Fig. 3: Sensitivity of extreme scaling to detection radius.
a, Analogous to Fig. 2a but for varying detection radius rcv (legend shows colour coding). b, Statistical consistency with previously published super-CC results: M1512, using a maximum intensity criterion over a precipitation event; mimicking B1323, using rcv = 50 km and τcv = 3 h; the light-blue curve is a replication of the rcv = 5 km curve shown in a (Methods). c, Mean slopes of the curves in a (calculated over 10 data points), which correspond to decreasing values of rcv, in shades of dark to light blue. Purely convective cases lie to the left of the plot, that is, small rcv. The green curve indicates the convective fraction (Conv. frac.) corresponding to each rcv. d, Analogous to c but for rst, that is, purely stratiform cases lie to the right of the plot. The green curve indicates the stratiform fraction (Strat. frac.) corresponding to each rst. Star symbols between panels c and d indicate the mean slopes for all precipitation (TOT; Fig. 2a, black curve), and M15 and B13 (shown in b). Note the logarithmic vertical axis scaling in a, b and c (right axis) and the logarithmic horizontal axis scaling in c and d. Data in c and d are represented as mean values ± an error estimate based on the mean of the squared residuals from a linear fit to 10 data points (Methods).
