项目主页：https://cmip6.science.unimelb.edu.au/
工具包地址：https://gitlab.com/netcdf-scm/netcdf-scm
文章PDF获取：(或者文末原文阅读)
链接：https://pan.baidu.com/s/1VdO4yjxHFMjXZdOM_YYeBA
提取码：zhe2
世界上最复杂的气候模型目前正在运行一系列经验，作为第六次耦合模型相互比较项目（CMIP6）的一部分。加上第五次耦合模式相互比较项目(CMIP5)的产出，总数据量将达到20PB。在此，我们提出了一个年度、月度、全球、半球和陆地/海洋手段的数据集，该数据集是由气候数据分析人员和降低复杂度的气候建模人员所选择的关键性参数得出的。衍生的数据集是验证、校准和开发复杂度较低的气候模型的关键部分，以对抗物理上更完整的模型的行为。除了用于复杂度降低的气候模拟者之外，我们的目标是使我们的数据能够被其他研究团体所使用。我们通过多种方式促进这一目标的实现。首先，由于我们的重点是年度、月度、全球、半球和陆地/海洋平均数量，我们的数据集比源数据小很多，因此不需要专门的 "大数据 "专业知识。第二，同样由于其较小的规模，我们能够以文本为基础的格式提供我们的数据集，极大地减少了使用CMIP输出数据所需的计算专业知识。第三，我们通过跟踪所有的源元数据和提供检查数据集是否已被收回的工具来实现数据的来源和完整性控制，这被认为是错误的。当新的CMIP6结果出现时，我们会更新所产生的数据集，并提供一个稳定的访问点，以允许自动下载。与我们的配套网站(cmip6.scien ce.unime lb.edu.au)一起，我们相信这个数据集提供了一个独特的社区资源，并允许非专家以一种新的、用户友好的方式访问CMIP数据。
大气-海洋-地球系统耦合模型是我们对气候系统最全面的表述。来自全球各地的地球系统模型目前正在运行，作为第六次耦合模型相互比较项目（CMIP6，Eyring等人（2016a））的一部分。CMIP6建立在第五次耦合模式相互比较项目(CMIP5，Taylor等人(2012))的基础上，该项目的产出仍在广泛使用。然而，CMIP5和CMIP6模型的计算成本很高，因此不能用于所有感兴趣的应用。为了填补这一空白，一些所谓的降低复杂性的气候模型(也称为 "简单气候模型")已经被开发出来(Nicholls等人，2020)。开发这些模型的一个重要部分是校准：推导出一套参数，使它们能够最好地复制参与CMIP实验的更复杂模型的行为。为了进行这种校准，必须首先处理CMIP5和CMIP6的输出.一旦完成，CMIP6档案将成为世界上最大的数据档案之一，预计总体积在18PB左右（Balaji等人，2018）。幸运的是，重新降低复杂性的气候建模者通常只需要半球和陆地/海洋尺度的年均值或月度模型输出。这大大减少了他们必须处理的数据量。然而，降低复杂性的气候模型通常包括一些不同的模块，涵盖了排放-气候变化因果链的所有范围。校准所有这些不同的模块需要处理多个数据集，因此，一个 "全面的 "降低复杂度的气候模型所需的CMIP原始输出总量仍将达到50TB。处理这么大的数据量，即使对于专家用户来说，也是一项艰巨的任务。首先是原始数据全部采用定制的、针对气候的netCDF数据格式(Unidata, 2020)。如果没有经过专业培训，就无法阅读，更不用说分析了。第二个因素是数据都是按照高度规范化的数据参考语法进行排序的(Balaji等人，2018)。这种正则化是使数据能够被机器处理所需要的；然而，对于非专家来说，它可能是混乱的。第三个因素是，数据通常以绝对值呈现。然而，降低复杂度的气候模型通常是扰动模型；也就是说，它们从一些参考状态而不是绝对值计算扰动。考虑到数据量和有时CMIP实验之间的复杂关系, 计算一个大数据量的扰动并不是一个简单的任务. 第四个因素是许可。所有的CMIP5和CMIP6文件都是在特定的许可下发布的，用户必须遵守，而且检索这些信息并不容易。最后一个因素是撤回，即删除后来被确认为错误的数据。为了避免错误的结果在科学文献中传播，这种撤回是必不可少的。然而，目前只有有限的工具可以让用户检查他们是否使用了一个回溯的数据集。本数据集的目标受众是复杂度降低的气候模型和建模者，这里的复杂度降低模型指的是侧重于气候系统的全球和年平均特性的模型。由于空间和时间分辨率有限，降低复杂性的模型在计算上非常有效。这些模型通常在更复杂的模型(如参加CMIP的模型)计算成本太高而无法使用时使用。例如，在许多综合评估模型中使用降低复杂性的模型来评估不同排放路径的气候影响（鉴于综合评估模型通常需要数百到数千次的气候实现，使用CMIP模型在计算上是不可行的）。再降低复杂性模型的一些突出例子是MAGICC（Meinshausen等，2011）、FaIR（Smith等，2018）和hector（Hartin等，2015）。关于降低复杂度模型的详细讨论和文献中可用模型的概述可以在降低复杂度模型相互比较项目的第一阶段中找到（Nicholls等人，2020）.我们的CMIP5-和CMIP6-衍生数据集是使用我们开发的开源工具netCDF- SCM（降低复杂度/简单气候建模者的netCDF处理，见第2.2节）提取的，并且已经准备好供降低复杂度气候建模者使用。它是解决上述所有复杂问题的结果，包括CMIP5和CMIP6输出的全球、半球、陆地/海洋年度和月度手段。考虑到每一次成形的处理，这个数据集比原始数据小了好几个数量级。数据量的减少意味着我们可以以文本格式提供数据。因此，虽然该数据集针对的是复杂性降低的气候模型的开发者，但其简单的基于文本的格式也允许气候科学界以外的非专家用户阅读和分析数据，因为他们不再需要使用气候特定的netCDF格式。（以上为DeepL机器翻译，未人工校验，如有不当见谅！）
The new climate projections
Climate change is one of the defining issues of our century. A new set of climate projections has been developed by the wide scientific community as part of Phase 6 of the Climate Model Intercomparison Project. This new dataset will also underpin the forthcoming Sixth IPCC Assessment Report (AR6), scheduled to be published in 2021 and 2022.
What these climate projections tell us is how warm the globe is going to get, if we continue to emit a lot of greenhouse gas emissions or how much we can limit warming, if we restrain future emissions. However, that is not the only thing. The data is much richer, with 100s of variables for each model, including variables like soil carbon content, deep ocean warming, cloud cover and many others.
For anyone interested in this new enormous dataset, there are various ways to access it. The primary portal for researchers are the Earth System Grid Federation servers. That is where all the latest datasets are stored and where a huge effort of the scientific community makes sure that the data is also maintained. For example, if an error is found, new versions of the datasets will be published there with documentation provided by es-doc.
For anyone interested in what these experiments actually try to model, and what the story is behind the various scenarios, like SSP5-8.5 or abrupt4x-CO2, it is best to look into the original scientific literature. For example, an overview of the scenario design is provided here with further details in O'Neill et al. (2016). We played our small part by contributing the greenhouse gas concentrations that underly all those different scenarios (see Greenhouse Gas Factsheets).
This website provides large scale averages, not gridded data. Sometimes terabytes of data are a bit unwieldy and a lot of researchers and the public are just interested in the large scale averaged timeseries. That is what this visualisation tool is about. We downloaded a lot of the monthly CMIP6 data and aggregated it to global, hemispheric and land and ocean averages. We also crunched averages for all of the AR6 regions defined in Iturbide et al., ESSD 2020. Thus, if that is what you are interested in, you are lucky, as all the data is now at your fingertips.
Data format
As described in Nicholls et al. (2021) (see citation below), we use the custom .MAG format for all of our data outputs. This format is a text-based format, designed to make life easy for all data processing programs (including those that can't handle binary data). A description of the data format can be found at pymagicc's documentation
We are developing examples of how this data can be used in a public GitLab repository, https://gitlab.com/netcdf-scm/calibration-data. In this repository, we currently have examples of how to download, read and plot the data in Python. Please feel free to use these as a starting point for your own analysis. If you do build new things (particularly in languages other than Python), please make a merge request so that we continue to share the knowledge.
Disclaimer
The data provided here is derived from the CMIP6 archive on the ESGF servers. If you are using this data, you must also abide by the CMIP5 and CMIP6 terms of use (found at https://pcmdi.llnl.gov/mips/cmip5/terms-of-use.html and https://pcmdi.llnl.gov/CMIP6/TermsOfUse/TermsOfUse6-1.html respectively). The tools we use to help us comply with these terms of use are demonstrated at https://netcdf-scm.readthedocs.io/en/latest/usage/using-cmip-data.html.
CMIP6 and CMIP5 model data that is aggregated here by the University of Melbourne's Climate Energy College is licensed under the same license as the raw CMIP data (typically, Creative Commons Attribution-ShareAlike 4.0 International License but the license of each file should be checked before use). The data producers and data providers make no warranty, either express or implied, including, but not limited to, warranties of merchantability and fitness for a particular purpose. All liabilities arising from the supply of the information (including any liability arising in negligence) are excluded to the fullest extent permitted by law.
About Us and Citation
This data portal has been developed by Zebedee Nicholls, Jared Lewis, Malte Meinshausen from the Climate & Energy College as well as Melissa Makin from Science IT and Usha Nattala, Geordie Zhang, Simon Mutch and Edoardo Tescari from the Melbourne Data Analytics Platform (MDAP) at the University of Melbourne. If you use this data, please cite (further citation formats available at https://doi.org/10.1002/gdj3.113)
Nicholls Z, Lewis J, Makin M, Nattala U, Zhang GZ, Mutch SJ, Tescari E and Meinshausen, M. Regionally aggregated, stitched and de-drifted CMIP-climate data, processed with netCDF-SCM v2.0.0. Geosci Data J. 2020; 00:000-000. https://doi.org/10.1002/gdj3.113
Other data formats, missing or erroneous data
If you would like to discuss other possible data formats, think that any data is missing or believe we have made an error in our processing, please check our issue tracker and if your issue is not already there, please raise one. We will aim to respond as quickly as possible.
Website failures
If you have any issues with the website (e.g. it fails in some way), please contact znicholls@unimelb.edu.au, malte.meinshausen@unimelb.edu.au, jared.lewis@unimelb.edu.au and melissa.makin@unimelb.edu.au.
API
A read-only API is available to search the archive and download zipped .MAG files.
The search and download endpoints all take the following search parameters
- experiment_id
- member_id
- normalised
- source_id
- timeseriestype
- mip_era
- variable_id
Values for these can be determined from the search form with the exception of mip_era, normalised and timeseriestype which use the mappings below. Note that variable_id always uses the short form, e.g. "tas" instead of "Near-Surface Air Temperature".
The search endpoint provides a way to programmatically search the archive in the same way that the Search page works.
代码语言：javascript复制curl -i "https://cmip6.science.unimelb.edu.au/api/v1/search?experiment_id=ssp585&variable_id=tas"
HTTP/1.0 200 OK
Content-Type: application/json
...
{
"count": 25,
"results": [
{
"activity_id": "ScenarioMIP",
"experiment_id": "ssp585",
"grid_label": "gn",
"institution_id": "NUIST",
"member_id": "r1i1p1f1",
"mip_era": "CMIP6",
"normalisation_method": "",
"source_id": "NESM3",
"table_id": "Amon",
"time_range": "185001-210012",
"timeseriestype": "monthly",
"mip_era": "CMIP6",
"url": "https://cmip6.science.unimelb.edu.au/magdownload?path=CMIP6/mag/monthly/CMIP6/ScenarioMIP/NUIST/NESM3/ssp585/r1i1p1f1/Amon/tas/gn/v20190728/netcdf-scm_tas_Amon_NESM3_ssp585_r1i1p1f1_gn_185001-210012.MAG",
"variable_id": "tas",
"version": "v20190728"
},
...
Bulk downloading files/api/v1/download_zip
All the files for a set of filters can be downloaded as a zipped archive.
代码语言：javascript复制wget -q -O ssp585_tas.zip "https://cmip6.science.unimelb.edu.au/api/v1/download_zip?experiment_id=ssp585&variable_id=tas"
HTTP/1.0 200 OK
Content-Type: application/zip
...
