Precalibrating an intermediate complexity climate model
Precalibrating an intermediate complexity climate model
复制标题
DOI：
10.1007/s00382-010-0921-0
复制
发表时间：
2011-10-01
期刊：
影响因子：
4.6
通讯作者：
Rougier, Jonathan
中科院分区：
文献类型：
--
作者：
Edwards, Neil R.;Cameron, David;Rougier, Jonathan
可信的气候预测需要对不确定性进行合理的量化，但完整的贝叶斯校准需要对先验概率分布和协方差进行详细估计，这在实践中很难获得。我们描述了一种简化的程序，称为预校准，它提供了气候预测不确定性的近似量化，并且只需要确定某些输入和输出的无可争议的不可信值。将该方法应用于大西洋经向翻转环流（AMOC）的中等复杂模式模拟，证实了淡水强迫输入空间中存在一个悬崖边缘突变。当考虑到另外14个参数的不确定性时，一个难以置信的AMOC-off区域仍然是模型动力学的鲁棒特征，但发现其位置强烈依赖于其他参数的值。
Credible climate predictions require a rational quantification of uncertainty, but full Bayesian calibration requires detailed estimates of prior probability distributions and covariances, which are difficult to obtain in practice. We describe a simplified procedure, termed precalibration, which provides an approximate quantification of uncertainty in climate prediction, and requires only that uncontroversially implausible values of certain inputs and outputs are identified. The method is applied to intermediate-complexity model simulations of the Atlantic meridional overturning circulation (AMOC) and confirms the existence of a cliff-edge catastrophe in freshwater-forcing input space. When uncertainty in 14 further parameters is taken into account, an implausible, AMOC-off, region remains as a robust feature of the model dynamics, but its location is found to depend strongly on values of the other parameters.