CMIP6 exceedance likelihoods for climate extremes

Full description

This collection consists of climate extremes data derived from several CMIP6 large ensembles (models with five or more runs), representing how the likelihood of exceeding various historical thresholds changes over the 1850-2100 period.

Climate metrics relating to heat (Warm Spell Duration Index; WSDI), drought (Standardised Precipitation Evapotranspiration Index; SPEI) and fire weather (Forest Fire Danger Index; FFDI) are included for four different future emission scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5), aggregated for each state and territory of Australia.

The data was produced to support macroeconomic modeling conducted by the Reserve Bank of Australia.
Lineage: The WSDI and SPEI were calculated at each model grid point for each run of six CMIP6 models (ACCESS-CM2, ACCESS-ESM1-5, CanESM5, IPSL-CM6A-LR, MPI-ESM1-2-LR, UKESM1-0-LL). To generate the final state and national aggregated values, a weighted mean was then calculated where the weight for each grid cell was the area of the cell multiplied by the fraction of the cell that overlaps with the geopgraphic shape (e.g. state) of interest. The FFDI was calculated at each model grid point for each run of one CMIP6 model (ACCESS-ESM1-5). For that spatial aggregation, grid points in arid climate zones were excluded since the FFDI isn't as appropriate / relevant in those zones.

Once the spatial aggregation was complete, the likelihood of exceeding a series of reference percentiles (calculated over the 1950-2014 period) was calculated. A likelihood was calculated for every year from 1860-2091, using a 20-year sliding window centered on those years. For instance, the ACCESS-CM2 model performed 10 runs of the SSP3-7.0 experiment. The likelihood of exceeding the reference (1950-2014) 98th percentile in the year 2050 for that model under that emission scenario was calculated empirically as the fraction of all 200 values across all runs from 2040-2059 that exceeded that 98th percentile threshold.

Further details, including the python software used to perform the data processing and a description of the data files included in this collection, is available at: https://github.com/AusClimateService/rba