Incorporating Community Science Research into Bayesian Maximum Entropy modeling improves depth to groundwater mapping in a remote rural region

Abstract:

In groundwater-dependent supply systems, continuous depth to groundwater data is challenging to acquire, especially in remote regions. Community science research is a potential strategy to improve the understanding of groundwater systems. In community science projects, actively engaged participants provide valuable qualitative knowledge that describes environmental processes, but this can be challenging to incorporate in quantitative analysis. Here, we combined community science research with a Bayesian Maximum Entropy geostatistical approach to map weekly depth to groundwater in a low-elevation portion of the River Man watershed, located in rural Colombia. We built three geostatistical models, one spatial and two spatiotemporal with and without descriptive qualitative data provided by the community. We found that the spatiotemporal model with probabilistic data, built using qualitative data, reduced model mean square error by a factor of 15 when compared to the spatial model that incorporated only quantitative data. Our weekly depth to groundwater results were used to map the probability of shallow depth to groundwater and to identify how the aquifer responded to precipitation in space and time. We found the portion of the watershed that was more likely to have shallow depth to groundwater increased from 12.6% (average year) to 56.6% (La Niña wet year), especially in headwaters of the main tributaries and in topographic depressions. This difference suggests that after the region receives an excess of precipitation, its flow regulation capacity can decrease, which is further threatened by local land-use activities. Our work shows how local descriptive knowledge can be incorporated into modeling by combining community science research with Bayesian Maximum Entropy modeling and emphasizes the benefits of an active community role in scientific projects.