Taylor and Francis Group, Acta Agriculturae Scandinavica, Section B - Soil and Plant Science, 1(64), p. 56-70, 2014
DOI: 10.1080/09064710.2014.885076
Full text: Download
A crop model and environmental data were used to simulate genotype-environment interactions for commercial forage maize cultivars. Genotype parameters defined by the MAISPROQ model were calibrated to observed aboveground dry matter (DM) yield and quality (concentrations of DM and starch) data from Swedish field experiments 2009–2011 on four forage maize cultivars with different maturation rates (Avenir, Isberi, Jasmic and Burli). The model calibration predictability (coefficient of determination, R 2) ranged from 0.18 to 0.45 for yield and 0.40 to 0.86 for quality. The corresponding values for validation were in a similar range for the growth model but less for the quality model (0.36–0.38 and 0.25–0.54, respectively). Thereafter the model was used to assess the cultivar performance for different locations and future climate conditions. The simulated DM yield averaged for 2003–2009 varied between 5% and 25% among nine locations in Sweden due to differences in weather conditions, depending on cultivar. The proportion of years with successful harvest (34% DM concentration being achieved by 31 October) varied between 60% and 100% for the early cultivar (Avenir) and 0% and 70% for the late cultivar (Jasmic). Under future climate conditions, harvest of the early-maturing cultivar (Avenir) will occur earlier (by up to 19 days in Lund [55.6°N] and 24 days in Uppsala [59.8°N] by 2085), but with unchanged or even slightly decreased DM yields compared with current levels. The starch concentration will remain almost unchanged in Lund but increase in Uppsala, especially for the late-maturing cultivar. We regard the model predictions of quality to be reasonably satisfactory, whereas those of DM yields are less reliable due to observations for calibration being available only for the period after flowering. Therefore, more frequent sampling in the early growing season is required to improve the predictive power of the model, especially for DM yield.