Data storage for "Understanding CMIP6 biases in the representation of the Greater Horn of Africa long and short rains"

Rainfall statistics, climatological circulation metrics, and correlations calculated for the research in Schwarzwald et al., published in Climate Dynamics.

Structure of all_stats.csv

all_stats.csv contains all precipitation statistics, circulation metrics, and correlations calculated for "Understanding CMIP6 biases in the representation of the Greater Horn of Africa long and short rains".

The first two columns are indices:

• model: The name of the data source of the statistics. Items are either the name of a CMIP6 model, or "obs" for observations or reanalysis. For "obs", the source for precipitation statistics is CHIRPS, for SSTs, OISST, and for $u$ and $\omega$ ERA5.
• season: Whether the statistic is calculated or refers to the 'long' or 'short' rains.

The next seven columns are statistics of the rainy seasons. They are all named "pr_[statistic name]". All statistics are calculated at the pixel level, and then averaged over every pixel in the study area (see paper Section 3.1 for details).

• onset: Season onset
• demise: Season demise
• duration: Season duration (onset - demise + 1)
• peak_timing: Day of peak rainfall
• peak_amount: Amount of rainfall on peak day
• integrated_amount: Total amount of rainfall in a season
• seas_ratio: The ratio of the 1st to the 2nd harmonic (necessarily < 1, since the study area is defined as all pixels where this ratio is < 1)

The next set of columns are the climatological circulation metrics $M_c$ (see Section S2) - "timing" of peak and "amount" for each model/obs, season. Columsn are named "[varname]_[timing/amount]" In addition to the variables used in the paper (WIOSSTs as "tos", Indian Ocean Dipole mode index as "dmi", $u$ as "ua250", $\omega$ as "wap250" for 250 hPa and "wap500" for 500 hPa), other variables listed include:

• hus1000/hus850: 1000 hPa and 850 hPa specific humidity, respectively
• mse1000/mse850: 1000 hPa and 850 hPa moist static energy, respectively

The final set of columns are the calculated correlations $\rho^{IM,mod}$ and $\rho^{IO}$ (see Section S4). Column names are "[varname]_[timing/ammount]_corr". "Timing" correlations are the timing of the peak of a given variability and the timing of peak rainfall, "amount" correlations are the peak of the circulation variable and the total rainfall in a season.

Structure of peaks_alt*.nc

peaks-alt*.nc contains the raw calculated peaks for each model for a particular variable in each file, for both the climatological peaks and the peaks in every year. Filenames are of the form peaks-alt_[varname]_allmods_historical_*.nc.

Each file contains four variables:

• clim_timing: the day of peak for a variable for either the first half of the year (season 1) or the second half of the year (season 2); see Section S2 for details
• clim_amount: the peak amount of a variable for each half of the year
• byyr_timing: the day of peak for each year in the sample
• byyr_amount: the peak amount for each year in the sample

Structure of prstats_ann_allmods_1981-2013_HoAfr-bimod-land.nc

prstats_ann_allmods_1981-2013_HoAfr-bimod-land.nc contains the year-by-year precipitation stats (see Section S1 for details of calculations) for CMIP6 models and CHIRPS observations (listed as obs in the model dimension). Climatological precipitation stats are in the pr_* columns in all_stats.csv. Stats are calculated at the pixel level, and then averaged across all pixels in the bimodal region to create one value for each model, year, and season.