基于国家气候中心气候系统模式(Beijing Climate Center Climate System Model,BCC_CSM1.1m)和美国NCEP/NCAR的气候预测模式(The NCEP Climate Forecast System Version 2,CFSv2)分别建立针对长江流域汛期降水的动力与统计相结合的降尺...基于国家气候中心气候系统模式(Beijing Climate Center Climate System Model,BCC_CSM1.1m)和美国NCEP/NCAR的气候预测模式(The NCEP Climate Forecast System Version 2,CFSv2)分别建立针对长江流域汛期降水的动力与统计相结合的降尺度预测模型,并比较两模式对应模型的预报技巧和差异来源。分别选择两模式2月起报的500 hPa及200 hPa全球位势高度场为预报因子,结合年际增量及经验正交分解(EOF)迭代法建立降尺度模型(分别简称DY_CSM1.1m和DY_CFSv2),研究发现:(1) EOF迭代法中截断解释方差的递增增加了预报因子的协同性和稳定性,从而显著提高预报技巧,并由此确定98%的截断解释方差为模型的最优参数。(2)两模型基于最优参数的预测效果均优于模式原始的降水预测,其中DY_CSM1.1m预测技巧更高,对应29 a距平相关系数(ACC)平均评分可达0.43,尤其在长江干流区域预报效果显著提高。将两模型预测的降水年际增量百分率转换为降水距平百分率时,ACC多年平均评分降为0.27和0.22,仍高于模式原始预测。(3) DY_CSM1.1m的ACC历年评分和长江流域汛期降水年际增量均与西太平洋副热带高压的一系列指数具有高相关性(以西太平洋副高脊线位置指数为例,DY_CFSv2则无此关系),因此BCC_CSM1.1m在西太平洋地区模拟性能优于CFSv2是导致该模式降尺度后预报技巧更高的重要原因,这一点在典型洪涝年1998和2020年中得以佐证。展开更多
本文利用国家气候中心气候系统模式(Beijing climate center climate System Model,BCC_CSM1.1m)提供的1991—2014年海表温度回报数据,将逐步回归模态投影方法(stepwise Pattern Projection Method,SPPM)应用到改进BCC_CSM1.1m模式El N...本文利用国家气候中心气候系统模式(Beijing climate center climate System Model,BCC_CSM1.1m)提供的1991—2014年海表温度回报数据,将逐步回归模态投影方法(stepwise Pattern Projection Method,SPPM)应用到改进BCC_CSM1.1m模式El Nino和南方涛动(ENSO)预报研究。SPPM是一种经验性模式误差订正方法,其主要思路是在大尺度模式预报因子场中找寻出与格点观测预报变量相关性高的信号,通过投影将这种信号反演出来,然后建立回归方程得到订正后的预报结果。本文交叉检验和滚动独立样本检验的结果表明,利用SPPM可以有效地提高BCC_CSM1.1m气候系统模式的预报技巧,尤其是在热带太平洋地区以及印度洋海区,24年交叉检验Nino3.4指数提前6个月预报的相关系数技巧可以提高8%~10%,预报误差得到显著降低。不同季节SPPM订正效果略有不同,其中对秋季的预报技巧提升最为显著。与此同时,交叉检验结果还显示,SPPM对El Nino中心纬向位置的预报也有一定程度的改进。展开更多
The prediction skill of Arctic Oscillation (AO) in the decadal experiments with the Beijing Climate Center Climate System Model version 1.1 (BCC_CSM1.1) is assessed. As compared with the observations and historical ex...The prediction skill of Arctic Oscillation (AO) in the decadal experiments with the Beijing Climate Center Climate System Model version 1.1 (BCC_CSM1.1) is assessed. As compared with the observations and historical experiments, the contribution of initialization for climate model to predict the seasonal scale AO and its interannual variations is estimated. Results show that the spatial correlation coefficient of AO mode simulated by the decadal experiment is higher than that in the historical experiment. The two groups of experiments reasonably reproduce the characteristics that AO indices are the strongest in winter and the weakest in summer. Compared with historical experiments, the correlation coefficient of the monthly and winter AO indices are higher in the decadal experiments. In particular, the correlation coefficient of monthly AO index between decadal hindcast and observation reached 0.1 significant level. Furthermore, the periodicity of the monthly and spring AO indices are achieved only in the decadal experiments. Therefore, the initial state of model is initialized by using sea temperature data may help to improve the prediction skill of AO in the decadal prediction experiments to some extent.展开更多
文摘基于国家气候中心气候系统模式(Beijing Climate Center Climate System Model,BCC_CSM1.1m)和美国NCEP/NCAR的气候预测模式(The NCEP Climate Forecast System Version 2,CFSv2)分别建立针对长江流域汛期降水的动力与统计相结合的降尺度预测模型,并比较两模式对应模型的预报技巧和差异来源。分别选择两模式2月起报的500 hPa及200 hPa全球位势高度场为预报因子,结合年际增量及经验正交分解(EOF)迭代法建立降尺度模型(分别简称DY_CSM1.1m和DY_CFSv2),研究发现:(1) EOF迭代法中截断解释方差的递增增加了预报因子的协同性和稳定性,从而显著提高预报技巧,并由此确定98%的截断解释方差为模型的最优参数。(2)两模型基于最优参数的预测效果均优于模式原始的降水预测,其中DY_CSM1.1m预测技巧更高,对应29 a距平相关系数(ACC)平均评分可达0.43,尤其在长江干流区域预报效果显著提高。将两模型预测的降水年际增量百分率转换为降水距平百分率时,ACC多年平均评分降为0.27和0.22,仍高于模式原始预测。(3) DY_CSM1.1m的ACC历年评分和长江流域汛期降水年际增量均与西太平洋副热带高压的一系列指数具有高相关性(以西太平洋副高脊线位置指数为例,DY_CFSv2则无此关系),因此BCC_CSM1.1m在西太平洋地区模拟性能优于CFSv2是导致该模式降尺度后预报技巧更高的重要原因,这一点在典型洪涝年1998和2020年中得以佐证。
基金National Natural Science Foundation of China (41790471, 41175065)National Key Research and Development Program of China (2016YFA0602200, 2012CB955203, 2013CB430202).
文摘The prediction skill of Arctic Oscillation (AO) in the decadal experiments with the Beijing Climate Center Climate System Model version 1.1 (BCC_CSM1.1) is assessed. As compared with the observations and historical experiments, the contribution of initialization for climate model to predict the seasonal scale AO and its interannual variations is estimated. Results show that the spatial correlation coefficient of AO mode simulated by the decadal experiment is higher than that in the historical experiment. The two groups of experiments reasonably reproduce the characteristics that AO indices are the strongest in winter and the weakest in summer. Compared with historical experiments, the correlation coefficient of the monthly and winter AO indices are higher in the decadal experiments. In particular, the correlation coefficient of monthly AO index between decadal hindcast and observation reached 0.1 significant level. Furthermore, the periodicity of the monthly and spring AO indices are achieved only in the decadal experiments. Therefore, the initial state of model is initialized by using sea temperature data may help to improve the prediction skill of AO in the decadal prediction experiments to some extent.
