Advancing NWP Wind Forecasts in Complex Terrain by Lidar Technology

Doppler lidars are capable of providing accurate information on wind conditions at levels of operating wind turbines, which continue to grow in size. In addition to the measurement of conventional quantities such as mean wind speed, wind direction, and turbulence, lidars provide data on wind speed shear and directional veering through the rotor layer which cannot be obtained from turbine nacelle anemometers. Deployment of tall towers with several levels of measurements to complex terrain sites can be a time consuming and costly effort. Lidar data collected at high vertical and time resolution provide a comprehensive dataset to validate NWP model wind forecasts used by wind farm operators for optimal grid management and effective balancing between electricity consumption and wind power generation.

To improve wind forecasts in complex terrain, an 18-month experiment, the Second Wind Forecast Improvement Program (WFIP-2), was held in the Columbia River Basin region in Oregon and Washington. Measurements from a variety of in-situ and remote-sensing instruments deployed to the study area were used to investigate atmospheric phenomena that affect model accuracy and to advance model physics. Among these instruments, two NOAA scanning, pulsed Doppler lidar systems were deployed to sites near the Wasco and Arlington, Oregon airports from September 2015 to April 2017, providing real-time measurements of wind velocity along the lidar beam with a precision of 20 cm/s. Another scanning Doppler lidar, the Halo Photonics, was deployed by the University of Notre Dame to the Boardman airport site.

High-resolution lidar data allow capturing wind variability at hub-height or through the rotor layer at different time scales: from minutes and hours to daily cycles, which may be substantial due to various factors such as daily temperature changes; and up to seasonal and yearly variations, which may be due to variability in pressure gradients and large-scale synoptic systems. In addition to quantifying the temporal and vertical variability of wind flow at both sites, measurements from three identical scanning lidars separated by ~40 km provide a unique opportunity to understand the horizontal variability between sites due to the complex terrain. Analysis of monthly, seasonal, annual, and experiment-long distributions of hub-height and rotor-layer wind directions show two modes of prevalent winds at both sites with more frequent westerlies for summer months and almost equal frequency of occurrence for westerly and easterly winds for the winter months. Observed wind speeds at all sites fluctuate in the range of 0-20 m/s with stronger winds during summer months compared to winter months.

Lidar measurements at three sites are used to validate NWP model wind forecasts under various meteorological conditions and to evaluate model accuracy over days, months and seasons, as well as for periods of interesting meteorological events observed in the study area. Verification metrics such as bias, RMSE, MAE, and the correlation coefficient between observed and modeled wind variables, analyzed as a function of height, time, and forecast lead hour, provide insight into potential model improvements. A discussion will be provided on validation uncertainty, which is the combined effects of lidar measurement errors, data averaging techniques, and extrapolation of modeled variables to the location of instruments.