Estimation and validation of a function describing the rate at which Mycosphaerella graminicola causes yield loss in winter wheat

Summary An experiment designed to generate a wide variety of epidemics of Mycosphaerella graminicola without deliberate replication was done in two years. Disease severity was estimated at frequent intervals during the life of the crop and yields measured at harvest. Absolute estimates of disease severity were derived by regression of visual estimates on subsamples of leaves on which severity was estimated objectively using an image analysis system. Yield was predicted best by the integral of the square root of M. graminicola severity over the normal lifetime of each leaf, measured in thermal time. Only the top two leaves contributed to yield loss; no influence of the third leaf on yield was detected. Thousand grain weight was best predicted by the integral of the square root of M. graminicola severity on the flag leaf alone. Parameter estimates were similar in the two years. The prediction equations were consistent with yields observed in an experiment done in a third year using two sowing dates and two rates of nitrogen fertilisation, despite a much greater range of disease severity. Although critical point models could describe each year's results adequately, neither parameter estimates nor the growth stage at which the best relation occurred were consistent across years. The equations to predict loss may be useful for farmers in decision‐support systems which are based on prediction.