Leaderboard

We provide a leaderboard for both deterministic and probabilistic models/metrics. This interactive page is inspired by the WeatherBench 2 design.

We showcase select variables (t-850, z-500, q-700) on select metrics e.g., RMSE, ACC for deterministic models or RMSE, CRPSS for ensemble models. Learn more about the metrics here and access the complete metrics across variables here. Unless stated, all results are evaluated for the year 2022.

Deterministic Models

Deterministic models make a single (control) forecast. This method is particularly sensitive to initial condition, and is therefore not recommended for long-term chaotic prediction, such as in the S2S case.

Ensemble Models

Ensemble models make multiple forecasts (members) at a time to give us a collection of possible future outcomes. This (plus a well-dispersed ensemble) is particularly important for skillful S2S forecasting.