cb-oura-1.0 : Generic climate scenarios from bias-adjusted CMIP5 global models

<strong>Context </strong> The need to adapt to climate change is present in a growing number of fields, leading to an increase in the demand for climate scenarios for often interrelated sectors of activity. In order to meet this growing demand and to ensure the availability of climate scenarios responding to numerous vulnerability, impact, and adaptation (VIA) studies, Ouranos is working to create a set of operational multipurpose climate scenarios. The initial version of “Scénarios Génériques” (generic scenarios, acronym cb-oura-1.0) is used mainly in Ouranos’ work to provide a consistent image of the changing climate over the North East of North America, principally the province of Québec. Cb-oura-1.0 was produced in 2016 by downscaling and bias-adjusting a selection of global climate model simulations available through the CMIP5 program. <strong>Climate simulations </strong> Climate simulations in the ensemble Modeling center Acronym Model RCP Status* <strong>College of Global Change and Earth System Science, Beijing Normal University</strong> GCESS BNU-ESM 4.5 s 8.5 s <strong>Canadian Centre for Climate Modelling and Analysis</strong> CCCMA CanESM2 4.5 a 8.5 s <strong>Centro Euro-Mediterraneo per I Cambiamenti Climatici</strong> CMCC CMCC-CMS 4.5 a 8.5 s <strong>Commonwealth Scientific and Industrial Research Organization (CSIRO) and Bureau of Meteorology (BOM), Australia</strong> CSIRO-BOM ACCESS1.3 4.5 s 8.5 a <strong>Institute for Numerical Mathematics</strong> INM INM-CM4 4.5 s 8.5 a <strong>Institut Pierre-Simon Laplace</strong> IPSL IPSL-CM5A-LR 4.5 a 8.5 s IPSL-CM5B-LR 4.5 s 8.5 s <strong>Met Office Hadley Centre</strong> MOHC HadGem2 4.5 s 8.5 s <strong>Max-Planck-Institut für Meteorologie (Max Planck Institute for Meteorology)</strong> MPI-M MPI-ESM 4.5 s 8.5 s <strong>Norwegian Climate Centre</strong> NCC NorESM 4.5 a 8.5 s <strong>NOAA Geophysical Fluid Dynamics Laboratory</strong> NOAA-GFDL GFDL-ESM2M 4.5 s 8.5 s From the complete ensemble of RCP 4.5 and 8.5 driven CMIP5 climate simulations, a selection of 22 simulations (11 per RCP) was made using a clustering ensemble reduction methodology (Casajus et al. 2016). This objective selection method identifies a reduced number of simulations that best represent the overall ensemble. Input criteria for the reduction were the monthly changes between the present (1981-2010) and two future horizons (2041-2070 and 2071-2100), at 15 regions distributed across Canada, for three variables (mean daily maximum temperature, mean daily minimum temperature and total precipitation). An initial selection of 16 simulations shows a distribution projected changes for the 12 (months) x 2 (horizons) x 15 (regions) x 3 (variables) indices that is not statistically different from the complete ensemble. A small number of simulations were subsequently added to have a complete set with both RCPs represented equally (11 members for each emission scenario). <strong>Reference dataset </strong> The bias-adjustment reference (or target) is a gridded observation dataset produced by Natural Resources Canada (McKenney et al., 2011; et Hutchinson et al. ,2009). It uses the ANUSPLIN interpolation method over station observations to derive daily grids of minimum and maximum temperature, as well as total precipitation for the Canadian landmass. The grid has a resolution of 10 km x 10 km and cover the time period from 1950 to 2013. As this dataset is not available over the United States, it was merged with another observation interpolation dataset produced by Livneh et al. (2015) in order to enable the production of bias-corrected climate scenarios covering a portion of the northern United States. <strong>Coverage </strong> The final version of this dataset covers a region covering the Atlantic provinces, Québec, Ontario, Manitoba and Saskatchewan and part of the northern United States: From 120°W to 54°W and from 40°N to 62°N. It contains the daily minimum temperature, daily maximum temperature and daily precipitation flux, covering the period 1950 to 2100. <strong>Bias-adjustment </strong> The global simulations where downscaled to the reference grid using bilinear interpolation and then bias-adjusted with a 1-D quantile mapping method, as described by Gennaretti et al. (2015). A moving window of 31 days was used to adjust each day of the year, using 50 quantiles to define the statistical distributions to match. The long-term linear trends of the temperature variables were preserved explicitly. <strong>Climate indicators </strong> This dataset is used to in the first versions (up to 1.3) of Ouranos’ Climate Portraits website. A selection of 26 seasonal and annual climate indicators were computed from the daily scenarios, using the xclim software package (Logan et al. 2022). The "virtual indicator module" used for the computation is made available here in the "indicators.yml" file. On the Climate Portraits website, the information is presented from three aspects: spatial, temporal and summary. This repository stores the reduced ensemble data as shown on the website. Filenames are constructed as "{aspect}_{indicator}_{season}.nc". Maps (files "spatial_*") : Climate indicators for each bias-adjusted climate simulation and for a given RCP emission scenario are averaged over 30-year horizons. Ensemble percentiles are computed in order to summarize climate model uncertainty. In particular the 10, 25, 50, 75 and 90th percentiles over the 11 members are calculated for each RCP. Timeseries (files "temporal_*") : Climate indicators for each bias-adjusted simulation are averaged spatially over each region for every time step (annual or seasonal). The ensemble statistics are computed by first pooling all regional average values for the 11 members using within a centred 30-year window and then calculating percentile values (same as above) on the pooled data. Summary (files "summary_*") : The indicators are averaged over each region and then over 30-year horizons. The ensemble statistics (same as above) are then computed. In versions 2.x of the app, this data will be presented as "CMIP5". <strong>Data availability </strong> This repository stores the climate indicator ensemble statistics as shown on the Climate Portraits website and described above. The complete daily dataset is too large for this platform. The complete daily dataset is available through the public THREDDS server of the PAVICS platform maintained by Ouranos. This data might be removed in the future. When this is the case, please contact us for data requests.<br> https://pavics.ouranos.ca/twitcher/ows/proxy/thredds/catalog/datasets/simulations/bias_adjusted/cmip5/ouranos/cb-oura-1.0/catalog.html The annual and season indicators of the Climate Portraits website are available on the same server, along with a few more indicators not shown on the app. https://pavics.ouranos.ca/twitcher/ows/proxy/thredds/catalog/birdhouse/ouranos/portraits-clim-1.3/catalog.html <em>Terms of use</em>: Use of this dataset should be acknowledged as 'Data produced and provided by the Ouranos Consortium on Regional Climatology and Adaptation to Climate Change'. Furthermore, the modeling groups from which the bias-adjusted climate scenarios were constructed must also be acknowledged, please refer to: The Coupled Model Intercomparison Project https://pcmdi.llnl.gov/mips/cmip5/citation.html.<strong>背景</strong> 越来越多的领域亟需适应气候变化，这使得针对诸多相互关联活动领域的气候情景需求持续增长。为满足这一日益增长的需求，并确保为大量脆弱性、影响与适应（Vulnerability, Impact, and Adaptation, VIA）研究提供可用的气候情景，Ouranos正致力于构建一套业务化多用途气候情景集。其首个版本“Scénarios Génériques”（通用情景，缩写cb-oura-1.0）主要用于Ouranos的相关工作，以呈现北美东北部（主要为魁北克省）气候变化的一致性图景。Cb-oura-1.0于2016年通过对耦合模式比较计划第五阶段（Coupled Model Intercomparison Project Phase 5, CMIP5）项目下的全球气候模式模拟结果进行降尺度与偏差校正得到。 <strong>气候模拟</strong> 集合气候模拟的相关参数如下：模拟机构、缩写、模式、典型浓度路径（Representative Concentration Pathway, RCP）、状态* <strong>北京师范大学全球变化与地球系统科学学院（College of Global Change and Earth System Science, Beijing Normal University）</strong> GCESS BNU-ESM 4.5 s 8.5 s <strong>加拿大气候模拟与分析中心（Canadian Centre for Climate Modelling and Analysis）</strong> CCCMA CanESM2 4.5 a 8.5 s <strong>地中海气候变化研究中心（Centro Euro-Mediterraneo per I Cambiamenti Climatici）</strong> CMCC CMCC-CMS 4.5 a 8.5 s <strong>澳大利亚联邦科学与工业研究组织（CSIRO）与澳大利亚气象局（BOM）</strong> CSIRO-BOM ACCESS1.3 4.5 s 8.5 a <strong>数值数学研究所（Institute for Numerical Mathematics）</strong> INM INM-CM4 4.5 s 8.5 a <strong>皮埃尔-西蒙·拉普拉斯研究所（Institut Pierre-Simon Laplace）</strong> IPSL IPSL-CM5A-LR 4.5 a 8.5 s IPSL-CM5B-LR 4.5 s 8.5 s <strong>英国气象局哈德利中心（Met Office Hadley Centre）</strong> MOHC HadGem2 4.5 s 8.5 s <strong>马克斯·普朗克气象研究所（Max-Planck-Institut für Meteorologie）</strong> MPI-M MPI-ESM 4.5 s 8.5 s <strong>挪威气候中心（Norwegian Climate Centre）</strong> NCC NorESM 4.5 a 8.5 s <strong>美国国家海洋和大气管理局地球物理流体动力学实验室（NOAA Geophysical Fluid Dynamics Laboratory）</strong> NOAA-GFDL GFDL-ESM2M 4.5 s 8.5 s 从RCP 4.5与RCP 8.5驱动的完整CMIP5气候模拟集合中，通过聚类集合缩减方法（Casajus等，2016）筛选出22个模拟结果（每种RCP对应11个）。该客观筛选方法可识别出最能代表整体集合的少量模拟结果。缩减过程的输入准则为：针对加拿大境内15个分布区域的3个变量（日最高气温均值、日最低气温均值与总降水量），计算当前时段（1981-2010年）与两个未来时段（2041-2070年、2071-2100年）之间的月际变化。初步筛选出的16个模拟结果，其12（月）×2（时段）×15（区域）×3（变量）指数的预估变化分布与完整集合无统计学差异。后续又补充了少量模拟结果，以实现两种RCP情景各包含11个成员的完整集合。 <strong>参考数据集</strong> 偏差校正的参考（或目标）数据集是加拿大自然资源部（Natural Resources Canada）制作的网格化观测数据集（McKenney等，2011；Hutchinson等，2009）。该数据集通过ANUSPLIN插值方法对台站观测数据进行插值，以生成加拿大陆地区域的日最低气温、日最高气温与总降水量的网格化数据。其网格分辨率为10 km×10 km，时间跨度为1950年至2013年。由于该数据集无法覆盖美国地区，因此与Livneh等（2015）制作的另一套观测插值数据集进行融合，以生成可覆盖美国北部部分区域的偏差校正气候情景。 <strong>覆盖范围</strong> 本数据集的最终版本覆盖区域包括大西洋省份、魁北克省、安大略省、曼尼托巴省与萨斯喀彻温省，以及美国北部部分区域：经度范围为120°W至54°W，纬度范围为40°N至62°N。数据集包含日最低气温、日最高气温与日降水通量，时间跨度为1950年至2100年。 <strong>偏差校正</strong> 全局模拟结果通过双线性插值降尺度至参考网格，并采用Gennaretti等（2015）描述的一维分位数映射方法进行偏差校正。校正过程采用31天的移动窗口对全年每日数据进行调整，使用50个分位数定义待匹配的统计分布。气温变量的长期线性趋势被显式保留。 <strong>气候指标</strong> 本数据集被用于Ouranos的Climate Portraits网站的早期版本（最高至1.3版本）。研究人员使用xclim软件包（Logan等，2022）从每日情景中计算出26个季节与年度气候指标。本次计算所用的“虚拟指标模块”已以“indicators.yml”文件的形式存储于本仓库中。在Climate Portraits网站上，相关信息从三个维度进行展示：空间维度、时间维度与汇总维度。本仓库存储了网站上展示的缩减集合数据。文件名格式为`{aspect}_{indicator}_{season}.nc`。 - 地图文件（命名格式为`spatial_*`）：针对每个偏差校正后的气候模拟结果与给定RCP排放情景，将其气候指标在30年时段内进行平均。为总结气候模式的不确定性，计算了集合分位数：针对每种RCP情景，在11个成员中分别计算10%、25%、50%、75%与90%分位数。 - 时间序列文件（命名格式为`temporal_*`）：针对每个偏差校正后的模拟结果，将每个时间步（年度或季节）的气候指标在各区域内进行空间平均。集合统计量的计算方式为：首先将11个成员的所有区域平均值在中心化30年窗口内进行合并，随后对合并后的数据计算上述分位数。 - 汇总文件（命名格式为`summary_*`）：先将指标在各区域内进行平均，随后在30年时段内进行平均，再计算上述集合统计量。在该应用的2.x版本中，此类数据将以“CMIP5”的形式进行展示。 <strong>数据可用性</strong> 本仓库存储了Climate Portraits网站上展示并前文所述的气候指标集合统计量。完整的每日数据集体积过大，无法在此平台存储。完整的每日数据集可通过Ouranos维护的PAVICS平台公共THREDDS服务器获取：https://pavics.ouranos.ca/twitcher/ows/proxy/thredds/catalog/datasets/simulations/bias_adjusted/cmip5/ouranos/cb-oura-1.0/catalog.html。本数据集可能会在未来被移除，若出现此类情况，请联系我们以获取数据请求支持。 Climate Portraits网站的年度与季节指标数据也可在同一服务器上获取，此外还包含部分未在应用中展示的额外指标：https://pavics.ouranos.ca/twitcher/ows/proxy/thredds/catalog/birdhouse/ouranos/portraits-clim-1.3/catalog.html <em>使用条款</em>：使用本数据集时需注明“数据由区域气候学与气候变化适应联盟（Ouranos Consortium on Regional Climatology and Adaptation to Climate Change）制作并提供”。此外，还需注明构建偏差校正气候情景所用的建模团队，具体可参考：耦合模式比较计划官网：https://pcmdi.llnl.gov/mips/cmip5/citation.html。