Atmospheric Correction for multiple Landsat-8 scenes

Question

I need to perform an atmospheric correction for Landsat-8 data and can't find a solution.

There is a very limited amount of information dealing with L8-correction methods in a comprehensive way, that's why I created this post. There's some existing questions here on StackExchange dealing with details related to this post but they are outdated. E.g: "how can i correct atmospherically Landsat 8 images?" considers i.atcorr as a solution which doesn't work anymore (see below). The answers to "How to atmospherically correct Landsat 8 data?" deal with single image Dark Object Subtraction which is not suitable due to reasons given below. (Sorry for not providing direct links, StackExchange prohibits this for new users)

So far I have tried:

• ATCOR toolbox in ERDAS - costs a lot of money and, having worked with it for several years, I'm not content with it's performance
• i.atcorr in Grass 6.4 does not support Landsat-8
• i.atcorr in Grass 7.0.3 does not work
• i.landsat.atcorr does not work and is not updated anymore
• ESPA-SR is very complex and I fail to set it up properly or run any commands. github.com/USGS-EROS/espa-surface-reflectance
• Dark Object Substraction is not feasible for a large amount of scences (as it requires user interaction) and has a questinable performance

Does anybody have other suggestions or solutions?#

Edit/Update: A relevant post: Open Source Alternatives to ERDAS IMAGINE ATCOR?

Answer 1

An open source package for atmospheric correction of a number of sensors, including Landsat 8, is ARCSI (http://rsgislib.org/arcsi/) from the Earth Observation and Ecosystem Dynamic group at Aberystwyth University.

It is written in Python (with some functions written in C++) and uses the 6S codes through Py6S. It is a command line tool and has some functions for processing a lot of scenes at once.

Basic usage would be something like:

arcsi.py -s ls8 -f KEA --stats -p RAD SREF --aeropro Maritime \
--atmospro MidlatitudeSummer --aot 0.25  -o ./OutputImages \
-i LC80430342013291LGN00/LC80430342013291LGN00_MTL.txt

More details are available in the following post: https://spectraldifferences.wordpress.com/2014/05/27/arcsi/

Binaries are only available for Linux and OS X (see details in above post) so if you need to run on Windows you would need to do so through a virtual machine.

One thing to point out is there isn't much support available if you do have problems with it. Also some level of UNIX / Python skills are required - in particular if you want to process multiple scenes.