ECOGRAPHV 22: 213-218. Copenhagen 1999

El Nino events, precipitation patterns, and rodent outbreaks arestatistically associated in semiarid Chile

Mauricio Lima, Pablo A. Marquet and Fabian M. Jaksic

Lima, M.. Marquet. P. A. and ,laksic, F. M. 1999. El Niiio events, precipitationpatterns, and rodent outbreaks are statistically associated in .semiarid Chile. -Eeography 22: 213-218.

In the last two decades, several researchers have noted rodent population outbreaksin semiarid South America, in association with unusually high precipitation thatseemingly concurs with El Nifio events. To date, no studies have been conducted todetermine the statistical relationships between ENSO (El Nino Southern Oscillation)events, increased precipitation, and rodent irruptions. Here we show that: 1) there isa statistical association between ENSO events and inereased precipitation in thesemiarid region of northern Chile: 2) the occurrence of rodent outbreaks in thatregion is statistically related with the precipitation levels of the same year: 3) themulti-annual patterns of the total annual precipitation levels and population abun-dance of those rodents during tlie summer are positively associated. The putativechain of effects seems to start with unusually high rainfall brought by ENSO tosemiarid environments, which thus respond with inereased primary productivity(herbage and seeds), which then fuels the rodent outbreaks.

M. Lima (mlima@ahello.dic.iichile.el). Dept de Cieneias Eeologieas, Univ. de Chile.Casilla 653. Santiago. Chile. - P. A. Marquet. and F. M. Jaksic. Dept de Feologia.Pontificia i'uiv. Catuliea de Chile. Casilla II4-D. Santiago. Chile.

Although rodent populations in the Neotropical regionhave been mueh less studied than those in the Holarcticregion (Hansson and Henttonen 1988, Stenseth and Ims1993. Norrdahl and Korpimaki 1995), there is eircutn-stantial evidence that the fomier tindergo marked pop-ulation fiuctuations (O'Connell 1986). Their populationpeaks often reach so high that they are eonsideredagricultural pests that cause heavy economic losses inmost of the Neotropics (Rodriguez 1993). Untbrtu-nateiy, much of the inlormation is anecdotal, or at bestshort-spanned, and thus there are no adequate data-bases that may enable a better understanding of themeehanisms that underlie sueh population tluctuations.Still, some patterns are evident. O'Connell (1986). inher review of rodent population variability patterns inthe Neotropics, noted that it was crieetid rodents whichdisplayed the most dramatic peaks, in particular thosefrom western South America. Indeed, historieal t"eeordshave long existed about rodent outbreaks in that re-

gion, either in association with masting shrubs or un-usual preeipitation. As far baek as 196U"s, thoseoutbreaks or irruptions were considered to be similar orequivalent to those displayed by microtines in theNorthern Hemisphere (Hershkovitz 1962).

A case in point is Chile, where sinee 1780 there havebeen anecdotal records of rodent outbreaks around thesouthern town of Valdivia. in association with mastingbamboo Clwsquca spp. (Philippi 1879, Het"shkovitz1962, Murua et al. 1986. 1996). Better documented yet.have been the rather more frequent rodent itTuptions insemiarid zones of southern Peru and northern Chile, inapparently close association with periods of high pre-eipitation (Pearson 1975, Pefaur et al. 1979. Fuentesand Campusano 1985, Meserve and Le Boulenge 1987.Meserve et al. 1995). In this latter case, the speeiesinvolved are essentially three cricetids: PhyUofis darwini(50 g). Akodoit oliiaceus (27 g), and Oligoryzomys long-icawkttits (28 g).

Accepted 4 September 1998

Copyright C ECOGRAPHY 1999ISSN 0906-7590Printed in Ireland all rights reserved

ECOGRAPHY 22:2 213

In particular, the hcrbivore-granivore leaf eared-mouse PItyilatis danvini has been reported for experi-encing the most dramatic population outbreaks, and ithas nol gone unnoticed (Pefaur et ;il. 1979. Jimenez etal. 1992. Jaksic et al. 1993. 1996, 1997, Meserve et al.1995) that they seemingly concur with ENSO events (ElNiiio Southern Oscillation), The putative chain of ef-fects would be increased primary productivity (herbageand seeds) associated with unusually high rainfallbrought by ENSO to semiarid environments (Armestoet al. 1993, Gutierrez et al. 1993. Jaksic et al. 1997).This assumed causal relationship has led some re-searchers to consider some rodent populations as non-regulated and driven by extrinsic climatic factors(Meserve and Le Boulenge 1987, Meserve et al. 1995).By contrast, we have determined strong delayed den-sity-dependent factors operating at demographic (Limaand Jaksic 1998a) and population dynamic (Lima andJaksic 1998b, 1999) levels in this species. In sum, todate no studies have been conducted to determine thestatistical relationships between ENSO events, in-creased precipitation, and rodent fluctuations. This iswhat we do in our paper.

First, we statistically analyse the relationships be-tween ENSO events and intcr-annua! variability ofprecipitation, as well as the relationship of both phe-nomena with reliable historical records of rodent popu-lation irruptions in semiarid Chile. Secondly, on thebasis of a long-term database on a mctapopulation ofthe rodent Phyllolis danvini (Lima et al. 1996. Torres-Contreras et al. 1997) in a specific locality of northernChile, we analyse the relationship between populationabundance and precipitation, at a multi-annual tempo-ral scale.

Material and methods

Meteorological and climatological data

One of the indices used to estimate the occurrence andintensity of the El Niiio Southern Oscillation (ENSO) isthe Southern Oscillation Index (SOI), that comparesmeteorological conditions in two tropical areas of thePacific Ocean (Allen et al. 1996). The SOI is calculatedby subtracting from the monthly-standardised atmo-spheric pressure measured in Tahiti that recorded inDarwin. Australia. The SOI is reported as a 5-monthmoving average. Periods characterised by low indexvalues (negative) correspond to ENSO events -wherein warm tropical waters intrude into the westerncoasts of the Americas - whereas those periods charac-terised by high index values (positive) correspond toevents denominated La Nina - when the cold Hum-boldt and California currents bathe the Americas"coastlines. For our analyses we used the SOI indexvalues obtained from 1943 to 1997. Values of this index

over the last century are accessible through the Internetweb site of the National Oceanographic and Atmo-spheric Administration (NOAA) of the USA.

Monthly precipitation data were obtained Ironi theclimatological stations of Hurtado (3O°17'S; 70°4rW;1200 m elevation: period 1943-1997), Pichasca(30°23'S: 70°52'W: 750 m elevation; period 1946- 1997)- both located in the Limari River basin - and lllapel(3r38'S: 7riO'W; 290 m elevation; period 1987-1997)

located in the Choapa River basin. Data from theformer iwo locations were deemed as adequate repre-sentations of climatological conditions in northernChile, largely because they provided the longest timeseries. Data from the third one - although a short timeseries - were deemed as representative of local condi-tions experienced in our primary study site at Auco,located approximately 18 km away of the lllapelstation.

Rodent irruptions at regional and local levels

At the regional level, we determined the occurrence andfrequency of rodent outbreaks in northern Chile bymeans of historical records published in local newspa-pers from several towns and cities in the region, partic-ularly those in the basin of the rivers Elqui and Limari.Presence and absence of rodent irruptions was trackedfrom 1943 to 1980 (see Fuentes and Campusano 1985,for details).

At the local level, we applied capture-mark-recapture(CMR) procedures from 1987 to 1997 to four mammalsubpopulations (two in equatorial-exposed, xeric habi-tats, and two in polar-exposed, mesic habitats) in twocreeks of the Reserva Nacional Las Chinchillas, atAuco {3r3O'S; 7I°O8'W). Live-trapping was conductedin El Cobre and El Grillo creeks (separated by 2 km instraight line) from October 1987 to December 1991 inEl Cobre and from November 1987 to .lanuary 1992 inEl Grillo. every other month, alternating between onecreek and the other (that is, six times a year in eachcreek). Erom July 1992 to October 1997. live-trappingwas relaxed to every three months (that is. in January,April. July and October) and effected only in the twosubpopulations of El Cobre creek.

Each grid had 7 x 7 Hve-trapping stations spaced 15m and equipped with a single Shennan-type trap, cov-ering an area of 105 x 105 m (which includes a periph-eral boundary strip of 7.5 m), thus representing 49 trapsin 1.1 ha per grid. In either creek, both xeric and mesicgrids were activated simultaneously during live consec-utive nights. All traps remained permanently in thefield, and were closed when not in use. Rolled oats wasthe bait, and all activated traps were checked everymorning. Each mammal individual captured wasmarked with a metallic ear tag. and the species, sex,body mass, and reproductive status was recorded. An

214 ECOORAPHY 22:

abundance index was estimated as the minimum num-ber of individuals known alive (Krebs 1989). The rela-tive abundances of P. danvini catches are higher duringthe breeding seasons (spring and summer; Meserve andLe Boulenge 1987. Jimenez et al. 1992) and resulted inpronounced tluctuations. The overall lower variationsduring autumn and winter result in iess pronouncedvariations, which presumably are more sensitive torandom sampling variations. The less consistent varia-tions of the autumn and winter did not permit consis-tent analyses of these data. Hence, only summer countsfrom El Cobre creek (eleven years) have been consid-ered for the analysis.

Statistical analyses

With the aim of assessing any statistical relationshipsbetween El Niiio events, precipitation records, androdent outbreaks, we used tools for the analysis of timeseries. Total annual precipitation records obtained inthe Hurtado and Pichasca climatologica! stations wereanalysed with reference to the SOI summed up yearly,by means of the Cross Correlation Function (CCE)(Chattield 1989).

To detemiinc the relationship (if any) between pre-cipitation and rodent irruptions, we applied a logisticregression analysis (Trexler and Travis 1993) betweenthe occurrence of rodent outbreaks (presence-absence)and yearly cumulative precipitation records from Hur-tado and Pichasca. separately. Precipitation was log-transformed in order to normalise its distribution. Weevaluated the contribution of each parameter (iog-transformed precipitation) using a log likelihood X"ratio test, which compares the log likelihood of a givenmodel to a model with its only parameter being aconstant (Trexler and Travis 1993).

The relationship between total annual precipitationrecords and the summer population abundance of therodent Phyllotis danvini in Auco was analysed tor theyears 1987 1997. With the aim of determining anystatistical relationship between rodent abundance andprecipitation, we applied linear regression between totalannual precipitation and summer population abun-dance of Phyilotis danviid during the summer season ofevery year.

Results

El Nino events and precipitation patterns

Low values of the SOI (i.e.. El Nino events) wereassociated with increased precipitation in the twoChilean localities with long-term data, while La Ninayears (high values of SOI) were associated to below-av-erage precipitation levels (Fig. I). The Cross Correla-

tion Function yielded a significant negative relationshipat interval 0 between the Southern Oscillation Indexand the yearly cumulative precipitation in the twolocalities considered (Fig. 2). This indicates that at leaststatistically there is evidence that El Nino events bringassociated increased precipitation in the semiarid regionof Chile. However, the rather low (but significant) co-efficients between ENSO and precipitation indicate anot very strong relationship.

Precipitation patterns and rodent ontbreaks

According to Fuentes and Campusano (1985) there ishistorical evidence for eight rodent irruptions in north-ern Chile between 1940 and 1980. Unfortunately, therearc no records of the magnitude of those outbreaks orof the rodent species involved. Nevertheless, the logisticregression model was significant when we consideredthe climatological data from Hurtado {y} = 19.29;DF = 1; p < 0.0001; Fig. 3a), and when using data fromPichasca {yj = lA.lA; DF - 1; p<O.OOOl; Fig. 3b). Inboth cases the regression model was significant, whichpoints out the close relationship between precipitationand probability of occurrence of rodent irruptions.

Precipitation patterns and rodent abundances

The precipitation pattern in northern Chile (as exem-plified by the lllapel climatological station) is highlyseasonal, with rainfall concentrated during wintermonths and a virtual absence during the summer (Fig.4), The summer population abundance of Phyllotisdanvini shows a clear relationship with the total annualprecipitation pattern (Fig. 5). When comparing thetotal annual precipitation levels and the respective pop-

500,-

•g^ 4 0 0 -

I 300

^ 200

s too

0

-

• 1 •' !l •' '•' f l l ' :

i\ fl; kWk •iIr

* 1 * •

-,20

- - 1 0

1942 1953 1964 1975Year

1986 1997

(/5

-20 2

Fig. 1. Flucluations of the aiinuiil cumulative precipitationlevels at two climatological stations in semiarid Chile; Hurtado(closed dots and solid line), and Pichasca (open dots andbroken line), and the Southern Oscillation Index (SOI; dottedline).

E C O C T R A P H Y 22 2 (1999) 215

Cross Correlation Plot Cross Correlation Plot

1.0

0.5 -

- 0.0 - -•-••-I

-0.5 -

-1.0

1

j

1. —I

.

1

1

11.1.1. 1

III"

—-

• III

1

a

h

-

1.0

0.5 -

H 0.0 -

-0.5 -

-30 -20 -10 0Lag

10 20 30-1.0

1

. — •

1

1

-" — ^ ^

LiiLi

.——

1

1

-^ - .—.

.|ili.--™.

1

1

b

, •

|lll|l.l ..II _

-.

1

-30 -20 -10 0Lag

10 20 30

1 2Log annual rainfall (mm)

Fig. 2. Cross Correlation Function between the annual cumulative precipitation levels at two climalological stations in semiaridChile, a) Hurtado station, b) Pichasca station. The horizontal lines encompass the 95"A> confidence Bartlett band ( ± 2/T' -; whereT is the length ofthe time series). The significant negative coefficients at lag = 0 indicate the relationship between ENSO eventsand increased annual precipitation.

Liialion abundances of Phyllotis darwini during the sum-mer, there is a significanlly positive relationship be-tween these variables in both habitat types (xeric:R- = 0.73: D F - I . 9; F = 23.88; p = 0.0009: mesic:R- = 0.66: DF = 1, 9; F = 16.99; p = 0.003).

Discussion

We have shown that: 1) there is a statistical associationbetween El Nino events and increased precipitation inthe semiarid region of northern Chile; 2) the occurrenceof rodent outbreaks in that region is statistically relatedwith the precipitation levels of the same year: 3) themulti-annual patterns of total annual precipitation lev-els and population abundance of Pliylloris darwini dur-ing the summer are positively associated.

The rodent Phyllolis darwini exhibited a seasonaldynamics, with increased population size toward springand summer (Meserve and Le Boulenge 1987. Jimenezet al. 1992). Therefore, we believe that there are reason-ably solid statistical links that imply some form ofcausality from El Nirio events through precipitationpatterns to rodent irruptions. We speculate on thiscausality below.

Because water is a limiting resource in semiarid re-gions, it does not come as a surprise that Increasedprecipitation brings about increased primary produc-tion in the form of both herbage and seeds (Gutierrez etal. 1993). In turn, because precipitation is concentratedduring winter, both reproduction and recruitment oflocal rodents occur during the subsequent spring andsummer months {Meserve and Le Boulenge 1987), atime when rodents may use the increased primary pro-duction that results from plant growth and reproduc-tion {Armesto et al. 1993). This scenario was proposedby Jaksic et al. (1997) for the processes linking precipi-

Log annual rainfall (mm)

Fig. 3. Logistic regression of the occurrence of lodeni out-breaks iu northern Chile (taken from Fuentes and Campusano1985) in years between 1943 and 1980 and the tolal annualrainfall in a) Hurtado station (logistic regression model

exp[-27.19 +{12.15 *.v]

•̂ il + exp[-27.19 +(12.15) *A-]i

and b) Pichasea station logistic regression rnodei

exp[-36.28 + 116.24 *.v]

• " " { ! + exp[ - 36.28 + {16.24) * x]}'

216 ECOORAPHY 22:2

40 r

30

^ 20

_ =

Month

Fig. 4. Mean monthly rainfall in the locality of Illapel for theperiod 1987 1997 (n = 11), indicating ihf strong seasonal pal-Icrn of the precipitation.

tation iind outbreaks in northern Chile, but il has alsobeen invoked as an explanation for rodent irruptions inother semiarid regions, such as those of Mus museulusin south-eastern Australia (Saundcrs and Giles 1977.Mutze et al. 1990), and of Mastomys spp. in centralAfrica (Leirs et al. 1996).

As shown for Phyllotis darwini in our study site, theabundance of this species track 5 closely precipitationlevels. Therefore, a large popuiation increase in face ofunusually high precipitation is not unexpected. Moreinteresting is the fact that Phyllotis darwini outbreaksare so quickly triggered, given that they shoot up withinthe same year of high preeipitation (Jimenez et al. 1992.Meserve et al. 1995). With regard to this same species.

120

•s 80 -

© 40

a.

1

o

1

I

o

1

1

1

n o 220 330 440Annual rainfall (mm)

Fig. 5. Linear regression between the summer population size(minimum number of individuals alive in a study plot of I.Iha) of Phyllotis darwini in the locahty of Aueo and the totalannual rainfall in the locality of Illapel for the period 1987-1997 ( n = l l ) . Closed dots and solid line are for the xcricsubpopulation (R^ = 0.7.1; DF = L 9; F = 23.88; p = 0.0009)and open dots and broken line for the mesic subpopulation{R- = 0.66: D F = I, 9: F = 16.99; p = 0.003).

but in southern Peru. Pearson (1975) had concludedthat at least two years with higher than average precip-itation were needed to trigger an outbreak. We testedthis hypothesis with the long-term climatological datafrom Hurtado and Pichasca by means of a step-by-steplogistic regression analysis, incorporating the cumula-tive precipitation fallen the year before to the eightrodent irruptions analysed. In none of the cases did wefind a significant relationship, which reinforces ourconclusion that it is the precipitation fallen the sameyear which triggers rodent population outbreaks.

In this study we have shown a close relationshipbetween rodent irruptions and precipitation levels. Con-sequently, this result may be construed as supportingthe hypothesis that density-dependent factors do notaffect these population systems and that only density-independent factors drive them (Meserve and Le Bou-lenge 19t̂ 7, Meserve et al. 1995). This hypothesis is theclassic "'climatic control theory"' (Bodenheimer 1938.Andrewartha and Birch 1954), applied to the strongsign of precipitation infiuencing population fluctuationsof rodents in semiarid Chile. However, we have de-tected delayed density-dependent effects operating atdemographic (Lima and Jaksie 1998a) and populationdynamic (Lima and Jaksie 1998b, 1999) levels in P.darwini. These results strongly suggest that populationfluctuations ofthis rodent species are driven by density-dependent factors and rainfall in combination.

In sum. we have determined some statistical relation-ships between environmental factors, rodent outbreaks,and population dynamics of rodent species in semiaridChile. Nevertheless, to disentangle the final causes ofthese population irruptions, we should be able to con-neet specific demographic processes with populationdynamie patterns. In this vein, we are eurrently study-ing the temporal structure and the role of density-dependent and density-independent factors on thebasic demographic processes (survival, reproduction,and recruitment) of the speeies involved in the out-breaks.

Ack/iunieilgeniciiti - M. Lima acknowledges a fellowship fromthe Vicerrectoria Academica of P. Univ. Catolica de Chile forsupporting his graduate studies. This research was funded byFondo Naeiona! de Ciencia y Tecnoloaia grants FONDBCYT296-0034 to ML, 195-1036 to PM. and""196-0319 to FMJ. FMJacknowledges the support of the Mellon Foundations and ofthe Presidential Chair in Science.

550 ReferencesEl Nino SouthernCSIRO. Colling-

Allen, R.. Lindesay. J. and Parker, D. 1996.Oscillation and Climatic Variability, gwood, Australia

Andrewartha. H. G. and Birch. L. C. I9.'i4. The distributionand abundance of animals. LIniv. Chicago Press.

ArmesLo. J. A., Vidiella. P. and Gutierrez, J . ' R , 1993. Plantcommunities of the fog-free coastal desert of Chile: plantstrategies in a fluctuating environment. - Rev. Chil. Hist.Nat. 66; 271-282.

ECOGRAPHY 21:2 217

Bodenheimer, F. S. i938. Problems of animal ecology. -O.xford Univ. Press.

Chatfield. C. 1989. The analysis of time series. - Chapmanand Hall.

Fuentes. E, R. and Campusano. C. 1985. Pest outbreaks andrainfall in the semi-arid region of Chile. J. Arid Environ.8: 67-72.

Gutierrez. J. R. et al. 1993. Structure and dynamics of vegeta-tion in a Chilean arid thorn scrub community. - ActaOecol. 14; 271-285.

Hansson. L. and Henttonen, H. 1988. Rodent dynamics as acommunity process. - Trends Ecol. Evol. 3: 195-200.

Hershkovitz. P. 1962. Evolution of Neotropical cricetine ro-dents (Muridae). with special reference to the Phyllotinegroup. - l-ieldiana Zoo!. 46; I -524.

Jak.sic, F. M., Feinsinger. P. and Jimenez, .1. E. 1993. Along-term study on lhe dynamics of guild structure amongpredatory vertebrates at a semi-arid Neotropical site. -Oikos 67; 87-96.

Jaksie. F. M.. Feinsinger. P. and Jimenez. J, E. 1996. Fxolog-ieal redundancy and long-term dynamics of vertebratepredators in semiarid Chile. - Conserv. Biol. 10; 252-262.

Jaksie, F. M. et al. 1997. A long-term study of vertebratepredator responses to an El Nino (FNSO) disturbance inwestern South Ameriea. - Oikos 78; 341-354.

Jimenez. J, E.. Feinsinger. P. and Jaksie. F, M. 1992. Spa-tiotemporal patterns of an irruption and decline of smallmammals in northcentral Chile. J. Mammal. 73: 356364.

Krebs. C. J. 1989. Ecological methodology. Harper andRow.

Leirs, H. et al. 1996. Forecasting rodent outbreaks in Africa;an ecolotjical basis for Mastomys control in Tanzania. J.Appl. Ecol. 33; 937-943.

Lima. M. and Jaksie, F. M. 1998a. Delayed density-dependeniand rainfall effects on reproductive parameters of an irrup-tive rodent in semarid Chile. Acta Theriol. 43; 225-234.

Lima, M. and Jaksie, F. M. 1998b. Population variabilityamong three small mammal species in the semiaridNeotropics: the role of density-dependent and density-inde-pendent factors. - Ecography 21: 175 180.

Lima. M. and Jaksie. F. M. 1999. Population dynamics ofthree Neotropical small mammals: time series models andthe role of delayed density-dependence in population irrup-tions. - Aust. J. Ecol. 24: in press.

Lima. M.. Marquet. P. A. and Jaksie. F. M. 1996. Extinctionand colonization processes in subpopulations of five

Neotropical small mammal species. - Oecologia 107; 197-203.

Meserve, P. L. and Le Boulenge, E. 1987. Population dynam-ics and ecology of small mammals in the northern Chileansemiarid region. - Fieldiana Zool. New ser. 39: 413 431.

Meserve. P. L. et al. 1995. Heterogeneous responses of smaHmammals to an Bl Nino southern oscillation event innorthcentral semiarid Chile and the importance of theecological scale. - J. Mammal. 76: 580-595.

Muriia, R.. Gonzalez, L. A. and Meserve. P. L. 1986. Popula-tion eeology of Oryzomys longicaudatus phiiippii (Rodentia:Cricetidae) in southem Chile. - J. Anim. Lcol. 55; 53-78.

Murua. P. L., Gonzalez, L. A. and Jofre. C. 1996. Efectos delflorecimiento del arbusto Chusquca quiia Kunth (Poaceae)sobre la demografia de poblaciones de roedores de losbosques templados frios del sur chileno. - Bol. Soc. Biol.Concep. 67: 37-42.

Mutze. G. J.. Veitch, L. G. and Miller, R. B. 1990. Mouseplagues in south Australian eereal-growing areas. 2. Anempirical model for prediction of plagues. - Aust. Wildl.Res. 17: 313 324.

Norrdahl, K. and Korpimaki. E. 1995. Effects of predatorremoval on vertebrate prey populations: birds of prey andsmall mammals. - Oecologia 103; 241-248.

O'Connell. M. A. 1986. Population variability of Neotropiealrodents; influence of body size, habitat, and food habits. -Rev. Chil. Hist. Nat. 59: 169-178.

Pearson. O. P. 1975. An outbreak of mice in the coastal desertof Peru. - Mammalia 39; 375-386.

Pefaur. J. E.. Yanez. J. L. and Jaksie, F. M. 1979. Biologicaland environmental aspects of a mouse outbreak in thesemi-arid region of Chile. - Mammalia 43; 313-322.

Philippi, F. 1879. A plague of rats. - Nature 20: 530.Rodriguez. J. F. 1993. Roedores plaga: un problema perma-

nente en Ameriea Latina y el Caribe. - FAO, OfieinaRegional para America Latina y el Caribe. Santiago, Chile.

Saunders. G. R. and Giles. J. R. 1977. A relationship betweenplagues of house mice, Mus mu.sciihis {Rodentia: Muridae)and prolonged periods of dry weather in south-easternAustralia. Aust. Wildl. Res. 4: 241 248.

Stenseth. N. C. and Ims. R. A. 1993. The biology of lemmings.Academic Press.

Torres-Conireras. H. et al. 1997. Spatiotemporal variability ofrodent subpopulations at a semiarid Neotropieal locality.- J. Mammal. 78: 505-513.

Trexler. J. C. and Travis. J. 1993. Nontraditional regressionanalyses. - Ecology 74; 1629-1637.

218 bCOGRAPHY 22:2 (1999]