In JGCRI/RCMIP5: Tools for Manipulating and Summarizing CMIP5 Data

Need for future climate projections and CMIP5

It has been noted previously that we are in the middle of one of the largest experiments of mankind. How the Earth system will respond to the addition of CO2 to the atmosphere by various anthropogenic source is of interest not just from an academic prospective but also socital. Earth system models allow us to both explore teh societal impacts of climate change but also the connections between dispargent systems within a broader Earth system context.

One of the most common source of such data is the archive of the 5th Coupled Model Intercomparison Project (CMIP5). CMIP5 consisted of a series of coordinated ESM experiments addressing three key questions: identifying the mechanisms responsible for model differences in feedbacks, particularly within the carbon cycle and clouds; quantifying how well models predict climate on multiple time scales; and examining why similarly forced models produce a range of responses (Taylor et al., 2012). CMIP5 produced multiple simulations from more than 20 modeling groups worldwide, including decadal hindcasts and predictions, long-term simulations, atmosphere-only, and four future scenario simulations. More than 20 modeling groups from all over the world used more than 50 models for CMIP5. The resulting model outputs allows have enabled model evaluation, fundamental science exploration, and understanding impacts and policy issues under changing climate scenarios.

CMIP5 variables are provided at a range of temporal resolutions depending on the modeling center discression. In addition, beyond a few key variables, not every proposed variable is simulated and/or supplied by each model. This can make managing downloads and model intercomparion difficult. As a result there are a few key function which are repeated over and over again when studies work with CMIP5 simulations, reinventing wheel each time. This is both time intenstive and nontrivial, because of the many quirks of the CMIP5 database (despite best intentions of modeling teams, whom we applaud):

• different grids, resolutions, and years
• inconsistencies in netcdf file formats (ie. missing variables coded differently, variable split into multiple files based on time)
• inconsistant variable aggrigation (ie. some models report land fraction which impacts global land means/totals)

RCMIP5 package

To help researchers deal with these issues, here we introduce the RCMIP5 package--collection of software tools--for working with CMIP5 data. This package aims to provide a reproducible, robust, and high-performance set of functions to (i) explore what data have been downloaded from CMIP5 archives, (ii) identify missing data, (iii) spatial/temporal slicing and average (or apply other mathematical operations) across experimental ensembles, (iv) produce both temporal and spatial statistical summaries, and (v) produce easy-to-work-with graphical and data summaries. The package also includes pre-summarized global means for a variety of frequent model outputs such as atmospheric CO2 and global mean temperature.

RCMIP5 is written in R (http://www.r-project.org), the open-source statistical language increasingly used for data processing and analysis across many scientific disciplines {Tippmann, 2014 #4171}. Open source is particularly importance in climate and global change-related fields {Easterbrook, 2014 #4137}, but increasingly practiced and demanded in general. RCMIP5 supports reproducible research in computation science {Peng, 2011 #4187}{Boettiger, 2014 #4185}, first by being freely available and open source; in its explicit support for tracking data provenance; and by including cryptographic hashes (checksums) of its data processing steps.

Example

We touch on a few of the basic functions included in this package. A more detailed vingette is available through CRAN.

library(RCMIP5)
library(plyr)
mypath <- "../sampledata"

library(RCMIP5)
mypath <- "~/my/path/to/data"

CMIP5 includes a large number of simulations which can result in thousands of files that need to be managed for model intercomparison studies. We provided several functions to assist with this including the following example where a single variable-experiment is extracted from the list of downloaded CMIP5 files.

c5files <- getFileInfo(mypath)
print(knitr::kable(head(c5files)))

nbpfiles <- subset(c5files, variable=="nbp" & experiment=="historical")
print(knitr::kable(nbpfiles))

TODO Slice, world plot and global seasonal mean or annual mean

Conclusions

We hope RCMIP5 will be a useful, flexible, and high-performance tool for researchers working with CMIP5 data, enabling faster, better, more reproducible science. We welcome contributions and ideas...

References

Boettiger, C.: An introduction to Docker for reproducible research, with examples from the R environment, arXiv, http://arxiv.org/abs/1410.0846, arXiv:1410.0846, 2014.

Easterbrook, S. M.: Open code for open science?, Nature Geoscience, 7, 779-781, 10.1038/ngeo2283, 2014.

Taylor, K. E., Stouffer, R. J., and Meehl, G. A.: An Overview of CMIP5 and the Experiment Design, Bulletin of the American Meteorological Society, 93, 485-498, 10.1175/BAMS-D-11-00094.1, 2012.

Tippmann, S.: Programming tools: Adventures with R, Nature, 517, 109-110, 10.1038/517109a, 2014.

Wolkovich, E. M., Regetz, J., and O'Connor, M. I.: Advances in global change research require open science by individual researchers, Global Change Biol., 18, 2102-2110, 10.1111/j.1365-2486.2012.02693.x, 2012.

JGCRI/RCMIP5 documentation built on May 7, 2019, 7:43 a.m.