Indian Ocean Climate Initiative Stage 3 (IOCI3) - Very High Resolution Modelling of Hot Spell Trends and Projections for South-West and North-West Western Australia

IOCI3, a climate research collaboration between CSIRO, the Bureau of Meteorology (BoM) and the Western Australian Government, produced maps of mean hot spell intensity, frequency and duration for the 1958-2010 period using estimates derived from statistical models. They also produced maps of trends in hot spell intensity, frequency and duration for this time period. In addition they provided maps of mean hot spell thresholds, intensity, frequency and duration for the 1981-2010 period using estimates derived from statistical models, and projections of these characteristics for the 2070-2099 period under the A2 greenhouse gas (GHG) emissions scenario (described in the IPCC Special Report on Emissions Scenarios [SRES]), as well as the difference between these two periods." Results are provided in the JPEG file format.\nLineage: High quality station data as well as quarter-degree gridded (0.25°× 0.25° resolution) daily maximum temperature data from BoM Australian Water Availability Project (AWAP) were used to produce these results. Hot spell temperature thresholds were selected using statistical methods. Hot spell occurrence (frequency) was modelled by a Poisson process, hot spell intensity by a generalized Pareto distribution, and hot spell duration through a geometric distribution. The Generalized Linear Model framework was used to estimate the parameters in the model for hot spells. This method was applied to daily maximum temperature data simulated from the CSIRO Cubic Conformal Atmospheric Model (CCAM) for both the present-day and possible future climate under the SRES A2 GHG emissions scenario. The CCAM was nested in the CSIRO Mk3.0 Global Climate Model host for the SRES A2 scenario.\nCaveats & limitations: The hot spell projections should be seen as initial estimates only, and they should not be used for making impact, vulnerability and risk assessments. They were made using only one climate model (CCAM); more work using an ensemble of global and regional climate model results is required to provide more robust projections of hot spells in Western Australia.\n\nExtreme events are by definition rare, and analysis relies on partial (extreme) datasets (e.g., daily maximum temperatures higher 35 °C). In addition, estimating extremes necessitates extrapolating beyond such relatively small observed records. Consequently, the uncertainty associated with these projections of extremes is large, especially when extrapolating from a small dataset. To produce these projections we used AWAP data was used to overcome data shortages. However, the methods used to construct the AWAP dataset (interpolation) may smooth out some extreme values; this may lead to an underestimation of extremes in some cases. To these uncertainties are added the uncertainties inherent in the use of climate models. \n

IOCI3是由澳大利亚联邦科学与工业研究组织（CSIRO）、澳大利亚气象局（Bureau of Meteorology, BoM）与西澳大利亚政府合作开展的气候研究项目。该团队基于统计模型估算结果，绘制了1958-2010年期间热浪强度、发生频次与持续时长的均值分布图；同时生成了该时段内热浪强度、频次与持续时长的变化趋势分布图。此外，团队还基于统计模型估算结果，提供了1981-2010年期间热浪阈值、强度、频次与持续时长的均值分布图；并针对2070-2099年时段，在政府间气候变化专门委员会（IPCC）《排放情景特别报告》（SRES）中定义的A2温室气体（GHG）排放情景下，给出了上述热浪特征的预测结果，以及两个时段间的差异分布图。所有结果均以JPEG（联合图像专家组格式）文件格式提供。
数据溯源：本研究采用了源自澳大利亚气象局澳大利亚水资源可利用性项目（Australian Water Availability Project, AWAP）的高质量台站数据与0.25°×0.25°分辨率的逐日最高气温格点数据，以生成上述结果。热浪温度阈值通过统计方法确定。热浪发生频次采用泊松过程（Poisson process）建模，热浪强度采用广义帕累托分布（Generalized Pareto Distribution）建模，热浪持续时长则通过几何分布（Geometric Distribution）拟合。研究采用广义线性模型（Generalized Linear Model）框架估算热浪模型的参数。针对SRES A2温室气体排放情景下的当代与未来气候场景，本研究将CSIRO三次共形大气模式（CSIRO Cubic Conformal Atmospheric Model, CCAM）模拟得到的逐日最高气温数据应用于上述方法；其中，在SRES A2情景中，CCAM嵌套于CSIRO Mk3.0全球气候模式宿主系统中。
注意事项与局限性：本研究提供的热浪预测结果仅为初步估算值，不得用于影响评估、脆弱性分析与风险评估。本次预测仅采用了单一气候模式（CCAM），后续需基于全球与区域气候模式集合的结果开展更多研究，以得到西澳大利亚地区更可靠的热浪预测结果。
极端事件在定义上属于稀有事件，其分析依赖于偏态的极端数据集（例如日最高气温高于35℃的样本）。此外，极端值估算需要在相对有限的观测记录之外进行外推。因此，这些极端事件预测结果伴随的不确定性较大，尤其当基于小型数据集进行外推时，不确定性更为显著。为生成本研究的预测结果，我们采用了AWAP数据集以弥补数据短缺问题。然而，AWAP数据集的构建方法（插值法）可能会平滑部分极端值，这在部分场景下可能导致极端值被低估。除上述不确定性外，气候模式本身固有的不确定性也会对结果产生影响。