全球气候模式中气候变化预测预估的不确定性

The uncertainty in climate change projections by global climate models

人类活动造成的以全球变暖为主要特征的气候变化对生态系统和人类社会造成严重的影响。全球气候模式正日益成为研究当前气候特征和现象、了解过去气候演变规律及预估未来气侯变化不可替代的、最具潜力的工具。气候模式已被广泛运用于全球和区域未来气候变化的研究中。未来情景的不确定性、气候系统内部的自然变率的不确定性和表征气候过程的不确定性是造成气候预测预估不确定性的主要来源,而概率分布是一个很好地表示气候变化预测不确定性的方式。介绍了贝叶斯多模式推断方法来描述气候变化预估不确定性的理论框架,并以中国区域为例,利用IPCC-AR5的气候模式数据,通过贝叶斯多模式推理方法预估未来中国区域的南北方两个典型流域(海河和珠江流域)未来气候变化情况。结果表明:中国区域都将呈现出变暖的趋势,在RCP2.6、RCP4.5和RCP8.5情景下,温度变化趋势分别为0.91±0.30°C/100a、2.41±0.77°C/100a、6.08±1.01°C/100a;降水也呈现出增加的趋势,三种情景下的趋势分别为(5.58±2.96)%/100a、(10.30±4.30)%/100a和(15.90±6.68)%/100a;中国海河流域的年降水量在2020s和2040s都将出现增加的趋势,珠江流域则在2020s略有降低,2040s开始增加。并且在2020s和2040s发生干旱和极端暴雨等极端降水事件的概率同时增加。

Global warming caused by human activities has devastating impacts on the Earth＇s eco-system and the human society. Climate models are the primary tools available for investigating the response of the climate system to various forcings and for making climate change projections into the future. Climate change projections are plagued by various sources of uncertainties, including the greenhouse gases emission scenarios, the internal variability of the climate system, and the representation of the climate processes. To cope with future climate changes, one must quantify those uncertainties properly. Probability distribution is an excellent way to describe the uncertainties. We presented the Bayesian multi-model inference methodology to quantify uncertainty in the climate change projections. We applied this Bayesian framework to assess the climate change projections contained in IPCC-AR5 in the continental China and in two typical large basins in China （Haihe and Pearl River）. The results showed that warming is expected all over China under all emissions scenarios. The warming trend from 2006 to 2099 in China is 0.91±0.30 ℃/100a, 2.41 ±0.77 ℃/100a, and 6.08± 1.01 ℃/100a under RCP2.6, RCP4.5 and RCP8.5 scenarios, respectively. Precipitation in China is also projected to be increasing during the 21st century by （5.58±2.96）%/100a, （10.30±4.30）%/100a, and （15.90±6.68）%/100a for the three RCP scenarios, respectively. Under climate change, extreme temperature and precipitation events are projected to be more probable in the future with the probability distribution shifting to the right for both temperature and precipitation.