Fig. 4: The changes in SST responses to ENSO.

a Change in regression coefficients between December(0)–February(1) (D(0)JF(1)) Niño3.4 sea surface temperature (SST) anomaly and SST during maximum R² period (τ) relative to March–May(0) (MAM(0)), based on Present-day (PD) simulations using CESM1 (shade; K K⁻¹). Stippling indicates regions with statistically significant changes between PD and 2085–2115 (2100), where the size of the dots represents confidence levels (90%, 95%, and 99%) and the color of the dots indicates whether the regression coefficient increased (red) or decreased (blue). b–e Temporal evolution of regression coefficients of area-averaged SST anomaly and latent heat flux with respect to D(0)JF(1) Niño3.4 SST anomaly from MAM(0) to June–August(1) (JJA(1)), during PD (blue line) and the 2100 period (red line), for each region: East China Sea (125°E–145°E, 20°N–35°N), U.S. East Coast (60°W–80°W, 25°N–40°N), South Indian Ocean (50°E–90°E, 40°S–20°S), and the subtropical North Pacific (150°W–165°W, 12°N–22°N). Red dots indicate statistically significant at the 95% confidence level using the bootstrap test. Vertical line marks τ (see “Methods” for details). The hatches show the net difference between the 2100 and PD El Niño-Southern Oscillation (ENSO)-induced latent heat flux curves (starting at the last moment before τ when the two curves are equal and ending at τ). A hatched area indicates an enlarged net heat uptake/loss during that window.