Reducing uncertainty in the calibration and validation of the INCA-N model by using soft data

• Published in: Hydrology Research 2014-01, Vol.45 (1), p.73-88
• Main Authors: ETHERIDGE, J. Randall, LEPISTÖ, Ahti, GRANLUND, Kirsti, RANKINEN, Katri, BIRGAND, François, BURCHELL, Michael R
• Format: Article
• Language: English
• Subjects: Agricultural land; Agricultural practices; Agricultural production; Agriculture; Calibration; Catchment area; Catchment models; Catchment scale; Catchments; Computer simulation; Data; Data processing; Earth sciences; Earth, ocean, space; Engineering and environment geology. Geothermics; Exact sciences and technology; Experimentation; Heterogeneity; Hydrology; Hydrology. Hydrogeology; Land use; Mathematical models; Mineral nutrients; Nitrates; Nitrogen; Nutrients; Pollution, environment geology; Rivers; Scale models; Surface water; Uncertainty
• ISSN: 1998-9563; 0029-1277; 2224-7955
• DOI: 10.2166/nh.2013.039

Process-based nutrient models are increasingly used to determine the impact of future changes in land use, agriculture production practices and climate on the quantity and timing of nutrients reaching surface waters. Calibration of catchment-scale models to observed conditions can be difficult due to parameter uncertainty and the heterogeneity of catchment processes. Soft data, i.e. knowledge of processes gained through experimentation, have been suggested as one method of reducing uncertainty and producing a more accurate model of the processes that occur in a catchment. In this work, the Integrated Catchment model for Nitrogen was calibrated and validated for the Ylaeneenjoki catchment in south-western Finland by incorporating soft data. The calibration for 2003-2008 produced an adequate model of the in-stream nitrate concentrations (R2 = 0.45, NS = 0.42). However, model validation using data from 1997-2002 showed that the simulated in-stream nitrate concentrations were above the observed concentrations throughout the entire period (R2 = 0.34, NS < 0). These results show that soft data can be used to constrain model parameters, resulting in a more accurate model of the catchment, but do not guarantee the best validation results as the simulated processes may not occur at the same time and rate as they did in the catchment.