The objective of this project is to build upon previous work in finding and categorizing flares in data from the Kepler telescope.
Sponsored By: PSU Department of Astronomy and Astrophysics
Team Members
Ben Wortman | Elijah Reber | Shao Hui Lee | Nick Gabrovsek | Andrew Casey | | | | | | |
Project Poster
Click on any image to enlarge.
Watch the Project Video
Download the Project Summary
Project Summary
Overview
From 2009-2013, NASA operated the Kepler mission that monitored the brightness of ~150,000 stars. Although the primary mission was to search for exoplanets, previous work has been able to identify flare events on approximately 4,000 stars. Studying these flares is of interest to astronomers as it helps them better understand how the habitability of planets orbiting these stars. Our project was to repeat the previous study and also improve upon their results to expand the existing catalogue of flares.
Objectives
We had two primary objectives:
– Work towards recreating Richard Davenport’s dissertation in finding flares in this data
– Develop new methods using existing libraries in R
Approach
Team 1: Repeating Davenport’s Study
– Our sponsor wanted us to work to reproduce results of ~4,000 flaring stars that Davenport had found in the Kepler data.
– The replication mostly consisted of signal processing techniques, such as subtracting out long-term and periodic dependencies through polynomial and sine curve fitting.
– The progress made can be seen in the reduction of the inter-quartile range for the light curves, as well as visually seeing the reduction of random peaks.
Team 2: Developing New Methods
– Our sponsor provided us with ARIMA residuals to work with. ARIMA is a time series model used to reduce noise in the data.
– We tested several outlier detection libraries against a set of stars with known flare activity to see how effective they were.
– The outliers detected were normally distributed and had a slight negative bias which is the opposite of what we would expect to see for flare activity.
– After seeing this we developed synthetic data to better understand how ARIMA impacts the flare shape after being passed through the model.
– With this we could calculate the signal to noise ratio for flares of varying intensities and determine at what point these flares start getting “eaten” by the ARIMA model.
Outcomes
– The work done towards our first goal can be used by the sponsor to run a change point analysis on, in order to identify flare candidates.
– Additionally the sponsor will gain some insight into how ARIMA influences flare profiles passed through it and will have a pipeline for testing the profiles of other phenomena in the future.