Top panel: Predictor departure from average of westerly wind stress (see footnote 1) in August (x-axis) vs the eastern Pacific warming (E) in the following January (y-axis). Observations are shown in red, while the CM2.1 model forecasts (with many ensemble members—i.e., repeated model runs with different starting conditions—are shown in grey with their 10%, 50%, and 90% percentiles shown by the black sloping curves to summaries the positions of most of the gray dots. Adapted from Takahashi & Dewitte (2015). Middle and bottom panels: SST deviation from average (°C) (colors, with color key below) and departures from average of surface wind stress (10-2 N/m2 units) (arrows, showing stress direction and strength by the length of the arrow line) in August 1982 (middle panel) and January 1983 (bottom panel). The red box outlines the averaging region for the wind stress predictor for judging the probability of occurrence of an extreme condition in the Eastern Pacific 5 months later in January. Images adapted from Ken Takahashi.