterkuma/ccplot/archive/master.zip

You can try it out, but I can't guarantee it gives good results in all
situations, as I haven't done much testing. If you use Windows, I will
prepare a build for you (the code from github needs to be compiled).

Attached is a plot produced with the new algorithm. It is still not as
clear as the original, though. Tuning the colormap might help a little.
It was produced with the command:

ccplot -o calipso532.png -d 80 -c calipso-backscatter.cmap -a 90 -x
4:09:13..4:22:42 -y ..30000 calipso532
CAL_LID_L1-ValStage1-V3-01.2007-06-12T03-42-18ZN.hdf

Regards,

Peter