J1.1 Automated Climate and Weather Data Analysis with "aospy"

Tuesday, 24 January 2017: 1:30 PM
Conference Center: Chelan 5 (Washington State Convention Center )
Spencer A. Hill, Princeton University, Princeton, NJ; and S. Clark

Handout (2.6 MB)

The vast amount of data generated by weather and climate models often makes performing all of one's desired analyses impractical without some automation, but such efforts are routinely stymied by idiosyncrasies such as model-specific variable names or differing simulation timespans.  Even once automation is attained, the resulting output quickly becomes too large to manage without some well-structured organizational methods.  We present "aospy", an open-source Python library (https://github.com/spencerahill/aospy) that attempts to remedy these problems.

The aospy package separates user code pertaining to any model- or simulation-specific metadata from that used to compute physical quantities, enabling users to write their scientific functions in a general way that is easy to write, read, and share.  Once the user's data characteristics have been described using aospy's simple data structures, a simple user-interface enables specifying a nearly arbitrary combination of simulations, physical quantities, time periods, and many other options.  aospy then iterates over the specified combinations and outputs the resulting data into a database for later retrieval.  In addition to improving user's climate and weather data workflows, aospy promotes open science and reproducible research by making scientists' code more modular and shareable and by imbuing the resulting output with more metadata specifying how it was generated.  We present visions for aospy's future and welcome both new users and contributors.

Supplementary URL: http://aospy.readthedocs.io/en/latest/

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner