Machine learning-based analyses using surface detector data of the Pierre Auger Observatory

The Pierre Auger Observatory is the largest detector for the study of extensive air showers induced by ultra-high-energy cosmic rays (UHECRs). Its hybrid detector design allows the simultaneous observation of different parts of the shower evolution using various detection techniques. To accurately understand the physics behind the origin of UHECRs, it is essential to determine their mass composition. However, since UHECRs cannot be measured directly, estimating their masses is highly non-trivial. The most common approach is to analyze mass-sensitive observables, such as the number of secondary muons and the atmospheric depth of the shower maximum. An intriguing part of the shower to estimate these observables is its footprint. The shower footprint is detected by ground-based detectors, such as the Water-Cherenkov detectors (WCDs) of the Surface Detector (SD) of the Observatory, which have an uptime of nearly 100%, resulting in a high number of observed events. However, the spatio-temporal information stored in the shower footprints is highly complex, making it very challenging to analyze the footprints using analytical and phenomenological methods. Therefore, the Pierre Auger Collaboration utilizes machine learning-based algorithms to complement classical methods in order to exploit the measured data with unprecedented precision. In this contribution, we highlight these machine learning-based analyses used to determine high-level shower observables that help to infer the mass of the primary particle, with a particular focus on analyses using the shower footprint detected by the WCDs and the Surface Scintillator Detectors (SSD) of the SD. We show that these novel methods show promising results on simulations and offer improved reconstruction performance when applied to measured data.