What can we know about future precipitation in Africa? Robustness, significance and added value of projections from a large ensemble of regional climate models
What can we know about future precipitation in Africa? Robustness, significance and added value of projections from a large ensemble of regional climate models
复制标题
DOI：
10.1007/s00382-019-04900-3
复制
发表时间：
2019-11-01
期刊：
影响因子：
4.6
通讯作者：
Hewitson, Bruce
中科院分区：
文献类型：
--
作者：
Dosio, Alessandro;Jones, Richard G.;Hewitson, Bruce
我们利用一个大的区域气候模式（RCMs）的集合，从协调区域气候降尺度实验，探讨两个问题：（1）我们可以知道什么关于未来的降水特征在非洲？以及（2）这些信息是否与驱动全球气候模式（GCMs）的信息不同？通过考虑变化的统计意义和模型对其符号的一致性，我们确定了预测气候变化信号稳健的区域，表明降水特征将发生变化的信心，以及降水统计数据变化不显著的区域。结果表明，当空间平均时，RCMs中值变化通常与GCMs集合一致：即使季节平均降水的变化可能不同，在某些情况下，其他降水特征（例如，强度，频率和持续时间的干旱和潮湿的法术）显示出相同的趋势。当鲁棒变化（即，在比较了全球气候模型和区域气候模型（仅在变化稳健的陆地点上平均变化值）之后，发现两者有惊人的相似之处，这表明，尽管在地理范围上存在一些不确定性，但全球气候模型和区域气候模型预测的未来是一致的。缩小尺度的未来气候预测的潜在增值（即，在埃塞俄比亚高原等地发现了不可忽略的细尺度信息（在较低分辨率模拟中不存在），与GCM结果相比，RCM集合显示平均降水量大幅减少。这种差异可能与更好地代表地形的细节，在大尺度大气环流模型中丢失。GCM-RCM矩阵的异质性对结果的影响也进行了研究;我们发现，对于大多数区域和指数，结果是稳健的或不显著的，它们是如此独立于RCM或GCM的选择。然而，在有些情况下，特别是在中非和西非部分地区，结果是不确定的，即大多数区域协调机制预测的变化具有统计意义，但它们对其符号不一致。在这些情况下，特别是在结果根据RCM明确聚类的情况下，没有简单的方法对模型集合进行二次采样，以减少不确定性或推断出更稳健的结果。
We employ a large ensemble of Regional Climate Models (RCMs) from the COordinated Regional-climate Downscaling EXperiment to explore two questions: (1) what can we know about the future precipitation characteristics over Africa? and (2) does this information differ from that derived from the driving Global Climate Models (GCMs)? By taking into account both the statistical significance of the change and the models' agreement on its sign, we identify regions where the projected climate change signal is robust, suggesting confidence that the precipitation characteristics will change, and those where changes in the precipitation statistics are non-significant. Results show that, when spatially averaged, the RCMs median change is usually in agreement with that of the GCMs ensemble: even though the change in seasonal mean precipitation may differ, in some cases, other precipitation characteristics (e.g., intensity, frequency, and duration of dry and wet spells) show the same tendency. When the robust change (i.e., the value of the change averaged only over the land points where it is robust) is compared between the GCMs and RCMs, similarities are striking, indicating that, although with some uncertainty on the geographical extent, GCMs and RCMs project a consistent future. Potential added value of downscaling future climate projections (i.e., non-negligible fine-scale information that is absent in the lower resolution simulations) is found for instance over the Ethiopian highlands, where the RCM ensemble shows a robust decrease in mean precipitation in contrast with the GCMs results. This discrepancy may be associated with the better representation of topographical details that are missing in the large scale GCMs. The impact of the heterogeneity of the GCM-RCM matrix on the results has been also investigated; we found that, for most regions and indices, where results are robust or non-significant, they are so independently on the choice of the RCM or GCM. However, there are cases, especially over Central Africa and parts of West Africa, where results are uncertain, i.e. most of the RCMs project a statistically significant change but they do not agree on its sign. In these cases, especially where results are clearly clustered according to the RCM, there is not a simple way of subsampling the model ensemble in order to reduce the uncertainty or to infer a more robust result.