The topic of this PhD study was modeling of spatially distributed nitrate transport and reduction at catchment scale in order to delineate so-called nitrate sensitive and nitrate robust areas with respectively low and high nitrate reduction potential. The research firstly focused on some of the main challenges in this topic: to estimate the depth of the redox interface and to accurately simulate local scale water flow patterns. These two issues are important since they control how much and where nitrate is reduced in the saturated zone. Finally, the study also looked into the uncertainty on the estimated nitrate reduction potentials and evaluated on the predictive capability of catchment scale models. The PhD research has resulted in three papers.
In the first paper the objective was to evaluate whether the simulation of local scale water flow patterns in the tile drained Lillebæk catchment could be improved by including small-scale tile drain discharge observations in the calibration of a catchment-scale model. The study showed that including tile drain data did improve model performance on catchment scale. However, the
model performance on tile discharge was poor whether or not tile drain observations were included. It is hypothesized that the poor representation of tile drain flow in the model is due to lack of geological heterogeneity in the model.
In the second paper a concept for estimating depth of the redox interface was developed based on a process-based understanding of how the redox interface has developed to its current location. The concept was tested by applying it for the clayey till Norsminde catchment and comparing estimated and observed redox depths. The redox concept was found to be able to estimate the general location of the redox interface in Norsminde catchment. However; when
comparing the estimated depths with observed redox depths at individual wells, the estimated depths did not fit well with observations. It is hypothesized that one of the main reasons for the lack of predictive capability on small scale for the application of the concept in Norsminde is the lack of spatial variation in sediment redox capacity within the till.
In the third paper spatially distributed nitrate reduction potential in the saturated zone was estimated and the uncertainty on the estimate due to geological uncertainty was evaluated using multiple geological realizations. Uncertainty on the geology was found to give rise to large uncertainty on the predicted nitrate reduction at grid scale, but the uncertainty decreased with increasing scale. The study showed, that the decrease in uncertainty was largest in the beginning and then leveled off at a scale corresponding to the mean length of sand units in the
till, indicating that the spatial resolution of the geology is constraining at what spatial scale a distributed model has predictive capabilities. Furthermore, the geological uncertainty and thereby also the uncertainty on predicted nitrate reduction was found to decrease when using geophysical data in combination with borehole data in the generation of geological realizations.
The main outcome from this PhD research was that nitrate sensitive and nitrate robust areas can be predicted using a physically-based distributed model, but since catchment models most often lack predictive capabilities at grid scale the uncertainty on the estimated nitrate reduction potential at grid scale is large. It should therefore be evaluated at which spatial scale the predicted nitrate reduction potential can be used. The reason for the lack of predictive
capability at grid scale is insufficient description of spatial patterns of parameters and input data in the models. In particular, it is important to describe the spatial variation in the location of the redox interface and the local scale geological heterogeneity, which control the local scale groundwater flow patterns, in the model.