Fifty-year trends in a box turtle population in Maryland

BIOLOGICAL CONSERVATION Biological Conservation 88 (1999) 165±172 Fifty-year trends in a box turtle population in Maryland Russell J. Hall a,*, Paula F.P. Henry b, Christine M. Bunck c a U.S. Geological Survey, Florida Caribbean Science Center, 7920 NW 71 Street, Gainesville, FL 32653, USA b U.S. Geological Survey, Patuxent Wildlife Research Center, Laurel, MD 20708, USA c U.S. Geological Survey, National Wildlife Health Center, Madison, WI 53711, USA Received 20 September 1996; received in revised form 22 August 1998; accepted 26 August 1998 Abstract A survey conducted in 1995 investigated long term declines reported in a population of box turtles Terrapene carolina monitored each decade since 1945 in bottomland hardwood forest at the Patuxent Wildlife Research Center, Maryland. Methods duplicated past surveys in most respects, but were supplemented by radiotelemetry and a survey of dominant vegetation. Seventy di€erent turtles were found on the 11.8 ha study area, a decline of >75% since peak populations were recorded in 1955. Searchers were less ecient in 1995 than in 1945±1975 for a variety of possible reasons. Among turtles recorded, approximately equal numbers persisted from each of the past ®ve decades, with some individuals surviving >70 years. A sex ratio strongly favoring males was ®rst recorded in 1975 and continued in 1995, but juveniles and subadults were found in greater proportion in 1995 than in any other survey. Six of nine radio-marked turtles left the bottomland study area and migrated to the adjoining blu€s to hibernate, suggesting more extensive movements and perhaps less stable home ranges than formerly thought. Age structure of trees indicated a gradual change to more shade-tolerant species. Examination of rates of change from survey data suggested that major losses probably resulted from changes in hydrology that exacerbated ¯ooding in 1972, with recovery only beginning in 1995 and perhaps limited both by repeated ¯ood events and successional changes in the forest. Slow recovery from losses may indicate that populations of this species would respond poorly to exploitation. # 1999 Elsevier Science Ltd. All rights reserved. Keywords: Box turtle; Terrapene; Populations 1. Introduction Soon after establishment of the US Department of the Interior's Patuxent Wildlife Research Center in 1939, biologists began wide-ranging studies of many resident species (Hall, 1988). Beginning in 1942, box turtles were permanently marked by a standardized system (Cagle, 1939) and in 1944 Lucille Stickel concentrated research e€orts in part of the alluvial ¯oodplain of the Patuxent River that was to serve as the control area for a study examining e€ects of applications of the pesticide DDT on populations (Stickel, 1951). Continued emphasis on the population of an 11.8 ha study area led to a comprehensive account of ecology and behavior of the population (Stickel, 1950). Count data from Stickel's multi-year study were particularly complete in 1945, and surveys were repeated in 1955, 1965, and 1975 to examine long term trends in population size. The results (Stickel, 1978) indicated persistence * Corresponding author. Tel.: +352-378-8181x305; fax: +352-3784956; e-mail: russ_hall@uscs.gov of some individuals over the three decades of study, but an overall population decline, despite protection of all biota throughout the period. Stickel was unsure of reasons for the decline, but speculated that changes in hydrology resulting from upstream impoundment of the Patuxent River might be involved. Another less complete survey in 1985 suggested continuing decline (Hallgren-Scadi, 1986). Additional reports (Stickel, 1989; Stickel and Bunck, 1989) dealt with detailed aspects of home range and growth. Other studies of box turtles have extended over decades (e.g. Schwartz et al. 1984; Williams and Parker, 1987), but none approaches the Patuxent studies in length. A survey of this box turtle population was conducted in 1995 to continue surveys conducted in each decade since 1945 in order to track changes in populations and update records of persistence of individuals. Eastern box turtles resemble other terrestrial chelonians, which are extreme in some aspects of life history and demography (Germano and Bury, 1994). Individuals appear sedentary, some occupying the same c.3 ha home ranges for more than four decades. Adults 0006-3207/99/$Ðsee front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S0006 -3 207(98)00107 -4 166 R.J. Hall et al./Biological Conservation 88 (1999) 165±172 are long lived, with some turtles surviving >50 years. Maturity is late and rates of reproduction are low. Rateconstants of change are low and, despite the unusually long interval over which data are available on the Patuxent population, the nature and causes of changes in populations remain poorly understood. All box turtles of the genus Terrapene were recently included in a proposed rule (US Fish and Wildlife Service, 1996) under terms of the Convention on International Trade in Endangered Species because it was believed that increasing exploitation for the pet trade might lead to unsustainable losses from populations. 2. Methods 2.1. Conduct of surveys Our primary objective was to conduct a survey resembling as closely as possible those of past decades (Stickel, 1978) to permit comparison with count data reported in previous surveys. Additional information on age structure in the population, individual histories of turtles marked in past decades, responses of the population to environmental factors, and changes in the alluvial ¯oodplain forest that has been heavily used by turtles was sought to help explain the nature, direction, and causes of change. On the basis of exhaustive studies spanning several years, Stickel (1950) was able to classify turtles as likely resident turtles, transients, or animals occupying ranges outside the study area. Singleyear surveys in subsequent decades (Stickel, 1978) did not permit such classi®cation, however, and we followed her practice of using the total number of turtles observed on the study area as an index of abundance intended to detect trends in the population using the area. Prior surveys relied on frequent systematic searches of the study area throughout the activity season. While attempting to replicate past counts, we also used other techniques to permit evaluation of the error inherent in the primary survey method. These included conducting sweep searches, with 10 or more searchers attempting to cover the entire study area in a single period, and attachment of radio transmitters to a subset of turtles found to estimate the eciency of searches in detecting animals present. Information on age, sex, location, and behavior of turtles found was recorded. Standardized methods for measuring, marking, and recording data on Patuxent box turtles developed in the 1940s were followed with only minor modi®cations. Individuals can be aged reliably from growth rings on scutes for about the ®rst 20 years of life (Ewing, 1939; Stickel and Bunck, 1989), and recaptures can extend known ages by decades. Turtles with 10 or fewer growth lines are here regarded as juveniles, those with 11±20 as subadults, and those with 21 or more lines as adults; data are not available to correlate these categories with the onset of reproductive maturity, but animals with fewer than 21 growth lines rarely had secondary sexual characters fully developed and could not reliably be identi®ed as to sex. Locational data were reported with reference to the 330 foot (100.56 m) US Geological Survey (USGS) grid system established in 1940 and marked with permanent monuments. Investigators resurveyed the study area in 1993 and 1994, relocating permanent grid markers and replacing or refurbishing intermediate sight reference markers at &33.5 m intervals. Data on the occurrence and identity of turtles on the study area were collected, and repeated locations were used to evaluate residency status. The study area (Stickel, 1950) and adjacent areas in the alluvial ¯oodplain of the Patuxent River were the focus of surveys, but as in past studies, data collected on box turtles from other locations on the Patuxent Wildlife Research Center were routinely recorded to supplement and help evaluate data from the study area. 2.2. Information on environmental change A survey of dominant vegetation of the study area begun in the Fall of 1994 and completed in 1995 was undertaken to better understand environmental change. A modi®ed point-quarter methodology (Cottam and Curtis, 1956) was employed. Points were located at the intersections of the 33.5 m grid lines used to subdivide the study area, and at each point the species and diameter at breast height (dbh) of the closest tree >10 cm in dbh were recorded, as were these data and the distance from the central tree of the closest tree in each of the four sectors of the compass. Data were used to calculate relative frequency, relative density, and relative dominance of each species. Available information on land-use and hydrologic change, including analysis of USGS data on daily and peak ¯ows of the Patuxent River at Laurel, was evaluated as it related to turtle population trends. 3. Results 3.1. Populations and persistence In all, 103 turtles were recorded a total of 320 times in searches and radiolocations. Of primary interest in evaluating population trends were the 70 turtles located 190 times on Stickel's study area in searches extending from April to November of 1995. Level of searching e€ort was comparable to that of previous decades, and captures tended to con®rm population trends observed over past decades (Table 1, Fig. 1). Nevertheless, eciency of searches indicated by the mean number of 167 R.J. Hall et al./Biological Conservation 88 (1999) 165±172 Table 1 E€ort expended and turtles recorded, 1945±1995 Year Hours in searches Days in ®eld Turtles recorded Total records Captures/turtle 1945 1955 1965 1975 1985 1995 257 172 304 236 232 260 a 77 57 81 75 34 69 284 291 230 117 58 70 1002 1184 941 443 87 190 b 3.53 4.07 4.09 3.79 1.50 2.71 a This total is minimal; it includes e€ort of primary searchers only and does not include ‹80 additional hours contributed by relatively inexperienced volunteers. b Captures in searches only; does not include radiolocations or captures o€ the primary study area. Fig. 1. Numbers and age structure of turtles recorded in six surveys. Animals reported as juveniles and subadults include all turtles with fewer than 21 growth rings. Turtles with growth rings near 20 are adult-sized, but generally lack the full array of secondary sexual characters. Individuals with fewer than ®ve growth rings are seldom seen on the study area and are represented in the above totals only by two individuals recorded in 1965 and one recorded in 1975. captures per turtle was less in 1995 than in the 1945± 1975 surveys. Recaptures (Table 2) seemed to be proportionally distributed among old residents and newly marked turtles, among young and old turtles, and those in central and marginal regions of the study area, with perhaps a slight tendency for old residents with home ranges near the center of the study area to be recorded more frequently. Adult males averaged 3.3 captures, adult females 2.5, and subadults and juveniles combined averaged 1.6 captures, but the di€erences were not statistically di€erent in a chi-square analysis of frequencies (p=0.064). Among 26 animals captured only once, 10 were adult males, 6 were adult females, and 10 were subadults and juveniles; adult males were most likely (67%) and females least likely (54%) to be captured more than once. Of the 39 previously marked turtles, seven had been ®rst recorded in the 1940s, ®ve in the 1950s, six each in 1960s, 1970s, and 1980s, and eight in the early 1990s. One apparently old turtle could not be traced to earlier records owing to shell damage. Forty-one turtles were judged to be >20 years old, and eight of these had not been marked previously, indicating that they were either missed in earlier surveys or were recent immigrants; three of these turtles were recorded only once during 1995, suggesting that they may have been transients. Age structure and overall population trends (Fig. 1) indicate little change in adult numbers since 1985, but apparent increases in the number of turtles estimated to be <20 years old. These young comprised 22.2% of populations in 1945±1965, and increased to 32.1% for Table 2 Distribution of captures in 1995 Number of captures Number of turtles a Number previously unmarked Died during study Number with <10 growth lines Number on marginal areas 1 2 3 4 5 6 7 8 9 Totals 26 16 11 4 4 4 3 0 2 70 12 11 4 2 0 2 0 ± 0 31 1 2 1 5 5 1 0 0 0 0 ± 0 11 13 8 5 1 1 2 0 ± 1 31 a 4 33 additional turtles recorded on Patuxent but o€ the study area, 7 of which were bottomland residents and possible transient visitors to the study area. 168 R.J. Hall et al./Biological Conservation 88 (1999) 165±172 Table 3 Ratios of juveniles and subadults to adult females recorded in surveys Year Adult females Juveniles and subadults Number of young per adult female a 1945 1955 1965 1975 1985 1995 122 119 85 37 12 13 55 78 58 39 17 29 0.45 0.66 0.68 1.05 1.42 2.23 a These ratios are an index based on observations only. They permit comparisons among di€erent surveys, but probably do not accurately represent relative numbers of adult females and immature turtles in populations; juveniles are rarely observed and almost certainly have greater relative abundance than indicated here. the 1975±1995 period, reaching a maximum contribution of 41.4% in the 1995 sample. Number of young recorded per adult female shows a consistent pattern through time (Table 3). Turtles nearing maturity may be confused with older individuals which ceased growing after 17 or 18 distinct growth lines had been laid down. To conservatively estimate the relative abundance of young individuals in the population, comparisons were made using only juveniles with 5±10 growth lines. They comprised 4.2% of the sample in 1945, 6.2% in 1955, 4.8% in 1965, and 15.7% in 1995. Known-age turtles, those with fewer than 21 growth lines at ®rst capture, included only 44 of the 70 turtles recorded. Minimum ages (Fig. 2) include all available data, but almost certainly underestimate average ages in the population because those with 21 or more growth Fig. 2. Minimum ages of turtles, based on known age turtles possessing fewer than 21 growth rings when ®rst recorded, and known minimum ages of turtles with 21 or more growth rings when ®rst recorded. If ages of all turtles were known, numbers in columns would shift to the right for all categories, beginning with the >20 group. lines at ®rst capture may have been decades older than the 21±30 year class. Also, these ®gures may misrepresent recruitment because no turtles <5 years old were found, continuing a pattern observed in past decades. Among adults recorded, 28 were males and 13 were females, but the probability of captures suggested by recapture records suggests that the true relative abundance of females may be greater than indicated by these numbers. Individuals observed to have died during the 1995 survey included two old turtles and two relatively young turtles. A female estimated to be >71 years old died of injury, possibly caused by a predator, and another female >65 years old was suspected to have died of heat exhaustion. The young turtles, estimated to be 9 and 18 years old, died apparently of disease. Reanalysis of published (Stickel, 1978) and unpublished data on earlier surveys on numbers of adult turtles (Table 4) permitted tracking the persistence of individuals and some overall comparisons of changes in the population (Table 5). There were inadequate data available from 1985, but 1995 data indicated that 77 adults were lost and 17 adults were added between 1975 and 1995. Average persistence rates (Table 5) of all marked turtles show a slight decline for females in the 1955±1965 interval, and signi®cant losses for both males and females in the 1965±1975 interval. The rates shown for the 1975±1995 interval are close to those of the preceeding decade, but they re¯ect losses over an interval twice as long. 3.2. Eciency of surveys Radio transmitters were attached to three turtles on 8 June 1995, but appropriate receivers were not available until 12 August; additional transmitters were attached on 14 August, 10 September and 13 September. Nine turtles (5 males, 1 female, and 3 subadults tentatively identi®ed as males) were initially located by radio signals and then visually sighted 96 times. A primary ®nding was that turtles are far less visible than formerly assumed; even when the location of an individual could be narrowed to a few square meters, it often took observers great e€ort to sight a turtle that was only thinly concealed by vegetation or debris. On days when turtles were known from radiolocations to be in the open and potentially able to be found, a single searcher had on the average only slightly more than a 10% chance of ®nding them in a normal search session without the aid of radio signals. A second ®nding was that movements appear to be more extensive and more complex than indicated by past studies of home range and movements (Table 6). The signal from one radioed turtle was lost after 6 weeks while the animal was moving away from the study area. Another radio failed after only 4 weeks, but was recovered 2 weeks later when the 169 R.J. Hall et al./Biological Conservation 88 (1999) 165±172 Table 4 Turtles recorded and persistence of adult turtles; 1945±1995 Year 1945 1955 1965 1975 1995 Cohort 1945 1945 1955  1945 1955 1965  1945 1955 1965 1975  1945 1955 1965 1975 1995  Males Females Both sexes 107 122 229 66 62 128 54 57 111 120 119 239 52 32 84 25 17 42 39 36 75 116 85 201 16 13 29 12 3 15 6 6 12 27 15 42 61 37 98 2 3 5 5 2 7 6 0 6 5 1 6 10 7 17 28 13 41 Table 5 Average persistence of all marked adult turtles, 1945±1995, based on the data of Table 4 Interval 1945±1955 1955±1965 1965±1975 1975±1995 a Males Females Both Sexes 0.62 0.51 0.56 0.64 0.41 0.53 0.29 0.26 0.28 0.30 0.16 0.24 a This is a 20-year period rather than the 10-year periods represented in earlier intervals; calculated average annual persistence rates of adult turtles are as follow: 1945±1955=0.96; 1955±1965=0.95; 1965±1975=0.93; 1975±1995=0.96. Table 6 Movements of turtles followed by radiotelemetry Serial number Identity Number of locations Inclusive dates Diameter of movements within home range a (m) Travel outside home range after 1 October linear distance (m) 1117 1207 1318 1322 1429 1438 2406 2445 2446 Subadult Adult Adult Adult Adult Adult Subadult Subadult Adult 18 19 15 22 18 17 6 15 10 6/8±10/9 5/27±10/9 6/30±11/5 6/20±11/5 5/29±10/9 6/8±10/9 8/29±10/6 8/12±10/9 8/12±10/3 85 256 111 174 134 57 151 469 225 390 323 0 0 335 356 0 408 258 a ``Home range'', as used here refers to observed home ranges within the bottomland and does not account for annual migrations to nesting or hibernation sites. turtle was found in a visual search. Radios were removed from six of the remaining turtles on 9 October. Transmitters were left on two turtles and they were located until 5 November, when they were in apparent hibernation sites. Other information may bear on the eciency of surveys. The number of turtles newly encountered in each survey year should decline progressively through the season as more turtles are found, eventually reaching zero when all turtles have been encountered. Graphing the occurrence of newly found turtles in each year against the cumulative number recorded should provide an estimation of those remaining unrecorded at the end of the season, and an indirect measure of sampling eciency. Estimates based on this method indicate that 91% of the population was sampled in 1955, 90% in 1965, 86% in 1975, and 92% in 1995. New turtles showed up erratically in the 1945 survey, and the resulting graph does not support estimation of the completeness of sampling. This approach probably does not lead to a good estimate of total populations owning to the uncertain number of transient individuals, but correspondence among years may indicate comparable e€ectiveness in sampling. 3.3. Changes in the study area As in 1945 (Hotchkiss and Stewart, 1947), the forest of the study area is dominated by two tree species, tuliptree Liriodendron tulipifera and American beech Fagus grandifolia. Fifteen other species were recorded, including in order of decreasing overall importance Liquidambar styraci¯ua, Ulmus americana, and Fraxinus americana. But Liriodendron (45.4% of total basal area in 1994) appears to have reproduced infrequently in recent decades, whereas Fagus (19.9% of total basal area in 1994) is reproducing strongly; in time therefore, the forest may become a nearly pure stand of beech. Only 24 of 88 (27%) of Liriodendron were <50 cm dbh, whereas 71 of 83 (86%) of Fagus were smaller trees. 170 R.J. Hall et al./Biological Conservation 88 (1999) 165±172 Moreover, the smallest size class, 10±19 cm dbh, includes 29 Fagus and only one Liriodendron. Frequent damage to trees by beavers Castor canadensis in the vicinity of creeks and sloughs suggests, however, that succession may be arrested in these parts of the study area and that limited open areas may be maintained. Data on peak and daily ¯ows of the Patuxent River at Laurel, c.10 km upstream from the study area, were available from beginning of operation of USGS gauging station 01592500 in late 1944 through 1994. Average daily ¯ows (recorded as cubic feet per second and here converted to cubic meters per second ± m3/s) declined from c. 5 m3/s in the 1945 to 1954 interval to c. 3 m3/s thereafter, presumably as the result of two upstream impoundments. Peak ¯ows, which never exceeded 146 m3/s before 1954, reached 330 m3/s on 21 July 1956, 728 m3/s on 22 June 1972 (Hurricane Agnes) and 476 m3/s on 26 September 1975. Presumably these greater peak ¯ows in later years occurred with massive releases of stored water from impoundments. Average daily ¯ow on 18 July 1945 at the peak of the ¯ooding that inundated the study area (Stickel, 1950) was 42.8 m3/s, whereas the comparable peak ®gure for 22 June 1972 was 364 m3/s. The onset and dissipation of more recent ¯ood events (Fig. 3) seems more abrupt than the 1945 ¯ood. The seasonal pattern of peak ¯ows seems to have changed also; in 1945 through 1954 only three out of 10 years saw peak ¯ows occurring at times (April through October) when turtles would be expected to be on the study area, whereas peak ¯ows in 25 of the past 40 years have occurred during the season when most turtles would be in the bottomland. 4. Discussion 4.1. Changes in the population Fewer captures per turtle in 1995 than in 1945±1975 may have re¯ected the relative inexperience of primary searchers in ®nding turtles, reduced activity during a record mid-summer drought, lack of ¯exibility to schedule surveys during times of presumed greatest turtle activity as done in 1975 (Stickel, 1978), reduced visibility due to closure of the forest canopy, or some fundamental change in turtle use of the study area. Young turtles are generally regarded as less visible and more dicult to sample than adults, and the fact that 1995 searches revealed more young turtles than earlier surveys suggests that the cause may lie in changes in the turtle population rather than di€erences in the ability of searchers or in the immediate environmental conditions during searches. Despite possible di€erences resulting from lower eciency in ®nding turtles or from interpretation of ®eld data, real declines in numbers of box turtles recorded in surveys spanning 50 years are evident. Discounting possible errors in counts, the population in 1995 numbered no more than 23% of the peak population of 1955. 4.2. E€ects of altered hydrology and ¯ooding on populations Fig. 3. Record of four ¯ood events before (solid line) and since (broken lines) operation of upstream impoundments on the Patuxent River. The 14±21 July 1945 ¯ood was reported by Stickel (1948, 1950) to have completely inundated the study area, but to have had only minor and reversible e€ects on the turtle population. De®nitive reasons for changes have been elusive, but evidence increasingly points to changes in hydrology. Environmental changes in the surrounding region have been dramatic, but the Patuxent Center lands in general and the study area have been protected throughout the period. Three separate lines of evidence suggest that population declines have resulted from hydrologic changes and ¯ooding, especially from the historic ¯ood occurring in 1972. Available evidence includes: (1) ¯ooding since 1956 and especially in 1972, which has resulted in daily ¯ows as much as eight times greater than those observed to have totally inundated the study area and mildly a€ected turtles in 1945; (2) severity of ¯ooding and losses of adult turtles were far greater between 1965 and 1975 than in any other period; (3) increases in relative numbers of young turtles in the population since 1975 suggest incipient recovery, or at least no pattern of decline in productivity. Stickel (1978) speculated that increased severity and altered periodicity of ¯oods owing to upstream water R.J. Hall et al./Biological Conservation 88 (1999) 165±172 171 control might be responsible. She lacked direct evidence that the massive ¯ooding caused in 1972 by Hurricane Agnes had lasting e€ects on the population, and the ¯ood in 1945 that she documented (Stickel, 1948Stickel, 1950) had little e€ect on resident turtles. Even after the record 1972 ¯ood some long term resident turtles continued to occupy traditional home ranges, but this ¯ood was nearly an order of magnitude greater than the 1945 ¯ood and almost certainly would have had severe e€ects on all turtles present in the ¯oodplain. Age structure within the population may have di€erent interpretations, but seems to support the idea that losses of turtles, particularly adults, are responsible for population declines. Our observation of nearly equal numbers of adults persisting from each of the ®ve decades of study may result either from declining recruitment or changes in survival rates during the period. The greater proportion of young observed in 1995 than in other decades may have di€erent interpretations, but suggests strongly that recruitment has not declined dramatically. Adults on the study area may not be the parents of the juveniles and subadults observed with them, but the simplest explanation for changes over the decades may be changes in the population structure of the turtles that use the area. The numbers of young turtles observed in 1995 were 53% of those recorded in 1945 even though overall populations in 1945 exceeded by fourfold those of 1995. One interpretation might be that in 1995 the population was growing through increased recruitment, but it might also indicate changed selective pressures resulting in losses of adults that would change population structure regardless of the trajectory of population size. Assuming that adults and young represent the same population, total numbers of young turtles recorded and ratios of subadults and juveniles observed relative to observations of adult females (Table 3) suggest that losses of adults are the problem and that declining populations are not the result of decreased recruitment of young. Stickel (1989), whose evidence suggested that most turtles hibernate within their summer ranges. Combined with evidence that gravid females migrate to the uplands for nesting (Stickel, 1989), these new ®ndings suggest less stability of home ranges than reported earlier. If some males leave their home ranges once each year, and females undertake such movements as many as two times each year, estimates of population density may be made more dicult. Moreover, turtles with fewer than ®ve growth rings have been observed only three times in over 3800 captures in all decadal surveys of this area spanning 50 years, suggesting that very young turtles do not use the bottomland study area. 4.3. Di€erential habitat use among sexes and ages 4.5. E€ects of losses on populations A declining proportion of females over ®ve decades of study is dicult to interpret, although potential consequences for recruitment seem obvious. Unbalanced sex ratios in favor of males have been observed in other box turtle populations (e.g. Dodd et al., 1994), but the apparent reversal in sex ratio over the decades of surveys at Patuxent seems unique. Stickel (pers. comm.) suggested that migration to the uplands for nesting may make females more vulnerable to the regular mowing of ®elds and to progressively increased vehicular trac in upland areas surrounding the study area as activities at the Patuxent Center have increased. Movement of six of nine radio-marked turtles outside their home ranges in October di€ers from the ®ndings of Catastrophic losses resulting from death or permanent displacement of turtles by episodic ¯ooding would have a€ected those age classes present on the ¯oodplain, which are primarily adults and subadults. Evidence from past surveys (Table 4) indicates reduction of the breeding population by one-half between 1965 and 1975. Had the trends observed in 1965±1975 continued, the 1995 population would be only about one-half of the number recorded. Turtles born in the years immediately following the 1972 ¯ood had only barely reached adulthood in 1995, and growth of the population may accelerate gradually as increased numbers of mature adults join the breeding population. Indications of some decline by 1965, of slow change in dominant vegetation, 4.4. E€ects of vegetation change Stickel (1978) stated that the forest canopy appeared not to have closed by 1975, but quantitative data presented here indicate an apparent slow replacement of the dominant tree species by a much more shade-tolerant species. Schwartz et al. (1984) presented evidence that populations of the three-toed box turtle, a geographic race of Terrapene carolina, declined in an area undergoing ecological succession to deciduous forest. Williams and Parker (1987) found that T. carolina in Indiana tended to remain in mature forest, but the population under study declined >50% over a 25 year period. Research on habitat selection on Long Island (Madden, 1975) indicated that box turtles shunned both dense forest and large openings, and concentrated most activity in forest margins. Similar research on the threetoed subspecies in Arkansas (Reagan, 1974) indicated seasonal change in habitats selected, with grasslands selected over forest in cool, moist periods. Changes in the upland habitats that are demonstrated here to be of importance to the Patuxent population may have been more dramatic than those on the ¯oodplain, and could exert an important e€ect on those parts of the life cycle not known from our studies. 172 R.J. Hall et al./Biological Conservation 88 (1999) 165±172 and of reduction of box turtle populations with ecological succession elsewhere suggest that, unless reduced by major ¯ood events, populations may grow and stabilize at lower levels than those attained in 1955. Should tentative conclusions about catastrophic loss of individuals and slow recovery be con®rmed in the future by gradual recovery and/or continuing losses from ¯ood events, implications for the conservation of box turtle populations could be signi®cant. Adler (1969) argued that prehistoric consumption of box turtles extirpated populations from large parts of the former range. The dynamics of long lived vertebrates are highly sensistive to changes in adult survival. Since their rates of increase are low, turtles can be expected to recover very slowly from catastrophic mortality. Acknowledgements The Division of Research and the Oce of Inventory and Monitoring of the National Biological Service made the study possible, primarily by granting the principal investigators time o€ from other duties to work on the study. The Oce of Scienti®c Authority of the US Fish and Wildlife Service provided ®nancial support covering the expenses of a full-time volunteer. Patuxent Wildlife Research Center provided access to research sites and buildings during the study. Lucille F. Stickel provided us with updated summaries of data used in preparation of her 1978 paper. Numerous volunteers helped in di€erent phases of the study, most notably Nina Haramis, who participated throughout the study and Rebecca Raynor of the University of Chicago, Will Staube of the University of Maryland and Peg Hall of Gallaudet University, who contributed frequently. Staff at the Upper Mississippi Science Center and the California Science Center of the National Biological Service generously loaned radiotelemetry equipment. Sta€ at the Patuxent Wildlife Research Center helped by reviewing study plans, loaning equipment, providing technical advice, and bringing in turtles found in their ®eld studies. O.J. Reichman and Lucille Stickel critically reviewed the manuscript and o€ered helpful suggestions. References Adler, K., 1969. The in¯uence of prehistoric man on the distribution of the box turtle. Ann. Carnegie Mus. 41, 263±280. Cagle, F.R., 1939. A system of marking turtles for future recognition. Copeia 1939, 170±173. Cottam, G., Curtis, J.T., 1956. The use of distance measures in phytosociological sampling. Ecology 37, 451±460. Dodd Jr., C.K., Franz, R., Smith, L.L., 1994. Activity patterns and habitat use of box turtles (Terrapene carolina bauri) on a Florida island, with recommendations for management. Chelonian Conservation and Management 1, 97±106. Ewing, H.E., 1939. Growth in the eastern box-turtle, with special reference to the dermal shields of the carapace. Copeia 1939, 87±92. Germano, D.J., Bury, R.B., 1994. Research on North American tortoises: a critique and recommendations for the future. In: Bury, B.M., Germano, D.J. (Eds.), Biology of North American Tortoises. US National Biol. Surv. Fish and Wildl. Resear. 13, pp. 187±204. Hall, R.J., 1988. An annotated bibliography of ®eld studies conducted on the Patuxent Wildlife Research Center: 1940±1987. US Fish Wildl. Ser. Biol. Rep. 88, 1±16. Hallgren-Scadi, L., 1986. Habitat, home range and population study of the eastern box turtle (Terrapene carolina). Masters Thesis, University of Maryland-College Park. Hotchkiss, N., Stewart, R.E., 1947. Vegetation of the Patuxent Research Refuge. Maryland. Amer. Midl. Natur. 38, 1±75. Madden, R.C., 1975. Home range, movements, and orientation in the eastern box turtle, Terrapene carolina carolina. Masters thesis, The City University of New York. Reagan, D.P., 1974. Habitat selection in the three-toed box turtle Terrapene carolina triunguis. Copeia 1974, 512±527. Schwartz, E.R., Schwartz. C.W., Kiester, A.R., 1984. The three-toed box turtle in central Missouri, Part II: a nineteen-year study of home range, movements and populations. Missouri Department of Conservation: Terrestrial Series 12. Stickel, L.F., 1948. Observations on the e€ect of ¯ood on animals. Ecology 29, 505±507. Stickel, L.F., 1950. Populations and home range relationships of the box turtle: Terrapene c. carolina (Linnaeus). Ecol. Monogr. 20, 351±378. Stickel, L.F., 1951. Wood mouse and box turtle populations in an area treated annually with DDT for ®ve years. J. Wildl. Manage. 15, 161±164. Stickel, L.F., 1978. Changes in a box turtle population during three decades. Copeia 1978, 221±225. Stickel, L.F., 1989. Home range behavior among box turtles (Terrapene c. carolina) of a bottomland forest in Maryland. J. Herpetol. 23, 40±44. Stickel, L.F., Bunck, C.M., 1989. Growth and morphometrics of the box turtle Terrapene c. carolina. J. Herpetol. 23, 216±223. U.S. Fish and Wildlife Service, 1996. Export of box turtles from the United States in 1996. Federal Register 61, pp. 3894±3898. Williams, E.C., Jr., Parker, W.S., 1987. A long-term study of a box turtle (Terrapene carolina) population at Allee Memorial Woods, Indiana, with emphasis on survivorship. Herpetologica 43, 328±335.