**POSTER**

**Title:
**Hylatis, A Platform for Hyperspectral Image Analysis in the Cloud

**Presenting Author: **Anne Wilson

**Organization: **University of Colorado, LASP

**Co-Author(s): **Doug Lindholm, Odele Coddington, Peter Pilewskie

**Abstract:
**Hylatis is an AIST 2016 award to develop a cloud-based platform for hyperspectral imagery analysis. Hylatis is building foundational capabilities that leverage big data engines to subset hyperspectral imagery on time, spatial range, pixel and wavelength range, and apply operations over them. We demonstrate platform capabilities uniform interface over four disparate datasets: HySICS, GOES, MODIS, and POLDER with simple data fusion operations.

Hylatis explores the deeper informatics question of modeling scientific data. Data representation is foundational to any analysis system and impacts how easy or hard it is to perform arbitrary operations. A foundational principle of Hylatis data modeling and operations is to adhere to mathematical principles. Hylatis models data using a domain agnostic, mathematical function of domain and range variables. Hylatis basic types are scalars, tuples, and functions. Basic operations are mathematical operations, such as selection, projection, join, groupBy, pivot, transpose, currying, and function composition.

Higher lever abstractions are built on this basic layer. For Hylatis, that is in the form of a domain specific language (DSL) for hyperspectral imagery analysis. (A DSL is a language that provides operations specific for a domain.)

Hylatis leverages functional programming. Functional programming is a valuable technique for science because when its rigor is followed the code that is produced is easier to understand, maintain, show correctness, and parallelize.