# Comparing Apples with Apples: Robust Detection Limits for Exoplanet High-Contrast Imaging in the Presence of non-Gaussian Noise

**Authors:** Markus J. Bonse, Emily O. Garvin, Timothy D. Gebhard, Felix A. Dannert, Faustine Cantalloube, Gabriele Cugno, Olivier Absil, Jean Hayoz, Julien Milli, Markus Kasper, Sascha P. Quanz  
**arXiv:** [2303.12030](https://arxiv.org/abs/2303.12030)  
**Date:** 2023-03-21  
**Summary page:** https://www.summarizepaper.com/en/paper/2303.12030v1/

## Key points

- Significant efforts made in recent years to detect new exoplanets through high-contrast imaging (HCI) using large telescopes
- Two main objectives of the research: gaining insights into exoplanet characteristics and differentiating between planet formation theories, as well as studying planets within our Solar neighborhood
- Accuracy of metrics crucial for quantifying planet detections and non-detections in HCI research
- Current standards in HCI often assume noise follows a Gaussian distribution without rigorous verification
- Team led by Markus J. Bonse revisited how detection limits are quantified in HCI, focusing on impact of violated assumptions on error budget
- Proposed novel metric based on bootstrapping to account for non-Gaussian noise distributions in HCI data analysis
- Applying the method to archival NACO-VLT instrument data, researchers found that current standards may provide overly optimistic estimates in scenarios dominated by speckle noise
- Importance of robust methodologies for quantifying detection limits highlighted to enhance reliability and comparability of findings across different datasets and instruments

## Summary

In recent years, significant efforts have been made to detect new exoplanets through high-contrast imaging (HCI) using the world's largest telescopes. This research has two main objectives: gaining insights into the characteristics of the exoplanet population and differentiating between various theories of planet formation and evolution, as well as identifying and studying planets within our Solar neighborhood. The success of these goals relies heavily on the accuracy of metrics used to quantify planet detections and non-detections. However, current standards in HCI often make assumptions about noise that are not rigorously verified. For example, it is commonly assumed that residual noise following data post-processing follows a Gaussian distribution. Failure to validate these assumptions can introduce systematic biases, making it difficult to compare results across different datasets or instruments with varying noise profiles. To address this issue, a team led by Markus J. Bonse has revisited the fundamental question of how detection limits are quantified in HCI. Their analysis focuses on assessing the impact of violated assumptions on the error budget. In response, they propose a novel metric based on bootstrapping that extends existing standards to account for non-Gaussian noise distributions. Applying their method to archival HCI data from the NACO-VLT instrument, Bonse et al. derived detection limits for different types of noise. Their findings reveal that current standards may provide overly optimistic estimates in scenarios dominated by speckle noise, potentially leading HCI surveys to overlook planets that could still exist. This study underscores the importance of robust methodologies for quantifying detection limits in HCI research. By introducing a more comprehensive approach that considers non-Gaussian noise characteristics, researchers can enhance the reliability and comparability of their findings across different observational datasets and instruments.