Abstract <p>Understanding water flow dynamics through karst aquifers is critical for managing one of the world's major water resources. However, the inherent heterogeneity and complex geometry of these aquifers pose significant challenges to water management. Current hydrogeological methods lack a reliable surface-based tool for detecting and quantifying flow dynamics. We propose that ambient noise monitoring offers the best solution to: (1) detect hidden subsurface flows, particularly through karst conduits beneath dry valleys; (2) estimate discharge within these conduits; and (3) monitor the activation and deactivation of such flows in real-time. This study presents in situ ambient noise measurements taken at the ground surface upstream of a small karst spring, connected to an underground conduit. By combining two years of continuous ambient noise data with hydrogeological records, we identify the seismic signature of subsurface water flow. Ambient noise spectrograms in the 7–12&#xa0;Hz frequency range show a strong correlation with discharge peaks exceeding 1.2 m³/s during flood events, with amplitude variations between 1 and 3&#xa0;dB, depending on the flood and sensor location. Additionally, long-term variations in baseline noise levels suggest a link to the seasonal dynamics of the karst system. A power-law relationship between seismic signals and discharge allows us to predict discharge time series effectively.</p> Highlights <p>• Subsurface karst flow was detected via ambient noise surface monitoring.</p> <p>• A power law quantitative relation predicts the seismic signals and discharge correlation.</p> <p>• Seasonal and long-term ambient noise variations reflect on the karst system dynamics and recharge.</p>