摘要：After a prolonged drought period in the early 2000s, the Canadian prairie experienced a remarkably wet year in 2010. Five stations near the edge of the Saskatchewan boreal forest recorded historically high cumulative precipitation (from April to September). The exceptional wet year causes the public concerns on flood controls and land use management in the region. Using the Canadian National Climate Data Achieve, characteristics of six-month cumulative precipitation sums over Saskatchewan prairie are investigated by the Generalised Extreme Value (GEV) Theory. Based on the unconstrained GEV distribution, the 2010 event is outside the estimated 95% confidence intervals for the five Canadian prairie stations. On the contrary, the exceptional high 2010 cumulative perception sums for the five stations are still bounded by the estimated confidence bounds if the GEV distribution is constrained to the Gumbel distribution (i.e. setting the shape factor of the GEV distribution to be zero). These results demonstrate that the classical extreme analysis is useful for planning unprecedented extreme events in the Canadian Prairie, if the GEV distribution is constrained to the Gumbel distribution with the estimated uncertainty bounds based on the order statistics.
关键词：Generalised Extreme Value (GEV) theory; Saskatchewan; prairie; confidence intervals; Gumbel distribution
