Detection of Upper-Level Troughs and Ridges Using Deep Learning - Application in the Mediterranean

Ofir Ariel; Omer Sela; Hadas Saaroni; Baruch Ziv

Physics-informed axis detectionERA5 500 hPa · Mediterranean domain

Detection of upper-level
troughs and ridges
using deep learning

Application in the Mediterranean

Ofir Ariel¹ Omer Sela² Hadas Saaroni¹ Baruch Ziv³

1

Department of Environmental and Earth Sciences, Tel Aviv University

2

Blavatnik School of Computer Science and AI, Tel Aviv University

3

Department of Natural Sciences, The Open University of Israel

We present a physics-informed neural network for detecting trough and ridge axes from ERA5 500 hPa geopotential height and horizontal wind fields, evaluated against a new expert-labelled benchmark.

Explore 600 expert-labelled trough scenes, 200 expert-labelled ridge scenes, 33,604 archived model analyses, and the 1979–2024 seasonal climatology.

Project resources

PaperDOI link forthcomingCodeTraining and inference DatasetBenchmark, archive, climatology CheckpointsProduction and CV folds

Detection of upper-leveltroughs and ridgesusing deep learning

Detection of upper-level
troughs and ridges
using deep learning