Modelling stream phosphorus in rural areas: improving on current approaches

Jackson-Blake, L. A. (2016) Modelling stream phosphorus in rural areas: improving on current approaches. PhD thesis, University of Reading.

Abstract/Summary

Process-based, catchment-scale water quality models are important tools to help answer key questions facing water managers (oday. However, for model results to be useful models must capture the dominant processes, it must be possible to calibrate them using available data and good modelling practice guidelines must be followed. The work presented here contributes to the development of robust phosphorus (P) modelling tools, starting with a detailed examination of one popular and representative model, INCA-P. INCA-P was applied in a Scottish agricultural catchment, where manual calibration and sensitivity testing were combined with auto¬calibration and uncertainty analyses to investigate: (J) the suitability ofthe process-representation; (2) whether model parameters could be constrained using available data; and (3) uncertainty in model output. Problems were identified with the particulate P and soil dissolved P processes, leading to the development of a new version of INCA-P, and suggestions were made for simplifications to the model structure. Model calibration using routine regulatory monitoring data led to highly uncertain output, suggesting that the model is (00 complex given the data availability in most areas, reducing its usefulness for water management. The same is likely to be true for other popular water quality models. To investigate issues of over complexity, a new simple model was developed, SimplyP. SimplyP performed as INCA-P in the study catchment, despite substantially simpler process-representation (e.g. 28 parameters, compared to INCA-P's 148), supporting the hypothesis that the current generation of dynamic water quality models are too complex. Simpler models, such as SimplyP, have the potential to be more useful water management and research tools, building blocks for future model development, or performance benchmarks against which more complex models could be evaluated.