Modelling climate change impact on Septoria tritici blotch (STB) in France: Accounting for climate model and disease model uncertainty

We calculate the impact of climate change on the effective severity of Septoria tritici blotch (STB) of winter wheat (Triticum aestivum L.) at three representative locations in France. The calculation uses climate models for climate prediction, and a disease model to link disease severity to weather. Four impact criteria are considered: the change in average (over years) severity, the change in interannual variance of severity, the change in number of years with particularly high severity and the change in the number of years with particularly low severity.We also calculate the uncertainty associated with those impact criteria. Three different uncertainty sources are considered: uncertainty in predicting climate, uncertainty in the values of the disease model parameters and uncertainty due to residual error of the disease model. Uncertainty in climate is considered by using different global climate models and downscaling methodologies to produce five different climate series for greenhouse gas emission scenario A1B, for a baseline period comprising harvest years 1971–1999 and a future period spanning 2071–2099. A Bayesian approach, using a Metropolis Hastings within Gibbs algorithm, is used for parameter estimation. This gives a posterior distribution both for the 17 model parameters that were considered and for the variance of residual error.Climate change is predicted to reduce the average severity of STB by 2–6%, depending on location, and to result in more low severity years and fewer high severity years. There is appreciable uncertainty. For example, the probability that average severity will increase rather than decrease is 40%, 18% and 45% for the three locations. We calculated first order sensitivity indices for climate model, parameter vector and residual error considered as three factors. The climate model factor has by far the largest sensitivity index. However, interactions between factors also make a major contribution to overall variance.