Dynamic analysis of stream flow and water chemistry to infer subsurface water and nitrate fluxes in a lowland dairying catchment

Abstract

Summary The use of process-based, dynamic and spatially-explicit models to describe water and nitrogen fluxes at the catchment-scale is often hampered by a shortage of detailed land use, hydrological and biogeochemical information. Accordingly, such complex models tend to be restricted to a small number of well investigated catchments, often associated with research projects. On the other hand, stream flow and stream water chemistry time series data are available for a much larger number of catchments, e.g. for many catchments that are routinely monitored by government agencies for state-of-the-environment reporting. It was the main aim of this study to provide a spatially lumped model that allows meaningful analysis of catchment-scale water and nitrate fluxes based on such data sets. Based on stream flow time series data, catchment hydrodynamics are often analysed using approaches derived from the linearised Boussinesq equation, which has analytical solutions for dynamic groundwater discharge expressed in terms of eigenvalues and eigenfunctions (eigenmodel approach). Calibrated Boussinesq models generally yield a good reproduction of stream flow dynamics, and stable estimates for aquifer parameters such as hydraulic conductivity and mean aquifer depth. By linking a soil water balance model with two Boussinesq groundwater eigenmodels linked in series, and assuming constant solute concentrations discharging from each source, a dynamic catchment model predicting stream flow and water chemistry at the catchment outlet (“StreamGEM”) was developed. Compared with previous approaches, inclusion of water chemistry in this model both aided hydrological understanding, and allowed assessment of catchmentscale nitrate fluxes. Simultaneous calibration of the model to stream flow and nitrate concentration data from a small lowland dairying catchment yielded good predictions to both variables (Nash–Sutcliffe Model Efficiency of 0.90 and 0.84), and the fitted parameters were able to be used to estimate annual flow and nitrate fluxes through near-surface, shallow groundwater, and deeper groundwater reservoirs conceptually present in the catchment. The calibration was cross-validated using an independent time series from the same catchment. The results support the hypothesis, based on groundwater observations, that stream flow in the catchment is the result of mixed discharge from a shallower, rapidly draining zone of oxidised groundwater carrying relatively high loads of agricultural nitrate, with a relatively deeper and slower draining zone of reduced groundwater that is essentially nitrate free. The proportions of stream flow discharging from the near-surface, shallow groundwater, and deeper groundwater reservoirs were estimated to be 5%, 80% and 15%, respectively. In spite of its small contribution to total stream flow, the deeper groundwater reservoir sustained stream flow during summer and dominated stream water chemistry 61% of the time. By combining the flow and nitrate concentration estimates derived from model calibration, it was estimated that discharge of shallow groundwater was responsible for 91% of the nitrate load entering the stream. However, the predicted nitrate concentration in this reservoir was significantly lower than the predicted nitrate concentration of near-surface flow and root zone leachate concentrations estimated using a nutrient budgeting model. This indicates that denitrification occurs within this reservoir. On the basis of the calibrated model, it was estimated that 36% of the nitrate recharged from the vadose zone gets denitrified within the shallow groundwater reservoir, and up to 9% in the deeper groundwater reservoir.