CLUES model calibration: residual analysis to investigate potential sources of model error. Issue 3 (3rd July 2021)

• Record Type: Journal Article
• Authors: Semadeni-Davies, Annette F.; Jones-Todd, Charlotte M.; Srinivasan, M. S.; Muirhead, Richard W.; Elliott, Alexander H.; Shankar, Ude; Tanner, Chris C.
• Is Part Of: New Zealand journal of agricultural research. Volume 64:Issue 3(2021)
• Page Start: 320
• Page End: 343
• Publication Date: 2021-07-03
• Subjects: Overseer -- SPARROW -- SPASMO -- total nitrogen -- total phosphorus -- E. coli -- annual loads -- uncertainty -- bias -- error; Agriculture -- Periodicals; Agriculture -- New Zealand -- Periodicals; Animal culture -- Periodicals; Animal culture -- New Zealand -- Periodicals; 630.99305
• DOI: 10.1080/00288233.2019.1697708
• Languages: English
• ISSNs: 0028-8233

Abstract

ABSTRACT: This paper presents the results of the latest calibrations of the CLUES model for each of total nitrogen, total phosphorus and Escherichia coli . The model provides estimates of in-stream annual loads of these contaminants for every river reach in the New Zealand River Environments Classification and has been used in New Zealand for catchment planning and policy development. CLUES has been calibrated nationally against loads estimated from data collected from water quality monitoring stations located across the country which could lead to parameter bias. To ascertain whether there is any systematic bias, the calibration residuals were evaluated against regional (geographical) location and a range of upstream catchment characteristics, namely land use, soil drainage properties, slope, network order, proportion of baseflow in total stream flow, climate, geology and source of flow. We found that CLUES gives reasonable load estimates at the catchment scale (Nash-Sutcliff efficiencies >0.8 for all the contaminants). However, there was significant uncertainty in the parameterisation. While several significant relationships were found between upstream catchment characteristics and the model residuals, these relationships were weak and are unlikely to point to any systematic bias in the calibration.