EuLerian Identification of ascending AirStreams (ELIAS 2.0)

The ML-model ELIAS2.0. using a U-Net convolutional neural network predicts conditional probabilities of WCB inflow, ascent, and outflow footprints from instantaneous gridded fields, offering a computationally efficient alternative to trajectory-based approaches. The ML-model was introduced by Quinting and Grams (2022).

The WCB inflow is defined as the part of the WCB in the lower troposphere, below 800 hPa, the WCB ascent as the part of the WCB between 800 and 400 hPa, and the WCB outflow the region above 400 hPa. ELIAS2.0. is trained for each WCB stage separately and uses a total of five predictors: four are derived from temperature, geopotential height, specific humidity, and the horizontal wind components at 1000, 925, 850, 700, 500, 300, and 200 hPa isobaric surfaces and the fifth is the 30-day running mean trajectory-based climatology of WCB occurrence frequency.