Identifying Clouds in Panoramic SETI Data with Machine Learning

The Panoramic SETI (PANOSETI) observatory offers the capability to instantaneously observe 4,450 square degrees for optical transients occurring between sub-nanosecond-to-second timescales. This observatory will greatly enlarge the current SETI phase space by increasing sky area searched, wavelengths covered, number of stellar systems observed, and duration of time monitored. However, a consequence of PANOSETI’s large solid angle is a high chance of observing sources of interference such as clouds, aircraft, and LIDAR satellites, resulting in contaminated data that must be discarded. Additionally, daily data volumes on the order of terabytes make manual identification of this contaminated data infeasible, implying an automatic approach is required. Here, we present a machine learning system capable of identifying the vast majority of contaminated data in PANOSETI observations.