Extended Data Fig. 3: Climate-change detection using the ridge regression method.
From: Anthropogenic fingerprints in daily precipitation revealed by deep learning
a, Time series of the observed AGMT anomaly from 1980 to 2020 (black line) and the annual average of the estimated AGMT using daily precipitation fields from the MSWEP (blue line), IMERG (purple line), GPCP (orange line) observations and ERA5 reanalysis (red line) as inputs in the ridge regression model. The blue and red dots denote the daily estimated AGMT using the MSWEP and ERA5 precipitation data, respectively. The dashed black horizontal lines denote a 95% confidence range of internal variability of the AGMT estimates, defined as the 2.5th–97.5th percentile of the daily estimated AGMT obtained from historical CESM2 LE simulations during 1850–1950. b, Fractional number of EM days within a corresponding year from 1980 to 2020 for which the estimated AGMT is greater than the upper bound of the 95% confidence range. Dashed line denotes an upper bound of the 95% confidence range of internal variability of fractional EM days, which is 10.9%. c, Linear trend of the number of EM days during 1980–2020 in ERA5 and MSWEP and 2001–2020 in ERA5, MSWEP, IMERG and GPCP. The dashed lines denote the upper bounds of the 95% confidence level based on the bootstrap method estimated using the historical CESM2 LE simulations (see Methods).
