Evaluation of the stochastic weather generators LABS-WG and AAFC-WG for climate change impact studies

There is a need to know how well stochastic weather generators can produce daily climate scenarios for climate change impact studies. In this study, 2 stochastic weather generators (LARS-WG and AAFC-WG) were assessed, based on an experiment with historical daily climate data. The experiment was conducted for 3 stations in Canada, using 1911-1940 as the baseline climate period and 1931-2000 as the changed climate period. Weather generators were calibrated with the baseline data. Daily climate scenarios were then generated for 1971-2000, using parameters that were adjusted to reflect climate change. Different schemes for modifying weather generator parameters were assessed in relation to their capability to reproduce statistical properties of daily climate data for a changed climate. Using changes in the statistics of daily climate data between the baseline climate and the changed climate period, it became feasible to modify parameters for the weather generators that use empirical distributions. For reproducing changes in frequency of wet and dry spells, modification schemes with a full adjustment (all 2nd order transition probabilities being modified separately) for the 2nd order Markov chain appeared more effective than the modification to mean length of wet and dry spells. Separate adjustments for maximum (Tx) and minimum (Tn) temperature, rather than modification on the basis of changes in daily mean temperature, might also be necessary. The probability distribution of daily precipitation amount (P) in a changed climate seemed more difficult to simulate well, although the means and variances were produced better. In general, daily climate scenarios developed by weather generators can be reasonably reliable for agricultural impact studies, provided the changes in the statistics of daily weather variables from daily GCM output are reliable. Otherwise, downscaling may be required to obtain reliable changes in the statistics of local daily weather variables.