The post-processor code is a bit of a mess. This update addresses some of the most obvious problems/confusing aspects of the code, while adding some useful new functionality:
(a) Code should be generally more readable, better commented and easier to understand now.
(b) Removed duplication of Bayesian selection code (and depreciated USE_NEW_BAYESIAN_SELECTION option).
(c) Only one input file worth of data is stored in memory at a time, regardless of which selection approach is used.
(d) Added the ability to run both Bayesian (for aerosol) and Pavalonis (for cloud) selections (for combining both cloud and aerosol files into a single product).
(e) Added the option to apply a simple weighting to the Bayesian selection method. The weights are passed via the BAYESIAN_WEIGHTS option in the driver file, and are treated as an a priori probability for each phase/type.
The post-processor code is a bit of a mess. This update addresses some of the most obvious problems/confusing aspects of the code, while adding some useful new functionality: (a) Code should be generally more readable, better commented and easier to understand now. (b) Removed duplication of Bayesian selection code (and depreciated USE_NEW_BAYESIAN_SELECTION option). (c) Only one input file worth of data is stored in memory at a time, regardless of which selection approach is used. (d) Added the ability to run both Bayesian (for aerosol) and Pavalonis (for cloud) selections (for combining both cloud and aerosol files into a single product). (e) Added the option to apply a simple weighting to the Bayesian selection method. The weights are passed via the BAYESIAN_WEIGHTS option in the driver file, and are treated as an a priori probability for each phase/type.