Contamination of Wind Profiler Data by Migrating Birds: Characteristics of Corrupted Data and Potential Solutions

View More View Less
  • 1 Environmental Technology Laboratory, ERL/NOAA, Boulder, Colorado
  • | 2 U.S. Army Research Laboratory, White Sands Missile Range, New Mexico
  • | 3 Department of Atmospheric and Oceanic Sciences, McGill University, Montreal, Quebec, Canada
  • | 4 CIRES, University of Colorado, Boulder, Colorado
  • | 5 Sonoma Technology, Inc., Santa Rosa, California
© Get Permissions
Full access

Abstract

Winds measured with 915- and 404-MHz wind profilers are frequently found to have nonrandom errors as large as 15 m s−1 when compared to simultaneously measured rawinsonde winds. Detailed studies of these errors which occur only at night below about 4 km in altitude and have a pronounced seasonal pattern, indicate that they are due to the wind profilers' detection of migrating songbirds (passerines). Characteristics of contaminated data at various stages of data processing are described, including raw time series, individual spectra, averaged spectra, 30- or 60-s moments, 3- or 6-min winds, and hourly averaged winds. An automated technique for the rejection of contaminated data in historical datasets, based on thresholding high values of rnoment-level reflectivity and spectral width, is shown to be effective. Techniques designed for future wind profiter data acquisition systems are described that show promise for rejecting bird echoes, with the additional capability of being able to retrieve the true wind velocity in many instances. Finally, characteristics of bird migration revealed by wind profilers are described, including statistics of the spring (March–May) 1993 migration season determined from the 404-MHz Wind Profiler Demonstration Network (WPDN). During that time, contamination of moment data occurred on 43% of the nights monitored.

Abstract

Winds measured with 915- and 404-MHz wind profilers are frequently found to have nonrandom errors as large as 15 m s−1 when compared to simultaneously measured rawinsonde winds. Detailed studies of these errors which occur only at night below about 4 km in altitude and have a pronounced seasonal pattern, indicate that they are due to the wind profilers' detection of migrating songbirds (passerines). Characteristics of contaminated data at various stages of data processing are described, including raw time series, individual spectra, averaged spectra, 30- or 60-s moments, 3- or 6-min winds, and hourly averaged winds. An automated technique for the rejection of contaminated data in historical datasets, based on thresholding high values of rnoment-level reflectivity and spectral width, is shown to be effective. Techniques designed for future wind profiter data acquisition systems are described that show promise for rejecting bird echoes, with the additional capability of being able to retrieve the true wind velocity in many instances. Finally, characteristics of bird migration revealed by wind profilers are described, including statistics of the spring (March–May) 1993 migration season determined from the 404-MHz Wind Profiler Demonstration Network (WPDN). During that time, contamination of moment data occurred on 43% of the nights monitored.

Save