AbstractContinuous depth to groundwater (DTG) data collection is challenging in remote regions. Community participation offers a way to both increase data collection and involves the local community in scientific projects. Local knowledge, which is often descriptive, can be difficult to include in quantitative analysis; however, it can increase scientists' ability to formulate hypotheses or identify relevant environmental processes. We show how Community Science Research can add useful descriptive information for a study based in rural Colombia. To estimate the spatiotemporal distribution of DTG, the community collected water level measurements during a wet (La Niña) year and an average year. We built one spatial and two spatiotemporal models (with and without probabilistic data) using Bayesian Maximum Entropy. Due to the inclusion of local knowledge, the spatiotemporal model with probabilistic data reduced its mean square error by a factor of 15 compared to the spatial model. Using this model, we found that 13% of the study area has a high probability of very shallow DTG (<0.1 m) during an average year, whereas during La Niña, this area increases to 56%. The difference in shallow DTG between the average and wet year implies that after reaching a precipitation threshold, the study region may lose its flow regulation capacity, contributing to flooding during extreme precipitation events. Our approach presents a method to incorporate local knowledge in data‐driven models by combining qualitative and quantitative information.