Abstract NASA's InSight mission investigates the interior structure of Mars. The data is characterized by multiple non‐seismic signals with varying attributes, including high‐energy instrumental noise, known as glitches, which frequently exhibit large linear polarization. Short‐duration, high‐frequency spikes often accompany these glitches, further complicating the detection of actual seismic event signals. In contrast, marsquakes signals are typically low in amplitude and can exhibit weak or poorly defined polarization, particularly at high frequencies. Thus, detecting marsquakes remains challenging even under generally low‐noise conditions. To address these issues, we developed an efficient denoising strategy that targets the nighttime periods of reduced noise levels. The method involves two key stages: despiking and deglitching. First, spikes are detected by analyzing the first derivative of the trace's energy, identifying outliers that deviate significantly from the surrounding signal, and then applying a polynomial replacement. After despiking, the method uses a rotation‐based filtering technique to concentrate the energy of the glitches into a single component through moving windows along the traces. The algorithm returns the rotated traces to their original orientation and subtracts the filtered components containing the glitches from the original data. This process produces denoised traces and noticeably mitigates glitches and spikes, preserving only ambient noise and potential Martian seismic events. This automated approach enhances the signal‐to‐noise ratio and facilitates subsequent event detection, offering a scalable solution for processing large data sets. We processed the nighttime hours of 3 months of data to compare the results of automatic detections between raw and denoised traces using an STA/LTA algorithm.
Zampieri et al. (Sun,) studied this question.