GAMM’s as a New Tool for Evaluating Spatiotemporal Distributions of Nitrate and Bacteria in an Agricultural Watershed

Abstract:

Generalized additive mixed models (GAMM’s) are flexible regression models that are increasingly used in ecological and environmental studies to assess spatial and temporal trends in complex monitoring data. GAMM’s hold promise for analysis of spatially and temporally correlated hydrogeologic data, but have been used only sparingly. Here we employed GAMM’s to investigate the spatiotemporal distribution of pathogen indicators (E. coli and total coliform [TC]) and nitrate in the vulnerable Abbotsford-Sumas aquifer (ASA), and to explore potential relationships with hydrologic and climatic drivers, such as precipitation, streamflow, and groundwater level and temperature. A total of 46 wells sampled over a one year period showed more than 50% of samples exceeded 10 mg-N/L for nitrate. E. coli detections in groundwater were infrequent (4 of 385 total samples) and attributed mainly to surface water-groundwater connections. TC was detected frequently in groundwater (70% of samples) and the widespread TC distribution across the ASA could not be attributed solely to surface water connectivity. GAMM’s showed that increased TC values in the wet season were most strongly related to groundwater temperatures and levels, while precipitation and well location were weaker (but still significant) predictors. In contrast, seasonal trends in nitrate were not significantly related to hydrologic forcings. Instead, nitrate concentrations across the aquifer were controlled by well location and depth, likely due to spatially variable nitrogen loading and localized geochemical attenuation. Major differences in nitrate and bacterial loading to the ASA were apparent in this study, and management strategies specific to each nonpoint source contaminant are recommended for improved source water protection.