This article describes how bird migrations in Western New York and Southern Ontario produce errors in the radar-derived winds on Buffalo's WSR-88D weather radar.  For more information on how our weather radar detects bird migrations and early morning "bird rings" produced by feeding fowl, you can click on our "Birds and Radar Page". 

Thomas A. Niziol
NWSFO Buffalo, NY
The implementation of the Weather Surveillance Radar -1988 Doppler (WSR-88D) network across the United States has introduced forecasters to an array of new meteorological products (Klazura and Imy, 1993). One of the most important contributions of the radar is its ability to provide real-time wind information from the surrounding atmosphere (Federal Meteorological Handbook No. 11, 1990). However, the technology is not totally foolproof. Raindrops and snowflakes aren't the only reflectors in the atmosphere. Swarms of insects and flocks of birds are also excellent reflectors of radar energy. Because birds are not necessarily carried along by the prevailing wind, they can introduce significant errors in the radar algorithms that calculate winds (O'Bannon, 1995). This paper will present a case study of bird contamination at the WSR-88D in Buffalo, NY, and suggest ways to alert the forecaster to the potential for a bird contaminated Velocity Azimuth Display (VAD) Wind Profile.

The WSR-88D measures Doppler radial velocity within a sample volume and converts the information through a series of algorithms into many derived products. Wind information is processed by the VAD algorithm. The processed information is displayed as wind velocity and direction in the VAD Wind Profile (VWP) product. The algorithm assumes horizontal uniformity of the wind and precipitation fields and does not account for deformation fields such as a frontal boundary. Significant deviations from this uniformity can bias the estimate, or increase RMS velocity and decrease symmetry enough to discard such wind estimates (Federal Meteorological Handbook No. 11, 1991).

The VWP product displays an estimate of the mean horizontal wind as a series of wind barbs on a time vs. height scale for up to 30 altitudes in the vicinity of the radar. The specific altitudes are adaptable in 1000-ft. MSL increments and are plotted at levels for which there is sufficient signal to be processed by the algorithm. The data from the VWP product provides wind information that aid the forecaster in many aspects of the forecast process. Examples include aviation forecasts for wind sensitive operations such as hot air ballooning and soaring, short term forecasts for dispersion of airborne pollutants during hazardous materials (HAZMAT) releases, low level wind forecasts to predict location and movement of lake effect snow bands (Niziol et al., 1995), and classification of thunderstorm type based on the low level wind shear profile (Weismann et al., 1982). Finally, VWP data is expected to eventually be incorporated into the initialization of operational numerical forecast models as well as the ever increasing suite of mesoscale models. It is extremely important therefore to be aware of the variables that can influence and directly affect the quality of derived wind data from the radar.

There is a wealth of published information about the sensitivity that many types of radars have to moving point targets such as birds (Eastwood, 1967). Recently, a number of case studies have been presented that show the possible contamination of WSR-88D VAD winds by birds (O'Bannon, 1995). In order to gain some understanding of the relationship between weather radar return signals and birds, it is important to know a bit about bird behavior and migration. Wilczak (1995) summarized favored patterns of migration across North America that can be categorized both seasonally and diurnally. Most bird migration occurs at night during the Spring and Fall. Typical air speeds range from 16 to 30 kts. and they tend to fly at a level where winds are most favorable. Maximum heights of migration are generally around 7000 ft. MSL but heights to 15,000 ft. MSL are possible if winds are favorable. In addition, the rate of migration, especially in the Fall, is often related to the passage of synoptic frontal systems. Because birds provide excellent reflectivity cross sections (Federal Meteorological Handbook No. 11, 1990), they are easily detected by radar. Birds, however, do not drift with the prevailing wind as insects do. Therefore, birds can introduce large errors into the VAD wind algorithm.

A number of WSR-88D sites have reported instances when radar-derived winds have differered considerably from nearby radiosonde observations (O,Bannon 1995). Often, these anamolies were consistent with the seasonal and diurnal patterns of bird migration, when large flocks of birds are all flying in the same direction. At times, the contaminated winds were very difficult to identify with current algorithms because the erroneous data set provides a of a very coherent signal.

Western New York State is located in close proximity to Lakes Erie and Ontario. The shores of both lakes are noted as staging grounds for many species of birds that migrate north and south each Spring and Fall. In addition, the strip of land that separates both lakes, known as the Niagara Peninsula, serves as a "flyway" for many species of birds that would prefer over-land migration routes (Rising 1997, personal communication). There are also a number of sites in Western New York, such as the Iroquois and Montezuma Wildlife Refuges, that are renowned as landing and nesting habitats for many types of waterfowl. Because of these features, Western New York State is considered a prime area for bird migration during the Spring and Fall. It follows that the weather radar at Buffalo is particularly vulnerable to potential contamination of its VAD wind profile. This is especially true under certain conditions such as the superrefraction of the radar beam due to nocturnal inversions that exaggerates the depth of the bird echoes (O'Bannon 1995).

One such case of coherent, erroneous winds occurred during early November 1996 in the vicinity of Buffalo, NY. By this time of the year, many of the migratory songbirds have already left for warmer conditions in southern latitudes. However, waterfowl flights are in full swing (Gathreaux 1980). During the period 5-10 November 1996, a pronounced change in the synoptic weather pattern occurred across the Great Lakes Region. During 6-8 November, a very strong southerly flow developed over the Great Lakes in advance of slow moving surface cold front over the Midwest (Fig 1). The surface cold front crossed the eastern Great Lakes on 8 November and winds shifted to north that evening (Fig 2). The evening also was characterized by nearly continuous rain showers. A secondary low and associated cold front crossed the region early on the 9 November. The wind maintained a north component during most of 9 November before backing to a weak westerly flow that night (Fig 3).

The change in the wind field at different levels for Buffalo, NY during the days before and during the migration is displayed in Fig 4. Of particular note is the extended period of very strong southerly flow during 7-8 November when the winds taken from the Buffalo radiosonde were as high as 60 kts between 2000 and 7000 ft. MSL.

During the evening of 9 November, approximately one half hour after sunset, the radar began to display an area of anomalous echoes along the north shore of Lake Ontario, about 60 km north of Buffalo (Fig 5). The echoes showed a continuous movement due south during the next two hours. At the same time, however, the area well to the south of the Buffalo radar was covered with echoes that looked more like distinct precipitation cells which were moving toward the northeast.

Prior to the appearance of the anomalous echoes north of the radar, the VWP generally indicated a southwest wind at 15 to 20 kts at levels between 2000 and 8000 ft, corresponding to isolated precipitation echoes south of Buffalo (Fig 6a). As the anomalous echoes north of the radar site appeared, the VWP product began to display north winds at 15 to 25 kts between approximately 2000-5000 ft. MSL (Fig 6b). The winds eventually became well aligned out of the north at 15 to 25 kts (Fig 6c). The pattern presented by the reflectivity loop and the VWP product immediately aroused the suspicion of the forecasters who postulated that what they were really seeing north of the radar were flocks of birds migrating south across the radar site from the north shore of Lake Ontario.

From an overall point of view, the bird migration scenario seems quite reasonable. The synoptic scale weather pattern a few days prior to the event produced an extended period of strong southerly winds, accompanied by precipitation, for a number of days before the flight. Lincoln (1939) noted that headwinds of high velocity, as well as periods of rain or snow, can force succeeding flocks of birds down at some point. It would seem reasonable that the north shore of Lake Ontario would make a good staging area for birds to wait for a few days until the weather conditions improved. The "foul" weather grounded a number of migrating birds until the winds and weather became favorable after the passage of the second cold front. It should be noted that even though the winds became more favorable for migration on the evening of 8 November, nearly continuous rain showers would have discouraged flight (Lincoln 1939).

When conditions finally did improve, the resulting migration on the evening of 9 November was more like a "bird wave", as discussed by Lincoln (1939), that progressed southward across the radar site. The large flocks of migrating birds produced anomalous VWP winds on the Buffalo WSR-88D. Because there were no precipitation echoes at the same elevation angle and distance from the radar, the reflected energy was dominated by the bird echo returns. Therefore, the VWP data showed only minor RMS errors and variablity to the mean radial velocity of the wind, a likely signature for birds all moving roughly in the same direction.

This event was in progress at the time of the evening radiosonde launch, so it was possible to evaluate the radar-derived winds against those measured directly by the radiosonde. The radiosonde winds were significantly different from the VWP data between 2000 and 5000 ft., indicating a weak southwest flow compared to the stronger north flow indicated by the radar (Fig 7). The radiosonde data strongly suggested that the north winds on the VWP were actually a signature resulting from the movement of the birds.

It was fortuitous that the bird migration occurred at the time of the radiosonde release and at a location in close proximity to the radiosonde site. At times other than the standard radiosonde release times (i.e. 0000 UTC and 1200 UTC) or in locatons far removed from the radiosonde site another quality control check can be done by comparing the VWP data with hourly forecast soundings from numerical model guidance (Mahoney et. al., 1997). In Figure 6, the VWP and radiosonde data from the evening of 9 November were compared to a 12-hr wind forecast for Buffalo, NY, from the 1200 UTC 9 November eta model run. The model forecast winds compare very well with the actual radiosonde data and, in this case, act as a good "first guess" quality control check against the VWP winds. Because the forecast soundings are available on an hourly basis, forty-eight hours into the future, this type of data set can be used by forecasters to perform a preliminary check on VWP data at times the are far removed from radiosonde releases. It also allows sites that are far removed from the standard radiosonde sites to quality control their VWP.

A check for bird contamination on wind profilers was devised by F. M. Ralph and D. W. Van de Kamp based on work reported in Wilczak et al. (1995). Although the somewhat rigorous set of algorithms developed for the wind profilers cannot be employed on Doppler weather radars, these more generalized ideas and patterns can be used to alert forecasters to the potential for contaminated wind data. The rules noted by Wilczak (1995) have been compiled in
Table 1.  Minor changes have been made to correspond to sunset in the eastern U.S.

Spring Fall
Time of Year 15 February - 15 June 15 August - 30 November 
Time of Day 2300 - 1100 UTC 2100 - 1200 UTC
Wind direction Southerly Northerly
Height below 10,000 ft below 10,000 ft
 Table 1: Patterns associated with bird migration that should alert forecasters to the potential for the
               contamination of radar derived winds.
Until more rigorous algorithms are developed for the WSR-88D to identify potential errors introduced by bird contamination, these simple rules, coupled with data checks from radiosonde releases and model soundings, can alert the forecaster to the conditions under which bird migration may produce contaminated wind data.

It has been shown that under certain conditions the VWP data from the WSR-88D can be significantly contaminated by biological targets such as birds. In fact, it is possible that the contaminated winds can have a serious impact on the forecast effort including erroneous forecast model initialization, wind-sensitive aviation events, dispersion of airborne hazardous materials, thunderstorm and lake effect snow morphology. However, a number of procedures can be employed to cross-check the accuracy of the winds.

If the event occurs near the location and time of radiosonde launches, which are roughly near sunrise and sunset during the Spring and Fall months, the VWP data in question can be cross-checked with actual radiosonde winds. At locations that are some distance from radiosonde sites, or at times other thatn the standard radiosonde release times, the VWP winds can be checked against hourly forecast profiles from the operational numerical models. Finally, because bird migrations follow identifiable patterns related to the season, time of day, direction of movement, and height above the ground, a number of general rules can be used to alert forecasters to the conditions under which VWP contamination from migratory birds can occur.

The author would like to thank Tim O'Bannon of the WSR-88D Operational Support Facility and Gerry Rising of the Buffalo Museum of Science for their helpful suggestions and insight into bird migration and radar contamination. Also thanks to Warren Snyder of the NWS office at Albany, NY for supplying the synoptic scale weather maps.

Eastwood, E., 1967: Radar Ornithology. Muethen & Co., Ltd. 278 pp.

Federal Meteorological Handbook No. 11 (Interim Version One), 1990: Doppler Radar Meteorological Observations, Part B, Doppler radar theory and meteorology. FCM-H11B-1990, Office of Federal Coordinator for Meteorological Services and Supporting Research, Rockville, MD, 228 pp.

_____, 1991: Doppler Radar Meteorological Observations. Part C, WSR-88D products and algorithms. FCM-H11B-1991, Office of Federal Coordinator for Meteorological Services and Supporting Research, Rockville, MD, 210 pp.

Gathreaux, S. A., 1980: Direct visual and radar methods for the detection, quantification, and prediction of bird migration. Special publication No. 2. Department of Zoology, Clemson University, Clemson, SC, 67 pp.
Klazura, G. E., and D. A. Imy, 1993: A description of the initial set of analysis products available from the NEXRAD WSR-88D system. Bull. Amer. Meteor. Soc., 74, 1293-1311.

Lincoln, F. C., 1939: The Migration of American Birds. Doubleday, Doran & Co., Inc., 189 pp.

Mahoney, E. M. and T. A. Niziol, 1997: BUFKIT: A software application toolkit for predicting lake effect snow. Preprints, 13th International Conference On Interactive Information . and Processing Systems. (IIPS) for Meteorology, Oceanography and Hydrology, Long Beach, Amer. Meteo. Soc. 388-91.

Niziol, T. A., W. R. Snyder, and J. S. Waldstreicher, 1995: Winter weather forecasting throughout the Eastern United States. Part IV: Lake effect snow. Wea. Forecasting, 10, 61-77.

O' Bannon, T,. 1995: Anomalous WSR-88D Wind Profiles - Migrating Birds? Preprints 27th Conference on Radar Meteorology, Vail, Amer. Meteor. Soc., 663-665.

Weisman, M.L., and J.B. Klemp, 1982: The structure and classification of numerically simulated Convective Storms in Directionally Varying Wind Shears. Mon. Wea. Rev., 112, 2479-2498.

Wilczak, J. M., R. G. Strauch, F. M. Ralph, B. L. Weber, D. A. Merritt, J. R. Jordan, D. E. Wolfe, L. K. Lewis, D. B. Wuertz, S. A. McLaughlin, R. R. Rogers, A. C. Riddle, and T. S. Dye, 1995: Contamination of wind profiler data by migrating birds: Characteristics of corrupted data and potential solutions. J. Atmos. Oceanic Technol., 12, 449-467.