Efficient high resolution heat mapping cuts climate simulation costs
Less is more: error-distance scaling relation for data-efficient kilometer-scale downscaling of extreme heat
Machine Learning
Summary
Predicting extreme heat in cities requires detailed weather data at very small scales, which usually needs expensive simulations. The authors developed a method called CASPER that uses a type of neural network to convert low-resolution climate data into high-resolution temperature and wind maps. They found that the quality of predictions depends on how close the training data climate is to the target climate, and training can be done with much less simulation than expected. This makes detailed heat mapping accessible without huge computing resources, helping cities better prepare for heatwaves.
What this means in practice
- •For urban climate planners: Generate detailed city-scale heat maps efficiently using limited climate simulations to aid heatwave adaptation plans.
- •For regional weather forecasters: Improve temperature and wind detail in local forecasts with fewer costly simulations by training models only on relevant climate periods.
Authors
Ahmed Marey, Henry Lu, Abhishek Gaur, Liangzhu Leon Wang, Sherif Goubran, Malek Aloui, Theodore Potsis, Alex Hernandez-Garcia, David Rolnick
Abstract
Extreme heat is where urban adaptation needs kilometer-scale data the most, but the simulations training a downscaler can cost more than they save, and how much is needed has not been identified. We measured it with CASPER, a U-Net with a structure-preserving loss downscaling 32 km reanalysis to 1 km temperature, humidity and wind, across 24 configurations of one to eight months. Held-out error grows linearly with climatological distance to the training data, RMSE = 0.83 + 2.95 d, explaining 90% of its variance against 7% for volume and predicting unseen months in advance. On held-out extreme summer weeks CASPER preserves the fine-scale structure and cross-variable physics that matched-budget baselines degrade, and matches station observations during documented heat waves to within 1.8 K. Transfer to a new region degrades geographically; 11 days of local simulation cuts Vancouver's held-out error from 3.8 to 1.3 K. Training periods should span the target climate: the same accuracy for four times less simulation, putting kilometer-scale downscaling of extreme heat within reach of groups without large computing facilities.