NONLINEARITY IN CLIMATE-CHANGE IMPACT ASSESSMENTS

Crop growth simulation models have been used to predict the consequenc es of climate change for cereal growth and yield (Adams et al., 1990). Two significant uncertainties in impact assessments result from the u se of possible future climate scenarios as inputs to crop models. Firs t, although based on the same physical principles, General Circulation Models (GCMs) provide individual estimates of the global climate and thus GCMs predict a variety of expected climate changes. We question t he rationale of formulating average or 'composite' climate change scen arios when used in combination with the many non-linear responses of c rops to their environment. Secondly, uncertainty also arises from the way in which weather scenarios are constructed from GCMs. Future clima tic scenarios, derived from GCMs, have described changes in mean weath er (Kenny, Harrison & Parry, 1993) but non-linear crop models require explicit incorporation of changes in climatic variability to assess th e risks to agricultural production from climate change. Accordingly, w e coupled a wheat crop simulation model (AFRCWHEAT2) with a stochastic weather generator (LARS-WG) and report that modelled changes in tempe rature variability may have more profound effect on simulated grain yi eld than changes in its mean value.