A high-resolution wind dataset over North America

The Science

Observation datasets of wind speeds and directions across North America are sparse. Developing and evaluating high-resolution, long-term gridded wind datasets that include both speed and direction across North America is critical for creating resilient infrastructure. For this study, researchers created a 20-year, 4 km-resolution dataset of wind speeds and directions across North America, including Alaska and Puerto Rico, and surrounding oceans. They created this dataset and other atmospheric variables using a method called dynamical downscaling. This method uses physics-based numerical models over a limited area. Researchers compared the results to real measurements from ground stations and towers. The study found that dynamical downscaling better captured wind patterns than coarser resolution reanalysis data. This was especially true over mountains and in coastal areas. The study also explored the uncertainties of wind simulations with different model setups and their year-to-year variability.

The Impact

The new wind dataset improves understanding of wind resources over complex terrains like hills and coastal regions. It also provides high spatial resolution information needed at the infrastructure scale. This means researchers can now better predict how wind behaves at different times of day and across multiple years. By taking a closer look at wind speed and direction with a focus on their variability from year to year or from one model configuration to another, the dataset helps planners lower the risk of placing energy infrastructure in unproductive locations. This improvement in understanding can reduce errors, improve spatial accuracy, and provide more useful information for stakeholders. The data can also support research in downstream areas, e.g., storm surge and wave modeling for coastal science.

Summary

Accurate information about wind conditions is important for deciding where and how to develop wind energy projects. However, direct wind observations are limited in both space and time, creating a need for detailed, long-term datasets that can provide wind information across large areas. Researchers developed a wind dataset ranging from 2001 to 2020 at a fine grid of 4 km. The modeling domain extends over most of North America, including ocean-adjacent regions. The team ran the Weather Research and Forecasting (WRF) model, using ERA5 reanalysis as initial and boundary conditions.

They evaluated the dataset by comparing it with measurements from meteorological towers and Automated Surface Observing System (ASOS) stations, as well as with the widely used ERA5 global reanalysis dataset. The high-resolution dataset generally represented wind conditions better than ERA5, particularly in mountainous regions and coastal areas, where local terrain strongly influences wind. It more closely reproduced observed wind-speed distributions and daily wind patterns both near the surface and at wind-turbine hub heights. The dataset also captured the large-scale weather patterns that influence seasonal changes in wind direction. In addition, the team evaluated uncertainty in the model by running simulations with different starting conditions and model settings. Together, these analyses provide a detailed picture of long-term wind resources and their uncertainty, making the dataset useful for applications in energy infrastructure, coastal resilience, and planning for extreme events, such as high wind gusts.