Proc. Natl. Acad. Sci. USA Vol. 90, pp. 10871-10875, November 1993 Population Biology Times to extinction for small populations of large birds (crow/owl/hawk/population lifetime/population size) STUART L. PIMM*, JARED DIAMONDt, TIMOTHY M. REEDt, GARETH J. RUSSELL*, AND JARED VERNER§ *Department of Zoology, University of Tennessee, Knoxville, TN 37996; tDepartment of Physiology, University of California Medical School, Los Angeles, CA 90024-1751; tJoint Nature Conservation Committee, Monkstone House, Peterborough, PE1 1UA, United Kingdom; and §U.S. Department of Agriculture, Forest Service, Pacific Southwest Research Station, Fresno, CA 93710 Contributed by Jared Diamond, August 9, 1993 ABSTRACT A major practical problem in conservation such counts. European islands provide the most extensive biology is to predict the survival times-"lifetimes"-for small existing data set, and hence the best surrogates for 'Alala and populations under alternative proposed management regimes. Spotted Owl populations for the foreseeable future. The data Examples in the United States include the 'Alala (Hawaiian consist of counts of nesting birds tabulated in the annual Crow; Corvus hawaiiensis) and Northern Spotted Owl (Strix reports of bird observatories on small islands around the occidentalis caurina). To guide such decisions, we analyze coasts of Britain and Ireland, and on the German North Sea counts of ail crow, owl, and hawk species in the most complete island of Helgoland. Subsets of these data have been ana- available data set: counts of bird breeding pairs on 14 Euro- lyzed previously (2-9). pean islands censused for 29-66 consecutive years. The data set Our paper expands the previously analyzed data set to yielded 129 records for analysis. We derme the population include more islands and more census years. We summarize ceiling as the highest number of breeding pairs observed from these records and extract estimates oftimes to extinction. We colonization to extinction, within a consecutive series of counts then consider what these data tell us about the utility of for a given species on a given island. The resulting distributions models in the extensive literature on modeling extinction of population lifetimes as a function of population size prove to (refs. 10, 11 and other chapters in ref. 12, and 13; refs. 8 and be highly skewed: most small populations disappear quickly, 14 provide reviews). but a few last for a long time. Median (i.e., 50th percentile) lifetimes are calculated as only 1-5 yr for hawk, owl, and crow THE ISLANDS AND THEIR BIRDS populations with ceilings of one or two breeding pairs. As expected if demographic accidents are the main cause of We chose 14 islands from a much larger set because the birds extinction for small populations, lifetimes rise by a factor of of those 14 islands have been counted almost every year for 3-4 for each additional pair up to three pairs. They rise more at least 29 yr. (We shall expand on "almost every year" lifetimes of below.) The islands and the lengths of the counts are as slowly thereafter. These observations suggest that follows: Bardsey (39 yr), Calf of Man (32 yr), Copeland (35 the 'Alala (now reduced to about three pairs in the wild), and yr), Fair Isle (33 yr), the Farne Islands (31 yr), Handa (29 yr), of populations of Northern Spotted Owl in the smallest forest Helgoland (56 yr), Hilbre (35 yr), Isle of May (66 yr), Lundy fragments, will be short unless active management is imple- (32 yr), Skokholm (64 yr), Scolt Head (55 yr), Skomer (29 yr), mented. and Steep Holm (47 yr). Refs. 7 and 8 give the islands' locations, sizes, distance to the nearest mainland, and other How long are small populations of animals or plants likely to statistics. last before becoming extinct? This question is of both theo- We define a "record" to be the sequence of numbers of retical and practical interest, because conservation biologists breeding pairs of a given "population" (i.e., one species on are often asked to predict the fates of small populations under one island), from the population's colonization until its (local) alternative management options. For example, two species in extinction. There can be more than one record for each the United States have generated litigation as well as scien- species on each island ifa species becomes locally extinct and tific debate. One, the Northern Spotted Owl (Strix occiden- then recolonizes. The species analyzed consist of four hawk talis caurina), is a subspecies largely restricted to old-growth species [Buzzard Buteo buteo (11 records), Kestrel Falco forests in the Pacific Northwest. The controversy centers on tinnunculus (14 records), Peregrine Falco peregrinus (16 the number of forest fragments, and the number of owl pairs records), and Sparrowhawk Accipiter nisus (1 record)]; two per fragment, required to ensure owl survival after the owls [(Little Owl Athene noctua (3 records) and Short-eared intervening forest has been logged. The second case, that of OwlAsioflammeus (12 records)]; and five species in the crow the 'Alala or Hawaiian Crow (Corvus hawaiiensis), involves family [Carrion and Hooded Crows Corvus corone (27 rec- a distinctive species reduced to three breeding pairs in the ords), Jackdaw Corvus monedula (8 records), Raven Corvus wild (1). A key question in both cases is, How long is the wild corax (20 records), Chough Pyrrhocorax pyrrhocorax (6 population likely to last without intervention? Times to records), and Magpie Pica pica (11 records)]. These analyzed extinction for small populations of Spotted Owls are un- species constitute all species of hawks, owls, and crows on known, while the aim of managing the 'Alala is to avoid these 14 islands, except that we excluded Little Owls on obtaining such a time. Skokholm and Jackdaws on Skomer because attempts were In this paper we analyze empirical data on times to made to eradicate them. extinction for small populations of owls, hawks, and crows Crows, owls, and hawks are sufficiently similar to warrant on European islands. Such data are tedious to obtain because their joint analysis. All are large-bodied, long-lived, at least entire populations must be counted repeatedly for many partly carnivorous, and terrestrial rather than aquatic. None years. Moreover, comparative analyses require a large set of is a long-distance migrant, though they may move to and from the adjacent mainlands. Other large-bodied, long-lived birds The publication costs of this article were defrayed in part by page charge differ by being either herbivorous (doves, partridges, and payment. This article must therefore be hereby marked "advertisement" pheasants), long-distance migrants (some cuckoos and wad- in accordance with 18 U.S.C. §1734 solely to indicate this fact. ers), or aquatic (ducks and some waders). 10871 Downloaded by guest on September 23, 2021 10872 Population Biology: Pimm et al. Proc. Natl. Acad. Sci. USA 90 (1993) ANALYSES ceiling. We discuss later how to deal with those alternative estimates. Basic Data. The yearly reports from each island usually Measures of Time to Extinction. We relate this measure of recorded the number ofbreeding pairs or of groups offledged population size to the number of years that the population young. Infrequently, the presence but not the actual number generates of breeding birds was reported. lasts from colonization to extinction. Each record Inferences of Breeding. In 7 of 129 cases we inferred one value of the population ceiling and one corresponding breeding without proof, because birds were observed to be time to extinction. For instance, the two records of example present throughout the breeding season in a given year and 3 yield times to extinction of 4 and 1 yr, respectively. For several pairs were known to have bred in previous and example 1 and the two records ofexample 2, we can certainly subsequent years, so we assumed that unrecorded breeding say that the times to extinction are at least 12, 5, and 3 yr, had also occurred in the given year. To formalize this respectively. inference, we assumed breeding only when (number ofbreed- However, we can go further and refine the minimum ing pairs in previous year) + (number of breeding pairs in estimates of example 2. Consider the problem of waiting for subsequent year) 2 2 x (number of missing years + 1). Thus, a bus. Arriving at a bus stop at a random time every day, you for a gap of 1 yr, we assumed nesting in that year ifthere were calculate the average length of time until the next bus arrives at least two pairs in both the previous and subsequent years, to pick you up. Ifthe bus actually runs once every 30 min, the or else three pairs in one year and one in the other. For a gap expected value ofthe mean wait is 15 min, because you might of 2 yr, there had to be three pairs in both the previous and arrive just after the bus departs, just as it is about to arrive, subsequent years, or else two and four, or five and one. or at any time in between. Hence conversely, if you record Skokholm was not counted from 1940 through 1945. Only your waiting times over many days and get an average of 15 Carrion Crow was sufficiently abundant on Skokholm before min, you can calculate that the true interval between buses is and after those years for us to infer its presence from 1940 to 30 min. More generally, if we "arrive at" an island bird 1945. population without knowing when the population itself "ar- Table 1 shows three hypothetical results of counts of an rived," we should estimate the time to extinction as twice the island population in 12 consecutive years. In example 1 the observed record length.