A Bayesian sequential updating approach to predict phenology of silage maize

[摘要] Crop models are tools used for predicting year-to-yearcrop development on field to regional scales. However, robust predictionsare hampered by uncertainty in crop model parameters and in the data usedfor calibration. Bayesian calibration allows for the estimation of modelparameters and quantification of uncertainties, with the consideration ofprior information. In this study, we used a Bayesian sequential updating(BSU) approach to progressively incorporate additional data at a yearlytime-step in order to calibrate a phenology model (SPASS) while analysing changes inparameter uncertainty and prediction quality. We used field measurements ofsilage maize grown between 2010 and 2016 in the regions of Kraichgau andthe Swabian Alb in southwestern Germany. Parameter uncertainty and modelprediction errors were expected to progressively be reduced to a final,irreducible value. Parameter uncertainty was reduced as expected with thesequential updates. For two sequences using synthetic data, one in which themodel was able to accurately simulate the observations, and the other inwhich a single cultivar was grown under the same environmental conditions,prediction error was mostly reduced. However, in the true sequences thatfollowed the actual chronological order of cultivation by the farmers in thetwo regions, prediction error increased when the calibration data were notrepresentative of the validation data. This could be explained bydifferences in ripening group and temperature conditions during vegetativegrowth. With implications for manual and automatic data streams and modelupdating, our study highlights that the success of Bayesian methods forpredictions depends on a comprehensive understanding of the inherent structure in the observation data and of the model limitations.

[学科分类] 大气科学