THE SOUTHWESTERN NATURALIST 51(1):28–34 MARCH 2006
IS THE COACHELLA VALLEY FRINGE-TOED LIZARD (UMA INORNATA)
ON THE EDGE OF EXTINCTION AT THOUSAND PALMS PRESERVE
XIONGWEN CHEN,* CAMERON W. BARROWS, AND BAI-LIAN LI
Department of Botany and Plant Sciences, University of California, Riverside, CA 92521-0124 (XC, B-LL)
Center for Conservation Biology, University of California, Riverside, CA 92521 (XC, CWB, B-LL)
Center for Natural Lands Management, 53298 Avenue Montezuma, La Quinta, CA 92253 (CWB)
ABSTRACT Reliable estimates of extinction time for small populations of threatened and en-
dangered species based on long-term ﬁeld surveys provide crucial information for species conser-
vation. We estimated population parameters and extinction times for Coachella Valley fringe-toed
lizard (Uma inornata), which is listed as a federally threatened species and a California state en-
dangered species. We used ﬁeld survey data from 2 transects (TPP1 and TPP2) at Thousand Palms
Preserve, California, from 1986 to 2003. We also used data from another subpopulation (TPP3)
of this species 6 km from TPP1 and TPP2, where this species became extinct in 2001, to estimate
uncertainty of extinction time. Our results indicated that the difference between modeled extinc-
tion time and the real extinction time of the subpopulation on TPP3 was about 2 y. The extinction
times for U. inornata at TPP1 and TPP2 would be about 23 to 50 y. The extinction time estimate
has good relationship with the habitat area; thus, in larger effective habitat area, there will be an
increase in the estimated time to extinction. Assuming all available habitats have population con-
nectivity, the estimated extinction time would be about 78 y. There was a trend toward a decrease
in the mean reproductive productivity of the lizard during the study. Although there were cycles
in population dynamics, the population density trajectories on phase diagram became close to 0.
For fragmented sand-dune habitats 100 to 200 ha, the persistence of subpopulations of U. inor-
nata is doubtful. The main cause for the decrease of this subpopulation might be the shortage of
food resources during the frequent, severe droughts.
RESUMEN Las estimaciones conﬁables de tiempo de extincion basadas en estudios de campo a
largo plazo, en pequenas poblaciones de especies amenazadas y en peligro de extincion, proveen
´ ´ ´
informacion crucial para la conservacion de las especies. Estimamos parametros poblacionales y
tiempos de extincion para la lagartija del Valle de Coachella (Uma inornata), la cual esta enlistada
como especie amenazada a nivel federal y en peligro de extincion para el estado de California.
Utilizamos datos de campo de dos transectos (TPP1 y TPP2) en Thousand Palms Preserve, Cali-
fornia, de 1986 al 2003. Para estimar con exactitud el tiempo de extincion, tambien utilizamos
datos de otra subpoblacion de la misma especie (TPP3) a 6 km de TPP1 y TPP2, donde esta ´
especie se extinguio en el ano 2001. Nuestros resultados indicaron que la diferencia entre el
´ ´ ´
tiempo de extincion modelado y el tiempo de extincion real de la subpoblacion en TPP3 fue de
aproximadamente dos anos. Los tiempos de extincion para U. inornata en TPP1 y en TPP2 podrıan ´
˜ ´ ´
ser alrededor de 23 a 30 anos. La estimacion del tiempo de extincion tiene buena relacion con ´
´ ´ ´ ´ ´
el area del habitat, ya que a mayor area efectiva del habitat habra un incremento en el tiempo
estimado de extincion. Suponiendo que todos los habitats disponibles tienen conectividad de
´ ´ ˜
poblaciones, el tiempo estimado de extincion podrıa ser alrededor de 78 anos. Durante los anos ˜
del estudio, hubo una tendencia hacia el decremento en el promedio de la productividad re-
productiva de las lagartijas. A pesar de que hubo ciclos en la dinamica poblacional, las trayectorias
de la densidad poblacional en un diagrama de fase se acercaron a cero. La persistencia de sub-
poblaciones de Uma inornata es dudosa en habitats de dunas de arena fragmentadas 100 a 200
ha. La causa principal por el decremento de esta subpoblacion podrıa ser la escasez de los recursos
alimenticios durante las frecuentes y severas sequıas.
March 2006 Chen et al.—Coachella Valley fringe-toed lizard 29
Estimations of expected time to extinction natural processes that create and maintain the
for small populations are often needed by dune ecosystem and a diverse community of
managers, ecologists, and conservation biolo- aeolian sand-adapted species (Barrows, 1996).
gists. Extinction risk evaluations for endan- Since 1986, the status of populations of Coach-
gered and threatened species are required by ella Valley fringe-toed lizard has been moni-
United States Federal Endangered Species Act tored. Based on these long-term data sets at
of 1973 and International Union for Conser- different locations, we attempt to estimate the
vation of Nature and Nature Resources (Unit- population parameters and make an assess-
ed States Senate, 1983; Mace and Lande, ment of the extinction time for this species at
1991). Shaffer (1981) indicated that minimum different locations.
viable population analysis depends on at least
4 stochastic processes affecting population ex- METHODS Study Sites Three plots were estab-
tinction: demographic stochasticity, environ- lished in the central Coachella Valley within the
mental stochasticity, catastrophes, and random Thousand Palms Preserve (TPP). Two plots (TPP1
genetic drift. However, most published models and TPP2) were located approximately 2 km apart
are either too hard to analyze with precision (33 47 N, 116 20 W) on separate active dunes, each
more than 50 ha in area, with sparse vegetation dom-
or too impractical to apply to real populations.
inated by creosote bush (Larrea tridentata) and salt-
Foley (1994) proposed a model based on the
bush (Atriplex) vegetation. Data were collected from
classical diffusion model to predict extinction 1986 to 2003 (TPP1) and from 1990 to 2003 (TPP2).
times from environmental stochasticity and A third plot (TPP3) was on an isolated small dune
carrying capacity, and he applied this model to (33 51 N, 116 19 W) approximately 1 ha in size, 6
estimate the extinction times for populations km north of TPP1 and TPP2. This plot differed from
of checkerspot butterﬂy (Euphydryas editha), TPP1 and TPP2 in that it was adjacent to a natural
grizzly bear (Ursus horribilis), wolf (Canis lupus), palm oasis and included phreatophytic vegetation,
and mountain lion (Felis concolor). Recent ex- such as honey mesquite (Prosopis glandulosa var. tor-
perimental investigations also suggest an un- reyana) and arrowweed (Pluchea sericea). Data were
collected here from 1996 to 2000. After that time,
recognized important role for nonlinear dy-
Coachella Valley fringe-toed lizards disappeared
namics (Belovsky et al., 1999). However, for
from this plot. We used information from this plot
most species and situations we lack detailed da- to estimate uncertainty about the extinction time.
tasets (Morris et al., 1999) to evaluate such dy- Field Survey The TPP1 and TPP2 plots each con-
namics. Furthermore, the classical diffusion sisted of a belt transect (1,000 m 10 m). Surveys
model for population growth and extinction is started in TPP1 and TPP2 in 1986 and 1990, respec-
considered as the least sensitive to random ob- tively. Transects were surveyed at least 6 times per
servation error and is a reasonable method for year in an early summer (May–June) census. The
conducting risk analysis (Wilson, 2000; McNa- number of repetitions required was determined us-
mara and Harding, 2004). ing a power analysis. Six repetitions allowed us to
statistically distinguish population changes 50%
Coachella Valley fringe-toed lizards (Uma in-
(t-test, 0.05, 0.80). Surveys consisted of 2
ornata) were once a common inhabitant of a
biologists walking slowly along transects searching
500-km2 sand-dune system near Palm Springs, for active lizards and tapping shrubs to elicit move-
California, but now are restricted to remaining ment from resting lizards. Each survey was conduct-
fragments of active sand dunes and sand hum- ed between 0800 and 1100 h, when sand surface
mocks in the Coachella Valley, Riverside Coun- temperatures were conducive to lizard activity (35 to
ty (Stebbins, 1944; Norris, 1958; Barrows, 45 C). Survey data were reported as the mean num-
1997). Over the past 3 decades, increases in ber of lizards observed per transect per count peri-
the human population and urban develop- od. Beginning in 1990, a second survey period,
ment have resulted in a 95% loss of its habitat again with 6 repetitions per transect, was added in
the fall (September–October) to count the number
(Barrows, 1996). As a result, the lizard was list-
of hatchling lizards produced that year. Because the
ed as federally threatened and a state endan-
available habitat at TPP3 was too small to include a
gered species in 1980. The listing of the lizard 1,000-m 10-m belt transect, a complete census of
as a threatened and endangered species set in the dune was conducted. This included traversing
motion a series of actions that resulted in the back and forth until all available habitats had been
creation of the Coachella Valley Preserve Sys- observed, and this census was repeated 8 to 10 times
tem in 1986, which focuses on protecting the each year during the fall census period.
30 The Southwestern Naturalist vol. 51, no. 1
Mean Hatchling Number Mean hatchling number
refers to the mean number of newly hatched lizards
observed on each transect each year. By dividing the
number of hatchling lizards by the number of adult
lizards seen the same year provided a measure of
productivity for that year.
Estimation of Population Parameters Simple mathe-
matical models based on ecologically realistic and
measurable parameters are important for compari-
son of species conservation strategies. Optimization
and cost-beneﬁt analyses depend on such compari-
sons. Some models including age structure and oth-
er biological details need extensive parameter infor-
mation, such as RAMAS (Burgman et al., 1993).
They can be made to do the job of comparing strat-
egies, but at some cost in time, generality, and con- FIG. 1 The relationship between the possible ex-
ceptual clarity. In this study, based on the population tinction time and habitat area for Uma inornata in
data collected from transects, we used Foley’s (1994) California.
model, which is based on diffusion analysis and in-
corporated both density dependence and carrying
capacity, to estimate parameters in the following way: RESULTS The estimated extinction times for
a population of size Nt at time t grows in this way the subpopulations of U. inornata at the Thou-
toward time t 1: Nt 1 RtNt, where Rt is the growth sand Palms Preserve were about 30 and 23 y
rate from time t to t 1 and it can be estimated by from 2003, based on information from the
count number of the lizard on a transect in this TPP1 and TPP2 transects, respectively. Spatial
study. After transforming the data by taking natural variation of extinction time existed within the
logarithms, we obtained the following: n ln N, k
same habitat. The extinction time was about 50
ln K, r ln R, where K is the environmental car-
rying capacity, and r is considered as an approxi-
y if TPP1 and TPP2 were connected. Although
mately normal (Gaussian) distribution, such that r this species became extinct at TPP3 in 2001, it
N(rd, vr), where rd is the expected change in n; here would have expected to persist just 2 y beyond
we use the average of r to estimate rd; vr is the vari- 2001 (to 2003) based on the above equation.
ance of the random effect for r. K is estimated as Serial autocorrelation in environmental effects
the maximum of N, although it might change with between consecutive years ( ) at TPP1 and
habitat conditions. These assumptions are based on TPP2 were about 0.376 and 0.304, respec-
the observation that (i) density dependence oper- tively.
ates effectively only close to k, and (ii) demographic Using the average information from TPP1
stochasticity is swamped by environmental stochastic-
and TPP2, we estimated the possible extinction
ity over most of the range (0, k). The extinction time
of a population (Te) can be estimated by (Folley,
time with the change of area assuming that
1994): other parameters are not changed (Fig. 1).
There was a signiﬁcant relationship between
1 habitat area and possible extinction time: ex-
Te (n0) [e 2sk (1 e 2sn0) 2sn 0]
2srd tinction time 16.735 ln(area) 17.969, R 2
0.993. The extinction time decreased dra-
where s rd/vr and n0 is the current abundance of
matically when the habitat area was below 100
the population. Serial autocorrelation in environ-
mental effects between consecutive years is estimat- to about 200 ha. Assuming that total habitat
ed by (Box and Jenkins, 1976; Chatﬁeld, 1989): area does not decrease in aerial extent or qual-
c1/c0 where the covariance cx is given by: ity over time and connectivity between patches
is maintained, the estimated extinction time
1 would be about 78 y for the entire study area.
cx (rt ¯
r )(rt x ¯
T t 1 Trends in estimates of mean hatchling num-
where T and x are time (e.g., year), ¯ is the average
bers are shown in Fig. 2. There seemed to be
of r. If 0, the extinction time will be overesti- a reduction in the reproductive productivity
mated, because bad luck for the lizard population during the study period.
will be more common; if 0, those extinction Although the U. inornata population seemed
times will be underestimated (Foley, 1994). to be quite dynamic, based on the phase dia-
March 2006 Chen et al.—Coachella Valley fringe-toed lizard 31
FIG. 2 The mean hatchling numbers of Uma in-
ornata in 2 plots at Thousand Palms Preserve, Cali-
gram of population density trajectories on the
TPP1 and TPP2 transects (Fig. 3), the popu-
lation of U. inornata gradually decreased to a
low level. By using the population phase dia-
gram we can better understand the modeled
extinction time from a population dynamics
perspective. If population density is above a
FIG. 3 The population density (individuals/ha)
minimum threshold, it is possible for the pop-
trajectories of Uma inornata on phase diagram at 2
ulation be restored to previous high levels. If
transects (time is indicated around each point). (A)
population density is continuously below a TPP1; (B) TPP2.
minimum threshold (here the population den-
sity is close to 0), it is possible that this species
will be extinct soon.
can be further divided into systematic misesti-
DISCUSSION Unlike the lizards on TPP3, U. mates (sampling bias) and non-systematic er-
inornata on the TPP1 and TPP2 dunes are not ror (random sampling error); the sampling
at risk of imminent extinction based on the bias (e.g., time to sample) does not affect the
estimate of extinction time, mean hatchling extinction risk estimate signiﬁcantly, but ran-
number, and phase diagram of population dy- dom sampling error inﬂuences extinction risk
namics. However, when evaluated as isolated signiﬁcantly for some models (Meir and Fagan,
habitat patches, these subpopulations were 2000). Model error: parameter drift will inﬂu-
predicted to disappear in 23 to 30 y. The TPP1 ence the uncertainty in extinction estimates
and TPP2 dunes are part of a larger 1,300-ha (e.g., Ellner, 2003). Increased habitat isolation
matrix of sand dunes and intervening sand and fragmentation will make estimated times
hummocks that allow for connectivity of sub- shorter. Lifespan: an individual of this species
populations. Within that matrix, there are can live 8 y in the wild (http://www.cvmshcp.
nearly 300 ha of sand dunes. Nearly 250 ha of org/sp 06.htm). Recolonization: if this site can
those dunes occur in 3 large polygons, 2 of be connected with sites with sources of this spe-
which include the TPP1 and TPP2 dunes. Us- cies or if habitat quality is improved over time,
ing this area as a base for the effective habitat such as the effects of variable precipitation, re-
available to the lizards, estimated time to ex- colonization will be possible. Minimum viable
tinction would be 78 y. population size is not known. If Allee effect
There are, of course, many uncertainties. works, this species will disappear more quickly
Measurement error: observation errors are re- from this site.
lated to the accuracy of the sampling itself and Populations in isolated habitats are vulnera-
32 The Southwestern Naturalist vol. 51, no. 1
ble to extinction through demographic sto- marily leaves and ants. The dietary content
chasticity, environmental stochasticity, catastro- also differs between breeding and non-breed-
phes, and human disturbances (Shaffer, 1981; ing seasons for males, but does not differ sig-
Goodman, 1987; Burkey, 1989). Demographic niﬁcantly for females. During late summer, the
stochasticity in ﬁnite populations causes pop- diets of the 2 sexes are indistinguishable
ulations in many small patches to be more (www.cvmshcp.org/sp 06.htm). The vegetation
prone to extinction than populations in a sin- growth in deserts is mainly dependent on suf-
gle large patch with the same total area (Bur- ﬁcient precipitation (C. W. Barrows, pers. ob-
key, 1989). Environmental stochasticity and ser.). Annual rainfall of 45 to 50 mm is a crit-
random catastrophes, if they are sufﬁciently ical threshold to determine whether there will
large and spatially uncorrelated, might be able be positive or negative growth for fringe-toed
to counteract the effect of demographic sto- lizard (C. W. Barrows, pers. obser.). However,
chasticity and might render populations in during recent years, precipitation in Coachella
fragmented systems less vulnerable to extinc- Valley has decreased dramatically, especially in
tion than populations in continuous systems 2002, when the annual precipitation was below
(Goodman, 1987; Burkey, 1989). Burkey 10 mm. The recent frequent droughts inhibit-
(1995) studied the extinction rates on islands ed vegetation growth. Not enough food re-
in archipelagoes and estimated the relative sources might be a cause for the declined
probability of extinction per species on single mean hatchling numbers and population den-
large island and sets of smaller islands with the sities. Drake and Lodge (2004) indicated that
same total area. His results indicated that spe- increasing variation in food resource increases
cies (lizards, birds, and mammals) were likely the probability of extinction, decreases the
to go extinct on all the small islands before probability of establishment, and decreases the
they went extinct on the single, large island. time to extinction. Our results conﬁrmed that
Observations of recent fringe-toed lizard ex- metapopulations might be susceptible to local
tinctions on 10 smaller and isolated habitat extinction due to regional climate and global
patches support this result (C. W. Barrows, climate change (Epps et al., 2004).
pers. obser.). Other empirical evidence has On the large scale, the dune systems histor-
shown that local animal populations persist ically covered more area of the Coachella Val-
longer in large patches of suitable habitat than ley. During the recent decades, 95% of dunes
in small patches and in patches close to other have been lost to urban and agricultural de-
suitable patches than in patches more isolated velopment. Off-road vehicle abuse, illegal
from neighbors (Lande, 1987; Bolger et al., dumping, and invasion by exotic weeds also de-
1991; Stacy and Taper, 1991). Theoretical re- grade non-preserved habitat. The increasing
search also indicated the existence of a de- fragmentation of habitats made what was once
struction threshold for species (Loehle and Li, perhaps a metapopulation matrix into a set of
1996; Sole et al., 2004). In this study, the pos- isolated populations with little or no connec-
sible extinction times decrease slowly if the tivity to migrate as resources and habitat con-
connected habitat area is maintained above ditions changed (C. W. Barrows, pers. obser.).
100 to 200 ha. If the isolation level is high enough, each lo-
The mean hatchling number provided a cally isolated population is threatened with ex-
good indicator for the analysis of the persis- tinction, which in turn makes the extinction
tence time of this species. During this study, risk of the whole metapopulation higher (Shi-
the mean hatchling number decreased contin- mada and Ishihama, 2000). Our results might
ually. The causes for the decline in mean support a predictable pattern that subpopula-
hatchling number seem complicated, but they tions will go extinct at the most isolated, small-
were mainly related to the food resources and est, and driest sites.
environmental change (Barrows, 1996). The To alter the serious prognosis for small sub-
Coachella Valley fringe-toed lizard is omnivo- populations of U. inornata in this area, short-
rous, and its diet changes as a function of food term management practices could be imple-
availability. During normal to wet years, it eats mented, although the long-term viability of
primarily ﬂowers and plant-dwelling arthro- small, isolated populations needs to be evalu-
pods. During dry periods, the diet shifts to pri- ated along with level of resources required to
March 2006 Chen et al.—Coachella Valley fringe-toed lizard 33
sustain those populations in perpetuity. A long- vironmental variation on extinction and estab-
term management approach would focus on lishment. Ecology Letters 7:26–30.
maintaining connectivity among populations ELLNER, S. P. 2003. When does parameter drift de-
through preserve design and protecting the crease the uncertainty in extinction risk esti-
natural processes that maintain the dynamic mates? Ecology Letters 6:1039–1045.
character of the aeolian sand habitats that the EPPS, C. W., D. R. MCCULLOUGH, J. D. WEHAUSEN, V.
lizards require. Increasing the connectivity C. BLEICH, AND J. L. RECHEL. 2003. Effects of cli-
among different habitats would increase ‘‘res- mate change on population persistence of desert-
cue effects.’’ Furthermore, efforts should be dwelling mountain sheep in California. Conser-
vation Biology 18:102–113.
contributed to avoid effects from catastrophic
FOLEY, P. Predicting extinction times from environ-
events (e.g., storm and extended severe
mental stochasticity and carrying capacity. Con-
drought). Catastrophic events would easily re-
servation Biology 8:124–137.
sult in extinction of small populations.
GOODMAN, D. 1987. Consideration of stochastic de-
This research was partially supported by Center mography in the design and management of bi-
for Conservation Biology of the University of Cali- ological reserves. Natural Resource Modeling 1:
fornia at Riverside and the University of California 205–234.
Agricultural Experiment Station. Staff from the LANDE, R. 1987. Extinction thresholds in demo-
United States Bureau of Land Management, United graphic models of terrestrial populations. Amer-
States Fish and Wildlife Service (Refuges Division), ican Naturalist 130:624–635.
California Department of Fish and Game, and Cali- LOEHLE, C., AND B.-L. LI. 1996. Habitat destruction
fornia Department of Parks and Recreation aided in and the extinction debt revisited. Ecological Ap-
lizard surveys. Thanks to A. C. Murillo for the Span- plications 6:784–789.
ish abstract. MACE, G., AND R. LANDE. 1991. Assessing extinction
threats: towards a reevaluation of IUCN threat-
LITERATURE CITED ened species categories. Conservation Biology 5:
BARROWS, C. W. 1996. An ecological model for the 148–157.
protection of a dune ecosystem. Conservation Bi- MCNAMARA, J., AND K. C. HARDING. 2004. Measure-
ology 10:888–891. ment error and estimates of population extinc-
BARROWS, C. W. 1997. Habitat relationships of the tion risk. Ecology Letters 7:16–20.
Coachella Valley fringe-toed lizard (Uma inorna- MEIR, E., AND W. F. FAGAN. 2000. Will observation
ta). Southwestern Naturalist 42:218–223. error and biases ruin the use of simple extinction
BELOVSKY, G. E., C. MELLISON, C. LARSON, AND P. A. models? Conservation Biology 14:148–154.
VAN ZANDT. 1999. Experimental studies of extinc- MORRIS, W., D. DOAK, M. GROOM, P. KAREIVA, J. FIE-
tion dynamics. Science 286:1175–1177. BERG, J. GERBER, P. MURPHY, AND D. THOMSON.
BOLGER, D. T., A. C. ALBERTS, AND M. E. SOULE. 1991. 1999. A practical handbook for population via-
Occurrence patterns of bird species in habitat bility analysis. The Nature Conservancy, Washing-
fragments: sampling, extinction, and nested spe- ton, D.C.
cies subsets. American Naturalist 137:155–166. NORRIS, K. 1958. The evolution and systematics of
BOX, G. E. P., AND G. M. JENKINS. 1976. Time series the iguanid genus Uma and its relation to the
analysis: forecasting and control. Holden-Day, evolution of the other North American desert
Oakland, California. reptiles. Bulletin of the American Museum of
BURGMAN, M. A., S. FERSON, AND H. R. AKCAKAYA. Natural History 114:251–326.
1993. Risk assessment in conservation biology. SHAFFER, M. L. 1981. Minimum population sizes for
Chapman and Hall, London, United Kingdom. species conservation. Bioscience 31:131–134.
BURKEY, T. V. 1989. Extinction in nature reserves: the SHIMADA, M., AND F. ISHIHAMA. 2000. Asynchroniza-
effect of fragmentation and the importance of tion of local population dynamics and persis-
migration between reserve fragments. Oikos 55: tence of a metapopulation: a lesson from an en-
75–81. dangerous composite plant, Aster kantoensis. Pop-
BURKEY, T. V. 1995. Extinction rates in archipelagoes: ulation Ecology 42:63–72.
implications for populations in fragmented hab- SOLE, R. V., D. ALONSO, AND J. SALDANA. 2004. Habitat
itats. Conservation Biology 9:527–541. fragmentation and biodiversity collapse in neu-
CHATFIELD, C. 1989. The analysis of time series. tral communities. Ecological Complexity 1:65–
Chapman and Hall, London, United Kingdom. 75.
DRAKE, J. M., AND D. M. LODGE. 2004. Effects of en- STACY, P. B., AND M. TAPER. 1991. Environmental var-
34 The Southwestern Naturalist vol. 51, no. 1
iation and the persistence of small populations. States Government Printing Ofﬁce, Washington,
Ecological Applications 2:38–42. D.C.
STEBBINS, R. C. 1944. Some aspects of the ecology of WILSON, E. O. 2000. On the future of conservation
the iguanid genus Uma. Ecological Monographs biology. Conservation Biology 14:1–3.
UNITED STATES SENATE, COMMITTEE ON ENVIRONMENT
AND PUBLIC WORKS. 1983. The endangered spe- Submitted 11 May 2004. Accepted 18 February 2005.
cies act as amended by public law 97-304. United Associate Editor was Geoffrey C. Carpenter.