CHANCE 7Each fall and spring is heralded by largeflocks of birds migrating between theirsummer and winter territories. Annualmigratory behavior ranges from move-ment of the entire population to migra-tion of only a subpopulation andnonmigration. There are even speciesthat in some years show a greater pro-portion or longer distance of movementthan in other years — termed irruption.An understanding of when andwhere birds migrate provides informa-tion needed to understand features ofbird populations that are of interest toornithologists, environmental scientists,and nature lovers. Among these featuresare habitat use, responses to climaticchange, and the potential for long-dis-tance spread of diseases.Although much information hasbeen gained from studies that includetracking of banded or radio-collaredbirds, the expense involved prohibitswidespread use of these methods. How-ever, the availability of a large numberof knowledgeable amateur bird-watch-ers across North America who are will-ing to assist with data collection of manytypes enables North American ornithol-ogists to conduct studies that involvespatial and temporal information with alarge combined scope and high resolu-tion (at least compared to the data avail-able on terrestial animals). Suchnetworks of birders have provided infor-mation on such interesting features asthe introduction and spread of alienspecies and the growth or decline of pop-ulations in certain locales. Can the datacollected by these birders be used tomeasure migratory behavior?Project FeederWatchProject FeederWatch, initiated on a con-tinent-wide basis in 1987 by the CornellSapsucker Woods Laboratory ofOrnithology, has a large database ofinformation about several bird popula-tions from volunteer participants acrossthe United States and southern Canada.These data, collected each winter fromData collected by volunteers across the U.S. provide avaluable resource for ornithologistsMeasuring Bird Migration UsingSpatial and Temporal CountsElliot Bernstein, Kevin Cottrell, Naomi Altman, AndréDhondt, Wesley Hochachka, and Roger SlothowerNorthern CardinalF.W. FALLON8 VOL. 14, NO. 3, 2001November through March, are countsof the number of birds of each speciesreported by each participant on abiweekly basis. These data, with theirwide spatial coverage and fine temporalresolution, seem ideally suited to meas-ure migratory behavior, although thetiming of the data collection, from latefall to early spring, might miss the migra-tion periods for some species.The Amateur Scientist column of Sci-entific American, April 1997, provideda nice overview of Project FeederWatchand several other “Citizen Science” proj-ects run by the Cornell Lab of Ornithol-ogy. More details are provided in someof the works listed at the end of this arti-cle.Project FeederWatch participants areinstructed to observe their feeders fortwo consecutive days, biweekly, frommid-November to mid-March andreport the maximum number of birds ofeach species observed simultaneouslyat the feeder. The participants alsorecord a rough measure of observationeffort — an indicator of whether or notthey viewed during the morning and/orafternoon of each of the two viewingdays. There are also a number of con-comitant variables, such as the type ofbird food offered and data about thehousehold, such as location, number offeeders, and so on. The number of par-ticipants in Project FeederWatch hasincreased through time from over 2,000initially to over 8,000 across NorthAmerica this last winter. Climate infor-mation was added to the data usinginformation provided in the U.S. His-torical Climatology Network (http://www.ncdc.noaa.gov/ol/climate/research/ushcn/ushcn.html).Species abundance maps createdfrom the Project FeederWatch data canbe viewed on the Web at http://birds.cor-nell.edu/PFWMaproom/. A look at theanimated monthly maps gives a clearstory for some migratory species — forexample, the common grackle. TheDecember 1999 and March 2000 mapsare displayed in Fig. 1. Grackles are verysparse away from the southeastern U.S.coastline on the November, December,and January maps. More grackles appearnorth of the coast in February, and abun-dance is very high in March. Clearly, thedata-collection period has missed thesouthward migration in the fall butrecords the northward migration in earlyspring. By contrast, there is little changein the monthly maps for the northerncardinal, a nonmigratory species. The animated maps suggest thatthere might be some way to character-ize migratory behavior from the ProjectFeederWatch data. If a numerical indexof migration can be developed thatclearly delineates between migratoryand nonmigratory species, this could beuseful in several ways. Changes inmigratory behavior, which might beinduced by competition from an intro-duced species, a new disease, or otherchanges in the environment, could betracked by determining if the index ofmigration changes over time. Irruptionscould be quantified. And more subtlebehavioral changes, such as local devi-ations from otherwise widespread migra-tory behavior, may be detectable.This article traces our attempts todevelop an index of migration from theProject FeederWatch data, using speciesknown to be either migratory or nonmi-gratory. A simplistic analysis, averagingover observers, seems to work well. Sur-prisingly, a more sophisticated analysis,taking into account observer variability,does not discriminate well betweenmigratory and nonmigratory species. Developing an Index ofMigrationSuppose that we could obtain full dataon the weekly species abundance ateach locality for the entire year for amigratory bird species. In the summerrange of the species, we expect morebirds in the summer weeks and few ornone in the winter. In the winter rangeof the species, we expect more birds inthe winter weeks and few or none in thesummer. If both the summer and win-ter range are within the area covered byFigure 1. Project FeederWatch maps for winter 1999–2000 showing the mid-winter and spring distribution of the common grackle,a migratory species.CHANCE 9our data, we ought to be able to plotspecies abundance versus date at thetwo extremes of the range and observethe rise and fall of abundance with sea-son. In a nonmigratory species, wewould not expect any seasonal differ-ences. This is the intuitive basis for thedevelopment of a migration index. It isnot necessary to include the full migra-tory period or range as long as some pop-ulation movement can be