Sunday 2:50 PM–3:35 PM in Track 3

Building a User Maintainable Visualization Dashboard for the Large Synoptic Survey Telescope (LSST)

Dharhas Pothina

Audience level:
Novice

Description

Starting in 2023, the Large Synoptic Survey Telescope (LSST) project will image the entire visible sky every few days for ten years. When in production the LSST will generate 15 Terabytes of data per night. Critical to the success of this project is ensuring the quality of the data processing pipeline. We will demonstrate the complex multi page data visualization dashboard used by scientists.

Abstract

The Large Synoptic Survey Telescope (LSST) is the most ambitious ground-based optical astronomy project yet undertaken. A 3.2 gigapixel camera mounted on an 8-meter telescope in Chile will deliver deep, high-resolution images of the entire visible sky every few days for ten years. A large portion of the work leading up to the start of this survey (scheduled for 2023) has been developing the data analysis pipeline that will transform the raw images to useful catalogs of measurements of astronomical sources. Visual analysis of the catalogs produced by this pipeline from test precursor-survey datasets has been crucial to this development. Currently, this analysis is done by inspecting folders full of static plots. In this talk, we show how we built a new dynamic visualization dashboard that will be used by LSST scientists to explore the data much more efficiently. A key aspect of designing this dashboard was allowing scientists to compose desired visualizations in Jupyter and then move them into a complex dashboard view in a way that could be maintained and enhanced by non-javascript experts. This project is one of the more complex real world Panel applications built to date and has inspired several powerful and visually appealing templates that will be in the next major Panel release. The source code from the LSST dashboard along with a sample dataset will be available for folks to try out.

Subscribe to Receive PyData Updates

Subscribe