Evaluating model performance and constraining uncertainty using a processed-based framework for Southern African precipitation in historical and future climate projections
Evaluating model performance and constraining uncertainty using a processed-based framework for Southern African precipitation in historical and future climate projections
复制标题
DOI：
--
复制
发表时间：
2017-05
期刊：
影响因子：
--
通讯作者：
M. Lazenby
中科院分区：
文献类型：
--
作者：
M. Lazenby
本论文开发了一个创新的过程为基础的分析当代模式的性能南部非洲的降水。由于降水的各种影响和驱动因素的复杂性，这一地区通常研究不足，也没有得到充分的了解。历史模拟的降水进行评估，包括主要驱动程序，来源的偏差和主导模式的年际变化。南印度洋辐合带（SIOCZ），一个大规模的，南半球夏季降雨功能延伸到西南印度洋南部非洲，被评估为历史模拟的功能感兴趣。大多数CMIP 5模式模拟SIOCZ特征，但通常过于区域性，并且在陆地和邻近的印度洋之间中断。大陆上的过度降水可能与安哥拉低压周围的低空水汽通量过高有关，这几乎完全是由于模式环流偏差造成的。南部非洲降水的驱动因素包括三个主要的水汽通量输送途径，它们来自SIOHP和SAOHP周围的气流和季风。SIOCZ的年际变化表现为明显的偶极子型，表明SIOCZ的东北-西南运动。这种转变的驱动因素与观测中的厄尔尼诺南方涛动和副热带印度洋偶极子显著相关。然而，模型不能很好地捕捉这些遥相关，限制了模型表示变异性的信心。绝大多数人口的农业生产严重依赖南部非洲的降水。因此，降水量的时空变化对于识别和理解并最终提供有关该地区水安全的有用气候信息至关重要。本论文建立了南部非洲主要的气候变化信号（OND和DJF），量化了南部非洲降水变化的主要区域机制。这些变化的稳健性和可信度也被量化。南部非洲降水量最显著的预测变化是在夏季前明显的干燥信号（OND）。这意味着雨季推迟，影响到播种和收获时间。确定了SIOCZ的未来预测，这表明向北移动约200公里。一个偶极子模式的降水湿/干是显而易见的，在最大降雨量的气候轴的北部发生湿润，因此意味着ITCZ向北移动，与SIOCZ移动一致。使用分解方法，它建立了ΔP的偶极子图案出现在很大程度上从动力分量，其中最不确定性，特别是在西南印度洋。陆地降水量的变化不仅受动力学变化的驱动，而且还受热力学贡献的驱动，这意味着陆地和海洋区域的预计变化需要不同的方法。印度洋的SST变暖模式证实了西南印度洋的最暖变湿机制，这在两个关键季节都很强劲。通过主成分的跨模型相关图来理解相干模型行为，从而识别相干变暖模式的模式。跨模型的诊断变量的复合分析说明了驱动预测降水变化的模式。陆地上的干旱比西南印度洋上的干旱更严重。OND中清晰的稳健干燥信号，但幅度不确定。不确定性的驱动因素包括SST模式的变化，这会调节大气环流模式。因此，减少不确定性依赖于在气候模型中准确表示这些过程，以变得更加稳健。气候科学家和政策制定者都希望减少未来预测的不确定性。没有一个特定的方法是一致同意的，但在这篇论文中分析了一种方法。未来的降水预测的Understanding解决使用基于过程的模型排名框架。几个指标最适用于南部非洲气候的选择和排名，其中包括方面的平均状态和变化。通过蒙特卡罗方法对CMIP 5模型数据集中“顶级”性能模型的各个子样本进行敏感性测试。不确定性显着降低时，特定的子集的“顶级”表现模式被选中，但只有南方夏季在大陆。其结果的含义，潜在的价值是建立在执行一个基于过程的模型在南部非洲的排名。然而，在这种方法变得可行、足够可信和健全之前，还需要进行更多的调查。模型传播的减少是另外建立在SIOCZ预测，从而模型的变化过程表现出协议，尽管不同的初始SIOCZ条件。因此，模式过程的收敛性和一致性建立在SIOCZ的预测变化，无论初始气候偏差。
This thesis develops an innovative process-based analysis of contemporary model performance of precipitation over southern Africa. This region is typically understudied and not fully understood due to the complexity of various influences and drivers of precipitation. Historical simulations of precipitation are assessed including principal drivers, sources of biases and dominant modes of interannual variability. The South Indian Ocean Convergence Zone (SIOCZ), a large-scale, austral summer rainfall feature extending across southern Africa into the south-west Indian Ocean, is evaluated as the feature of interest in historical simulations. Most CMIP5 models simulate an SIOCZ feature, but are typically too zonally oriented and discontinued between land and the adjacent Indian Ocean. Excessive precipitation over the continent is likely associated with excessively high low-level moisture flux around the Angola Low, which is almost entirely due to model circulation biases. Drivers of precipitation over southern Africa include three dominant moisture flux transport pathways which originate from flow around the SIOHP and SAOHP and monsoon winds. Interannual variability in the SIOCZ is shown by a clear dipole pattern, indicative of a northeast-southwest movement of the SIOCZ. Drivers of this shift are significantly related to the El Nino Southern Oscillation and the subtropical Indian Ocean dipole in observations. However models do not capture these teleconnections well, limiting confidence in model representation of variability. A large majority of the population rely heavily on precipitation over southern Africa for agricultural purposes. Therefore spatial and temporal changes in precipitation are crucial to identify and understand with intentions to ultimately provide useful climate information regarding water security over the region. Key climate change signals over southern Africa are established in this thesis (OND and DJF), in which the dominant regional mechanisms of precipitation change over southern Africa are quantified. Robustness and credibility of these changes are additionally quantified. The most notable projected change in precipitation over southern Africa is the distinct drying signal evident in the pre-summer season (OND). This has the implication of delaying the onset of the rainy season affecting planting and harvesting times. Future projections of the SIOCZ are determined, which indicate a northward shift of approximately 200km. A dipole pattern of precipitation wetting/drying is evident, where wetting occurs to the north of the climatological axis of maximum rainfall, hence implying a northward shift of the ITCZ, consistent with the SIOCZ shift. Using a decomposition method it is established that ΔP’s dipole pattern emerges largely from the dynamic component, which holds most uncertainty, particularly over the south-west Indian Ocean. Changes in precipitation over land are not solely driven by dynamical changes but additionally driven by thermodynamic contributions, implying projected changes over land and ocean regions require different approaches. SST patterns of warming over the Indian Ocean corroborate the warmest-get-wetter mechanism driving wetting over the south-west Indian Ocean, which is robust in both key seasons. Coherent model behaviour is understood via across model correlation plots of principal components, whereby patterns of coherent warming patterns are identified. Composite analyses of diagnostic variables across models illustrate patterns driving projected precipitation changes. Drying is more robust over land than over the south-west Indian Ocean. Clear robust drying signal in OND, however magnitude is uncertain. Drivers of uncertainty include SST pattern changes, which modulate atmospheric circulation patterns. Therefore reductions in uncertainty rely on the accurate representation of these processes within climate models to become more robust. There is a desire from both climate scientists and policy-makers to reduce uncertainty in future projections. No one particular methodology is unanimously agreed upon, however one approach is analysed in this thesis. Uncertainties of future precipitation projections are addressed using a process-based model ranking framework. Several metrics most applicable to southern African climate are selected and ranked, which include aspects of both mean state and variability. A sensitivity test via a Monte Carlo approach is performed for various sub-samples of “top” performing models within the CMIP5 model dataset. Uncertainty is significantly reduced when particular sub-sets of “top” performing models are selected, however only for austral summer over the continent. The result has the implication that potential value is established in performing a process-based model ranking over southern Africa. However additional investigation is required before such an approach may become viable and sufficiently credible and robust. Reductions in model spread are additionally established in SIOCZ projections, whereby model processes of change exhibit agreement, despite differing initial SIOCZ conditions. Therefore model process convergence and coherence is established with respect to projected changes in the SIOCZ, irrespective of initial climatology biases.