Fig. 1: Illustration of the record-breaking probability metrics used in this study, based on observed records in annual maximum daily precipitation (Rx1d) in HadEX3 at four locations which differ in historical record quantile level and climate change trend in Rx1d.
From: Precipitation disaster hotspots depend on historical climate variability
a Upper half: local annual Rx1d anomalies relative to the full-period 1950–2015 average, records (red markers), and observational distributions of Rx1d. Lower half: temporal evolution of the standing record’s quantile level (right y axis) as projected by eight CMIP6 models forced with SSP2-4.5. b Pr(YNR = n): the probability of setting the next Rx1d record in year n, starting from the end of observations in 2015. c Pr(YNR≤N): cumulative probability of breaking the Rx1d record in any year between 2016 (the first unobserved year) and N. In all plots, coloured shading shows the model mean envelope resulting from differences in Rx1d trends across different climate models, dark grey shading shows the multi-model mean uncertainty range resulting from observational generalised extreme value distribution (GEV) fit uncertainties (see Sect. 4.2), light grey shading encompasses the range of observational GEV uncertainties combined with across-model differences. Vertical grey lines indicate the year 2050.
