Adaptation of a crop-growth model and its extension by a tuber size function for use in a seed potato forecasting system

Abstract

The crop-growth model for potatoes by Johnson et al. (1986) was adapted for use within a decision support and forecasting system for seed potato production and coupled to a soil water-balance model in order to simulate growth under suboptimal hydrological conditions. The original model produced outputs that deviated largely from field data for cultivars grown in Switzerland. To achieve a better fit of these data, we adapted the model by estimating a new set of parameters. Furthermore, we needed to calculate the tuber size distribution, which is not done in the original model. For this purpose we introduced a tuber size output function. These modifications together with the corresponding validation results are described in this study. To simulate plant growth of many cultivars with a minimum of different parameter values, the cultivars were grouped into three maturity classes, which differ in the tuber initiation age, the tuber dry matter content at maturity and the leaf age at senescence. Tuber dry matter content is calculated as a function of physiological time. The tuber weight distribution is modelled by a Weibull distribution; its shape parameter α is kept sonstant, whereas the scale parameter β is a function of the average tuber weight. Before calculating the weight distribution, the tuber grades are transformed into tuber weights by accounting for the cultivar-specific shape of the tubers. The overall agreement of the model generated output with the corresponding field data was satisfactory. The mean absolute deviation between simulated and observed data was 15% of the average of the field data in the case of total fresh yield, 17% for leaf dry weight, 25% for seed yield and 35% for ware yield. The agreements for stem and root dry masses were not as good, but simulated values were in the same order of magnitude as observed ones. We conclude that the model meets the objectives and can be used within the forecasting system.