Evaluation of the predictive capacity of five plant growth models for sugar beet

Abstract

A lot of plant growth models coexist, with different modelling approaches and levels of complexity. In the case of sugar beet, many of them are used as predictive tools, even when they were not originally designed for this purpose. We propose the evaluation and comparison of five plant growth models that rely on the same energetic production of biomass, but with different levels of description (per plant or per square meter) and different biomass repartition (empirical or via allocation): Greenlab, LNAS, CERES, PILOTE and STICS. The models were calibrated on a first set of data, and their predictive capacities were compared on an independent data set from the same variety and similar environmental conditions, using the root mean squared error of prediction (RMSEP) and modelling efficiency (EF) for the total dry matter production and the dry matter of root. All the models tended to overestimate both the total dry matter and the dry matter of root. Greenlab gave the best predictions for the root biomass, and CERES the best total biomass predictions. The overestimation was partly explained by a hail episode that caused a lot of damages to the leaves in the validation year. The five models also provided similar yield prediction errors.