A multi-institutional team funded by the US Department of Energy and NOAA published a 578-day meteorological event log covering 19 months of offshore atmospheric observations along the US East Coast. Collected between February 2024 and August 2025 as part of the Third Wind Forecast Improvement Project (WFIP3), the dataset focuses on the southern New England offshore region and is publicly available through DOE’s Wind Data Hub.
WFIP3 event log published and made publicly available
The event log spans exactly 578 days — February 1, 2024, through August 31, 2025 — and is openly accessible through DOE’s Wind Data Hub. Any researcher, model developer, or grid operator can download it without restriction.
Two complementary components make up the dataset: a manually curated daily weather discussion written by rotating members of the WFIP3 research team, and a set of automated detection algorithms that flagged specific atmospheric phenomena in near real time. Together, they form a structured, searchable record of nearly every day of the campaign. Funding came jointly from DOE’s Wind Energy Technologies Office and NOAA’s Atmospheric Science for Renewable Energy Program, with contributions from Lawrence Livermore and Pacific Northwest National Laboratory, as well as the Woods Hole Oceanographic Institution.
Why the log was created: gaps in offshore wind forecasting data
The southern New England offshore region hosts active and expanding wind energy development, yet it has historically been one of the most data-sparse areas for atmospheric observation along the US East Coast. Before WFIP3, no high-resolution vertical profiles of wind, temperature, and humidity existed for the marine atmospheric boundary layer here.
That gap carries real consequences. Forecast errors in offshore wind speed — particularly around ramp events and low-level jets — impose direct economic costs on grid operators balancing supply and demand in real time. Improving the numerical weather prediction models that underpin those forecasts depends on better observational records to work from.
WFIP3 generated the first measurements of their kind in this region. But a large volume of high-frequency sensor data is only useful if researchers can navigate it efficiently, which is exactly what the event log was designed to enable. It builds on a framework proven during the Second Wind Forecast Improvement Project (WFIP2) — focused on complex terrain in the US Pacific Northwest — where systematic daily logging demonstrably guided model evaluation.
What the log records: automated and manual components
The automated component drew on scanning lidars operating in six-beam vertical-profiling mode at three sites: RHOD (onshore), BLOC (nearshore Block Island), and NANT (offshore Nantucket). Each day, algorithms calculated the percentage of time that low-level jets, weak winds, strong winds, easterly winds, extreme wind veer, and wind ramp-up or ramp-down events were present at each location.
Low-level jets were flagged when a lidar profile showed a distinct wind speed maximum below 800 meters, a drop of at least 3 meters per second above that peak within 400 meters, and a peak speed of at least 10 meters per second. Extreme wind veer was defined as a direction difference exceeding 20 degrees between 100 and 300 meters altitude.
Manual entries documented synoptic conditions, fog, convection, precipitation, sea breezes, lightning, and the perceived accuracy of the HRRR model’s day-ahead forecast below 500 meters. Each day also received two usefulness ratings on a 1-to-5 scale — one for model development, one for observational research — because the most valuable days differ depending on the application. National Weather Service offices in Boston and Upton, New York, contributed expert entries alongside the core team.
Key findings from the 19-month dataset
Low-level jets emerged as the dominant feature of the dataset, occurring on more than two-thirds of days in some months. Activity peaked in spring and summer, with mean daily durations exceeding 20 percent of the day at the nearshore BLOC site in June. Winter months showed reduced activity across all three sites.
Extreme wind veer was most common at the onshore RHOD and nearshore BLOC sites, averaging more than 20 affected days per month. At the offshore NANT site, it occurred less frequently — a pattern pointing to a clear land-to-sea gradient in boundary-layer structure.
The co-occurrence statistics reflect physically meaningful relationships. Sea-breeze days carried a 56.8 percent probability of coinciding with a low-level jet, while only 10.2 percent of low-level jet days also featured a sea breeze. That asymmetry is driven by summer and early fall, when cool sea-surface temperatures sustain the stable marine layer supporting both phenomena. Fog co-occurred with precipitation on 62 percent of fog days, and convection was accompanied by precipitation on 90.7 percent of convective days — confirming that most precipitation in the region is stratiform or frontal rather than convective.
Background: the broader WFIP3 campaign and context for future use
WFIP3 is the third in the Wind Forecast Improvement Project series. WFIP1 and WFIP2 targeted land-based settings, with WFIP2 concentrating on complex terrain in the Pacific Northwest. WFIP3 shifts focus entirely offshore, where the marine boundary layer presents distinct forecasting challenges, and US wind energy capacity is expanding rapidly.
The event log’s standardized metadata and open-access format are built for downstream use. Researchers can filter by event type to build targeted case study sets, validate numerical models against specific atmospheric conditions, or identify periods of large forecast error for diagnostic work — isolating low-level jet days unaffected by sea-breeze forcing, for instance, requires filtering a single column.
The authors flag one significant limitation. Oceanographic processes were excluded from the automated log because the relevant sensors did not stream data in real time during the campaign. They argue this points to a clear need: future offshore field experiments will require real-time sensor-cloud-AI integration capable of automatically flagging oceanographic variability alongside atmospheric conditions.
What this dataset offers the research community
The WFIP3 event log is a 578-day, openly accessible record of offshore atmospheric conditions along the US East Coast, combining automated lidar-based detection with manually curated daily weather narratives. It documents the frequency and seasonality of low-level jets, wind ramps, extreme wind veer, sea breezes, fog, and convection across onshore, nearshore, and offshore sites. Usefulness ratings distinguish days most valuable for model development from those best suited to observational research. Available now on DOE’s Wind Data Hub, its standardized structure is built to integrate directly with the broader suite of WFIP3 high-frequency measurements archived there.
