How accurately do coupled climate models predict the leading modes of Asian-Australian monsoon interannual variability?
How accurately do coupled climate models predict the leading modes of Asian-Australian monsoon interannual variability?
复制标题
DOI：
10.1007/s00382-007-0310-5
复制
发表时间：
2008-05
期刊：
影响因子：
4.6
通讯作者：
Bin Wang;June‐Yi Lee;I. Kang;J. Shukla;J. Kug;Arun Kumar;J. Schemm;Jing‐Jia Luo;T. Yamagata-T.-Yama
中科院分区：
文献类型：
--
作者：
Bin Wang;June‐Yi Lee;I. Kang;J. Shukla;J. Kug;Arun Kumar;J. Schemm;Jing‐Jia Luo;T. Yamagata-T.-Yama
准确预测亚洲-澳大利亚季风（A-AM）季节变化是气候预测中最重要和最具挑战性的任务之一。为了了解目前A-AM降水预测精度低的原因，本研究力求确定十个最先进的大气-海洋-陆地耦合气候模型及其多模式集合（MME）在多大程度上能够捕获A-AM降水变化的两种观测到的主要模式——这两种模式占回顾性预测期间总年际方差的43%。 1981-2001。第一种模式与厄尔尼诺南方涛动 (ENSO) 中变暖/变冷的转变有关，而第二种模式则使变暖/变冷提前约 1 年，标志着 ENSO 的先兆条件。第一种模态具有很强的两年一次趋势，反映了热带两年一次振荡（Meehl in J Clim 6:31–41, 1993）。我们表明，MME 季节性降水异常的 1 个月超前预测捕获了前两种主要的变化模式，在季节性演变的空间模式和逐年时间变化以及它们与 ENSO 的关系方面具有高保真度。 MME 显示出在强厄尔尼诺现象成熟之前大约五个季节捕获第二模式 ENSO 前兆的潜力。然而，MME 低估了两种模式的总方差以及第一种模式的两年趋势。这些模型难以捕捉海洋大陆的降水和 ENSO 衰减阶段的沃克型遥相关，这可能在一定程度上导致季风“春季预测障碍”（SPB）。 NCEP/CFS模式后报结果表明，随着超前时间的增加，第一模态的分数方差增加，表明A-AM降雨的长超前可预报性主要来自ENSO可预报性。在 CFS 模型中，第一主成分的相关性技能在 6 个月内保持在 0.9 左右，然后迅速下降，但对于空间模式，它在整个北方春季表现出下降。这项研究发现了两个令人惊讶的发现。首先，耦合模型的 MME 预测比 ERA-40 和 NCEP-2 再分析数据集更好地捕获了降水变率的前两种主要模式，这表明将大气视为奴隶可能本质上无法模拟强降水地区的夏季季风降雨变化（Wang 等人，J Clim 17:803–818, 2004）。建议今后结合大气和海洋模型进行再分析。其次，虽然 MME 总体上优于任何单个模型，但 CFS 集合后报在两年趋势和异常幅度方面优于 MME，这表明 MME 预测技巧的提高是以高估主导模式的分数方差为代价的。还讨论了其他未决问题。
Accurate prediction of the Asian-Australian monsoon (A-AM) seasonal variation is one of the most important and challenging tasks in climate prediction. In order to understand the causes of the low accuracy in the current prediction of the A-AM precipitation, this study strives to determine to what extent the ten state-of-the-art coupled atmosphere-ocean-land climate models and their multi-model ensemble (MME) can capture the two observed major modes of A-AM rainfall variability–which account for 43% of the total interannual variances during the retrospective prediction period of 1981–2001. The first mode is associated with the turnabout of warming to cooling in the El Niño-Southern Oscillation (ENSO), whereas the second mode leads the warming/cooling by about 1 year, signaling precursory conditions for ENSO. The first mode has a strong biennial tendency and reflects the Tropical Biennial Oscillation (Meehl in J Clim 6:31–41, 1993). We show that the MME 1-month lead prediction of the seasonal precipitation anomalies captures the first two leading modes of variability with high fidelity in terms of seasonally evolving spatial patterns and year-to-year temporal variations, as well as their relationships with ENSO. The MME shows a potential to capture the precursors of ENSO in the second mode about five seasons prior to the maturation of a strong El Niño. However, the MME underestimates the total variances of the two modes and the biennial tendency of the first mode. The models have difficulties in capturing precipitation over the maritime continent and the Walker-type teleconnection in the decaying phase of ENSO, which may contribute in part to a monsoon “spring prediction barrier” (SPB). The NCEP/CFS model hindcast results show that, as the lead time increases, the fractional variance of the first mode increases, suggesting that the long-lead predictability of A-AM rainfall comes primarily from ENSO predictability. In the CFS model, the correlation skill for the first principal component remains about 0.9 up to 6 months before it drops rapidly, but for the spatial pattern it exhibits a drop across the boreal spring. This study uncovered two surprising findings. First, the coupled models’ MME predictions capture the first two leading modes of precipitation variability better than those captured by the ERA-40 and NCEP-2 reanalysis datasets, suggesting that treating the atmosphere as a slave may be inherently unable to simulate summer monsoon rainfall variations in the heavily precipitating regions (Wang et al. in J Clim 17:803–818, 2004). It is recommended thatfuture reanalysis should be carried out with coupled atmosphere and ocean models. Second, While the MME in general better than any individual models, the CFS ensemble hindcast outperforms the MME in terms of the biennial tendency and the amplitude of the anomalies, suggesting that the improved skill of MME prediction is at the expense of overestimating the fractional variance of the leading mode. Other outstanding issues are also discussed.