华南冬季极端低温事件的统计建模及未来预估

Statistical modelling and projection of winter extreme cold spells over South China

基于1961—2005年华南冬季逐日最低温度观测资料,以及参与CMIP5的IPSL-CM5A-MR、MPI-ESM-MR、CMCC-CMS 3个模式的历史模拟和未来排放情景预估数据,从过程角度定义和提取华南冬季极端低温事件及其特征指标,并通过引入极值理论构建极端低温事件的概率分布模型(Cold Spell Model,CSM),刻画其强度、频数和持续时间特征分布;进一步结合基于累计概率分布变换的偏差订正方法降低模式模拟的偏差,探讨了RCP4.5情景下全球升温1.5℃和2℃时极端低温事件的风险变化。研究表明:(1)CSM模型能够较好地拟合极端低温特征概率分布,大部分台站通过了K-S或卡方检验;(2)相对于观测,模式模拟极端低温存在一定偏差,经偏差订正后,各站点BS评分接近于0,SS评分普遍提高到0.88左右;(3)全球增暖背景下,华南地区极端低温呈现强度减弱、频数减少、持续时间缩短的特征。全球增暖1.5、2.0℃背景下,华南大多数地区极端低温事件平均强度减弱,频次降低,持续时间变化具有明显的空间差异;每年发生1次极端低温事件的概率增加,但每年发生3、5次的概率减少;增暖1.5℃下极端低温事件持续性特征变化不明显。但增暖2℃下,华南大部分地区持续1 d的极端低温事件更易发生,但更长持续时间(3、5 d)的极端低温事件发生的可能性显著减少。

Based on observed and simulated daily minimum temperature from three GMCs,the extreme cold spell taking into account three characteristics including the intensity,frequency,and duration was defined.The extreme value theory was applied to establish the Cold Spell Model(CSM).The bias correction method applied on the raw simulation and the future projection of CSM over southern China under global warming of 1.5℃and 2℃in the RCP4.5 scenario was studied.Results show that:(1)the CSM model can well represent the extreme cold spells based on observation,and most stations pass the K-S and chi-square test.(2)Compared to observation,climate model simulation exhibits quite large errors.The biases-corrected simulations can reduce the biases with the BS scores of each station closing to 0,and the SS score increasing to about 0.88.(3)Under 1.5℃and 2.0℃global warming,the intensity and frequency of extreme cold spell will decrease in most regions in the southern China.The large spatial variability was found for the changes in the duration of cold spell.The probability of events occurring once a year will increase,while the events occurring every 3 and 5 years will decrease.The direction of cold spell exhibit little changes under 1.5℃warming world,while the probability of events with longer duration(3 and 5 days per year)will decrease.