Assessing inter-annual variability in nitrogen sourcing and retention through hybrid Bayesian watershed modeling
Miller, J. W., Karimi, K., Sankarasubramanian, A., & Obenour, D. R. (2021, February 9). (Vol. 2). Vol. 2.
Abstract. Excessive nutrient loading is a major cause of water quality problems worldwide, including in North Carolina (NC), where reservoirs and coastal systems are often subject to excessive algae and hypoxia. Efficient nutrient management requires that loading sources are accurately quantified. However, loading rates from various urban and rural non-point sources remain highly uncertain especially with respect to climatological variation. Furthermore, statistical calibration of loading models does not always yield plausible results, given the noisiness and paucity of observational data common to many locations. To address these issues, we leverage data for two large NC Piedmont basins collected over three decades (1982–2017) using a mechanistically parsimonious watershed loading and transport model calibrated within a Bayesian hierarchical framework. We explore temporal drivers of loading by incorporating annual changes in precipitation, land use, livestock, and point sources within the model formulation. Also, different representations of urban development are compared based on how they constrain model uncertainties. Results show that urban lands built before 1980 are the largest source of nutrients, exporting over twice as much nitrogen per hectare than agricultural and post-1980 urban lands. In addition, pre-1980 urban lands are the most hydrologically constant source of nutrients, while agricultural lands show the most variation among high and low flow years. Finally, undeveloped lands export an order of magnitude (~ 7–13x) less nitrogen than built environments.