What’s new in Py6S
The last few months have seen a flurry of activity in Py6S – probably caused by procrastinating from working on my PhD thesis! Anyway, I thought it was about time that I summarised the various updates and new features which have been released, and gave a few more details on how to use them.
These have all been released since January 2014, and so if you’re using version 1.3 or earlier then it’s definitely time to upgrade! The easiest way to upgrade is to simply run
pip install -U Py6S
in a terminal, which should download the latest version and get it all set up properly.Â So, on with the new features.
A wide range of bugfixes
I try to fix any actual bugs that are found within Py6S as soon as they are reported to me. The bugs fixed since v1.3 include:
- More accurate specificiation of geometries (all angles were originally specified as integers, now they are specified as floating point values)
- Fixed errors when setting custom altitudes in certain situations – for example, when altitudes have been set and then re-set
- Fixes for ambiguity in dates when importing AERONET data – previously if you specified a date such as 01/05/2014 which could be interpreted either day-first (1st May) or month-first (5th January) then it assumed month-first, which was the opposite what the documentation specified. This now assumes day first – consistent with the documentation
- Error handling has been improved for situations when 6S itself crashes with an error – rather than Py6S crashing it now states that 6S itself has encountered an error
- Added the extraction of two outputs from the 6S output file that weren’t extracted previously: the integrated filter function and the integrated solar spectrum
Parallel processing support
Now when you use functions that run 6S for multiple wavelengths or multiple angles (such as the
run_vnir functions) they will automatically run in parallel. From the user’s point of view, everything should work in exactly the same way, but it’ll just be faster! How much faster, depends on your computer. If you’ve got a dual-core processor then it should be almost (but not quite) twice as fast. For a quad-core then it will probably be around three times faster, for an eight-core machine then it will probably be more like five times as fast. If you want to experiment then there is an extra parameter that you can pass to any of these functions to specify how many 6S runs to perform in parallel – just run something like:
run_landsat_etm(s, 'apparent_radiance', n=3)
to run three 6S simulations in parallel.
I’ve tested the parallel processing functionality extensively, and I’m very confident that it produces exactly the same answers as the non-parallel version. However, if you do run into any problems then please let me know immediately, and I’ll do whatever fixes are necessary.
Python 3 compatibility
Py6S is now fully compatible with Python 3. This has involved a number of changes to the Py6S source code, as well as doing some alterations to some of the dependencies so that they all work on Python 3 too. I don’t use Python 3 much myself, but all of the automated tests for Py6S now run on both Python 2.7 and Python 3.3 – so that should pick up any problems. However, if you do run into any issues, then please contact me.
Added wavelengths for two more sensors
Spectral response functions for Landsat 8 OLI and RapidEye are now included in the
PredefinedWavelengths class, making it easy to simulate using these bands by code as simple as:
s.wavelength = Wavelength(PredefinedWavelengths.LANDSAT_OLI_B1)
I’m happy to add the spectral response functions for other sensors – please email me if you’d like another sensor, and provide a link to the spectral response functions, and I’ll do the rest.
I’ve got lots of plans for the future of Py6S. Currently I’m finishing off my PhD, which is having to take priority over Py6S, but as soon as I’ve finished I should be able to release a number of new features.
Currently I’m thinking about ways to incorporate the building of Lookup Tables into Py6S – this should make running multiple simulations far quicker, and is essential to use Py6S for performing atmospheric corrections on images. I’m also considering a possible restructuring of the Py6S interface (or possibly a separate ‘modern’ Pythonic interface) for version 2.0 or 3.0. I’m also planning to apply to the Software Sustainability Institute Open Call, next year, with the aim of developing the software, and the community, further.
Categorised as: Academic, Programming, Py6S, Python, Remote Sensing
Thanks so much for this powerful explanation.
I have one question about the Py6Sã€‚
How could I add the spectral response functions for GaoFen-1?
The best thing to do is post a question on the Py6S Mailing List at https://groups.google.com/forum/?fromgroups#!forum/py6s and I’ll get back to you there.
Basically you’ll need to provide a spectral response function for the sensor at 2.5nm resolution – and then that can just be used directly in Py6S. Have a look at the code in wavelength.py (eg. here: https://github.com/robintw/Py6S/blob/master/Py6S/Params/wavelength.py#L110) to see how this is done for other sensors.
Thank you for your Py6S.
Let me have a question.
Did you incorporate LUT into Py6S?
Hi – unfortunately not yet, sorry.