What does COMDAST do?

The Climate Observations and Model Data Analysis and Synthesis Toolkit (COMDAST) is a collection of custom C programs written and devleloped by Jitendra Kumar and Bjorn Brooks for climatologists and modelers to upscale or downscale meteorological variables including shortwave and longwave radiation, precipitation, temperature, wind and more. COMDAST was initially created to do site-level temporal downscaling, but we also include in our repository R code, shell scripts, and example data for doing a regional space & time downscaling. COMDAST was created to be easy-to-use on any LINUX system and does not require programming expertise. Note however, that this is 'research code' and that to use COMDAST you should be familiar with working from the command line in a LINUX environment.

COMDAST is a machine learning algorithm that is simple and fast because it is only concerned with developing a statistical model based on patterns in a training dataset and using that model to extrapolate a limited number of values to a target resolution. COMDAST requires only two things:

  1. a series of training data in ASCII text (e.g., 50 years of continuous hourly temperature)
  2. a series of testing data in ASCII text (e.g., 500 years of monthly temperature).

The output for this example case would be 500 years of hourly temperature that fit the 500 years of monthly means used in training.

Although we originally created COMDAST to downscale CMIP5 GCM output, you can strip-out parts of our code and customize it to downscale any data set. Here is a schematic overview of how COMDAST works: 

   CRUNCEP                                                --------------
   NetCDF       ---------                \           \   | Mean         |
   Training____| Shell   |    -------    / normalize /   | Conservation |    --------
   Data        | Scripts |___| ASCII |___\   and     \___| Analysis     |___| NetCDF |
               | for     |   | Input |   /  neural   /   | ane Post     |   | output |
   Coarse______| Parsing |   | Data  |   \  networks \   | Processing   |    --------
   CMIP5       | Data    |    -------    /  (C code) /   | (R code)     |
   NetCDF       ---------                \           \    --------------
   Data

To inspect some example before-and-after data download and un-compress one of the following files: example_output.zip or example_output.tar.gz (9 MB). Inside are 2 NetCDF files both representing 1 year of temperature data over Eastern North America. One file has the original 2°, monthly mean air temperature, and the other is the downscaled 0.5°, 6-hourly version.

Usage permissions, license and copyright

Permission to use, copy, modify, and distribute this software and its documentation under the terms of the GNU General Public License is hereby granted. No representations are made about the suitability of this software for any purpose. It is provided "as is" without express or implied warranty. See the GNU General Public License for more details.

COMDAST has been developed for research purposes and is under active development. Authors are happy to share the code with anyone interested under the terms of the GNU General Public License without any gurantee or support.