A multi-instrument framework for boundary layer wind and turbulence profiling

Abstract. Characterizing how the planetary boundary layer varies in space and time requires reliable, continuous wind and turbulence profiles from the surface into the lower troposphere. No single instrument spans this height range. Although combining complementary instruments is well-established practice, there is still no widely adopted, transparent method for turning these measurements into analysis-ready profile products. We present WINDPROF, an open-source framework that combines measurements from scanning and profiling Doppler lidars, radar wind profilers, tower-mounted sonic anemometers, and surface meteorological stations. It produces 10-minute profiles of wind speed, direction, vertical velocity, turbulence intensity, turbulence kinetic energy, and velocity-component standard deviations on a standardized vertical grid. The framework applies instrument-specific quality control and retrieves winds using velocity-azimuth display fitting. Turbulence statistics are estimated using Reynolds decomposition, with lidar beam-geometry biases corrected in the mean wind speed used to calculate turbulence intensity. Profiles are then merged using a height-dependent hierarchical strategy. We demonstrate the framework using 19 months of observations from four coastal sites during the Third Wind Forecast Improvement Project (WFIP3). Comparisons between co-located lidars, pooled within each site, yielded wind speed root-mean-square errors of 0.36–0.78 m s^-1, while merged data availability below 200 m above ground level ranged from 84% to 98%. During coastal fog, median data availability at 100–300 m was 39.3% at a site relying exclusively on optical instruments, compared with 94.5% at a radar-equipped site, demonstrating the value of complementary measurement technologies. Comparison with an independently developed Bayesian optimal-estimation retrieval at Nantucket showed close agreement (mean difference = +0.160 m s^-1, root-mean-square error = 0.851 m s^-1, r = 0.990). Differences were larger near low-level jet maxima and during periods of rapid wind change, consistent with the methods' different averaging times.