Replication data for: "Revisiting the 'East African Paradox': CMIP6 models also struggle to reproduce strong observed Long Rain drying trends."

Intro

This repository contains replication data for Schwarzwald, Kevin and Richard Seager (2024), "'Revisiting the “East African Paradox': CMIP6 models also struggle to reproduce strong observed MAM long rain drying trends." (under revision at Journal of Climate). This includes the data necessary to replicate all main text figures and most figures in Supplementary Materials. Additional figures in Supplementary Materials require raw precipitation time series, detailed below. This repository also includes a copy of the code necessary to replicate the entire study; the latest version of the code, in addition to instructions on how to use it, is kept at this GitHub archive. 

Structure

The repository is structured as follows: 

code: Static / stable version of reproduction code for Schwarzwald and Seager (2024) that uses and creates these data (see here for more detailed instructions)

figures: Static / stable versions of main and supplemental figures for Schwarzwald and Seager (2024).

climate_raw: "raw" (often pre-processed) climate data files; only some files are included, see below for more details

climate_proc: Processed climate data files upon which the analysis is based, used by main and supplemental figure code

aux_data:  Certain auxiliary data files (fonts, critical values) and intermediate files for long code processes. Created and used by the replication code.  

Notes on raw climate data

Due to space limitations (and a desire to not create yet another cloud copy of CMIP6 data), this repository only contains processed CMIP6 data: the calculated linear trends of rainfall, sea surface temperatures, and 500 hPa geopotential height used in the analysis of Schwarzwald and Seager (2024). The raw data used to create these files can be downloaded from the ESGF, and consists of every available CMIP6 monthly rainfall, sea surface temperature (SST), and geopotential height file on the archive at the time of processing for the experiments detailed in the manuscript. For precipitation, only a bounding box (-3 S to 12.5 N, 32 E to 55 E) around East Africa was downloaded, and saved with the file suffix "_HoAfrica" (see replication code README for more details). One example precipitation file (one ensemble member of ACCESS-CM2 historical precipitation) is included in this repository for reference. 

Similarly, this repository only contains processed NMME data: calculated linear trends of rainfall used in the analysis of Schwarzwald and Seager (2024). The raw data used to create these files can be downloaded from the IRI Data Library and consists of every available monthly rainfall hindcast / forecast file from the NMME archive for the models used (see manuscript Table S1). As above, only a bounding box around East Africa was downloaded, and files were standardized to a partial CMIP* file format (one file per variable, with CMIP file and variable name conventions, but with forecast lead time as an additional dimension). Preprocessing is a bit more extensive than for CMIP6 models: first, hindcasts and forecasts were concatenated into a single file (since hindcasts are saved in the archive only up to the time when the model was operationalized), and then reindexed such that the "time" variable refers to the time for which the forecast is made, and not the time at which the forecast is made. One example precipitation file (CanCM4i rainfall hindcasts/forecasts) is included in this repository for reference. 

This repository does, however, contain preprocessed copies of the "raw" gridded observational rainfall data products used (in addition to the processed trends), since harmonizing and standardizing the data into a format easily compatible with the CMIP6 data was a nontrivial amount of work that would be tedious to replicate. For the ten gridded observational data products used, monthly rainfall was brought into the CMIP* file format (one file per variable, with CMIP file and variable name conventions), with one notable exception: all observational rainfall is saved in units of mm/day. 

Reanalysis and gridded ocean observations can be downloaded from each product's respective repositories. As before, the code assumes the data have been preprocessed into something akin to the CMIP* file format. 

Notes on processed climate data

Note that the repository includes a file (aux_data/pr_doyavg_CHIRPS_historical_seasstats_dunning_19810101-20141231_HoAfrica.nc) containing CHIRPS seasonal characteristics in East Africa, created as part of Schwarzwald et al., 2023, Climate Dynamics. This file is primarily used to set the boundaries of the study region, see the manuscript for details of how it was calculated. 

Licensing, citing, questions

These data are offered under a CC 4.0 license, which allows redistribution and reuse as long as they are correctly cited; note that for much of these data (especially for "raw" data), this requires citations to the original creators. 

For questions, please feel free to reach out to corresponding author Kevin Schwarzwald.