# AutoRegressive Planet Search: Feasibility Study for Irregular Time Series

## TL;DR

This study assesses the effectiveness of the AutoRegressive Planet Search (ARPS) method, combining ARIMA modeling, a Transit Comb Filter, and machine learning, for detecting exoplanets in irregular, noisy ground-based survey data.

## Contribution

It demonstrates that ARPS can be adapted to irregularly sampled ground-based light curves, predicting successful detection of planets with certain transit depths and periods.

## Key findings

- Detection rates improve for shorter orbital periods and deeper transits.
- ARPS can detect planets with transit depths greater than 0.1% and periods less than 40 days.
- Method shows promise for planet discovery in ground-based surveys with dense cadence patterns.

## Abstract

Sensitive signal processing methods are needed to detect transiting planets from ground-based photometric surveys. Caceres et al. (2019) show that the AutoRegressive Planet Search (ARPS) method --- a combination of autoregressive integrated moving average (ARIMA) parametric modeling, a new Transit Comb Filter (TCF) periodogram, and machine learning classification --- is effective when applied to evenly spaced light curves from space-based missions. We investigate here whether ARIMA and TCF will be effective for ground-based survey light curves that are often sparsely sampled with high noise levels from atmospheric and instrumental conditions. The ARPS procedure is applied to selected light curves with strong planetary signals from the Kepler mission that have been altered to simulate the conditions of ground-based exoplanet surveys. Typical irregular cadence patterns are used from the HATSouth survey. We also evaluate recovery of known planets from HATSouth. Simulations test transit signal recovery as a function of cadence pattern and duration, stellar magnitude, planet orbital period and transit depth. Detection rates improve for shorter periods and deeper transits. The study predicts that the ARPS methodology will detect planets with $\gtrsim 0.1$\% transit depth and periods $\lesssim 40$ days in HATSouth stars brighter than $\sim$15 mag. ARPS methodology is therefore promising for planet discovery from ground-based exoplanet surveys with sufficiently dense cadence patterns.