The warm season (May-September) LLJ events were identified using the observed floating lidar wind profiles and a combination of three different types of LLJ detection algorithms in the literature, whose criterion have been slightly modified from their original usage in prior coastal LLJ studies. An event was included in the warm season LLJ dataset, if at least two out of the three different algorithms identified a jet. The ERA5 reanalysis underpredicts wind speeds by 1 – 9% through the lowest 200 m profile during spring and summer, with April through June having the largest negative bias of 9.5-17.5% (1.0 – 1.8 m s-1). In contrast, NOW-23 has a more similar 60 – 200 m wind profile compared to the observed, with average underprediction of -0.08 m s-1 (0.9%). NOW-23 overpredicts wind speeds by 0.5 – 1.8 m s-1 in June-September, especially from 2000 – 0800 UTC. During warm season LLJ events, the ERA5 was found to have the poorest performance in terms of both detection (zero hours detected) and LLJ structural characteristics. On average, the lidar observed LLJ profile had a jet nose maximum of 11.2 ± 0.4 m s-1 at 100 m ASL, while ERA5 exhibited a more elevated and weaker maximum of 9.3 ± 0.4 m s-1 at 140-160 m. The jet for the HRRR analysis is on average 9.6 ± 0.9 m s-1 and also elevated to 140-160 m. Overall, this work highlights the challenges in using analyses with the limited offshore observations available for data assimilation in offshore wind prediction and wind resource assessment. This presentation will also highlight the synoptic patterns and environmental conditions when exceptionally large wind speed profile errors occur in the New York Bight.

