David Hein (acknowledgment to Dr. Carol McSweeney )

Presentation on theme: "David Hein (acknowledgment to Dr. Carol McSweeney )"— Presentation transcript:

1 Selecting QUMP sub-sets for Regional Modelling Experiments 选择区域模拟实验的QUMP子集
David Hein (acknowledgment to Dr. Carol McSweeney ) ACCC Meeting 3rd March 2010

2 Table of Contents 目录 Introduction to Ensemble Regional Prediction and QUMP 区域预测集合与QUMP的简介 Key Questions 主要研究问题 Selecting ‘best performing’ models from the Hadley Centre QUMP GCM ensemble for South East Asia 根据Hadley Centre关于东南亚研究的QUMP GCM集合 选择运行最佳的模型

3 Systematic sampling of members that represent the full range of QUMP regional sensitivity
Does the QUMP ensemble represent the same range of regional sensitivity/uncertainty for South East Asia in models assessed in IPCC AR4 (i.e. the CMIP3 models)? QUMP集合是否代表与IPCC AR4评定的东南亚模型中区域敏感性/不确定性有同一范围（如CMIP3模型） Some conclusions and arising issues 结论和所引发的问题

4 Ensemble regional prediction: Issues and approaches 区域预测集合：问题和方法
The major uncertainties in simulated broad-scale climate changes come from global climate models 模拟大规模气候变化的较大不确定性来自全球气候模型 Global climate models provide information which is often too coarse for applications--thus downscaling is required 全球气候模型提供的信息对于应用来说时常过于粗糙，因此要求缩小模型规模 To provide the best possible information currently available we need to: 要提供最佳现行可应用信息，我们需要 Consider the full range of simulated climate changes from the global climate models 考虑全球气候模型中各种模拟气候变化 Downscale these to provide information relevant to applications which accounts for this range of possible future climate changes 缩小模型中那些应用于影响未来气候变化相关信息的部分

5 Quantifying Uncertainty in Model Predictions: 量化模型预测不确定性
Basic explanation:基本说明 Take a global climate model (HadCM3). 应用全球气候模型（HadCM3） Ask experts which of the parameters in the model are (a) uncertain and (b) important. 咨询专家模型中哪些参数是不确定的和重要的 Run multiple simulations with different values of those parameters 在不同参数值下运行出各种模拟值 More technical explanation:更多技术说明 Parameter perturbation: Changing uncertain aspects of model formulation within plausible ranges to create new GCMs which can provide different, but hopefully plausible, projections of climate change 参数扰动： 在可信域内改变模型制定的不确定方面，创造一个可提供不同的且希望可行的气候变化预测的GCMs（大气环流模型） For example, the fall velocity of cloud ice crystals is assumed to be 2ms-1 in standard HadCM3, but is set to 1ms-1 in one of the perturbed GCMs, a value within the bounds of observational uncertainty 例如，冰晶云下降速度被假定为2米每秒（在HadCM3标准下），但在扰动GCMs模型下被设定为1米每秒，这是在能观测到不确定性范围内的一个值 This “perturbed physics” approach allows uncertainties in various components of the model to be systematically explored. 这种“扰动物理”方法承认在模型不同构成的不确定性是可系统探讨的。

6 QUMP data for downscaling 缩小QUMP数据规模
17 members of the QUMP ensemble QUMP集合中的17个子集 1 member (Q0) is the unperturbed physics member, acting as a control 一个子集（Q0）是非扰动物理子集，作为对照组 16 other members have a variety of perturbed physics parameters 其他16个子集拥有扰动物理参数 Each member is uniquely identified according the pattern Q_, e.g. Q1, Q2, Q13, Q16. 通过模式Q_唯一确定每个子集，例如Q1, Q2, Q13, Q16. Output data from all 17 is available for downscaling by regional climate models such as PRECIS 从17个子集中输出的数据都是可用于缩小区域气候模型规模的，例如PRECIS

7 Selecting QUMP sub-sets for Regional Modelling Experiments: a case study for Southeast Asia 选择区域模拟实验的QUMP子集：一个东南亚案例研究 Task: 任务 UNDP project for IMHEN Vietnam (Institute of Meteorology, Hydrology and Environment) 越南IMHEN（气象，水文，环境研究所） UNDP（开发计划署）项目 Recommend a subset of the 17 QUMP GCM members which span the range of sensitivities in the full QUMP ensemble for simulations over South East Asia, and exclude any members which simulate a less realistic control climate 推荐一个跨度为全部QUMP敏感性范围并包含17个QUMP GCM的成员的子集用以模拟整个东南亚，这个子集去除了模拟出较弱真实气候的子集 Develop/suggest criteria that other institutions might use to select QUMP subsets for regional modeling experiments 开发/建议一个标准，其他研究机构可以用来选择区域模拟实验的子集

8 Key Questions主要问题 Do some QUMP members simulate the climate of South East Asia (SEA) better than others?是否QUMP集合能够比其他参数更好的模拟东南亚气候？ Metrics for precipitation, seasonal development of Asian summer monsoon, temperature, typhoons?降水，亚洲夏季季风季节性发展，温度的度量 How do we exclude better/worse ensemble members?我们怎样去除好的/坏的整体性元素？ Which QUMP members are least/most sensitive in SEA?在东南亚哪个QUMP元素是最敏感/最不敏感的？ Metrics for precipitation, monsoon sensitivity, temperature, typhoons?度量降水，季风灵敏性，温度，台风的指标？ How do we select members to encompass a range of sensitivities and patterns of change? 我们怎样选择包含敏感性和制式改变的元素？ How does the QUMP ensemble compare with the CMIP3 ensemble in SEA and what does this mean for PRECIS users?在东南亚QUMP集合与CMIP3集合相比如何？这对PRECIS用户意味着什么？ Do sensitivities span the same range as CMIP3? If not, how do we approach this in our experimental designs?敏感性跨度是否和CMIP3同一范围？如果不是，我们在实验设计中应该如何操作使两者接近呢？ How does model performance compare between QUMP and CMIP3 in Southeast Asia? Do the QUMP models do a good job, and are there particular CMIP3 members which perform particularly well/badly in the region?在东南亚QUMP与CMIP3模型性能相比如何？QUMP模式更好呢还是存在某些特殊的CMIP3元素在一些地区性能更好/更差？ Broader Issue: Advising other regional modellers on selecting QUMP subsets to downscale扩展问题：建议其他地区建模人员在选择QUMP子集问题上尽量缩小规模

9 Selecting the ‘Best’ ensemble members for Southeast Asia为东南亚选择最优的整体参数

10 Which QUMP members simulate the most realistic climate for Southeast Asia)? 哪项QUMP元素模拟了最真实的东南亚气候
Generic Indicator of Monsoon performance Q15 Q11 Q6 Q2 Q8 Q12 Generic Indicator of precipitation performance

11 Regions… 地区

12 Monsoon Onset 季风爆发 Tendency to simulate too strong a monsoon flow (a known systematic error in Hadley centre models, e.g. Martin and Soman, 2000)模拟过强季风流量的倾向（ 在Hadley centre models一个已知的系统性错误，例如Martin and Soman, 2000 ） CMIP3 models have a tendency to underestimate monsoon strength (as does ECHAM4!)CMIP3模型有低估季风强度的倾向（如ECHAM4!） All do a reasonably good job at simulating rainfall in the region 在模拟区域降雨量的工作中都有不错表现 Tendency to over-estimate September rainfall高估九月降雨量的倾向 Those that best represent the characteristics of the monsoonal flow don’t necessarily also best represent the local rainfall…那些最能代表季风流量的特点不一定最能代表当地降雨量 Difficult to ‘pick off worst models’ 难以剔除最差的模型

13 Metrics? 指标？ Monsoon Indices as useful metrics for monsoon strength and variability?季候风指数是否能作为测量季候风强度和变化性的有用指标？ South East Asian Monsoon index (SEAM)东南亚季候风指数（SEAM） U850 (5-15N,90-130E) –U850 ( N, E) (Wang and Fan, 1999)

14 Metrics and ranking…? 指标和排序
Top Q0, Q13 Model SEAM Rank variance Var JJAS Pr RMSE (Vietnam) JJAS Pr Corr (SEA) SUM of RANKS Q0 8.6 5 10 1.9 2 0.89 3 20 Q1 13.5 16 6.5 1 2.1 9 0.86 35 Q2 12.7 13 6 4 2.3 15 0.87 7 39 Q3 12.6 12 8 2.0 0.9 25 Q4 13.4 4.7 33 Q5 11.9 11 5.1 1.8 23 Q6 7.4 27 Q7 15.5 17 0.84 14 45 Q8 8.7 7.7 2.2 38 Q9 5.9 11.5 2.4 0.85 43 Q10 32 Q11 10.1 11.1 Q12 9.2 Q13 7.6 0.88 21 Q14 6.4 16.1 41 Q15 9.8 13.9 37 Q16 6.1 0.83 46 Bottom Q7, Q9, Q16

15 Selecting QUMP models that sample the full range of regional sensitivities/uncertainties (all changes w.r.t , SRES A1B) 选择QUMP模型，以区域敏感性/不确定性的全程范围为样本（所有自2070到2100的变化，关于 的变化参照A1B）

16 Selecting QUMP members that span the full range of sensitivities
Selecting QUMP members that span the full range of sensitivities.选择跨度敏感性全程范围的QUMP子集合 Q13/Q14 consistently most sensitive in SEA 在东南亚Q13/Q14始终是最敏感的 Q1,2,3,4,5 least sensitive (less consistent) in SEA 在东南亚Q1,2,3,4,5最不敏感（也不那么始终一致） South East Asia Vietnam

17 Changes in the Asian Summer Monsoon in the QUMP ensemble 亚洲夏季季候风在QUMP集合内的变化
Q10/Q16 are the only 2 members that indicate a decrease in monsoon intensity (as indicated by SEAM) Q10/Q16是仅有的两个表明季候风强度有所减弱的子集合（由东南亚季候风表明） Q16 is also our ‘worst’ performing model, as indicated by ranking of metrics…? (Although both Q10 and Q16 do well on the SEAM metrics…) 指标的排序也表明，Q16是我们运行最糟糕的模型。（尽管Q10和Q16作为东南亚季候风指数的表现都很不错）

18 Changes in Monsoon behaviour -2070-2100 (A1B) 2070-2100季候风习性变化（A1B）
Monsoon response is clearly different in Q10/Q16季候风的反应在Q10与Q16中是明显不同的。 Although this doesn’t result in a clear difference in response in rainfall amount, it has an impact on spatial patterns of change…尽管这并不能导致在降雨量上的一个明显不同的回应，但它对变化的空间模式有显著影响…

19 Does the QUMP ensemble represent a similar range of uncertainty in Southeast Asia to CMIP3? QUMP集合是否代表了与CMIP3在东南亚有相似的不确定性范围？

20 QUMP vs. CMIP3 regional sensitivity (SEA / Vietnam) QUMP与CMIP3区域敏感性的对比（东南亚/越南）
QUMP tends to be warmer than CMIP (particularly in Vietnam), but sensitivities span similar range QUMP通常比CMIP暖和（尤其是在越南），但敏感度区间相似 More significantly, QUMP tends to be drier than CMIP3 in Vietnam – over come by ensuring we include Q11?更重要的是，在越南，QUMP通常比CMIP3更干旱，是否能通过确保我们计入Q11来加以克服？

21 Summary / Conclusions 小结/结论

22 ‘Conclusions’ so far… 目前得出的结论
Model performance within QUMP ensemble is generally reasonable with respect to climate in SEA, including simulation of Monsoon… 模型在QUMP集合内关于东南亚气候，包括模拟季候风的运行是大致合理的。 No members ‘stand-out’ as ‘poor’ – errors tend to affect members similarly (e.g. late onset of summer rainfall in Vietnam, over-active summer monsoon) 没有哪个子集合特别突出，同样也没有特别差的，误差常常影响子集合。（例如，晚发性的越南夏季降雨量，过度活跃的夏季季候风） Selection criteria might be based on ranking, but no clear winners/losers 选择标准可基于排序，但并没有明确的优劣

23 Perhaps we should include all variants and base selection on sensitivity?
或许我们应该包含所有变量并且基于敏感性做出选择？ Range of Sensitivities敏感性范围 Need to look at the magnitude of change, but also identify ensemble members that show different patterns of change. E.g. Q10/Q16… 需要依变化大小而定，但也要明确表明不同的变化模式的全部集合。例如Q10/Q16… Q13/Q14 characterise models that have high magnitude precipitation response, and a ‘typical’ pattern of change Q13/Q14刻画了有大量降水反应的模型，以及变化的一个典型模式 Q10/Q16 characterise models that have a moderate magnitude of precipitation response and a more ‘a typical’ pattern of change… related to differences in response of the monsoon…Q10/16刻画了有中度降水反应的模型，和变化的一个更典型的模式…与对季候风的不同反应有关。 QUMP vs. CMIP3 sensitivity QUMP与CMIP3敏感性的对比 QUMP gives a spread of regional sensitivities that is reasonably representative of CMIP3, although tend to be ‘warmer’, and tend towards drying in JJAS – overcome by including Q11 (wettest model in JJAS)? QUMP给出了作为CMIP3合理代表的区域敏感性的扩展，尽管有变得更暖和，并在JJAS变得干旱的趋势—能否通过计入Q11来加以克服（在JJAS的多雨模式）？

24 End终