Skip to main content
ARS Home » Plains Area » Fort Collins, Colorado » Center for Agricultural Resources Research » Rangeland Resources & Systems Research » Research » Publications at this Location » Publication #347029
Title: A modified F-test for evaluating model performance by including both experimental and simulation uncertainties

Author
item SIMA, NATHAN - COLORADO STATE UNIVERSITY
item Harmel, Daren
item FANG, QUAN - QINGDAO AGRICULTURAL UNIVERSITY
item Ma, Liwang
item ANDALES, ALLAN

Submitted to: Journal of Environmental Modeling and Software
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/15/2018
Publication Date: 4/8/2018
Citation: Sima, N.Q., Harmel, R.D., Fang, Q.X., Ma, L., Andales, A.A. 2018. A modified F-test for evaluating model performance by including both experimental and simulation uncertainties. Journal of Environmental Modeling and Software. 104:236e248. doi.org/10.1016/j.envsoft.2018.03.011.
DOI: https://doi.org/10.1016/j.envsoft.2018.03.011

Interpretive Summary: The uncertainty contributed by measured data and model representation of the real world have not been included in many of the statistics used in assessing agricultural model performance. The objectives of this study were to develop a statistical test (called an n F-test) that can be used to evaluate model performance considering experimental and simulation uncertainties, and identify the best datasets to use for model calibration. In this study, we used different water stress functions in a cropping system model. Data on irrigated maize in Colorado, USA, and the Root Zone Water Quality Model (RZWQM) were used as an example to demonstrate model calibration using the modified F-test along with other commonly used statistics. Compared to the d-index, the F-test provided a statistical test under a certain confidence level that better distinguished the goodness of model prediction for both biomass and yield while considering uncertainty. To obtain more robust model parameters, we recommend using two or more treatments across multiple years for model calibration.

Technical Abstract: Experimental and simulation uncertainties have not been included in many of the statistics used in assessing agricultural model performance. The objectives of this study were to develop an F-test that can be used to evaluate model performance considering experimental and simulation uncertainties, and identify the best datasets to use for model calibration using different water stress functions in a cropping system model. Data on irrigated maize in Colorado, USA, and the Root Zone Water Quality Model (RZWQM) were used as an example to demonstrate model calibration using the modified F-test along with other commonly used statistics. Compared to the d-index, the F-test provided a statistical test under a certain confidence level that better distinguished the goodness of model prediction for both biomass and yield while considering uncertainty. To obtain robust model parameters, we recommend using multiple treatments across multiple years for model calibration, regardless of water stress functions used.