A simple forecast of Arctic sea-ice extent is putting complex models to the test
A forecasting method built only from the historical record of Arctic sea-ice extent has achieved September hindcast skill comparable with models used by the Sea Ice Prediction Network, giving researchers a simple test of whether more complex physics-based and AI systems are actually extracting additional predictive information. The Random Analogue Predictor (RAP), developed by Faiq Raees and Francesco Paparella, was evaluated at lead times of up to nine months and showed negligible bias with root-mean-square error no greater than 0.6 million km² (232 000 mi²).

