作者: Jouni Räisänen , Olle Räty
DOI: 10.1007/S00382-012-1515-9
关键词: Projection (set theory) 、 Climatology 、 Uncertainty analysis 、 Future climate 、 Mean radiant temperature 、 Quantile 、 Cross-validation 、 Environmental science 、 Climate change 、 Bias correction
摘要: Because of model biases, projections future climate need to combine simulations recent and with information on observed climate. Here, 10 methods for projecting the distribution daily mean temperatures are compared, using six regional change Europe. Cross validation between models is used assess potential performance in Delta bias correction type show similar cross-validation performance, based quantile mapping approach doing best both groups due their apparent ability reduce errors projected time temperature change. However, as no single method performs under all circumstances, optimal might be use several well-behaving parallel. When applying various real-world projection late 21st century, largest intermethod differences found tails distribution. Although variation generally smaller than intermodel variation, it not negligible. Therefore, should preferably included uncertainty analysis projections, particularly applications where extremes important.