Atmospheric Rivers (ARs), defined as narrow bands of intense water vapor transport, are a major driver of wintertime heavy precipitation and flood events across Western Europe. Due to the high impacts of ARs, accurate forecasts are crucial, but their initialization and evaluation remains challenging due to a lack of observational data over the oceans. The “North Atlantic Waveguide, Dry Intrusion and Downstream Impact Campaign" (NAWDIC) collected unique dropsonde observations of several strong ARs that caused widespread flooding, disruption and economic damage over the Iberian Peninsula and Northern Africa in January and February 2026.</p> <p>Using this one-of-a-kind dataset, we performed the first systematic evaluation of the AR representation over the North Atlantic ocean basin in the IFS and the ICON models. Specifically, the vertical structure of temperature, specific humidity, horizontal wind speed, and moisture flux magnitude was analyzed using the model data assimilation outputs. While the overall representation of ARs in the IFS and ICON models is skillfull, results indicate large random errors in the horizontal wind speed estimates and a slight cold bias over North Atlantic AR regions. The IFS exhibits an underestimation and large random errors in the humidity, peaking between 700 and 750 hPa, which corresponds to the upper-level extension of ARs. The ICON model tends to slightly underestimate the humidity throughout the boundary layer and the entire lower troposphere up to 650 hPa. The large random errors in wind speed and the underestimation of humidity in both models lead to an underestimated and highly uncertain moisture flux magnitude in the IFS and ICON models. The results of this study highlight the importance of observational data from AR regions over the North Atlantic and the current model limitations in accurately representing intense ARs affecting Western Europe.