ILAMB 2.5

We are pleased to announce a new version of the ILAMB package. In addition to a new version, the ILAMB repository is now hosted at:

https://github.com/rubisco-sfa/ILAMB

This release includes some interface improvements as well as core technology enhancements increasing speed and reliability. See the following list for major changes:

• The landing page generated from ilamb-run has received an overhaul. We have merged the two tabs with the image and data into one dynamic table which can be clicked. Clicking on a row header will either expand the row or take you to the dataset page. Clicking on a particular model’s square will take you to the dataset page with that particular model highlighted. In addition to this, you can now select which scalar you wish to plot in the table (i.e. Overall Score, Bias Score) over any region included in the study.

• The appearance of the Data Information tabs on the dataset pages has been greatly enhanced. References can be included in the netCDF files in Bibtex format and will be rendered inside of ILAMB. Hyperlinks also will be detected and rendered as clickable links in the output pages. Thanks to Mingquan Mu for this addition.

• Added a soil carbon temperature sensitivity metric from Charlie Koven, added this to our curated configure file cmip.cfg.

• The CO2 emulation code will now account for ocean and fossil fuel fluxes when emulating the land model’s nbp. Thanks to Ke Xu for this contribution.

• We have added new datasets LORA for runoff and DOLCE for latent heat. While these datasets include uncertainty estimates, we are currently not making use of them in our analysis.

• We have replaced basemap in favor of cartopy as the tool for plotting on maps. Not only is this needed as basemap is being deprecated, but plotting is now approximately 10x faster. For the most part, this change will be invisible to the user.

• Added options and structure to ilamb-run to improve runtimes. If running a large set of models on a cluster, we recommend first running with the –skip_plots option and using a low number of processes per node. This is because memory utilization tends to dominate the analysis phase and you do not want to run out. Then you can submit a second job without –skip_plots and using a large number of processes per node.

• Intermediate files generated during ilamb-run will now include a complete flag, initialized to False and only flagged true if the file closed at the end of the analysis phase without error. This helps us reinitialize ilamb-run when a parallel run crashes and leaves file present and not corrupted, but neither complete.

• If the psutil python package is installed, ilamb-run will now log the peak memory being used during the analysis phase in the logfiles along with the node name and process rank. This is to help in memory debugging for when high resolution models are being run.

• In addition to this, ilamb-run now caches the model initialization process which should speed up re-initialization for when multiple jobs must be submitted.

• Added initial support for uncertainty bounds in the ILAMB.Variable. If uncertainty is included in the observations, such as in the Hoffman nbp dataset, then ILAMB will automatically operate of it and show it as a shading in plots without changes to your scripts.

• ilamb-fetch now will correctly try to decode server SSL certificates before downloading files. However, the authority that www.ilamb.org uses to create its certificates is not in the list that python supports. Your browser maintains a different list of authorities, which is why you can navigate to sites like this. You will likely need to run with the –no-check-certificate option which implies that you trust that we are who we say we are.