Site by Bennett Web & Design Co.
2006 SOUTHEASTERN NATURALIST 5(1):149–156
Comparison of Survival Estimates Using Age-Specific
Mortality and Radiotelemetry Data for Florida Key Deer
PATRICIA M. HARVESON1,*, ROEL R. LOPEZ1, NOVA J. SILVY1, AND PHILIP A. FRANK2
Abstract - Obtaining reliable survival estimates is important in the management of
wildlife populations, particularly for the construction of computer simulation models.
Many methods for estimating survival (e.g., radiotelemetry) are cost-prohibitive or
time consuming. Life tables can provide survival estimates using data routinely
collected by some management agencies. We calculated annual survival for
Odocoileus virginianus clavium (Key deer) using age-specific mortality data. We
compared our life-table estimates to those calculated from radiotelemetry data. Key
deer survival estimates derived from life tables were similar to rates calculated from
radiocollared deer. The only exception was for yearling/adult females on north Big
Pine Key, where the life-table estimate was only slightly outside of the 95% confidence
interval for the radiotelemetry estimate. Our results suggest that life tables based on
age-specific mortality data can be a useful tool in estimating survival for Key deer.
Comparing survival estimates from both methods allowed us to evaluate potential
biases due to violation of assumptions associated with life-table calculations. While
wildlife managers should be aware of the potential biases, age-specific mortality data
may provide an adequate and cost-effective alternative for estimating survival.
Estimating wildlife population demographics is an important component
in construction of simulation models (e.g., harvest models, population viability
analyses [PVAs]) used to predict population trends. Annual survival
is an important population parameter that influences population growth
(Krebs 1999, Rabe et al. 2002, White and Bartmann 1998) and is a key
component in the development of these models. For example, PVAs are
commonly used in endangered species management (Akçakaya 2000, Boyce
1992) and require precise survival estimates. Many methods for estimating
survival exist; however, each of these methods have their own benefits and
problems (Krebs 1999). Estimating survival from radiotelemetry or markrecapture
data, for example, can provide precise estimates under mild
assumptions, yet often at great expense (Krebs 1999). Limited or declining
budgets of many wildlife management agencies may prohibit the use of
radiotelemetry data in estimating survival (Rabe et al. 2002). Alternative
approaches to estimating survival might include the use of age-composition
data or life tables (Krebs 1999). However, these alternative approaches
require more restrictive assumptions than telemetry or mark-recapture
which, if violated, can bias estimates (Caughley 1977, Williams et al. 2002).
1Department of Wildlife and Fisheries Sciences, Texas A&M University, College
Station, TX 77843. 2US Fish and Wildlife Service, National Key Deer Refuge, Big
Pine Key, FL 33043. *Corresponding author - firstname.lastname@example.org.
150 Southeastern Naturalist Vol. 5, No. 1
Life tables can be used to estimate age-specific mortality or survival
from an assumed cohort using various methods including age at death, age of
remains, and age distribution of a population (Caughley 1977, Krebs 1999).
Although data collection for life tables also can be expensive (Caughley
1977), some agencies such as the US Fish and Wildlife Service (USFWS)
National Key Deer Refuge (NKDR) routinely collect deer mortality data that
can be used in the construction of life tables. Use of already collected data
could be a cost-effective way for agencies to estimate important population
parameters for use in managing wildlife populations.
Caughley (1977) cautioned against the improper use of these methods
and violations of assumptions in life-table construction. For example, to
estimate survival from carcasses or skulls, collected data must represent a
random sample of all population mortalities and the population must have
a stable age distribution and a known rate of increase (Caughley 1977).
Potential biases pertaining to data collection include the use of hunterharvest
mortalities, seasonal collection (i.e., winter or summer only), or
mortalities resulting from rare events such as catastrophes. Each of these
situations could produce biased survival estimates. Udevitz and Bellachey
(1983) presented a method in which the assumption of either stable age
distribution or known rate of increase may be removed by combining ageat-
death data and independently sampled standing-age-structure data.
Regardless of which method is used, care should be taken to evaluate the
accuracy of survival estimates based on life-table data, and, whenever
possible, these estimates should be validated with estimates derived by
other means such as radiotelemetry data. Use of erroneous survival estimates
in making management decisions could have potentially devastating
effects on a population, especially in the management of an endangered
species like Odocoileus virginianus clavium Barbour and G.M. Allen
(Florida Key deer).
Key deer are a sub-species of white-tailed deer endemic to the Florida
Keys (Hardin et al. 1984). Urban development and habitat fragmentation
threaten the Key deer population, with 50% of Key deer mortality attributed
to deer-vehicle collisions (Harveson et al. 2004, Lopez et al. 2003b). Since
1968, the USFWS NKDR has collected Key deer mortality data (Lopez et al.
2004) as part of a long-term monitoring program. Additionally, radiotelemetry
data have been collected during two separate studies from December
1968 to June 1972, and January 1998 to December 2000 (Hardin 1974,
Lopez 2001, Silvy 1975). Survival estimates for Key deer using radiotelemetry
data were recently reported (Lopez et al. 2003b), which offers a unique
opportunity to compare survival estimates from different sources (i.e., radiotelemetry
versus mortality data). Our research objectives were to evaluate
the use of USFWS Key deer mortality data as an alternative method for
calculating survival by (1) estimating Key deer survival using USFWS Key
deer mortality data, and (2) comparing these survival estimates to previously
published survival estimates calculated from radiotelemetry data.
2006 P.M. Harveson, R.R. Lopez, N.J. Silvy, and P.A. Frank 151
The Florida Keys are a chain of small islands approximately 200-km long
extending southwest from peninsular Florida. Big Pine Key (BPK; 2548 ha)
is within the boundaries of the NKDR in Monroe County, and supports
approximately 60% of the deer population (Lopez 2001). Soil types vary
from marl deposits to bare rock of the oolitic limestone formation (Dickson
1955). Island vegetation varies by elevation with Rhizophora mangle L. (red
mangrove), Avicennia germinans (L.) L. (black mangrove), and
Laguncularia racemosa (L.) Gaertn. f. (white mangrove), and Conocarpus
erectus L. (buttonwood) forests occurring near sea level (maritime zones).
As elevation increases inland, maritime zones transition into hardwood (e.g.,
Bursera simaruba (L.) Sarg. [gumbo limbo], Piscidia piscipula (L.) Sarg.
[Jamaican dogwood]) and pineland (e.g., Pinus elliottii Engelm. [slash
pine], Serenoa repens (Bartr.) Small [saw palmetto]) upland forests with
vegetation intolerant of salt water (Dickson 1955, Folk 1991).
Since 1968, USFWS NKDR staff have recorded deer mortality as part of
recovery efforts. Dead animals were located primarily by direct sightings,
citizen reports, and observation of Cathartes aura Linnaeus (Turkey Vultures).
Animals collected were held frozen prior to necropsy examination or necropsied
immediately. Carcass quality or ability to determine cause of death ranged
from good to marginal (Nettles 1981, Nettles et al. 2002). Age, sex, body mass,
and cause of death were recorded for each animal using procedures described
by Nettles (1981), and all mortality locations were recorded. Attempts were
made to collect all deer mortalities on BPK. The island’s small size (2548 ha),
high human population (4026; US Census Bureau 2000 population estimate),
and lack of predators make it likely that most deer mortalities were located.
While some carcasses may have been missed, it is reasonable to assume that
those found represented an unbiased sample of deer mortalities.
Life tables were constructed for Key deer by sex and age using the
USFWS mortality data collected from 1995–2000 on BPK. Lopez et al.
(2003b) reported differences in survival between the northern and southern
portions of the island, thus, we analyzed our data for two areas: north
BPK (NBPK; carcasses located north of Watson Boulevard) and south BPK
(SBPK; carcasses located south of Watson Boulevard). Age-specific survival
was estimated for each area by sex and age-class assuming a stable
age distribution and an instantaneous population growth rate of 0.00
(Caughley 1977, Krebs 1999, Lopez et al. 2004). We believe that the deer
population on BPK has a stable age distribution because there is no evidence
of major variation in mortality or recruitment rates. Lopez et al.
(2004) analyzed Key deer survival using radiotelemetry data and found no
difference in survival for data collected from 1968–1971 compared to data
collected from 1998–2000 (Lopez et al. 2003a). Further, Key deer are not
152 Southeastern Naturalist Vol. 5, No. 1
susceptible to hard-winter die-offs due to the tropical nature of the Keys,
and as an unhunted population, there have been no changes in mortality
risk due to increased or decreased hunting pressure. While the Florida
Keys are susceptible to potentially catastrophic hurricanes, there have been
no major impacts to the deer population from 1968–2000.
We estimated the Key deer population growth rate using the average
number of deer seen per year during USFWS survey counts. Spotlight
surveys were conducted monthly along an established survey route (same
beginning and ending point) to provide NKDR biologists with an index of
population size (Lopez et al. 2004). We calculated the mean (0.004) and
standard error (0.066) of yearly growth rates from 1995–2000. There was no
difference between life table calculations using r = 0.000 or r = 0.004, thus,
we assumed that the population was stationary from 1995–2000. We also
compared our estimated growth rate from 1995–2000 (r = 0.004) to the
growth rate from 1970–2000 reported in Lopez et al. (2004; R = 1.038 or r =
0.037) and found that the latter rate fell within the 95% confidence interval
of our estimate of the population growth rate, supporting our estimate of
population growth and stable age distribution.
The estimated number of deer dying for each age interval was calculated
using the equation
d´x = dxerx
where: dx = actual number of carcasses in each age class, r = instantaneous
population growth rate, x = age class, and e = base of natural logarithms (i.e.,
2.71828). Survival (px) was calculated using the equation (Caughley 1977)
⎟ ⎟ ⎟ ⎟
⎜ ⎜ ⎜ ⎜
Deer of unknown sex or age were not used in calculations. For comparison
purposes, we calculate survival for fawns (< 1 year of age) and
yearlings/adults using a weighted mean by grouping carcasses aged ≥ 1
year (Caughley 1977). Survival estimates, standard errors, and confidence
intervals were computed using the program SURVIV (Udevitz and
Ballachey 1998, White 1983).
Lopez et al. (2003b) recently reported Key deer survival estimates based
on 314 radiocollared animals by sex, age, and area (north BPK, south BPK,
and No Name Key [NNK]). Due to constraints from small sample sizes with
our mortality data, we were not able to construct life tables for NNK. Deer
were classified into three age groups, fawn (< 1 year old), yearling (1–2
years old), and adult (≥ 2 years old). However, yearling and adult age groups
were combined as model selection found no differences in survival for these
age groups (Lopez et al. 2003b). Annual Key deer survival was estimated
2006 P.M. Harveson, R.R. Lopez, N.J. Silvy, and P.A. Frank 153
using a known-fate model framework in program MARK (Lopez et al.
2003b, White and Burnham 1999).
We used survival estimates calculated from radiotelemetry data as a
benchmark for comparison under the assumption that these estimates best
reflected actual Key deer survival rates. Known-fate models estimate survival
with high precision since the status of each animal is known at each sampling
occasion (alive, dead, or censored; Lopez et al. 2003b, White and Burnham
1999). In addition, the assumptions of stable age distribution and known rate
of increase are not required with known-fate models as they are with life-table
estimates calculated from age-specific mortality data. Therefore, if our survival
estimates based on mortality data are biased due to violation of life-table
assumptions, then our estimates should differ from those calculated using
radiotelemetry data. We compared life table survival estimates to radiotelemetry
survival estimates for each area, sex, and age category (Lopez et al.
2003b) using 95%-confidence intervals (estimate ± 1.96SE). Survival estimates
calculated from mortality data and radiotelemetry data were considered
similar if 95%-confidence intervals overlapped (Johnson 1999).
A total of 506 deer (177 females, 329 males) mortalities was recorded by
USFWS biologists from 1995–2000. Key deer survival estimates derived
from life tables were similar to rates calculated from radiocollared deer
(Table 1). The only exception was for yearling/adult females on NBPK
where the life-table estimate was only slightly outside of the 95%-confidence
interval for the radiotelemetry estimate (Table 1). Overall, variability
was smallest for our life-table survival estimates compared to those based on
radiotelemetry data. Life table estimates differed by area (NBPK and SBPK)
and age group (fawn and yearling/adult), but were similar by sex (Table 1).
Table 1. Annual Key deer survival estimates by data source (mortality, radiotelemetry), sex, and
age group on north Big Pine Key (NBPK) and south Big Pine Key (SBPK), FL.
95% 95% 95% 95%
Sex Age Area Survival SE LCI UCI Survival SE LCI UCI
Fawn NBPK 0.667 0.058 0.553 0.780 0.726 0.109 0.512 0.940
Fawn SBPK 0.739 0.042 0.657 0.820 0.695 0.091 0.517 0.873
Y/Adult NBPK 0.707 0.037 0.634 0.780 0.848 0.033 0.783 0.913
Y/Adult SBPK 0.610 0.034 0.544 0.676 0.710 0.082 0.549 0.871
Fawn NBPK 0.683 0.051 0.582 0.784 0.668 0.091 0.490 0.846
Fawn SBPK 0.741 0.028 0.686 0.796 0.599 0.158 0.289 0.909
Y/Adult NBPK 0.678 0.035 0.609 0.748 0.583 0.060 0.465 0.701
Y/Adult SBPK 0.563 0.024 0.516 0.612 0.412 0.099 0.218 0.606
AMortality data collected by US Fish and Wildlife Service from 1995–2000.
BRadiotelemetry data collected from 1968–1972 (Silvy 1975) and 1998–2000 (Lopez 2001).
154 Southeastern Naturalist Vol. 5, No. 1
Construction of life tables with mortality data for white-tailed deer
requires several assumptions that may introduce bias (Caughley 1977); thus,
results should be viewed cautiously. For example, we found a small, although
significant, difference in estimated survival of yearling/adult females
between the two data sources (Table 1.) Accurate survival estimates for
adult females are particularly important as these estimates tend to have a
significant impact on large ungulate population trends (Rabe et al. 2002,
White and Bartmann 1998). Overestimating adult female survival could
have detrimental effects in both endangered species and game population
management. Our life table survival estimates, however, underestimated
yearling/adult female survival which would produce a lower, more conservative
estimate of population growth. The reason for this difference is
unknown. Possible explanations include biases with the life-table and/or
radiotelemetry estimates. It is possible that one of the life-table assumptions
(e.g., random sample, stable age distribution, or known-growth rate) may
have been violated.
However, with one slight exception, survival estimates between the two
methods were similar suggesting that estimating survival based on mortality
data may be an adequate alternative to collecting radiotelemetry data.
The purpose of this paper was to evaluate the utility of using mortality data
for estimating survival for Key deer. In the absence of known-survival
rates, we chose the best available estimates, those calculated using radiotelemetry
data (Lopez et al. 2003b). Thus, for comparison purposes, we used
the same age, sex, and area categories as reported in Lopez et al. (2003b).
Better models for estimating survival using mortality data may exist and
can be easily evaluated using the program SURVIV (White 1983) or
MARK (White and Burnham 1999), which calculate AIC, likelihood ratios,
and goodness-of-fit estimates for various user-defined models. For example,
we set survival equal for ages ≥ 1 year. Other models can be
specified such as equal survival across all ages, or equal survival for adults
only and separate survival for fawns and yearlings. The statistics generated
by the program SURVIV can be used to select the most appropriate model
for estimating survival of Key deer based on mortality data (Udevitz and
Ballachey 1998, White 1983).
We found life-table survival estimates to be similar to those derived from
radiotelemetry data, suggesting an alternative for estimating survival of Key
deer. The long-term monitoring of Key deer mortality by USFWS biologists
offers managers such an opportunity. In our study, comparing results from
both methods allowed us to evaluate potential biases due to violation of
assumptions associated with life-table calculations. Furthermore, alternative
methods exist which can eliminate some of the assumption associated with
2006 P.M. Harveson, R.R. Lopez, N.J. Silvy, and P.A. Frank 155
life-table calculations. For example, combining age-structure data with mortality
data permits the elimination of either the assumption of stable age
structure or known rate of increase (Udevitz and Bellachey 1998). While
wildlife managers should be aware of the potential biases associated with
life-table calculations, age-specific mortality data may provide an adequate
and cost-effective alternative for estimating survival.
We thank Texas A&M University graduate students and interns who assisted in
the collection of field data. We also thank the staff of the US Fish and Wildlife
Service National Key Deer Refuge. Funding was provided by the TAMU System, the
Hispanic Leadership Program in Agriculture and Natural Resources (US Department
of Agriculture), the Rob and Bessie Welder Wildlife Foundation, and the US Fish
and Wildlife Service. This manuscript is supported by the Welder Wildlife Foundation,
Contribution No. 646.
Akçakaya, H.R. 2000. Population viability analyses with demographically and spatially
structured models. Ecological Bulletins 48:23–38.
Boyce, M.S. 1992. Population viability analysis. Annual Reviews Ecology and
Caughley, G. 1977. Analysis of Vertebrate Populations. John Wiley and Sons, New
Dickson III, J.D. 1955. An ecological study of the Key deer. Technical Bulletin 3,
Florida Game and Fresh Water Fish Commission, Tallahassee, FL.
Folk, M.L. 1991. Habitat of the Key deer. Ph.D. Dissertation, Southern Illinois
University, Carbondale, IL.
Hardin, J.W. 1974. Behavior, socio-biology, and reproductive life history of the
Florida Key deer, Odocoileus virginianus clavium. Ph.D. Dissertation, Southern
Illinois University, Carbondale, IL.
Hardin, J.W., W.D. Klimstra, and N.J. Silvy. 1984. Florida Keys. Pp. 381–390, In
L.K. Halls (Ed.). White-tailed Deer: Ecology and Management. Stackpole
Books, Harrisburg, PA.
Harveson, P.M., R.R. Lopez, N.J. Silvy, and P.A. Frank. 2004. Source-sink dynamics
of Florida Key deer on Big Pine Key, Florida. Journal of Wildlife Management
Johnson, D.H. 1999. The insignificance of statistical significance testing. Journal of
Wildlife Management 63:763–772.
Krebs, C.J. 1999. Ecological Methodology. Second edition. Addison-Welsey, Menlo
Lopez, R.R. 2001. Population ecology of Florida Key deer. Ph.D. Dissertation.
Texas A&M University, College Station, TX.
Lopez, R.R., N.J. Silvy, R.F. Labisky, and P.A. Frank. 2003a. Hurricane impacts on
Key deer in the Florida Keys. Journal of Wildlife Management 67:280–288.
Lopez, R.R., N.J. Silvy, B.L. Pierce, P.A. Frank, M.T. Wilson, and K.M. Burke.
2004. Population density of the endangered Florida Key deer. Journal of Wildlife
156 Southeastern Naturalist Vol. 5, No. 1
Lopez, R.R., M.E.P. Vieira, N.J. Silvy, P.A. Frank, S.W. Whisenant, and D.A. Jones.
2003b. Survival, mortality, and life expectancy of Florida Key deer. Journal of
Wildlife Management 67:34–45.
Nettles, V.F. 1981. Necropsy procedures. Pp. 6–16, In W.R. Davidson, F.A. Hayes,
V.F. Nettles, and F.E. Kellogg (Eds.). Diseases and Parasites of White-tailed
Deer. Tall Timbers Research Station, Tallahassee, FL.
Nettles, V.F., C.F. Quist, R.R. Lopez, T.J. Wilmers, P. Frank, W. Roberts, S.
Chitwood, and W.R. Davidson. 2002. Morbidity and mortality factors in Key
deer, Odocoileus virginianus clavium. Journal of Wildlife Diseases 38:729–737.
Rabe, M.J., S.S. Rosenstock, and J.C. deVos, Jr. 2002. Review of big-game survey
methods used by wildlife agencies of the western United States. Wildlife Society
Silvy, N.J. 1975. Population density, movements, and habitat utilization of Key deer,
Odocoileus virginianus clavium. Ph.D. Dissertation, Southern Illinois University,
Udevitz, M.S., and B.E. Ballachey. 1998. Estimating survival rates with age-structure
data. Journal of Wildlife Management 62:779–792.
Williams, B.K., J.D. Nichols, and M.J. Conroy. 2002. Analysis and Management of
Animal Populations. Elsevier-Academic, San Diego, CA.
White, G.C. 1983. Numerical estimation of survival rates from band-recovery and
biotelemetry data. Journal of Wildlife Management 47:716–728.
White, G.C., and R.M. Bartmann. 1998. Mule deer management: What should be
monitored? Pp. 104–118, In J. deVos, Jr. (Ed.). Proceedings of the 1997 Deer-Elk
Workshop, 21–23 May 1997, Rio Rico, Arizona. Arizona Game and Fish Department,
White, G.C., and K.P. Burnham. 1999. Program MARK: Survival estimation from
populations of marked animals. Bird Study 46:S120–S138.