S2S prediction focuses on the “forecast gap” between 2 weeks and 2 months. This is an incredibly challenging timeframe because it is too long for atmospheric “memory” to provide a detailed forecast, yet too short for slow-moving ocean signals (like El Niño) to dominate the outcome entirely, requiring a mix of both atmospheric and oceanic physics.
The MJO is a massive cluster of rain and storms that travels around the equator. In 2026, its “phase” is the primary search for S2S forecasters because it acts as a “trigger” for weather patterns thousands of miles away. If the MJO is in a specific phase, a forecaster can predict a “higher-than-average risk” of West Coast flooding or East Coast cold snaps three weeks in advance.
Because S2S depends on finding subtle patterns in massive datasets, AI has proven exceptionally good here. Professionals search for “S2S AI benchmarks” to see how models like the Subseasonal Experiment (SubX) are using neural networks to identify “Teleconnections”—links between tropical rainfall and polar jet stream shifts—that traditional physics models often miss or wash out.
The energy sector is the biggest consumer of S2S data. In 2026, grid operators search for “S2S temperature anomalies” to hedge fuel purchases. Predicting a “colder-than-average” February in January allows utility companies to secure natural gas or manage hydropower reservoirs, potentially saving billions of dollars and preventing grid failures during extreme weather.