CMS report – April 2018: The DEVL in CMIP6

The first model submissions to CMIP6 are expected for later this year. It is then time to finalise the preparations to organise and maintain the CMIP6 dataset. NCI in collaboration with CLEX, the Bureau of Meteorology and the CSIRO in now working on the Climate and Weather Science Data Enhanced Virtual Laboratory (DEVL) project. This is an ANDS/Nectar/RDS project aimed at organising the Climate and Weather community together with NCI to prepare for and deliver the CMIP6 dataset in Australia. The experience gained managing CMIP5 has made it clear the community needs to prepare for CMIP6 and to better collaborate around a whole range of tasks.

The project plans to first tidy and reorganise the CMIP5 dataset to help maintaining this dataset at NCI for the long term. This will also enable CMIP5 and CMIP6 to follow the same organisation known as Data Reference Syntax (DRS). This data organisation is the default organisation for CMIP data.

Another goal of the project is to prepare for the publication of the ACCESS results to CMIP6 and the replication of international data locally at NCI. NCI will manage the replication of the first large bulk of the replication following the advice and priorities of the community. Further requests for additional local downloads of data will be handled by NCI and data managers at CLEX, BoM and the CSIRO.

In addition, to facilitate the work of researchers, we will develop and install tools to search for data locally and remotely and to analyse the CMIP6 data. The CMS will provide a new version of ARCCSSive to search the local dataset and all ESGF nodes. This version will be based on a database developed by NCI called Metadata Attribute Search (MAS). The aim is to enable researchers to easily setup complex searches for CMIP6 files. Moreover, searching all ESGF nodes will enable researchers to discover if there is additional data needed for their analyses on other ESGF nodes. Researchers should then be able to log in a request for additional downloads directly from ARCCSSive.

Finally, ESGF requires nodes to provide the Earth System Model eValuation Tool (ESMValTool) and the PCMDI Metrics Package (PMP). PMP calculates statistics on the model outputs compared to observations and reanalyses. This enables researchers developing ACCESS-CM2 for CMIP6 to easily assess ACCESS’s skills compared to other models submitted to CMIP6. As stated on the ESMValTool website, this “is a community diagnostics and performance metrics tool for the evaluation of Earth System Models (ESMs) that allows for routine comparison of single or multiple models, either against predecessor versions or against observations.” The tool already contains hundreds of diagnostics and is developed by tens of institutions internationally. It should enable researchers to perform the routine evaluations quicker, to ensure better reproducibility and to easily document provenance of the diagnostics. It should also facilitate sharing workflows between the Australian and overseas researchers. The tool is also built to encourage contributions from users.

Both packages rely on data organised in a DRS structure. This is true for both model outputs and observations or reanalyses. Hence, a related goal of the DEVL project is to provide observation and reanalysis datasets organised in a DRS structure.

The DEVL project also puts emphasis on developing proper documentation and training for all achievements of the project. We will develop centralised documentation between all institutions and different trainings around the organisation of the CMIP6 dataset at NCI and the several tools developed and supported during this DEVL project.