Coastal urban environments are facing an increasing risk of flooding from both surface water and groundwater due to globally elevated water levels, causing substantial socio-economic damages, threatening livelihoods, and hindering economic development. To date, assessing the risk of groundwater-related flooding has primarily relied on groundwater modeling and analytical solutions, but these approaches face the challenge of limited groundwater observations. In this context, spatio-temporal mapping of groundwater-related flooding in urban environments using satellite imagery offers a promising option to augment limited groundwater field measurements. However, the ability of satellite imagery to capture the dynamics of groundwater fluctuations, which can induce saturation excess flooding, has not been assessed previously. This study is the first to compare flood detection estimates derived from Sentinel-2 satellite imagery with those obtained by analyzing the difference between topography and the piezometric surface, using in-situ groundwater level measurements. This research uses the coastal city of Nouakchott, Mauritania as a case study, and covers the period from 2015 to 2023. The satellite images were processed to map flooded areas using a supervised classification (random forest algorithm) and water detection spectral indices (MNDWI and NDPI). The novelty of this study lies in the successful mapping of flooding in a data-scarce coastal urban environment. For the first time, this study demonstrated that satellite imagery, combined with machine learning methods, effectively captures the spatio-temporal dynamics of groundwater-related flooding and complements low-frequency in-situ groundwater level measurements. Results showed that flooded area peaked during the wet season (July to September) due to rainfall infiltration, and continuously decreased during the rest of the year. This work also demonstrated that mapping groundwater-related flooding based on groundwater level measurements may lack the resolution necessary to detect smaller flooded areas in complex urban environments, while the supervised classification produced more nuanced flooding estimates, both in space and in time. These results have promising applications for a wide range of projects linked to the characterization, modeling, and forecasting of groundwater-related flooding in urban coastal environments. In turn, this approach can directly benefit local stakeholders for developing hazard maps and, more broadly, for planning risk management strategies.