Future Weather Match: How Cities Can Predict Each Other’s Heat

Scientists have found a way to let one city look at its own past heat levels and guess what the temperature will be in another place years later. The method blends a deep‑learning tool called a Temporal Convolution Network with a statistical test that checks if one time series can help predict another. Using records of the highest daily temperatures, researchers studied four Indian cities: Coimbatore, Gwalior, Ludhiana and Kashmir.

They asked: when will the maximum temperature in Kashmir become similar to the temperatures that have already happened in the other three cities? By comparing past data, the model found clear matches up to the present and even indicated future years where the temperatures could line up. The accuracy is high: over a 75‑year period the average error was only about 0. 03 degrees, and the spread of errors stayed very tight. Such predictions can help farmers plan crops like maize and allow planners to set up irrigation and other infrastructure well ahead of time.

Actions