Modelling and evaluation of nitrate transit times in groundwater systems

Abstract

Transit time distribution is one of the fundamental descriptors of flow and transport behavior of catchments. Intensification of agriculture and consequent anthropogenic inputs of nitrate rich fertilizers in agricultural catchments has significantly increased the global groundwater nitrate stock. Transit time modelling is a way to determine the rate at which a contaminated catchment will self-cleanse itself. Multi-site water chemistry data is required to capture the proper spatiotemporal variability of groundwater dynamics, e.g.: piezometers beyond the stream riparian influence are helpful in getting proper insight into the deeper legacy stores. But at the same time, the modelling frameworks should also be parsimonious, flexible, accurate and free from numerical errors; and over parameterization and ill-posedness of calibration problems should be avoided. In this thesis, we explored the groundwater transit time modelling from the “catchments as reactors” standpoint. We used multi-site multi-objective calibration to test the scalability of a lumped linear reservoir model with well-mixed assumptions - which highlighted the importance of using long-term piezometer concentration data for segregation of catchment processes aggregated at stream scale. We tried to bridge the gap between simple lumped or semi-distributed hydrological response unit-based models and fully distributed finite element-based models by generating mathematical equivalence between conceptual and physical parameters via synthetic experiments. Furthermore, we created a novel framework for modelling non-uniform flow and transport in hillslope scale aquifers by applying tanks in series discretization to a flow field established using the concept of mean action time. Moreover, we established the authenticity of local nitrate flushing rate determination in a heterogeneous aquifer using lumped linear reservoir models by experimentally illustrating inverse correlation between localized Darcy flux vector density and localized concentrations of a larger heterogeneous flow-field. Through such exercises we were able to give calibration reliant lumped or semi-distributed conceptual models a forward modelling potential whilst being able to reach the desired golden mean between parsimony and accuracy of transit time models to some extent. Furthermore, we realized that there exist remarkable similarities in parameters of models across different philosophies, and remarkable equivalence in self-cleansing rates of nitrate in similar fissured rock formations across the world irrespective of the loading conditions