Fig. 6: Model forecast performance for AR landfalls in three Northern Hemisphere regions.
From: Global performance benchmarking of artificial intelligence models in atmospheric river forecasting
a Success rates of AR landfall forecasts in NAM at 1–10 days lead time. The sector shows each model’s performance, with equally divided sectors indicating identical performance. b IVT difference between model forecasts (1-, 5-, and 10-day lead times) and ERA5 reanalysis for AR landfall events in NAM. For each model, the left half-box shows the distribution of the values, and the right half-scatter plot displays individual values. The central line in each box represents the median of the values, the box edges denote the interquartile range (25th–75th percentiles), and the whiskers extend to 1.5 times the interquartile range. c Intersection over union (IOU) between model forecasts and ERA5 reanalysis for AR landfall events in NAM. d–f Same as (a–c), but for EU. g–i Same as (a–c), but for EA.
