BIOLOGICAL CONSERVATION 132 (2006) 88– 97

available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/biocon

Viability of an endangered population of ortolan buntings: The effect of a skewed operational sex ratio

Øyvind Steifetten*, Svein Dale

Department of Ecology and Natural Resource Management, Norwegian University of Life Sciences, P.O. Box 5003, NO-1432 A˚ s, Norway

ARTICLE INFO ABSTRACT

Article history: Population viability analyses (PVAs) are based on basic demographic parameters such as Received 4 August 2005 breeding success, juvenile and adult survival. However, no PVAs have included the opera- Received in revised form tional sex ratio (OSR). The Norwegian population of ortolan buntings (Emberiza hortulana)is 1 March 2006 characterized by a strongly skewed OSR (almost half of all males are unpaired). We Accepted 17 March 2006 assessed the effect of a skewed OSR on annual growth rate (k). Based on empirical data Available online 6 May 2006 the deterministic growth rate was negative (k = 0.82), and mean time to extinction was 21.8 years. Model parameters had to be increased substantially compared to the empirical Keywords: mean values to achieve k = 1.0 (clutch size from 4.25 to 8.06, juvenile survival from 17.6% to Population viability 33.5%, adult survival from 62.9% to 80.8%, and proportion of males breeding from 52.4% to Demography 99.8%). Using low, mean and high empirical parameter estimates, we examined 36 scenar- Operational sex ratio ios for population development. Only five resulted in values exceeding k = 1.0, of which four Unpaired males required an OSR of 1:1. Correlations between inter-annual population change and demo- Ortolan graphic parameters indicated that the proportion of males breeding matched the observed population trend best. Because breeding success, juvenile and adult survival were within the normal range seen among other species of similar body size, we suggest that the skewed OSR is the major factor limiting population growth. Skewed OSRs may be par- ticularly common in small and isolated bird populations, because of female-biased natal dispersal, and should therefore be considered in all PVAs of endangered bird species. Ó 2006 Elsevier Ltd. All rights reserved.

1. Introduction identified in many cases (Newton, 2004), and measures to halt or reverse these impacts have been taken, some populations Population declines and range contractions are currently seen still decrease in numbers, suggesting that other factors could in a number of bird species worldwide (e.g. Tucker and Heath, be preventing population growth. One of these factors could 1994; Fuller et al., 1995; BirdLife International, 2000; Chamber- be intrinsic demographic problems. Thus, an understanding lain and Fuller, 2000; Cumming et al., 2001), and has become of how demography affects population dynamics and popula- an issue of great conservation concern. Human-caused tion growth rates is essential when trying to recover small changes in the environment, such as habitat destruction and declining populations (Bro et al., 2000). and conversion, are likely to be the ultimate causes for many Basic demographic parameters such as breeding success, of these declines, and one of the most serious threats con- juvenile and adult survival have all been suggested as the fronting the long-term survival of many species (Holsinger, proximate mechanisms that underlie the declines of many 2000; Young and Clarke, 2000). However, even though the hu- bird species (Newton, 2004). For example, breeding success man-related factors causing population decline have been has been suggested to be below levels needed to sustain

* Corresponding author: Tel.: +47 64965780; fax: +47 64948502. E-mail address: [email protected] (Ø. Steifetten). 0006-3207/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.biocon.2006.03.016 BIOLOGICAL CONSERVATION 132 (2006) 88– 97 89

viable populations among grassland bird species in North fect on annual population growth. In addition, we performed America (Basore et al., 1986; Herkert et al., 2003), and also a population viability analysis using mean empirical values in among some farmland in Europe (Bradbury et al., order to assess the viability and long-term persistence of the 2000; Siriwardena et al., 2000), and juvenile and adult survival focal population. have been implicated in the declines of several species found in farmland habitat (Thomson et al., 1997; Peach et al., 1999; 2. Methods Siriwardena et al., 1999; Bradbury et al., 2000; Bro et al., 2000; Freeman and Crick, 2003; Robinson et al., 2004). 2.1. Study area and population Although knowledge of breeding success and survival are essential for understanding changes in population size (Para- The remaining population of the ortolan bunting in Norway is dis et al., 1998), other demographic parameters may be just as restricted to an area of nearly 500 km2 in central Hedmark important. One such parameter might be the operational sex County (60°290–60°530N, 11°400–12°180E), although a minor ratio (i.e., the ratio of males to females available for mating at sub-population exists approximately 60 km further south in any one time; see Krebs and Davies, 1996). Dale (2001a) argued Akershus County. The occur in about 40 well-defined that small and isolated bird populations are particularly habitat patches (raised peat bogs, forest clear-cuts on poor prone to skewed sex ratios, because of female-biased natal sand, land being cleared for cultivation, and one forest burn; dispersal, subsequently leading to a high proportion of un- Dale, 2001b), but always close to arable land, which is used for paired males, and hence, reduced potential for population foraging (Dale, 2000; Dale and Olsen, 2002). Dispersal among growth. To our knowledge the operational sex ratio has never habitat patches is frequently observed (Dale et al., 2005), but been specifically assessed in population viability analyses, but dispersal between neighbouring populations is probably lim- a male-biased sex ratio has been seen in several bird popula- ited, as the nearest population is located approximately tions that are fragmented or isolated (reviewed by Dale, 2001a; 250 km further east, in Sweden. The Norwegian population see also Walters et al., 1999; Bayne and Hobson, 2001; Zanette, lies at the edge of the species’ range. 2001; Fraser and Stutchbury, 2004), indicating that the opera- tional sex ratio should be given more attention in studies of 2.2. Data collection declining populations. The ortolan bunting (Emberiza hortulana), a long-distance The demography of the ortolan bunting was monitored annu- migrant , is one of several farmland birds that have ally during the breeding season (May–June) from 1996 to 2004. suffered major population declines in much of Europe (Cramp The first three years of observations were mainly from a single and Perrins, 1994; Kutzenberger, 1994). As a result, it is listed locality, whereas the last six years included the entire known as depleted (i.e., species with unfavourable conservation sta- population in Norway. To avoid biases in demographic param- tus; Tucker and Heath, 1994), but does not meet any European eters between the two periods, only data from the latter was or global IUCN red list criteria (BirdLife International, 2004). used for analyses unless otherwise stated. Data on adult sur- Agricultural changes, such as amalgamation of fields and vival was obtained by capturing males in mist nets with the the consequent loss of breeding habitat together with reduc- aid of playback of song. For individual identification each male tions in crop diversity, have been implicated as the primary was given a unique combination of three colour rings and one causes of the decline (Lang et al., 1990), but the use of mer- numbered metal ring. In any year at least 55% of all adult males cury-treated seed grains might also have played a role (Ree, had colour rings, but for most years the proportion of colour- 1992). In addition, about 50,000 migrant ortolan buntings were ringed males were between 70% and 84% (range 103–133 estimated to be trapped each year in south-west France as a males). Resightings were confirmed through regular visits ‘‘culinary delicacy’’ in the early 1990s (Stolt, 1996), and trap- (1–3 days interval) to all habitat patches which have been used ping of ortolan buntings is still going on today. In Norway by ortolan buntings. In addition, all other potentially suitable the ortolan bunting was a common farmland bird until habitat patches within the study area were visited frequently. around 1950, but has since declined dramatically (Haftorn, Because females behave cryptically throughout most of the 1971), and is now classified as endangered on the Norwegian breeding season and do not aggressively respond to playback red list (Størkersen, 1999). Although the primary causes of de- of song, the capture and subsequent resightings of females is cline have been halted to some extent the population is still difficult. Thus, little data on female survival is available, but decreasing, and current estimates of total breeding popula- is assumed to be similar to male survival. This assumption tion size are 159–166 males, but fewer breeding pairs. More- was based on the study by Dobson (1987) who showed that over, preliminary data on reproductive success, juvenile and there were no significant differences between male and female adult survival suggest that the population should have an an- survival in the majority of the fifteen species he examined, and nual rate of increase of 0.2% (Dale, 2001b). This indicates that a preliminary analysis of our own data showed that male and other factors than classical demographic parameters may be female survival rates did not differ (95% confidence interval: involved in the observed decline, and the large proportion of males 0.593–0.671, n = 581 male-years; females 0.378–0.625, unpaired males may play a role (Dale, 2001b). n = 63 female-years). Resighting probability of returning males In the present six-year study we determined the current was very high. Data on movements between years, showed demographic basis for the decline of the ortolan bunting. that for males recorded in at least two years, the time series Demographic parameters such as breeding success, juvenile were continuous without ‘‘holes’’ (e.g., not recorded in year y, and adult survival, in addition to the operational sex ratio, but recorded in years y 1 and y + 1), in 216 out of 217 cases. were incorporated into a population model to assess their ef- Hence, the use of mark-recapture modelling is not needed. 90 BIOLOGICAL CONSERVATION 132 (2006) 88– 97

The mating status of each male in the population was clas- relying on a single value. Thus, males that were observed in sified into one of the four following categories: (1) males with the same territory for 15 days or more were considered likely a female (i.e. nest found, feeding behaviour, prolonged bouts to be the same individual throughout the breeding season, of warning calls, or seen together with a female most of the and were included in the low estimate. On the other hand, breeding season), (2) males most likely with a female (i.e. re- males that were observed in the same territory for 5–15 days duced singing activity late in the breeding season, and/or could potentially have moved to a different locality after the warning calls, but the female or any sign of nesting activity last observation, increasing the likelihood of being counted was not observed directly), (3) males most likely without a fe- as a new individual. These males were included in the high male (i.e. seen with a female early in the breeding season, but estimate. Single observations of unmarked males, or males no signs of being mated during the egg-laying, incubation or observed less than five days within the same territory were nestling period), and (4) males without a female (i.e. never considered to have moved to other territories and excluded observed with a female, and sang frequently throughout the from the estimate altogether. However, unmarked males breeding season). Although categories 2 and 3 are connected showing feeding behaviour were included in the low estimate with uncertainties regarding male status, most males regardless of the time observed. (90.2%) were assigned to either category 1 or 4. As initial population sizes for the PVA we used the low, During the nestling period as many nests as possible were mean, and high estimates of total population size for 2004. located and both clutch size and brood size were recorded. There was a deficit of females in the population (see Section Each nest was visited on a regular basis until nestlings had 3), which we assumed to be due to female-biased dispersal. left the nest. Nestlings were ringed with one colour ring and In VORTEX one does not have the option to model male and one numbered metal ring, and returning male nestlings were female population size separately. Therefore, the deficit due recaptured and given two additional colour rings. In total 621 to dispersal was modeled as the proportion of females that nestlings were ringed (includes 11 fledged nestlings), but nes- did not breed in a given year, and this approach will lead to tlings from 1996 to 1998 and 2004 (n = 251) were excluded in exactly the same results as if one could model female popula- survival analyses. Thus, juvenile survival was based on 370 tion size separately. Thus, the number of females incorpo- nestlings, of which 185 were assumed to be males. rated into the population size estimates were equivalent to the number of adult males, but those that were assumed to 2.3. Model parameters and underlying assumptions have dispersed were modeled as non-breeders. In calculating the mean proportion of males breeding (equivalent to the In the present study, we explicitly wanted to examine the an- operational sex ratio), we first assumed that all pairs that nual growth rate (k) in order to use it as a measure of popula- were observed in a given year would also breed. Because we tion viability. For this purpose we used the computer program used low and high estimates for both population size and VORTEX (Lacy et al., 2003), which is an individual-based simu- the number of breeding pairs, this resulted in four different lation model for population viability analyses (PVA). Popula- estimates of the proportion of males breeding in a given year. tion growth under the influence of stochastic fluctuations The mean value of each year were then used to calculate the was calculated by random simulations repeated 1000 times, overall mean value used for analyses. whereas the deterministic growth rate was determined from As a measure of breeding success we used clutch sizes re- life table analysis. Although most data needed for analyses corded during 1996–2004. Clutch size was used because we were empirically available, some assumptions had to be made also wanted to incorporate mortality that occurs at the incu- for a few input parameters. First, since no data on maximum bation stage into juvenile survival estimates. In total 190 breeding age, or life expectancy, exists for the ortolan bunting, clutches were included in analyses. Note that although the value was arbitrarily set at 20 years to ensure that all males Cramp and Perrins (1994) cited one study which reported a and females in the population would be modeled as reproduc- clutch size of seven (1.1% of all clutches), the largest clutch tively active until death. Although studies have shown that size observed in the focal population was five. To be conserva- reproductive success decreases with age (e.g. Robertson and tive, we therefore set the maximum clutch size at six, which Rendell, 2001; Laaksonen et al., 2002), we have no reason to be- is also the highest value normally reported for the ortolan lieve that old males cannot mate and reproduce (see Holmes bunting (Cramp and Perrins, 1994). To define the proportion and Austad, 1995). Indeed, the oldest male in the focal popula- of individuals within various cohorts, we used the stable tion that has been observed breeding successfully was nine age distribution calculated by VORTEX for both males and fe- years old. Second, we assumed a 1:1 sex ratio at birth. For males (see Miller and Lacy, 2003 for a definition of and meth- the PVA we also assumed that no catastrophes would occur ods for calculating this distribution). Carrying capacity was during the 100 years of simulation, and that there was no set at 2000 individuals in order to avoid habitat availability inbreeding depression. Although two examples of close kin as a limiting factor. Environmental variation in carrying mating (brother and sister, father and daughter) have been ob- capacity was assumed to be unchanged during the simulated served, nestlings fledged successfully in both cases. time period. Although some habitats are lost to new cultiva- Estimates of total population size were based on the num- tion or other land uses, new breeding habitats are also created ber of marked and unmarked males in each year. Due to the through clear-cutting of forest, and the attractivity of some possibility of counting unmarked males twice because of peat bogs may have increased through cultivation of new extensive within-breeding season and between-patch dis- areas close to bogs. Thus, at present, habitat availability does persal by adult males (Dale et al., 2005), we operated with a not seem to be limiting (Dale, 2001b). Summary of the input low and high estimate in order to avoid the uncertainty of parameters used for analyses are shown in Table 1. BIOLOGICAL CONSERVATION 132 (2006) 88– 97 91

Table 1 – Summary of base parameter values for VORTEX population viability analysis

Mating system: monogamous Age at first reproduction: 1 year for both sexes Age at last reproduction: 20 years for both sexesa Maximum clutch size: 6 Sex ratio at birth: 1:1a Reproductive rates Percent breeding: 52.4 (range 45–63; SD = 4.5)b Exact distribution of clutch size within the population 1 = 0%; 2 = 2%; 3 = 14%; 4 = 43%; 5 = 41%; 6 = 0% Mortality rates (%)c Juvenile (0–1 year) for both sexes: 82.4 (range 73–89; SD = 6.3) Unhatched eggs: 4.7 (range 3–9; SD = 2.4) Egg predation: 0.5 (range 0–2; SD = 1.1) Nestling mortality: 4.0 (range 0–15; SD = 6.1)d Nestling predation: 5.4 (range 0–9; SD = 4.0) Post-fledging mortality: 67.8 (range 64–69; SD = 2.2) Adult (>1 year) for both sexes: 37.1 (range 28–47; SD = 7.1) Males in breeding pool: 100% Initial population size (set to reflect a stable age distribution): 325 (range 318–332) Carrying capacity: 2000a

a Parameter values based on assumptions (see text for further details). b Equivalent to the proportion of males breeding and the expected proportion of females breeding. c Values for mortality rates for both sexes are based on empirical data on male mortality. d Mortality due to starvation or severe weather conditions.

2.4. Sensitivity analysis 3. Results

In addition to modeling population persistence using the Based on the empirical data in Table 1 the deterministic cal- empirical mean demographic rates, we also undertook sensi- culations performed by VORTEX showed that population tivity analysis in which we increased or decreased several growth rate was negative (k = 0.82) even in the absence of variables one by one while all other variables were held con- any random fluctuations. When stochastic events were in- stant. For breeding success we modeled the population with cluded in the model the corresponding growth rate was clutch sizes in steps of one from 1 to 6. Juvenile and adult sur- 0.81. This resulted in a probability of extinction within 100 vival were stepwise changed by three percentage points from years of 1.0, and mean time to extinction was 21.8 years the empirical mean value, with extremes of 6% below and (range: 21.6–22.1 years). 15% above, whereas the proportion of males breeding was de- The effect of increasing breeding success to a mean clutch creased to 30% and increased up to 100%. We also calculated size of six was only of minor importance to population the specific value of each parameter that was needed to growth (k = 0.91; Fig. 1a). For breeding success to achieve achieve k = 1.0. k = 1.0 the mean clutch size had to be raised to 8.06, an almost To examine the possible range of growth rates intrinsic to twofold increase from the empirical mean value of 4.25. the population, the low, mean and high empirical estimates Increasing juvenile survival by 15%, on the other hand, had of juvenile and adult survival, and the proportion of males a stronger positive effect on population growth (k = 0.99; breeding, were modeled for all different combinations of each Fig. 1b). To achieve k = 1.0 the empirical mean value of 17.6% parameter. In addition, we included an operational sex ratio had to be increased to 33.5%. For adult survival a 15% increase of 1:1 because this is expected in most natural populations, resulted in a slightly lower growth rate than juvenile survival making the resulting values comparable to other studies. (k = 0.97; Fig. 1c). The empirical mean value of 62.9% had to be However, low and high estimates of clutch size were omitted increased to 80.8% to reach k = 1.0. The only parameter that from analyses because preliminary results showed that this resulted in a positive growth rate using the maximum preset parameter had least effect on population growth, and also be- values was the proportion of males breeding. By increasing cause annual fluctuations in clutch size was low (4.25 ± 0.28; the proportion to 100% the growth rate was slightly above range 3.75–4.83). We therefore used the mean clutch size as 1.0 (k = 1.01; Fig. 1d). To achieve k = 1.0 the corresponding va- a standard variable in all analyses. A total of thirty-six differ- lue was 99.8%. However, this is still an almost twofold in- ent combinations were analysed. crease from the empirical mean value of 52.4%. Finally, we tested whether the yearly mean empirical val- The current range of possible growth rates based on low, ues of each parameter could explain the observed change in mean and high empirical values of survival and the propor- population size from one year to the next using correlation tion of males breeding, in addition to the value reflecting an analysis. We assumed that the magnitude of the correlation operational sex ratio of 1:1, are shown in Fig. 2. When we used coefficients of the four demographic parameters, would indi- the low estimate of juvenile survival as a constant variable, cate their relative importance in explaining the observed pop- none of the different scenarios were able to bring k above ulation decline. 1.0 (range: k = 0.60–0.89; Fig. 2a). However, when the mean 92 BIOLOGICAL CONSERVATION 132 (2006) 88– 97

1.2 (a) 1.2

) (c) ) λ λ

( 1.1 1.1 (

1.0 1.0

0.9 0.9

0.8 0.8

0.7 0.7 Population growth rate 0.6 Population growth rate 0.6

0.5 0.5 123456 55 60 65 70 75 80 Mean clutch size Adult survival (%)

1.2 1.2 (d)

) (b) λ 1.1 1.1 ( ) λ 1.0 1.0

0.9 0.9

0.8 0.8

0.7 0.7 Population growth rate (

Population growth rate 0.6 0.6

0.5 0.5 10 15 20 25 30 35 30 40 50 60 70 80 90 100 Juvenile survival (%) Breeding males (%)

Fig. 1 – Estimated population growth rates for different values of: (a) mean clutch size, (b) juvenile survival, (c) adult survival, and (d) the proportion of males breeding. Open circles show the deterministic growth rate, and filled circles show the stochastic growth rate with ±SD resulting from the simulations. The arrows indicate growth rates based on empirical mean values.

value of juvenile survival was used as a constant variable we tive, and that the population is likely to go extinct within a found one scenario in which k was >1.0 (range: k = 0.70–1.10; short period of time. One important reason to make deter- Fig. 2b). For the high estimate of juvenile survival the corre- ministic projections is to obtain a general idea of whether sponding number of scenarios that exceeded k = 1.0 was four, or not the rates of reproduction and survival are minimally in which three were associated with scenarios that included adequate to allow for population growth (Miller and Lacy, an operational sex ratio of 1:1, and only one that was associ- 2003). The present results suggest that the current causes of ated with values based on empirical observations (range: the decline are more likely an effect of limiting demographic k = 0.78–1.26; Fig. 2c). rates, than by stochastic fluctuations alone. Among the There were no relationships between inter-annual popula- demographic parameters investigated, the operational sex ra- tion change and clutch size (Spearman rank correlation: tio had the largest impact on population growth, suggesting rs = 0.20, n =6, p = 0.65), juvenile survival (rs = 0.37, n =6, that the male-biased sex ratio observed in the focal popula- p = 0.41), and adult survival (rs = 0.37, n =6,p = 0.41). However, tion is the major force driving the decline. We stress the the proportion of males breeding contributed to explaining importance of including operational sex ratios in population the observed population trend, although not significantly so studies of declining bird species, in particular those that have due to small sample sizes (rs = 0.60, n =6,p = 0.18). This indi- small and isolated populations (Dale, 2001a). cates that the observed changes in population size from one year to the next (mostly decline) is more likely an effect of a 4.1. Breeding success skewed operational sex ratio, than by any of the other three demographic parameters. 4.1.1. Clutch size The mean clutch size of 4.25 is close to the maximum num- 4. Discussion ber of five observed in the focal population, and also close to mean clutch sizes of the ortolan bunting reported elsewhere The ortolan bunting, in common with a number of other pas- in Europe (see Cramp and Perrins, 1994). In addition, the ob- serine species that occur on farmland, has experienced large served distribution of clutch sizes matches to some degree population declines. Analyses of long-term field data from those found in other studies (see Cramp and Perrins, 1994). Norway showed that the deterministic growth rate was nega- Compared to other closely related species with a similar BIOLOGICAL CONSERVATION 132 (2006) 88– 97 93

Fig. 2 – Estimated population growth rates expressed as a function of adult survival and the proportion of males breeding for three estimates of juvenile survival: (a) low estimate (LE), (b) mean estimate (ME), and (c) high estimate (HE). The same three estimates are used for both adult survival and the proportion of males breeding, but for the latter the maximum value (MAX) is also given. All estimates are based on empirical observations.

population trend, i.e. reed bunting (E. schoeniclus), yellow- mean clutch size had to be raised to 8.06, a value that has hammer (E. citrinella), and corn bunting (E. calandra), the never been observed in the ortolan bunting. Correlation mean clutch size in our study was similar or even higher analysis also showed that there was no relationship between (Yom-Tov, 1992; Bradbury et al., 2000; Siriwardena et al., inter-annual population change and clutch size, again sug- 2000), indicating that clutch size is within the normal range gesting that fecundity does not contribute to the decline. seen among other bunting species. Also, an increase of the mean clutch size to the maximum number of six failed to 4.1.2. Nest mortality move k above 1.0, further supporting the view that female Mortality that occurs at the incubation and nestling stages fecundity is already high and unlikely to be the primary may be a significant factor determining breeding success and cause of the decline. Furthermore, to achieve k = 1.0 the the subsequent number of new recruits into the population 94 BIOLOGICAL CONSERVATION 132 (2006) 88– 97

(Ricklefs, 1969). High mortality rates have been found in sev- ably high, and therefore most likely not the major limiting eral studies of both grassland and forest-breeding birds in demographic parameter, we found no relationship between North America (Anders et al., 1997; King et al., 1996; Porneluzi inter-annual population change in abundance and juvenile and Faaborg, 1999; Pietz and Granfors, 2000; Herkert et al., return rates, indicating that population trend is unaffected 2003), and losses up to 80–90% is typical of many neotropical by changes in juvenile survival rates. Moreover, Weatherhead migrants occupying fragmented, urban, or agricultural land- and Forbes (1994) found that most migratory species that scapes (Robinson and Wilcove, 1994; Brawn and Robinson, exhibited high juvenile returns were from isolated popula- 1996). Low breeding success until fledging has also been sug- tions, although extrapolating these results to the study popu- gested to have driven the decline in the linnet (Carduelis canna- lation should be taken with great caution. Nonetheless, it is bina) in Britain (Siriwardena et al., 2000), and of being too low to possible that the juvenile survival estimate found in this maintain a stable population of the on lowland study could be somewhat higher, partly because a substantial farmland (Bradbury et al., 2000). Our data, however, show that number of ortolan buntings are trapped each year in France only a small proportion of eggs and nestlings are lost to either during migration (Stolt, 1996). Although both adults and juve- predation, starvation or severe weather conditions (mainly niles are susceptible to trapping, juveniles might be more na- heavy rainfall) (see Table 1). Although a sampling bias (i.e. local ive than adults, thus experiencing higher mortality. differences in predation or weather conditions) could have oc- curred due to small sample sizes during the first five years of 4.3. Adult survival study (13–37% of all nests were found), in 2004 we found be- tween 68–73% of all nests, and the different mortality rates Compared to other passerines of similar body size (both sed- were comparable with those in previous years. We thus argue entary and migratory species) the mean adult survival rate that the mortality rates shown in Table 1 reflects the true range (62.9%) was high (see Botkin and Miller, 1974; Sæther, 1989; of the focal population, and that they are normal or even below Dobson, 1990; Johnston et al., 1997; Siriwardena et al., 1998; values usually seen in ground-nesting species (see Ricklefs, Siriwardena et al., 1999; Peach et al., 2001 for reviews). How- 1969). Because mortality is already low the opportunity for ever, the compared data sets are mostly based on mark- any substantial increase in survival is absent, and an increase recapture studies faced with the problems of distinguishing in survival will therefore have only a marginal effect on popu- between permanent emigration and mortality, and may thus lation growth rate. underestimate survival rates. Although one can argue that we are faced with the same problems, data on movements be- 4.2. Post-fledging mortality tween years showed that we had an almost 100% resighting probability (Dale et al., 2005). Nonetheless, this suggests that Our mean estimate of post-fledging mortality (67.8%), here adult survival is unlikely to be the limiting factor for popula- defined as mortality that occurs after fledging until the return tion growth. This is further supported by the finding that to as first year breeder, is substantially lower compared to many achieve a positive growth rate in the population, mean adult other species of similar body size (see Weatherhead and For- survival had to be raised above 80.8%. To our knowledge such bes, 1994). Also, when we compared total juvenile mortality an estimate has never been recorded for a long-distance mi- (i.e. mortality imposed at the incubation stage and onwards grant passerine of small body size. Moreover, tropical passe- until the return as first year breeder) with the same dataset, rines, which are thought to have a higher survival rate than our estimate was still lower for all but two cases (n = 65). It species in northern temperate regions, rarely have survival should be noted, however, that natal dispersal can be exten- rates above 80% (Johnston et al., 1997; Peach et al., 2001). sive, thus some locally studied populations may have overes- The fact that there was no relationship between inter-annual timated mortality rates (Weatherhead and Forbes, 1994). population change and adult survival again show that this Although Siriwardena et al. (1998) found higher first-year sur- parameter is unlikely to have driven the recent decline. vival rates in a study of twenty-eight passerines, these esti- mates were based on survival rates of young birds ringed 4.4. Operational sex ratio after independence, and excludes nestling mortality and mor- tality that occurs immediately after fledging. Thus, their val- One striking feature of the focal population was the skewed ues cannot be compared directly with ours. It seems that operational sex ratio with almost twice as many males as fe- our estimates of juvenile mortality are well within, and even males. Several explanations may account for this apparent below, the range normally seen in small passerines. As such, discrepancy. First, there may be a skewed sex ratio at birth. an increase of total juvenile survival from 17.6% to 33.5%, Numerous studies have shown that individual birds make which is required to achieve k = 1.0, is probably unrealistic adjustments of primary sex ratios that appear to be adaptive and not normal for small long-distance migrant passerines. (e.g. Komdeur et al., 1997; Bradbury and Blakey, 1998; Nager A study by DiQuinzio et al. (2001) on the saltmarsh sharp- et al., 1999; Whittingham and Dunn, 2000), but this does not tailed sparrow (Ammodramus caudacutus) found annual return affect the population-level sex ratio. Furthermore, in the rates for juveniles ranging from 0% to 44%, hence showing study population a male-biased sex ratio would be maladap- that survival may be high in some years, but they argue that tive, because there already is an excess of reproductive males. the mean values (11.6–29.7% depending upon year and site) We therefore believe that the assumption of a 1:1 sex ratio at fall at the upper end of documented return rates for migrants birth is reasonable, and that the skewed operational sex ratio and well within typical return rates for residents. In support is due to some other cause. Second, adult mortality could dif- of the view that juvenile survival in our population is reason- fer between sexes due to differences in ecology (Siriwardena BIOLOGICAL CONSERVATION 132 (2006) 88– 97 95

et al., 1998). Although slightly higher mortality rates have ing stochastic growth rate was 0.95. Therefore, it is likely that been found for females than males in some species, very also juvenile survival may have an effect on the population few species-specific examples have been statistically signifi- development in the ortolan bunting. Sensitivity analyses cant (Dobson, 1987; Siriwardena et al., 1998). Moreover, even showed that when the proportion of males breeding was set if there is a sex-biased mortality in the focal population, it at 100%, four out of five scenarios required juvenile survival is unlikely to produce a sex ratio with almost twice as many to be set at the high estimate (27%) to achieve k > 1.0. Thus, males as females, which is the current situation. Further- for long-term stable population dynamics it seems likely that more, sex-biased mortality is expected to be more common an increase in both parameters are needed. Nonetheless, in in sexually size-dimorphic species than in sexually size- this study we have shown that a skewed operational sex ratio monomorphic species (Ellegren et al., 1996) such as the orto- has great impact on population viability. To our knowledge no lan bunting. A third possibility, proposed by Dale (2001a),is studies have used this as a model parameter in the evaluation that a male-biased sex ratio may be the consequence of fe- of population declines. One study by Cooper and Walters male-biased natal dispersal. In small and isolated popula- (2002) suggested that patch isolation was responsible for the tions some females may disperse away from the local high proportion of unpaired males in fragmented habitat, population through natal dispersal, subsequently being lost and related this to a lack of female immigration, which paral- from the breeding pool. As a result, this will lead to a skewed lels the situation for the small and isolated Norwegian popu- operational sex ratio with a high proportion of unpaired lation of ortolan buntings, where female emigration is males. A similar trend has been observed among other spe- probably much higher than female immigration. Also, Brook cies with a high degree of isolation (see Dale, 2001a). In the et al. (2000) argued that sex ratio was an important determi- ortolan bunting, we found that only 52% of all males paired nant for correctly predicting extinction risks between differ- with a female, and such a value is likely to be detrimental ent PVA packages. They pointed out that matrix-based for total reproductive output and the subsequent recruitment packages do not consider the possibility of a temporarily of new individuals into the population. Our data also showed skewed sex ratio, and may thus underestimate extinction that the proportion of males breeding had a large influence on risks compared to individual-based packages (e.g. VORTEX) population growth rate. Increasing the number of males that which do. They stress the importance of incorporating demo- breed to 100%, which is usually assumed to be the case in graphic uncertainty in the sex ratio in PVAs. Nevertheless, it most natural populations (Dale, 2001a), resulted in a stable seems that the operational sex ratio has been overlooked in growth rate (k = 1.01). Also, sensitivity analyses showed that many population studies in birds, and that skewed opera- this maximum value was needed to achieve k > 1.0 in four tional sex ratios may be particulary common in small and iso- of the five scenarios that gave a positive growth rate. Finally, lated bird populations because of female-biased natal we found a positive trend in the relationship between inter- dispersal. Future studies of declining species should therefore annual population change and the proportion of males breed- incorporate data on operational sex ratios, and analyse the ing. These results suggest that the skewed operational sex relative contribution of breeding success, survival and opera- ratio, through the proportion of males breeding, could be an tional sex ratios on population viability. important determinant of population growth rate in the focal population, and hence, population viability. 6. Implications for conservation

5. Conclusions The operational sex ratio is an important quality of any pop- ulation, and directly affects the reproductive output and the All PVA models are more or less simplified versions of real life, number of new recruits into the population. As such, it is a vi- and thus, susceptible to errors at a number of stages. To ob- tal demographic parameter in the evaluation of population tain meaningful data from any PVA-programme, one has to declines. Although skewed operational sex ratios may arise have extensive knowledge of the focal species’ biology, in due to different causes, we argue that female-biased natal addition to a lot of data, so that it represents the wild popula- dispersal is likely to occur in small and isolated bird popula- tion as accurately as possible. In this study we think that both tions (see also Dale, 2001). This means that because small criteria are fulfilled. Although some assumptions have been populations are usually also small in range and size, any fe- made in the model, most parameters (at least the critical male that disperse through natal dispersal has the potential ones) were based on extensive field data. We therefore believe of being lost from the breeding pool. Furthermore, because that the simulations reflect the demographic situation of the of isolation there will be limited immigration of females (or focal population well. At least for two of the four parameters none depending upon the degree of isolation). Thus, the num- investigated (i.e. breeding success and adult survival) it seems ber of females that emigrates will exceed the number that reasonable to conclude that they are not the cause of the cur- immigrates, leaving the population with an operational sex rent decline, because the estimated values falls within, or ratio that is male-biased. As a result, the reproductive output even above, the range seen among other species of similar of the population will decrease (equivalently to the opera- body size. We suggest that the proportion of males breeding tional sex ratio), which again will reduce population size. This is the principal factor limiting population growth. Even so, will subsequently make the population more vulnerable to increasing the proportion of males breeding to 100% only re- stochastic events, which in turn may lead to extinction. sulted in k being slightly above 1.0, thus making the popula- Conservation actions designed to restore the natural opera- tion potentially unstable in the long run when stochastic tional sex ratio can be classified into two categories: (1) mea- events are taken into consideration. In fact, the correspond- sures to prevent female emigration, and (2) measures to 96 BIOLOGICAL CONSERVATION 132 (2006) 88– 97

increase female immigration. Direct actions to prevent Bradbury, R.B., Kyrkos, A., Morris, A.J., Clark, S.C., Perkins, A.J., females from emigrating seems difficult, especially in highly Wilson, J.D., 2000. Habitat associations and breeding success mobile such as birds. However, creating suitable hab- of on lowland farmland. Journal of Applied Ecology 37, 789–805. itat patches around the population may reduce the number of Brawn, J.D., Robinson, S.K., 1996. Source-sink population females that are lost from the core areas, because these dynamics may complicate the interpretation of long-term patches may act as ‘‘conspecific male traps’’ that attracts census data. Ecology 77, 3–12. females to settle. In the long term, natural selection might in- Bro, E., Sarrazin, F., Clobert, J., Reitz, F., 2000. Demography and the crease female philopatry and reduce dispersal distances, but decline of the grey partridge Perdix perdix in France. Journal of we have no information on the time span involved, or, indeed, Applied Ecology 37, 432–448. if there is enough heritable variation in dispersal behaviour for Brook, B.W., Burgman, M.A., Frankham, R., 2000. Differences and congruencies between PVA packages: the importance of sex selection to act on. In highly isolated populations (which is the ratio for predictions of extinction risk. Conservation Ecology case for the Norwegian population) natural immigration of [online] http://www.consecol.org/vol4/iss1/art6/. females is not likely to occur. The only conservation measure Chamberlain, D.E., Fuller, R.J., 2000. Local extinctions and available then is the introduction of females from other changes in species richness of lowland farmland birds in populations. However, such actions are connected with uncer- England and Wales in relation to recent changes in tainties as to whether females will stay in the site of introduc- agricultural land-use. Agriculture, Ecosystems and Environment 78, 1–17. tion, and whether the females will return in future years. Cooper, C.B., Walters, J.R., 2002. Experimental evidence of dis- Moreover, this is still just a temporary solution, and will have rupted dispersal causing decline of an Australian passerine in no long-term effect on the operational sex ratio. In summary, fragmented habitat. Conservation Biology 16, 471–478. the scope for management actions to offset the negative ef- Cramp, S., Perrins, C.M., 1994. The birds of the Western fects of female emigration from small populations seems to PalearcticBuntings and New World Warblers, vol. IX. Oxford be disappointingly small. Thus, management actions need to University Press, Oxford. be implemented before the population size decreases below a Cumming, E.E., Hobson, K.A., Van Wilgenburg, S.L., 2001. Breeding bird declines in the boreal forest fringe of western Canada: threshold value where this kind of Allee effect starts to operate. Insights from long-term BBS routes. Canadian Field-Naturalist 115, 425–435. Acknowledgements Dale, S., 2000. The importance of farmland for Ortolan Buntings nesting on raised peat bogs. Ornis Fennica 77, 17–25. We thank V. Bunes, P. Christiansen, J.P. Cygan, A.K. Darrud, T. Dale, S., 2001a. Female-biased dispersal, low female recruitment, unpaired males, and the extinction of small and isolated bird Granerud, H. Gregersen, B. Hessel, H. Johansen, A. Lamb- populations. Oikos 92, 344–356. rechts, B. Lelaure, A. Lunde, N. Manceau, A.K. Narmo, B.F.G. Dale, S., 2001b. Causes of population decline of the Ortolan Olsen, T. Olstad, T. S. Osiejuk, K. Ratynska, T.I. Starholm, C. Bunting in Norway. In: Tryjanowski, P., Osiejuk, T.S., Kupczyk, Sunding and K. Sørensen for assistance in the field, and J. M. (Eds.), Bunting Studies in Europe. Bogucki Wydawnictwo Swenson for comments on the manuscript. The study was Naukowe, Poznan, pp. 33–41. financially supported by Selberg legacy/WWF Norway, Mr. Dale, S., Olsen, B.F.G., 2002. Use of farmland by Ortolan Buntings and Mrs. Sørlie’s Foundation, the Norwegian Directorate for (Emberiza hortulana) nesting on a burned forest area. Journal fu¨ r Ornithologie 143, 133–144. Nature Management and the Environmental Authorities of Dale, S., Lunde, A., Steifetten, Ø., 2005. Longer breeding dispersal Hedmark and Akershus Counties. than natal dispersal in the ortolan bunting. Behavioural Ecology 16, 20–24. DiQuinzio, D.A., Paton, P.W.C., Eddleman, W.R., 2001. Site fidelity, REFERENCES philopatry, and survival of promiscuous Saltmarsh Sharp-Tailed Sparrows in Rhode Island. AUK 118, 888–899. Dobson, A.P., 1987. A comparison of seasonal and annual Anders, A.D., Dearborn, D.C., Faaborg, J., Thompson III, F.R., 1997. mortality for both sexes of fifteen species of common British Juvenile survival in a population of neotropical migrant birds. birds. Ornis Scandinavica 18, 122–128. Conservation Biology 11, 698–707. Dobson, A.P., 1990. Survival rates and their relationship to Basore, N.S., Best, L.B., Wooley, J.B., 1986. Bird nesting in Iowa life-history traits in some common British birds. Current no-tillage and tilled cropland. Journal of Wildlife Management Ornithology 7, 115–146. 50, 19–28. Ellegren, H., Gustafsson, L., Sheldon, B.C., 1996. Sex ratio Bayne, E.M., Hobson, K.A., 2001. Effects of habitat fragmentation adjustment in relation to paternal attractiveness in a wild bird on pairing success of ovenbirds: importance of male age and population. Proceedings of the National Academy of Sciences floater behaviour. AUK 118, 380–388. of the USA 93, 11723–11728. BirdLife International, 2000. Threatened Birds of the World. Lynx Fraser, G.S., Stutchbury, B.J.M., 2004. Area-sensitive forest birds Edicions and BirdLife International, Barcelona and Cambridge. move extensively among forest patches. Biological BirdLife International, 2004. Birds in Europe: Population Conservation 118, 377–387. Estimates, Trends and Conservation Status. BirdLife Freeman, S.N., Crick, H.Q.P., 2003. The decline of the Spotted International, Cambridge. Flycatcher Muscicapa striata in the UK: an integrated Botkin, D.B., Miller, R.S., 1974. Mortality rates and survival of birds. population model. IBIS 145, 400–412. American Naturalist 108, 181–192. Fuller, R.J., Gregory, R.D., Gibbons, D.W., Marchant, J.H., Wilson, Bradbury, R.B., Blakey, J.K., 1998. Diet, maternal condition, J.D., Baillie, S.R., Carter, N., 1995. Population declines and range and offspring sex ratio in the zebra finch, Poephila contractions among lowland farmland birds in Britain. guttata. Proceedings of the Royal Society of London B Conservation Biology 9, 1425–1441. 265, 895–899. Haftorn, S., 1971. Norges Fugler. Universitetsforlaget, Oslo. BIOLOGICAL CONSERVATION 132 (2006) 88– 97 97

Herkert, J.R., Reinking, D.L., Wiedenfeld, D.A., Winter, M., Pietz, P.J., Granfors, D.A., 2000. Identifying predators and fates of Zimmerman, J.L., Jensen, W.E., Finck, E.J., Koford, R.R., Wolfe, grassland passerine nests using miniature video cameras. D.H., Sherrod, S.K., Jenkins, M.A., Faaborg, J., Robinson, S.K., Journal of Wildlife Management 64, 71–87. 2003. Effects of prairie fragmentation on the nest success of Porneluzi, P.A., Faaborg, J., 1999. Season-long fecundity, survival, breeding birds in the midcontinental United States. and viability of Ovenbirds in fragmented and unfragmented Conservation Biology 17, 587–594. landscapes. Conservation Biology 13, 1151–1161. Holmes, D.J., Austad, S.N., 1995. The evolution of avian Ree, V., 1992. Buskspurver. In: Hogstad, O. (Ed.), Norges senescence patterns: implications for understanding primary dyr – Fuglene No. 4.. J.W. Cappelen Forlag, Oslo, pp. 225–250. aging processes. American Zoologist 35, 307–317. Ricklefs, R.E., 1969. An analysis of nestling mortality in birds. Holsinger, K.E., 2000. Demography and extinction in small Smithsonian Contributions to Zoology 9, 1–48. populations. In: Young, A.G., Clarke, G.M. (Eds.), Genetics, Robertson, R.J., Rendell, W.B., 2001. A long-term study of Demography and Viability of Fragmented Populations. reproductive performance in tree swallows: the influence of Cambridge University Press, Cambridge, pp. 55–74. age and senescence on output. Journal of Ecology 70, Johnston, J.P., Peach, W.J., Gregory, R.D., White, S.A., 1997. Survival 1014–1031. rates of tropical and temperate passerines: a Trinidadian Robinson, S.K., Wilcove, D.S., 1994. Forest fragmentation in the perspective. American Naturalist 150, 771–789. temperate zone and its effects on migratory songbirds. Bird King, D.I., Griffin, C.R., Degraaf, R.M., 1996. Effects of Conservation International 4, 233–249. clearcutting on habitat use and reproductive success of the Robinson, R.A., Green, R.E., Baillie, S.R., Peach, W.J., Thomson, Ovenbird in forested landscapes. Conservation Biology 10, D.L., 2004. Demographic mechanisms of the population 1380–1386. decline of the song trush Turdus philomelos in Britain. Journal of Komdeur, J., Daan, S., Tinbergen, J., Mateman, C., 1997. Extreme Animal Ecology 70, 1014–1031. adaptive modification in sex ratio of the Seychelles warbler’s Siriwardena, G.M., Baillie, S.R., Wilson, J.D., 1998. Variation in the eggs. Nature 385, 522–525. survival rates of some British passerines with respect to their Krebs, J.R., Davies, N.B., 1996. An Introduction to Behavioural population trends on farmland. Bird Study 45, 276–292. Ecology, third ed. Blackwell Science Ltd., Oxford. Siriwardena, G.M., Baillie, S.R., Wilson, J.D., 1999. Temporal Kutzenberger, H., 1994. Ortolan Bunting. In: Tucker, G.M., Heath, variation in the annual survival rates of six granivorous birds M.F. (Eds.), Birds in Europe: their Conservation Status, BirdLife with contrasting population trends. IBIS 141, 621–636. Conservation Series No. 3. BirdLife International, Cambridge, Siriwardena, G.M., Baillie, S.R., Crick, H.Q.P., Wilson, J.D., 2000. The pp. 432–433. importance of variation in the breeding performance of Laaksonen, T., Korpima¨ki, E., Hakkarainen, H., 2002. Interactive seed-eating birds in determining their population trends on effects of parental age and environmental variation on the farmland. Bird Study 45, 276–292. breeding performance of Tengmalm’s owls. Journal of Animal Stolt, B.O., 1996. Ortolansparven jagas i Frankrike. Fauna og Flora Ecology 71. 91, 30–33. Lacy, R.C., Borbat, M., Pollak, J.P., 2003. VORTEX: A Stochastic Størkersen, Ø.R., 1999. Ny nasjonal rødliste over truete fuglearter. Simulation of the Extinction Process. Version 9. Chicago Va˚r Fuglefauna 22, 149–155. Zoological Society, Brookfield, IL. Sæther, B.-E., 1989. Survival rates in relation to body weight in Lang, M., Bandorf, H., Dornberger, W., Klein, H., Mattern, U., 1990. European birds. Ornis Scandinavica 20, 13–21. Breeding distribution, population development and ecology of Thomson, D.L., Baillie, S.R., Peach, W.J., 1997. The demography the Ortolan (Emberiza hortulana) in Franconia. O¨ kologie der and age-specific annual survival of song thrushes during Vo¨gel 12, 97–126. periods of population stability and decline. Journal of Animal Miller, P.S., Lacy, R.C., 2003. VORTEX: A Stochastic Simulation Ecology 66, 414–424. of the Extinction Process. Version 9 User’s Manual. Tucker, G.M., 1994. Birds in Europe: Their Conservation Status. Conservation Breeding Specialist Group (SSC/IUCN), Apple BirdLife Conservation Series No. 3. BirdLife International, Valley, MN. Cambridge. Nager, R.G., Monaghan, P., Griffiths, R., Houston, D.C., Dawson, R., Walters, J.R., Ford, H.A., Cooper, C.B., 1999. The ecological basis of 1999. Experimental demonstration that offspring sex ratio sensitivity of brown trecreepers to habitat fragmentation: a varies with maternal condition. Proceedings of the National preliminary assessment. Biological Conservation 90, 13–20. Academy of Sciences of the USA 96, 570–573. Weatherhead, P.J., Forbes, M.R.L., 1994. Natal philopatry in Newton, I., 2004. The recent declines of farmland bird populations passerine birds: genetic or ecological influences? Behavioural in Britain: an appraisal of causal factors and conservation Ecology 5, 426–433. actions. IBIS 146, 579–600. Whittingham, L.A., Dunn, P.O., 2000. Offspring sex ratios in tree Paradis, E., Baillie, S.R., Sutherland, W.J., Gregory, R.D., 1998. swallows: females in better condition produce more sons. Patterns of natal and breeding dispersal in birds. Journal of Molecular Ecology 9, 1123–1129. Animal Ecology 67, 518–536. Yom-Tov, Y., 1992. Clutch size and laying dates of three species of Peach, W.J., Siriwardena, G.M., Gregory, R.D., 1999. Long-term buntings Emberiza in England. Bird Study 39, 111–114. changes in over-winter survival rates explain the decline of Young, A.C., Clarke, G.M., 2000. Genetics, Demography and reed buntings Emberiza schoeniclus in Britain. Journal of Viability of Fragmented PopulationsConservation biology Applied Ecology 36, 798–811. series, vol. 4. Cambridge University Press, Cambridge. Peach, W.J., Hanmer, D.B., Oatley, T.B., 2001. Do southern African Zanette, L., 2001. Indicators of habitat quality and the reproductive songbirds live longer than their European counterparts? output of a forest songbird in small and large fragments. Journal OIKOS 93, 235–249. of Avian Biology 32, 38–46.