India has recently been the focal point of a groundbreaking study that harnessed the power of machine learning models to forecast maximum temperatures (Tmax) up to 10 days ahead during the scorching months spanning from March to June. This advanced research tested ten machine learning models to see how well they predict daily Tmax anomalies.
Surprisingly, among these models, the AdaBoost regressor, with a Multi-layer Perceptron as the base estimator, emerged as the optimal choice for forecasting Tmax anomalies during these crucial months. It significantly outperformed other models. It showcased its potential to enhance temperature predictions in this critical period.
The research involved a comprehensive evaluation of the machine learning models. The machine learning models compared the forecasting capabilities with two benchmarks. These were a 10-day persistence predictions benchmark and forecasts derived from the Climate Forecast System (CFS) reforecast.
The results were particularly noteworthy for April and May. The machine learning models exhibited superior performance compared to the benchmark of 10-day persistence. In these months, the machine learning models surpassed persistence as well as achieved results comparable to the CFS reforecast predictions.
However, March and June proved to be challenging for the machine learning models as they did not surpass the skill level of persistence during these periods. This dichotomy in performance highlights the season-specific nature of the models’ effectiveness.
In essence, this research highlights the potential of machine learning models in enhancing our capacity to predict surface air maximum temperature anomalies, particularly during the critical months of April and May. These models could serve as valuable complements to more advanced numerical forecasting systems in the future and help in the preparation of extreme temperature events in India.
