This study evaluates a physics-informed neural network for potential-temperature forecasting under incomplete thermal observations, constrained by a pressure-coordinate advection-source equation and a diabatic-source closure frozen after the preceding 12 hours. Validation uses hourly ERA5 reanalysis at three pressure levels for one-, two- and three-hour horizons against persistence, local-trend, and two neural baselines.
The approach shows how physical constraints complement data in short-horizon forecasting, which is relevant for climate-aware digital twins of cities and regions.
Translation and summary are machine-generated from the source. The full article is not republished; its rights belong to the source.
