Predicting maximum temperatures over India 10-days ahead using machine learning models.

Area of Science:

• Meteorology
• Climate Science
• Artificial Intelligence

Background:

• India faces significant heatwave risks during March-June due to high daytime temperatures (Tmax).
• Accurate prediction of Tmax anomalies is crucial for mitigating heatwave impacts.

Purpose of the Study:

• To evaluate machine learning models for predicting daily Tmax anomalies 10 days in advance for India.
• To identify the optimal machine learning model for this prediction task.

Main Methods:

• Evaluated 10 distinct machine learning models.
• Identified the AdaBoost regressor with Multi-layer Perceptron as the optimal model.
• Benchmarked predictions against persistence and Climate Forecast System (CFS) reforecasts.

Main Results:

• The optimal machine learning model showed higher skill than persistence and was comparable to CFS reforecasts in April and May.
• Model performance was limited in March and June, performing similarly to persistence.

Conclusions:

• Machine learning models are promising for predicting surface air maximum temperature anomalies in India during April and May.
• These models can supplement predictions from sophisticated numerical weather models, enhancing heatwave preparedness.