A new algorithm called the Random Analog Predictor (RAP) can forecast Arctic sea ice extent months in advance, providing a benchmark for climate modeling.

Researchers at the Mubadala Arabian Center for Climate and Environmental Sciences (ACCESS) at NYU Abu Dhabi have created an algorithm capable of forecasting Arctic sea ice extent up to nine months in advance. Published in the journal Scientific Reports, the tool, known as the Random Analog Predictor (RAP), uses historical data to identify patterns and generate forecasts while providing an estimate of uncertainty.

How the Algorithm Functions

Unlike physics-based models that simulate atmospheric and oceanic conditions, RAP relies exclusively on the historical record of Arctic sea ice extent. The algorithm identifies past conditions that resemble current data and uses subsequent historical outcomes to produce an ensemble of potential future scenarios. The variance within these generated forecasts allows users to gauge the level of confidence in each prediction.

Performance and Benchmarking

The research team found that RAP performs at a level comparable to more complex models currently used by the Sea Ice Prediction Network. Specifically, for September sea ice extent, the algorithm's forecast error was similar to that of 34 other models used for seasonal forecasting. The researchers suggest that RAP serves as a transparent, low-cost benchmark for evaluating the predictive value of more sophisticated AI-driven or physics-based approaches.

Climate Significance

Arctic sea ice is a critical component of the global climate system because it reflects solar energy back into space, whereas the darker ocean absorbs it. Because changes in the region can influence atmospheric and oceanic patterns globally, the ability to anticipate these shifts months in advance is considered increasingly important. The tool may also support the United Arab Emirates' expanding polar and Arctic research initiatives.