Interagency Data Stewardship/LifeCycle/Preservation Forum/TeleconNotes/2018-09-17

Meeting Agenda - DS Committee - 2018-09-17 2 p.m. EST / 12 p.m. MST / 11 a.m. PT

Link to the Meeting Notes: 2018 DS Committee Notes

──────────────────────────────────────────

Please join my meeting from your computer, tablet or smartphone. https://global.gotomeeting.com/join/453694565

You can also dial in using your phone. United States: +1 (408) 650-3123

Access Code: 453-694-565

First GoToMeeting? Let's do a quick system check: https://link.gotomeeting.com/system-check

───────────────

1. ESIP Program Committee & other updates - Matt Mayernik

2. Invited presentation - Hylatis, for Hyperspectral Image Analysis - Anne Wilson (LASP), Doug Lindholm, Chris Lindholm, Peter Pfister, the LASP Web Team

Presentation slides
Abstract: Hyperspectral imagery datasets are extremely large and their usage is only growing. Today, an analysis could take 100TB of space. Hylatis is a NASA AIST 2016 project addressing the need for tools to effectively handle these massive data products by developing a framework for efficient hyperspectral imagery analysis in the cloud. As a demonstration of the framework, and in the spirit of taking the computation to the data rather than the other way around, Hylatis will load four disparate multi and hyperspectral datasets into the cloud, allowing multiple simultaneous users to interact with a single shared instance of each. This capability could enable data-related centers to host these large datasets and provide access to many users in an efficient way. There are common server side operations such as subsetting on variables, time, location, wavelength, and other dimensions, as well as statistical methods, filtering, etc. Any algorithms can be implemented. The creation of new products across datasets via simple interpolation will demonstrate the ability to support arbitrary data fusion algorithms. Hylatis achieves interoperability by representing datasets mathematically and adhering to mathematical principles to manipulate data. That is, a dataset is represented as a function of independent and dependent variables. (Consider temperature as a function of time and location.) Operations on datasets are essentially function transformations. Being written in a functional programming style, these benefits accrue: thoughtful, rigorous, up front development based on mathematical principles produces code that is more correct, easier to maintain, and also more easily parallelizable. This talk will give a high level view of the Hylatis framework and progress to date, lessons learned, etc.

3. Data citation recommendation update - Mark Parsons

Presentation by Adam Shepherd (WHOI) on Schema.org, EarthCube project P418