Model comparison of soil processes in eastern Canada using DayCent, DNDC and STICS

• Published in: Nutrient cycling in agroecosystems Vol. 109; no. 3; pp. 211 - 232
• Main Authors: Guest, G., Kröbel, R., Grant, B., Smith, W., Sansoulet, J., Pattey, E., Desjardins, R., Jégo, G., Tremblay, N., Tremblay, G.
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.12.2017; Springer Nature B.V
• Subjects: Agriculture; Assessments; Biomedical and Life Sciences; Carbon; Climate system; Computer simulation; Emissions; Hydrology; Life Sciences; Nitrogen; Nutrient cycles; Original Article; Simulation; Soil improvement; Soil moisture; Spring wheat; Trace gases; Unsaturated flow; Wheat; Canada; STICS; DNDC; DayCent; N dynamics; Soil moisture
• ISSN: 1385-1314; 1573-0867
• DOI: 10.1007/s10705-017-9880-8

Process-based models are useful tools for estimating the complex interactions between plant, soil and climate systems, assessments which are necessary for improving nutrient cycling and reducing trace gas emissions. Incorporation of knowledge gained through new research is ongoing, thus there is a need for evaluation of model processes and process interactions. In this study, three sites in eastern Canada (St. Bruno, St. Jean, and Ottawa) planted with spring wheat during the years 1993–2007 were used to evaluate and compare the water and N process simulations of the models DayCent, DNDC, and STICS. The simulated soil moisture by all models was generally well represented with low ARE (< 8%) and an EF > 0.1. The unsaturated flow mechanism included in DayCent further improved soil moisture estimates compared to the other models. When sufficient replicate data was available measurement variability was considered, resulting in soil nitrogen being only slightly underestimated (ARE of −10, −1, and −22%, for DayCent, DNDC, and STICS, respectively). On average across the three sites, considering all statistics, the DNDC model proved to be most accurate for simulating mineral N, followed by DayCent and then STICS. Continued process model development is reliant on measurement datasets that can accurately represent carbon and nitrogen dynamics. Frequently, site specific biases convolute model mechanism evaluation and thus assessments have to be conducted across numerous sites to better benchmark model performance. On this premise a comprehensive multi-site inter-model mechanism evaluation was conducted and future model development needs were identified.