Home > Research > Publications & Outputs > Uncertainty assessment of a dominant-process ca...

Links

Text available via DOI:

View graph of relations

Uncertainty assessment of a dominant-process catchment model of dissolved phosphorus transfer

Research output: Contribution to journalJournal article

Published
Close
<mark>Journal publication date</mark>8/12/2016
<mark>Journal</mark>Hydrology and Earth System Sciences
Issue number12
Volume20
Number of pages17
Pages (from-to)4819-4835
Publication statusPublished
Original languageEnglish

Abstract

We developed a parsimonious topography-based hydrologic model coupled with a soil biogeochemistry sub-model in order to improve understanding and prediction of soluble reactive phosphorus (SRP) transfer in agricultural headwater catchments. The model structure aims to capture the dominant hydrological and biogeochemical processes identified from multiscale observations in a research catchment (Kervidy–Naizin, 5 km2). Groundwater fluctuations, responsible for the connection of soil SRP production zones to the stream, were simulated with a fully distributed hydrologic model at 20 m resolution. The spatial variability of the soil phosphorus content and the temporal variability of soil moisture and temperature, which had previously been identified as key controlling factors of SRP solubilization in soils, were included as part of an empirical soil biogeochemistry sub-model. The modelling approach included an analysis of the information contained in the calibration data and propagation of uncertainty in model predictions using a generalized likelihood uncertainty estimation (GLUE) "limits of acceptability" framework. Overall, the model appeared to perform well given the uncertainty in the observational data, with a Nash–Sutcliffe efficiency on daily SRP loads between 0.1 and 0.8 for acceptable models. The role of hydrological connectivity via groundwater fluctuation and the role of increased SRP solubilization following dry/hot periods were captured well. We conclude that in the absence of near-continuous monitoring, the amount of information contained in the data is limited; hence, parsimonious models are more relevant than highly parameterized models. An analysis of uncertainty in the data is recommended for model calibration in order to provide reliable predictions.