Bird Collisions with Windows on Swarthmore Campus

Guido Grasso-Knight

Michael Waddington

Abstract

Birds hit windows quite frequently, and this is thought to be a major source of bird mortality. A study was undertaken to establish the frequency and conditions under which bird collisions with windows occur on Swarthmore Campus. A route was selected around several academic buildings, and all dead and injured birds were recorded. Additionally, smudges on the windows thought to be left from bird collisions were recorded, and reports from other bird hits on campus were recorded. One dead bird and zero non-lethally injured birds were found on these walks, but several smudges were recorded on certain windows. There was a non-significant correlation between window frame size and number of collisions, as well as window pane size and number of smudges recorded. There was a tendency for reported bird collisions to occur on good migration days (winds from the South). It is suggested that future experiments find a more complete method of recording collision data, such as the use of vibration detectors on the windows.

Introduction

Increasing numbers and diversity of man-made structures create novel obstructions that pose lethal threats to flying birds. Incidence of bird collisions with human structures have shown that birds of all species are affected under a variety of weather conditions. The magnitude and scope of this problem receives little public attention, and remains largely unrecognized (Klem, 1998). However, the efforts of various researchers and volunteer organizations have revealed several major considerations. Through a brief literature review we describe general factors involved in these collisions; specific investigations of window as obstacles and attractants ; and analyses of the mechanisms underlying the behavioral responses of birds to man-made structures.

Birds face numerous dangers when flying in areas under heavy human development. Birds might become disoriented by man-made structures, such as lighted structures, or may not recognize the structures as obstacles, and then collide with often lethal consequences. While specifically addressing the hazards of powerlines, Bevanger explains that the impact of human artifacts on the perceptual environment of the birds depends on the behavior of the bird and the features of the environment (Bevanger, 1994). How the bird flies, whether it is soaring above small houses or darting between narrow passageways, may determine the risk that windows pose to the birds. More generally, the manner in which the bird interacts with its habitats, particularly obstructions, may influence the risk of collision just as greatly as the meteorological, technical , and topographical conditions of the environment.

Obstructions can fall into three groups: 1) passive- overhead wires, tall towers, and windows such as those in office buildings (the focus of our study); 2) active threats- aircraft, cars, trains, wind towers; 3) and trapping threats- lighthouses, and gas flames. Each type shapes the birds environment and responses in a different way. However, all introduce unnatural stimuli with which the birds have limited experience. He explains that the evolutionary success of flight among birds has not yet responded to the threats posed by man-made structures. Bevanger highlights the need for additional research and the importance of understanding how individual bird flight behavior determines collision frequency for different birds. Studies should incorporate detailed understanding of local bird populations, flight and migratory behaviors. Remedies and efforts to reduce the man-made threats will reflect knowledge of local bird communities. For example, recognizing that continuously lighted stuctures may disorient and trap birds, many communities have incorporated rotating or strobe lights of varied colors on tall structures. Yet, the effectiveness of these steps depends on the whether the birds involved actually perceive the colored light without being overly distracted (Bevanger, 1994).

The coordinated efforts of research and volunteers of FLAP (Fatal Light Awareness Program) have detailed the magnitude of collisions in major urban centers such as Toronto (Evans, 1996). They suggest that during the fall migration preventable collisions with man-made structures killed nearly 0.1 % of the fall migrating population. The sheer numbers of birds affeced, often in the thousands, assumes greater significance when we recognize that this was a preventable loss. They identified night time illumination and windows as the major hazards facing migrating birds (Evans, 1996).

Earlier studies by Klem have described the behavioral responses of birds to windows, as well as basis for window mortality (Klem, 1989). Klem identifies two types of windows: transparent and reflective windows. While they are both equally dangerous, they affect the birds' environment differently. Transparent windows remain invisible, while reflective windows mirror the surrounding landscape, and effectively cloak the presence of an obstruction. Birds encounter these windows with two possible results: non-lethal collisions involve tapping or bumping against the window with little harm suffered; lethal collisions kill the victim immediately through cerebral damage or leave the bird too injured to recover or escape predation (Klem, 1989, 1990a).

Most window strike victims do not survive, but they do not die immediately. Despite the widely held belief that birds die from cervical separation and fractures, Klem revealed that internal cranial hemorrhaging is the most common cause of death. In many cases these birds stunned or dazed immediately after impact. As the internal bleeding increases they are also less responsive, and fall easily to predation (Klem, 1990a).

In order for prevention to be effective, Klem considered two questions: why birds collide with windows, and what are the sources of window lethality. Survey information from museum curators and self-reporting from individual homes throughout North America highlighted possible relationships between meteorological and topographical factors in window mortality. Surprisingly, collisions were often equal or higher under favorable conditions as under inclement weather, Based on estimates of window strikes from individual houses and their distribution throughout the nation, the window mortality claims from 97.5 million to 975 million birds each year (Klem, 1989). These are mostly preventable losses.

Four experiments identified deficiencies in bird responses to windows that increase collision frequency and lethality. The first experiment compiled survey responses and personal observations over several years and under varied weather conditions. The second set of experiments during the late fall placed clear and reflective glass panes covering several square meters in various land patches near a small farm. These glass panes were on the ground and not associated with any buildings. The third series conducted in the early spring examined the role of windows in new structures. Window frames were built and placed with clear picture windows in areas previously devoid of any structures. Covering an area of over 8 m2, the windows created different impressions, with reflected, transmitted, or combined images of the surroundings. The final experiment inserted clear and reflected windows( combined surface area over 6 m2) into an old barn that had long stood without any windows. Klem concluded that windows are general threats; age, sex, season, time of day, weather, window type and setting of window do not significantly alter the patterns of collisions (Klem, 1989). Refuting the suggestions that window victims had defective or impaired vision, or were deceived, he maintains that birds simply do not see windows. Old and young birds struck windows with equal relative frequency suggesting that lifetime experience with window environments does not improve the bird's safety (Klem, 1989). Moreover, his work indicated that the orientation of the windows, in line with the migratory pathway axis or perpendicular, did not change strike rates (Klem, 1989).

From his findings, he suggested several considerations that would reduce overall window risk. The number of collisions were closely correlated with the density of the birds in the surrounding area. Reducing the number and type of attractants near windows, such as bird feeders and food-bearing vegetation, could save many bird lives. He alternatively suggests placing the attractants very close to the windows, as the birds do not build up enough momentum for a lethal hit under these conditions. Similarly, the location of the window might be altered favorably; windows near the ground or above 3 m in height were less likely to impair bird flight. Finally, he recommended designing structures which took in account recognized bird flight patterns for a given location (Klem, 1989).

Later works by Klem outlined preventive techniques directly related to window modifications. His research also highlighted the causes of death following window strikes. By comparing different coverings and barrier markings on windows, Klem demonstrated that birds will avoid windows when they are clearly marked (Klem,1990b). Completely covering the window with a cloth provided the most reliable deterrent. Obviously, this is an unappealing and impractical option for large buildings with numerous sheet glass surfaces. Vertical and horizontal strips across the windows, however, also reduced window strike frequency. When placed 5-10 cm apart vertical strips (2.5 cm wide) were more effective barriers than horizontal strips. Other alternatives include orienting the windows downward so that they reflect the ground (Klem,1990b).

Despite these research efforts, window mortality remains an under-publicized risk for birds. Working with independently reported observations from Feeder Watch programs in Canada and the U.S., Erica Dunn supported Klem's estimates of the death toll exacted by window collisions. Bird attractants such as feeders and vegetation may pose smaller risks to birds when they are as far away from windows as possible. Yet, her results do not indicate a minimum safe distance. She also calls for additional attention to the placement of windows. Unlike Klem, she found that South facing windows encountered more bird strikes than other windows, possibly because of bird flight patterns and/or sunlight. Despite the possibility that this trend might relate to the greater number of windows installed facing south at most houses, she recommends along with Bevanger and others that careful observation of local bird behavior and flight patterns provides crucial clues for prevention endeavors (Dunn, 1993).

These studies highlight many of the environmental and habitat factors that contribute to increased mortality. Analysis of bird flight and comparisons with human flight has provided models for understanding bird vulnerability to windows and lighted structures. Focusing on nocturnal birds A.D. Herbert develops a model from several assumptions. Aerodynamic forces behave similarly on plane and bird wings. The sense organs are used for similar ends in both birds and humans. Moreover, the responses to stimuli are also similar between birds and people. With this framework he demonstrates that birds become spatially disoriented when faced with unnatural environments such as night time illumination. Artificial light sources create a false horizon against which birds modulate their flight behavior. The birds fall into spiral dives as they attempt to correct discrepancies between their visual perception and the other senses (vestibular system, acceleration of visceral organs, displacement of feathers) by adjusting their movement according to the false horizons. The birds fail to recognize the false cues, and can not react accordingly during the very rapid process of disorientation. They experience a conflict between sensory stimuli, and have no method to compensate for the unnatural environment (Herbert, 1970).

Vehereigen (1985) follows with an extensive description of unnatural light and its reshaping of the bird's environment. Artificial light can have a negative effect on wildlife, thereby constituting photopollution. It impairs behavior and may lead to inopportune decisions and death by interfering with the sensory organs. Artificial light such as that emanating from indoor lighting through transparent windows can deceive birds and limit their flying ability and responsiveness (Veheijen, 1985).

In her review of the studies and programs in Toronto, Evans called for increased research in prevention of avian collision mortality. Volunteers and researchers surveyed many of the buildings in downtown Toronto, counting dead birds, and assisting injured ones whenever possible. The public outcry and increased awareness in Toronto allowed for small but helpful changes in daily practices. Many buildings in the downtown metropolis abstain from nighttime illumination, resulting in a marked decrease in reported window mortality. Other steps can increase the overall safety of birds in flight. Windows can be covered with netting or screens, or replaced with non-reflective (stained/frosted) glass. Recent advancements in glass industry suggest the use of window films that specific obstructions to birds, possibly reflecting in the UV spectrum. Drawing from studies by Dunn and Klem, Evans recommends placement of bird attractants within 1 m or at a minimum distance of 10+ m from windows. As seen in Toronto, light pollution and window mortality are interconnected. Several recommendations include directing light downwards with shielding, strobe or rotating lighting near towers and tall structures, development of sound deterrent systems, and perches to prevent fatigue related injuries or collisions of disoriented birds (Evans, 1996).

In this study, correlations between bird deaths and window design are investigated. We hypothesize that factors such as window size, type, direction, lighting, weather, and bird attractant locations will have an impact on both the number of deaths and the composition of the species killed. We propose to test this by monitoring bird deaths in the local area of Swarthmore College.

Methods

Bird death counts and window strikes will be established by doing morning walks around Swarthmore campus. Specifically, buildings on the main campus with the most prominent windows and/or the highest reported bird mortality rates will be monitored for bird mortality between 3/15/00 and 5/4/00. The Swarthmore College campus covers over 330 wooded acres, including a woodland area. The campus is a nationally recognized arboretum. In the center of campus there is a cluster of academic buildings that border a loosely defined quadrangle. These buildings comprise a wide array of window types, and are associated with different levels of vegetation.

Fig. 1. Survey region from campus map with compass direction and building labels: 3-Beardsely (not surveyed), 7-Cornell Science Library, 11-Dupont, 15- Kohlberg, 17-Lang Music Building, 18-LPAC, 19-Martin.

Our survey included Cornell, Kohlberg, Dupont, LPAC, and Lang as examples of large windows which might be invisible to birds. With similar-sized windows but much smaller individual panes, Martin served as our negative control.

Throughout our census, campus-wide support and involvement were solicited. Other students conducting regular counts of the birds at various feeders. The Ground and Horticultural Crews were willing to report and collect any dead birds that they encountered during their two early morning walks around campus. Our survey route divided the various sites into separate sections. We observed Kohlberg's Courtyard, East, and North Faces. Both Martin and Cornell had an East side facing the quadrangle and a West Side directed toward the Crum. Moreover, both Cornell and Martin have bird feeders placed within 2 meters of the windows. Dupont was divided into several sections because its design created distinct areas which could not bee monitored simultaneously. The South side was divided into the South West are near Cornell, and the South East section which included the Harry Wood garden. We also monitored Dupont's East, North (facing the water tower), and West sides (facing the Crum woods). Only the West side of both Lang and LPAC were surveyed despite the extensive window surface area on the other sides of the buildings. This decision was based on the accessibility of the remaining windows.

In order to compare window strike frequency with window type, the window panes of each site were measured or calculated from building design charts acquired from the Facilities department. A window or window arrangement was considered to be a series of glass panes separated only by the border of the pane. Individual widows were separated by the wall face itself.

This is especially relevant for some windows in Cornell, Dupont, LPAC, and Lang, where many large windows panes of different types are mounted nearby, creating a larger window area than calculated below.

Table 1. Kohlberg Courtyard and North individual and total window pane area for different window types

Courtyard	Pane 1 (m2)	Pane 2 (m2)	Pane 4 (m2)	Pane 3 (m2)	Total Area
1st Floor	1.4	5.84	1.90	0.45	9.59 m2
2nd Floor (a)	1.16	0.39	-----	-----	1.55 m2
2nd Floor (b)	9.07	0.83	0.39	0.90	4.19 m2
3rd Floor (a)	0.77	2.07	0.34	0.90	4.08 m2
3rd Floor (b)	1.23	0.37	-----	-----	1.60 m2

Table 2. Kohlberg East individual and total window pane area for different window types

East	Pane 1 (m2)	Pane 2 (m2)	Pane 4 (m2)	Pane 3 (m2)	Total Area
1st Floor	2.07	0.71	0.39	0.90	4.07 m2
2nd and 3rd Floor (a)	1.16	0.39	-----	-----	1.55 m2
2nd and 3rd Floor (b)	2.07	0.71	0.39	0.90	4.07 m2

Table 3. Martin East and West individual and total window pane area for different window types

East & West	Pane 1 (m2)	Pane 2 (m2)	Pane 4 (m2)	Pane 3 (m2)	Total Area
All Floors	0.11	0.19	-----	-----	0.30 m2

Table 4. Cornell East individual and total window pane area for different window types

Cornell East	Pane 1 (m2)	Pane 2 (m2)	Pane 4 (m2)	Pane 3 (m2)	Total Area
1st Floor	0.41	2.13	0.68	-----	3.22 m2
2nd Floor	1.59	0.56	-----	-----	2.15 m2

Table 5. Cornell West individual and total window pane area for different window types

Cornell West	Pane1(m2)	Pane 2(m2)	Pane 4 (m2)	Pane 3 (m2)	Total Area
Lower Level (LL) and 1st Floor near Feeders(a)	0.54	2.7	2.55	0.38	6.17 m2
LL and 1st Floor near Concrete Screen(b)	1.51	0.77	1.87	2.27	6.42 m2
LL and 1st Floor near Concrete Screen (c)	2.46	1.67	3.23	-----	7.63 m2
2nd Floor (a)	0.90	1.61	1.27	-----	3.78 m2
2nd Floor (b)	0.68	-----	-----	-----	0.68 m2
2nd Floor (c)	0.97	-----	-----	-----	0.97 m2

Table 6. Dupont South West and North individual and total window pane area for different window types

Dupont SW, W, N	Pane1(m2)	Pane 2(m2)	Pane 3 (m2)	Pane 4 (m2)	Total Area
1st and 2nd Floor (2nd only for N)	1.16	0.97	-----	-----	2.13 m2
1st floor (W only)	1.57	0.78	-----	-----	2.35 m2
Foot Level SW, W	not found
Foot Level N	1.04				1.04 m2

Table 7. Dupont SE and E individual and total window pane area for different window types

Dupont	Pane1(m2)	Pane 2(m2)	Pane 3 (m2)	Pane 4 (m2)	Total Area
1st floor (SE only) (a)	2.16	0.33	0.60	-----	3.09 m2
1st floor (SE only) (b)	2.40	0.84	-----	-----	3.24 m2
1st floor (E)	0.46	1.96	0.92	0.63	3.97 m2
1st floor (E)	1.57	0.78	-----	-----	2.35 m2
1st floor (E)	2.39	-----	-----	-----	2.39 m2
2nd Floor	1.55	0.52	0.33	-----	2.40 m2

Table 8. Lang and LPAC individual and total window pane area for different window types

Lang & LPAC	Pane1(m2)	Pane 2(m2)	Pane 3 (m2)	Pane 4 (m2)	Total Area
LPAC	0.58	1.08	-----	-----	1.66 m2
Lang Concert Hall Window	4.09	-----	-----	-----	4.09 m2
2nd Floor	1.40	1.60	-----	-----	3.00 m2
Roof Level	3.11	1.40	-----	-----	4.51 m2

Our survey route was traversed 4 times each week. Weather conditions were recorded prior to observations. The entire walk was paced so as to occur within the same time frame each day (approx 2 hours after sunrise). The number and species of birds injured or killed were recorded at each site. A brief 2 minute scan of the site counted the number of other live birds present. A window strike was recorded and rated according to the following criteria: 1) Lethal-strike: bird collided with window and died; 2) Sub-Lethal: bird collided with window and was unhurt or stunned, but did not die; 3) Imprint- bird collided with window and left a smudge imprint and/or feathers stuck on the window surface, but neither the bird nor the collision were actually seen. Injured birds were cared for when found, while dead birds were collected in specimens bags and stored in a freezer for later examination. The extent of window coverings or other obstructions was noted at each site.

In addition to the recorded data, students and faculty members were instructed to report all bird hits to us. We noted the specific window where the collision occurred and the bird was cared for. These data points were useful for verifying our findings and gave us more information about the composition of species hitting windows.

Weather was coded on a scale of 1 to 5, where 1 is least favorable for spring migration and 5 is most favorable (Figure 2).

Weather Code	Location	Wind direction	Wind speed	Temp
1	N or W of cold front, E of H	N component	Mod to strong	cooling
2	Center or E of H	W to variable	Weak	warming
3	W of H	S component	Moderate	warming
4	E of cold front or in a low	S component	Mod to strong
5	S of warm front	S component	Moderate to weak	warm

Fig. 2. Weather codes. Weather code 1 is least favorable to spring migration (wind from North) while weather code 5 is most favorable (wind from South). Barometric pressure rises from 1 to 2 and then falls to 3 and 4 or 5.

Results

We actually encountered only one dead bird and no injured birds over the seven weeks of observations. The dead bird was found on the north side of Lang, where it appears that it hit one of the very high windows that we could easily investigate for imprints/etc and therefore were not investigating in our study. There were, however, several cases of what we labeled #3 window strikes (where a feather or smudge was noted on the window). Additionally, reports from around campus gave us some additional information from the windows in very public places. Both our walks and reports from around campus pinpointed the Kohlberg courtyard and the west side of Cornell library as danger zones for bird hits (Table 9). Each of these locations have large windows, and many students inside (increased chances of reporting hits).

Table 9: distribution of hits across campus.

Location	# of external reports	#3 window strikes
Kohlberg S courtyard	2	3
Kohlberg E	*	2
Kohlberg N	0	0
Martin E	*	0
Cornell E	1	0
Dupont S courtyard	*	0
Dupont E courtyard	*	2
Dupont E	*	4
Dupont N	*	1
Dupont W	*	0
Cornell W	3	4
Martin W	*	0
Lang W	*	0
LPAC W	*	0

*Not enough people present at most times of the day to provide comparable data

When we compare the number of #3 hits to window size, window pane size, or total window area in the location, it is apparent that total window area is not an important determinant of bird collisions (negative correlation). Window size (r2=0.32, p>0.25) and window pane size (r2=0.15, p>0.6) each had non-significant correlations with #3 window strikes. Note that no hits occurred on the smallest windows (Martin), while some of the larger windows in the area monitored also consistently had no hits (Kohlberg North, Dupont Courtyard). Apparently factors other than window size play an important role in determining the frequency of bird collisions.

Weather was coded for each day of observations and for each day in which an injured bird was reported on campus. The results actually conflict between these two data sets. Number 3 strikes were four times more likely to be spotted on days of weather code 1 or 2 (vs 3, 4, and 5). On the other hand, of the birds that were actually observed hitting windows, seven collided on days of weather code 3, 4, or 5 and only 2 collided on days of weather code 1 or 2.

Discussion

Dead birds would probably have been found by us if they had been killed that morning. However, if a bird had died in the afternoon, it would probably either have been removed, sent to us, or eaten by an animal overnight. Between reported birds and those discovered, we would estimate that we have found about 30-50% of those that had died. Between the total number dead found (1) and the total number of dead reported (3), we would estimate that a total of 8-12 birds have died by hitting academic buildings on campus this spring.

Those on campus that reporting stunned birds have said that the animals usually recover in less than 30 minutes. Given that birds hit as frequently in the morning as other times of day, our chances of discovering a given injured bird would probably be less than 5% on a given day (less than 3% when you consider we only observed 4 days per week). It is therefore not too surprising that we did not find any injured birds on our birdwalks. From observing the reactions of people in the public areas (plus the fact that people aren't always around), it is likely that the injured birds reported to us constitute only a small fraction (perhaps 10-20%) of the total bird hits in the most busy public areas on our route (such as Cornell library and Kohlberg). In other areas of our study, the fraction of those reported over those that hit would be very close to zero, as they would go unnoticed.

The smudge/feather data is difficult to use effectively for several reasons. 1) there is some chance that smudges on the windows were caused by humans. Those found with feathers are likely to be from birds however, and the smudges on lower windows often closely match those on windows far out of people's reach. A close examination of the windows in Cornell confirmed that all of the smudges present were on the outside of the windows. 2) noticeability issues. It is very likely that some of the smudges could have originated on a day prior to the day of investigation. Though this wouldn't influence such things as frequency at a specific location or number of hits to window size/window pane calculations, it would compromise their effectiveness in statistics that are date-sensitive such as weather data. Additionally, the weather/sun conditions may play a significant role in how easy smudges are to see, thus further causing a bias. 3) Window cleaning. We could not practically control for factors such as which windows were cleaned when. Windows in areas such as Kohlberg were cleaned often, while windows in Cornell don't appear to have been cleaned all spring. This could cause significant underestimation of the number of hits in areas that are frequently cleaned. 4) Double counting. In windows not cleaned, it was difficult to keep track of which smudges were new and which had already been counted on a previous day.

Since all smudge data was recorded in a similar fashion, it is likely that it provides at least some sense of the relative frequency of bird hits (perhaps underestimating in regions such as Kohlberg). At the end of observations, a careful examination of Cornell's West windows (next to a bird feeder) was done on 5/18, approximately 2 weeks after the study. Since these windows have been rarely washed, many smudges remain. Approximately 65 smudges were counted from the inside of Cornell. Of these, 18 were on the panels above the lower level, so they could not have been easily reached by humans (and would not have been counted by us in our rounds). All appeared to be on the outside of the windows, further contributing to this likelihood. Most were the size of larger birds like morning doves. Since hummingbird imprints would not have been noticed, this figure may even be an underestimation of the total number of birds hit. Since this location is near a bird feeder, it is likely that most of the hits occurred during takeoff, resulting only in minor injuries (explaining why no dead birds were found at this site). If these data are any indication, however, our methodology for recording smudges may grossly underestimate the total number of hits (we only recorded 4 from this site, and only 3 were reported). This could have been largely due to difficulty in seeing such smudges clearly due to glare and other factors from the outside.

If this ratio [(observed feathers and smudges + reported bird hits)/ total hits] is about 1/10 for the Cornell windows, it is conceivable that it is even lower for less public windows (as there are lower reported hits in these regions). Many of these windows do not have upper levels out of our reach, however, so our ratio of noticed smudges may be higher (1/6, rather than 1/10) (Table 10).

Table 10: Projected total number of bird injuries.

Location	#reported	Conservative estimate: (x5-x10 #reported)	Smudges+reported	Estimated total injured on lower level: (smudge+ reported) x6
Kohlberg S courtyard	2	10-20	5	30
Kohlberg E	*		2	12+
Kohlberg N	0	0	0	0
Martin E	*		0	0+
Cornell E	1	5-10	1	6
Dupont S courtyard	*		0	0+
Dupont E courtyard	*		2	12+
Dupont E	*		4	24+
Dupont N	*		1	6+
Dupont W	*		0	0+
Cornell W	3	15-30	7	42
Martin W	*		0	0+
Lang W	*		0	0+
LPAC W	*		0	0+

*Not enough people present at most times of the day to provide comparable data

These data indicate that the total number of bird hits on this campus may be substantial. From the estimations above, one could conservatively state a range of 30-130 hits in the buildings observed. Of these, our data would suggest that the vast majority of collisions are not lethal.

The methods we are using to record bird hits, however, are less than optimal. Reported hits will be bias toward windows in populated areas, while smudge data will be bias toward windows that are easily inspected. Thus areas such as Lang and LPAC may experience many hits over the course of the day, but we have no possible way to estimate this given our current methodology.

The lack of significant correlation between window size and number of hits may be entirely due to issues in observing some of the larger windows clearly, such as those on the North side of Dupont or the West side of Lang and LPAC. Since smudge data is unreliable for date-dependent measures, this data must also be excluded from formal conclusions. The remaining 9 data points, however, appear to indicate that weather codes 3-5 cause more bird injuries. Since the n is so small, it would be useful to have more data points of this nature in the future (actually having some method to record when more of the birds hit).

Since the data recorded in this experiment are not very strong, it is recommended that a new approach is used for future study of this topic. It has been suggested that vibration sensors could be used to detect bird impacts on windows, thereby giving time and date for each major impact. There could be a problem identifying the birds, while using this methodology, and there may be some false positives, but this would alleviate many of the problems associated with identifying, remembering, and dating smudges observed.

Literature Cited

Bevanger, J. 1994. Bird interactions with utility structures: collision and electrocution, causes and mitigating measures. IBIS 136: 412-425

Dunn, E.H. 1993. Bird mortality from striking residential windows in winter. J. Field Ornithology 64(3): 302-309

Evans, L. J. 1996. Collision Course: The Hazards of Lighted Structures and Windows to Migrating Birds. World Wild Life Fund and Fatal Light Awareness Program

Herbert, A.D. 1970. Spatial Disorientation in Birds. Wilson Bull. 82(4):400-419

Klem, D. Jr. 1989. Bird-window collisions. Wilson Bull. 101(4):606-620

Klem, D. Jr. 1990b. Collisions between birds and windows: mortality and prevention. J. Field. Ornithology 61(1):120-128

Klem, D. Jr. 1990a. Bird injuries, cause of death, and recuperation from collisions with windows. J. Field. Ornithology 61(1):115-119

Klem, D. Jr. 1998. Presentation at North American Ornithological Conference, St. Louis, Wednesday 8 April 1998.

Verheijen, F.J. 1985. Photopollution: artificial light optic spacial control systems fail to cope with. Incidents, causations, remedies. Bird Study 44:1-18

Return to main E.S. page

Send message to Swarthmore College Environmental Studies

last updated 6/7/00