Received: 9 December 2016 | Revised: 23 August 2017 | Accepted: 30 August 2017 DOI: 10.1111/geb.12661

RESEARCH PAPER

Apparent annual survival estimates of tropical songbirds betterreflectlifehistoryvariationwhenbasedonintensive field methods

Thomas E. Martin1 | Margaret M. Riordan2 | Rimi Repin3 | James C. Mouton2 | William M. Blake2

1U.S. Geological Survey, Montana Cooperative Wildlife Research Unit, Abstract University of Montana, Missoula, Montana Aim: Adult survival is central to theories explaining latitudinal gradients in life history strategies. 2 Montana Cooperative Wildlife Research Life history theory predicts higher adult survival in tropical than north temperate regions given Unit, University of Montana, Missoula, Montana lower fecundity and parental effort. Early studies were consistent with this prediction, but 3Sabah Parks, Research and Education, Kota standard-effort netting studies in recent decades suggested that apparent survival rates in temper- Kinabalu Sabah, Malaysia ate and tropical regions strongly overlap. Such results do not fit with life history theory. Targeted marking and resighting of breeding adults yielded higher survival estimates in the tropics, but this Correspondence approach is thought to overestimate survival because it does not sample social and age classes Thomas E. Martin, Montana Cooperative Wildlife Research Unit, University of with lower survival. We compared the effect of field methods on tropical survival estimates and Montana, Missoula, MT 59812, U.S.A. their relationships with life history traits. Email: [email protected] Location: Sabah, Malaysian Borneo. Present address William M. Blake, MPG Ranch, Missoula, Time period: 2008–2016. MT 59801, U.S.A. Major taxon: Passeriformes. Funding information National Science Foundation, Grant/Award Methods: We used standard-effort netting and resighted individuals of all social and age classes of Number: DEB-1241041, DEB-1651283, 18 tropical songbird species over 8 years. We compared apparent survival estimates between IOS-1656120 these two field methods with differing analytical approaches.

Editor: Erica Fleishman Results: Estimated detection and apparent survival probabilities from standard-effort netting were similar to those from other tropical studies that used standard-effort netting. Resighting data veri- fied that a high proportion of individuals that were never recaptured in standard-effort netting remained in the study area, and many were observed breeding. Across all analytical approaches, addition of resighting yielded substantially higher survival estimates than did standard-effort net- ting alone. These apparent survival estimates were higher than for temperate zone species, consistent with latitudinal differences in life histories. Moreover, apparent survival estimates from addition of resighting, but not from standard-effort netting alone, were correlated with parental effort as measured by egg temperature across species.

Main conclusions: Inclusion of resighting showed that standard-effort netting alone can negatively bias apparent survival estimates and obscure life history relationships across latitudes and among tropical species.

KEYWORDS apparent survival, egg temperature, embryonic development, latitudinal variation, life history, lon- gevity, resighting

1386 | VC 2017 John Wiley & Sons Ltd wileyonlinelibrary.com/journal/geb Global Ecol Biogeogr. 2017;26:1386–1397. MARTIN ET AL. | 1387

1 | INTRODUCTION produced similar results to those of Karr et al. (1990), although some studies found somewhat higher rates of apparent adult survival (Table Adult survival is a major component of demography and lifetime repro- 1). Standard-effort netting was later recognized to sample transient and ductive success, and figures prominently in life history theory young individuals that disperse or have higher mortality (Pradel, Hines, (Charlesworth, 1994; Martin, 2002, 2015; Michod, 1979; Roff, 2002; Lebreton, & Nichols, 1997). Subsequently, a time-since-marking Williams, 1966). Life history theory predicts that higher adult survival is approach was used to allow separation of apparent survival estimates in associated with slower life history strategies that include lower the year after first capture versus all subsequent years, under the fecundity and parental effort across taxa (Charnov & Krebs, 1974; assumption that transient and young individuals would disperse or die in Ghalambor & Martin, 2001; Pianka, 1970; Roff, 2002; Williams, 1966). that first year (Johnston, White, Peach, & Gregory, 1997; Pradel et al., Higher survival and slower strategies were long thought to be preva- 1997). Apparent survival estimates increased when such models were lent in the tropics (Dobzhansky, 1950; MacArthur, 1972; Pianka, 1970). used, but not substantially (e.g., Blake & Loiselle, 2013; Brawn, Karr, The perception of life history differences between the tropics and Nichols, & Robinson, 1999; Francis, Terborgh, & Fitzpatrick, 1999; John- north temperate regions was facilitated by early work on birds (e.g., ston et al., 1997). Ultimately, the apparent survival rates across seven Fogden, 1972; Lack, 1954; Snow, 1962; Snow & Lill, 1974; Willis, standard-effort netting studies of 90 total species that used time-since- 1974). However, Karr, Nichols, Klimkiewicz, and Brawn (1990) sug- marking models averaged .62, which was moderately higher but often gested that tropical birds do not have substantially higher apparent sur- overlapped survival estimates of north temperate species (Table 1). vival rates than temperate species. Given the widely documented A perception of substantially higher apparent survival in the tropics lower fecundity and parental effort in tropical species (i.e., Jetz, Seker- (i.e., .74–.89) was based on early studies that resighted colour-marked cioglu, & Bohning-Gaese,€ 2008; Martin, 2015; Martin, Martin, Olson, birds (e.g., Fogden, 1972; Snow, 1962; Willis, 1974). A review of single- Heidinger, & Fontaine, 2000; Martin, Oteyza, Boyce, Lloyd, & Ton, species studies that used colour-banding and resighting indicated that 2015; Martin et al., 2006; Moreau, 1944; Skutch, 1949, 1985; Snow, apparent survival estimates were higher than those from standard- 1962), an absence of higher survival calls into question a central tenet effort netting studies (Sandercock, Beissinger, Stoleson, Melland, & of life history theory. Hughes, 2000). Yet this interpretation was complicated because com- The apparent survival rates for tropical birds estimated by Karr et al. parison of the two approaches was based on different species in differ- (1990) were based on 10 years of capture–recapture data collected with ent environments. In addition, higher apparent survival estimates from standard-effort netting in Panama and rigorous estimation methods that previous resighting studies may be biased because they often focus on accounted for detection probability (i.e., Pollock, Nichols, Brownie, & marking territorial breeding individuals and exclude social and age Hines, 1990). These methods have been used in other locations and classes that may have lower apparent survival and higher dispersal

TABLE 1 Studies of multiple songbird species that encompassed 5 years and estimated annual apparent survival probability with a Cormack–Jolly–Seber model

Used annual Used (A) or actual standard- Used a netting (N) Number Annual apparent effort time-since- Used intervals for Number Location of species survival probability netting marking model resighting estimates of years Reference

(a) Tropical Panama 25 .558 (.022) Y N N A 10 Karr et al. (1990) Costa Rica 11 .56 (.030) Y N N A 9 Blake and Loiselle (2002, 2008) Panama 11 .578 (.028) Y Y N A 20 Brawn et al. (1999) Brazil 31 .586 (.023) Y Y N A 4–10 Wolfe, Stouffer, and Seeholzer (2014) Ecuador 37 .596 (.016) Y Y N N 12 Blake and Loiselle (2013) Ecuador 5 .616 (.096) Y Y N A 7 Parker et al. (2006) Mexico 6 .617 (.048) Y Y N A 16 Ruiz-Gutierrez et al. (2012) Trinidad 17 .653 (.024) Y Y N A 9 Johnston et al. (1997) Peru 14 .676 (.024) Y Y N A 10 Francis et al. (1999) Puerto Rico 9 .680 (.033) Y N N A 18 Faaborg and Arendt (1995) Malaysia 14 .755 (.021) Y Y Y A 5 Martin et al. (2015) Venezuela 18 .764 (.025) Y Y Y A 7 Martin et al. (2015)

(b) North temperate Arizona, USA 16 .530 (.014) Y Y Y A 21 Martin et al. (2015) Maryland, USA 8a .520 (.032)b N N N A 8 Karr et al. (1990) North America 75a .542 (.011)c N N Y A Martin (1995)

N 5 no; Y 5 yes. aWoodpeckers excluded. bBirds were fed supplemental food for 6 months of the year and captured weekly for 3 months at or near feeding stations. cApparent survival rates derived from the literature and based on return rates (simple percentage surviving) instead of a Cormack– Jolly–Seber model. 1388 | MARTIN ET AL.

(Francis et al., 1999; Johnston et al., 1997; Sandercock et al., 2000). At estimates from standard-effort netting alone differed from estimates the same time, birds may learn to avoid nets in the tropics, where based on the addition of resighting in the same populations. We further many remain on permanent territories and possess long-term knowl- examined whether estimates differed among analytical approaches that edge of their territories. Resighting studies provide an opportunity to included annual versus actual capture intervals and inclusion versus identify whether individuals that are never recaptured actually remain exclusion of first-year birds under time-since-marking and Barker on the study area. Yet whether resighting applied to all social and age (1997) models. Finally, we examined whether different estimates from classes captured by standard-effort netting yields higher apparent sur- different field and analytical approaches affected statistical relation- vival rates than standard-effort netting alone is untested. ships between apparent survival estimates and life history traits. Analytical approaches also may influence estimates. For example, time-since-marking models yielded slightly higher apparent survival 2 | METHODS estimates (see above). Most tropical studies estimated survival on the basis of annual detection intervals (Table 1), whereas use of actual time 2.1 | Study area and data collection intervals between netting samples may improve accuracy of estimates (Blake & Loiselle, 2013). In addition, Barker (1997) developed models We studied 18 passerine species (Table 2) on mid-elevation sites that separately estimated recapture and resighting probabilities, which (1,450–1,950 m) in tropical rainforest in Kinabalu Park in Malaysian may improve accuracy of estimates when auxiliary information, such as Borneo (68 N) for 8 years (2009–2016) from early February to mid- resighting, is available. Nonetheless, the influence of these different June. Scientific names of our study species are provided in Table 2. field and analytical methods on the ability of estimates to reflect bio- In 2009, we established a study area of c. 560 ha that was divided logical relationships across species is untested. into seven contiguous plots of roughly equal size, in which we searched We examined the effects of standard-effort netting alone com- for nests. A total of 240 marked net sites in 20 subplots were distrib- pared with standard-effort netting combined with resighting of colour- uted across the seven plots. Each netting subplot consisted of 12 net banded birds on apparent survival estimates and their relationships sites, with nets 25–50 m apart. We sampled each subplot three times with life history traits among 18 species of tropical songbirds in Malay- per breeding season, with 5–6 weeks between sampling occasions. sian Borneo. We colour-banded birds captured in standard-effort net- One subplot was sampled per day, with nets deployed for 6 hr, starting ting to include all social and age classes. Our simultaneous use of both at dawn. One rotation in each 5- to 6-week sampling period takes 20 field approaches allowed us to examine whether apparent survival days of sampling, allowing extra days for skipping rain days or other

TABLE 2 Effective sample sizes quantified by the program MARK for each of 18 species marked through 2016 and including hatch-year birds for standard-effort netting alone versus adding resighting based on annual/actual netting sampling intervals

Species code Common name Scientific name Netting only Plus resighting

BOWH Bornean whistler Pachycephala hypoxantha 252/272 354/453

WTFA White-throated fantail Rhipidura albicollis 140/152 178/223

TESU Temminck’s sunbird Aethopyga temminckii 20/21 33/37

EJFL Eye-browed jungle flycatcher Vauriella gularis 219/247 307/388

WBSH White-browed shortwing Brachypteryx montana 44/45 67/73

SBFL Snowy-browed flycatcher Ficedula hyperythra 204/227 282/400

BWTH Bornean whistling thrush Myophonus borneensis 21/21 41/61

BOFO Bornean forktail Enicurus borneensis 26/28 37/57

OCBU Ochraceous bulbul Alophoixus ochraceus 115/123 159/190

CCYU Chestnut-crested yuhina Yuhina everetti 95/98 137/180

GTBA Grey-throated babbler Stachyris nigriceps 704/902 956/1,394

SULA Sunda laughingthrush Garrulax palliatus 58/60 91/106

CHLA Chestnut-capped laughingthrush Garrulax mitratus 31/31 50/54

TEBA Temminck’s babbler Pellorneum pyrrogenys 99/109 155/190

MWBA Mountain wren-babbler Napothera crassa 47/47 120/176

MOTA Mountain tailorbird Phyllergates cuculatus 25/27 34/47

YBWA Yellow-breasted warbler Seicercus montis 111/115 164/204

MLWA Mountain leaf warbler Phylloscopus trivirgatus 55/56 67/68 MARTIN ET AL. | 1389

purposes. Two people set up and managed nets to ensure short inter- years (a2) in a time-since-marking approach (Pradel et al., 1997) for a vals (c. 20 min) between net checks. All captured birds were banded for each field-method data group, whereas p differed only in field- with a numbered metal band and with unique combinations of three method data groups [a(g, a) p(g)]. In the fourth model, we modelled plastic colour-bands (two bands per leg). Colour-bands were resighted field-method data groups and separated first year of capture from all by seven nest-searchers who visited each nest-searching plot daily subsequent years for both a and p [a(g, a) p(g, a)]. We used Akaike’s throughout the season. T.E.M. and a field supervisor also resighted information criteria, with adjustment for small sample sizes (AICc; Burn- birds across all nest-searching plots each year. Thus, capture and ham & Anderson, 2002) for model selection, and used the program resighting effort was reasonably constant in our 8-year study. RELEASE (Burnham, Anderson, White, Brownie, & Pollock, 1987) to test goodness of fit. Model results are presented in Supporting Information 2.2 | Literature summary Tables S1–S4 and goodness-of-fit tests in Supporting Information Tables S5–S8. In the few cases necessary, we corrected for overdisper- We sought all studies that used standard-effort netting of tropical bird sion (ĉ > 1) on the basis of an adjusted ĉ and based model selection on communities (i.e., multiple species) to provide general background on quasi AICc (QAICc) (Burnham et al., 1987). In cases where time-since- apparent survival rates across a range of tropical studies (Table 1). We marking (g, a) was a top model, we report estimates for the second restricted studies to those in large continuous areas of habitat to mini- time class (a2). mize any restrictions on movements of birds that might occur in frag- Finally, we estimated apparent survival with Barker (1997) models mented habitats. We also restricted studies to those with 5 years of in the program MARK to estimate detection probabilities separately for capture–recapture data. recaptures and resighting. We compared estimates from Barker models For latitudinal comparisons, we used three studies that compiled with Cormack–Jolly–Seber estimates from the same dataset with actual apparent survival data for north temperate species (Table 1). North netting intervals and excluding hatch-year birds. The Barker model has temperate studies included resighting or targeted capture of birds, and seven estimable parameters, but some were not pertinent to our study. thus did not use standard-effort netting alone. In particular, r reflects recovery rate (probability that an individual was found dead and reported); this did not apply to our study and was set 2.3 | Statistical analyses at zero. F and F0 are fidelity probabilities. F is the probability that an individual remains in the study area and available for detection. F0 is Given that our study was focused on the breeding season, we were the probability that, on the basis of auxiliary detection outside the able to identify young of the year (i.e., hatch-year) birds on the basis of study area, an individual has emigrated from the study area, but returns gape colour, plumage characteristics and skull ossification. We analysed and is available for detection in a subsequent sample period. We did data with and without hatch-year birds to test whether estimates not have auxiliary detections outside the study area, and F and F0 are differed. confounded with p,sowesetF to one and F0 to zero. The four remain- We organized encounter data into eight subsets in two field-method ing parameters, S (survival probability), p (recapture probability), R groups. The first field-method group was based solely on captures and (probability that an individual was resighted) and R0 (the probability an recaptures from standard-effort netting alone and included four data animal that dies in a sample interval without being found dead is subsets of annual intervals (a) with and (b) without hatch-year birds ver- resighted alive in that interval before it died), were estimated in the sus the three actual netting intervals (c) with and (d) without hatch-year model. We built eight models with combinations of time-since-marking birds. The second field-method group was based on standard-effort net- 0 (a) or constant (.) structure for S, p, R and R parameters (Supporting ting plus resighting of colour-banded birds and included four data subsets Information Table S9). We again used Akaike’s information criteria, of annual intervals (e) with and (f) without hatch-year birds versus actual with adjustment for small sample sizes (AICc; Burnham & Anderson, netting intervals (g) with and (h) without hatch-year birds. 2002), or QAICc that corrected for overdispersion with an adjusted ĉ, – – We used Cormack Jolly Seber models for live encounter data for for model selection (Supporting Information Table S10). open populations in the program MARK to estimate apparent annual We used paired t tests to test whether datasets differed. We also a adult survival ( ) and detection (p) probabilities (Burnham & Anderson, tested whether parental effort, measured by average 24 h egg temper- 2002; White & Burnham, 1999). For each species, we built four models ature, was correlated with apparent survival estimates from the differ- 5 to compare the two groups of field-method data (g1 standard-effort ent field and analytical approaches. Life history theory predicts that 5 data alone; g2 standard-effort plus resighting data). We applied these parental effort decreases as survival increases (Charnov & Krebs, 1974; models to each of the four subsets of data: annual intervals with and Martin, 2002; Michod, 1979; Williams, 1966), and parental effort, without hatch-year birds and actual netting intervals with and without measured by egg temperature, decreases as apparent survival increases hatch-year birds. We set time intervals as 0.165, 0.165 and 0.67 for among diverse tropical and temperate songbirds (Martin et al., 2015). actual netting intervals. In the first model, we assumed a and p were Egg temperatures were based on extensive measurements and cor- constant [a(.) p(.)], indicating that estimates do not differ between field- rected for embryo age as reported by Martin et al. (2015). We cor- method data groups. In the second model, we tested whether esti- rected for possible phylogenetic effects (Felsenstein, 1985) in mates differed between field-method data groups [a(g) p(g)]. In the correlations with phylogenetic generalized least squares analyses third model, we separated first year of capture (a1) from all subsequent (PGLS). Pagel’s lambda (Pagel, 1992) was not statistically 1390 | MARTIN ET AL.

TABLE 3 Mean (6 1 SE) annual apparent survival plus detection probabilities for all 18 species when based on Cormack–Jolly–Seber models with and without first-year birds when based on standard-effort netting only versus standard-effort netting plus resighting, and estimated on the basis of annual versus actual netting intervals

Apparent survival probability Detection 6 resighting probabilities Field method Standard-effort Standard-effort Standard-effort Standard-effort Resighting (number of species) only (18) plus resighting (18) only (18) plus resighting (18) probability (18)

Annual sample intervals Annual sample intervals

CJS with first-year birds .543 (.062)1,a .783 (.024)1,b .219 (.034)1,c .524 (.034)1,d

CJS without first-year birds .570 (.052)1,a .779 (.020)1,b .227 (.035)1,c .500 (.037)1,d

Actual netting intervals Actual netting intervals

CJS with first-year birds .645 (.039)2,a .765 (.020)1,b .059 (.009)2,c .223 (.023)2,d

CJS without first-year birds .650 (.039)2,a .775 (.022)1,b .069 (.014)2,c .227 (.026)2,d

Barker without first-year birds .747 (.021)2 .056 (.013)3a .525 (0.026)b

CJS 5 Cormack–Jolly–Seber. Note. In addition, survival, detection and resighting probabilities were estimated from Barker models on the basis of actual netting intervals without including first-year birds. 1,2,3Different numbers within a column reflect a statistical difference (p < .05) on the basis of paired t tests, corrected for multiple tests. a,b,c,dDifferent letters within a row reflect a statistical difference (p < .05) on the basis of paired t tests, corrected for multiple tests.

distinguishable from zero in all tests, indicating no or few phylogenetic effort netting alone for the individual species (Figure 1), whether based effects among these species. Therefore, we report statistics from analy- on annual versus actual netting intervals or including versus excluding ses of raw data and provide PGLS results in Supporting Information first-year birds (Table 3). A top model (DAICc or DQAICc < 2.0) indi- Table S11. cated no difference in estimates between standard-effort netting alone versus standard-effort netting plus resighting for two of the 72 com- 3 | RESULTS parisons across the 18 species and four datasets: one dataset for mountain tailorbird and one dataset for Bornean forktail. However, in 3.1 | Apparent survival estimates both cases, another top model indicated differences between field- method data groups (Supporting Information Tables S2 and S4). Top Apparent survival estimates for the 18 species did not differ when models indicated that apparent survival was higher for standard-effort first-year birds were included or excluded (Table 3). The lack of differ- netting alone than when resighting was added for six datasets: three ence in estimates when including versus excluding first-year birds was for white-browed shortwing, two for mountain tailorbird, and one for consistent across the four comparisons: annual sample intervals versus Sunda laughingthrush (Figure 1; Supporting Information Tables S1–S4). actual netting intervals for standard-effort netting only versus Ultimately, however, apparent survival estimates were higher for standard-effort netting plus resighting (Table 3). standard-effort plus resighting than for standard-effort netting alone for 64 of the 72 datasets of the individual species (Figure 1; Supporting 3.1.1 | Annual versus actual netting intervals using Information Tables S1–S4). standard-effort netting alone Apparent survival estimates based on Barker models were slightly, The use of actual netting intervals yielded higher average apparent sur- but significantly, lower than Cormack–Jolly–Seber estimates based on vival estimates than those based on annual sample intervals from resighting data across the 18 species (Table 3). Moreover, slightly lower standard-effort netting alone (Table 3). This difference, however, was estimates were fairly consistent across the 18 species (Figure 1c). Yet attributable to much lower apparent survival estimates for Temminck’s apparent survival estimates based on Barker models were highly corre- sunbird and mountain leaf warbler from annual than actual netting lated (r 5 .95, p < .001, n 5 18) with estimates based on Cormack– intervals under standard-effort netting alone (Figure 1). Average appa- Jolly–Seber models from the same dataset (Figure 1c and Table 3). rent survival did not differ between annual and actual netting intervals Confidence intervals were substantially broader, and uncertainty for the other 16 species, whether based on inclusion (paired t test, greater, for estimates from standard-effort netting alone than when t 5 1.20, p 5.25) or exclusion (paired t test, t 5 1.70, p 5 .11) of first- resighting was added (paired t tests; annual intervals with hatch-year: t year birds. 5 3.7, p 5 .002; annual intervals without hatch-year: t 5 3.7, p 5 .002; actual netting intervals with hatch-year: t 5 7.1, p < .001; actual net- | 3.1.2 Standard-effort netting alone versus addition of ting intervals without hatch-year: t 5 6.3, p < .001). Average breadths resighting of confidence intervals across the 18 species on the basis of standard- Apparent survival estimates were substantially higher when resighting effort netting alone ranged from 0.47 6 0.07 to 0.60 6 0.06 for the of birds was added to standard-effort netting compared with standard- four data sets (annual intervals with and without hatch-year, and actual MARTIN ET AL. | 1391

Standard effort only (a) Annual sample intervals Standard effort + Resighting 1.0

0.8

0.6

0.4

0.2 Apparent survival probability

0.0

BFL FO YU WA EJFL S WTH O SULA TEBA WBA BOWHWTFATESU WBSH B B OCBUCC GTBA CHLA M MOTAYBWAML

(b) Actual netting intervals Standard effort only 1.0 Standard effort + Resighting

0.8

0.6

0.4

0.2 Apparent survival probability

0.0 U A TFA BFL FO EJFL S O SULA TEBA LW BOWHW TESU WBSH BWTHB OCB CCYUGTBA CHLA MWBAMOTAYBWAM

(c) Actual netting intervals Barker estimates 1.0 Cormack-Jolly-Seber estimates

0.8

0.6

0.4

0.2 Apparent survival probability

0.0 A BFL EJFL S WTH SULA TEBA BOWHWTFATESU WBSH B BOFOOCBUCCYUGTB CHLA MWBAMOTAYBWAMLWA

FIGURE 1 Apparent annual survival probability for 18 species of tropical passerines when based on standard-effort netting only versus standard-effort netting plus resighting of birds and excluding first-year birds on the basis of (a) annual or (b) actual netting intervals. (c) Apparent survival estimates were compared between Barker and Cormack–Jolly–Seber models. See Table 2 for names of species 1392 | MARTIN ET AL.

80 (a)

70

60

50

40

30

20 in a subsequent year

10 effort netting that were recaptured

% individuals marked via standard- 0

80 (b) 70

60

50

40

30

20 in a subsequent year

10 effort netting that were re-sighted

% individuals marked via standard- 0

EJFL SBFL SULA TEBA BOWHWTFATESU WBSH BWTHBOFOOCBUCCYUGTBA CHLA MWBAMOTAYBWAMLWA

FIGURE 2 Percentages of individuals, excluding hatch-year birds, that were captured and marked with standard-effort netting and (a) recaptured or (b) resighted at least once in a subsequent year after marking. Grey line 5 mean percentage recaptured in standard-effort net- ting. Black line 5 mean percentage resighted. See Table 2 for names of species netting intervals with and without hatch-year). In contrast, they aver- The average apparent survival rate when resighting was included aged from 0.24 6 0.03 to 0.27 6 0.04 among the four data sets with for our 18 Malaysian species (Table 3) was also substantially greater resighting added to standard-effort netting. The breadth of confidence than any of the north temperate estimates (Table 1). Indeed, standard- intervals based on Barker models did not differ from Cormack–Jolly– effort netting, resighting and target netting were used in a 21-year Seber models for the same dataset (paired t 5 0.09, p 5 .93). study in north temperate Arizona (Martin et al., 2015), which should not underestimate apparent survival. Yet apparent survival estimates for 16 songbird species at this temperate site averaged .53 6 .014 and 3.1.3 | Comparisons with literature were substantially below Malaysian apparent survival rates (Table 1). In The average apparent survival rate for the 10 tropical studies from the contrast, the average estimate for our temperate site was similar to literature that used standard-effort netting alone (Table 1a) did not dif- two other temperate studies, one of which included a summary of fer from the average estimate for our 18 species when based on studies of 75 species, most of which included resighting in estimates 5 5 standard-effort netting alone (F 0.33, p .57 based on annual inter- (Table 1b). vals and excluding hatch-year birds; F5 0.50, p 5 .48 based on actual netting intervals and excluding hatch-year birds). In contrast, the aver- 3.2 | Detection of individuals age apparent survival estimate for our 18 species was significantly higher than the average estimate for the 10 tropical studies from the Differences in apparent survival estimates when adding resighting literature when we included resighting with standard-effort netting (F5 appear to come from an increase in detection of individuals that are 33.5, p < .001 based on annual intervals and excluding hatch-year alive. An average of 20.8 6 3.04% of unique individuals were recap- birds; F5 27.4, p < .001 based on actual netting intervals and excluding tured in a subsequent year by standard-effort netting alone across the hatch-year birds; F5 20.1, p < .001 based on actual netting intervals 18 species (Figure 2a). In contrast, we resighted 48.7 6 2.75% of indi- and excluding hatch-year birds using Barker models). viduals marked through standard-effort netting in subsequent years MARTIN ET AL. | 1393

1.0 (a) Annual sample intervals

0.8

0.6

0.4

Detection probability 0.2

0.0

EJFL SBFL SULA TEBA BOWHWTFATESU WBSH BWTHBOFOOCBUCCYUGTBA CHLA MWBAMOTAYBWAMLWA

1.0 Standard effort only (b) Actual netting intervals Standard effort + Resighting

0.8

0.6

0.4

Detection probability 0.2

0.0 H EJFL SBFL SULA TEBA BOWHWTFA TESU WBS BWTHBOFOOCBUCCYUGTBA CHLA MWBAMOTAYBWAMLWA

FIGURE 3 Estimates of detection probabilities derived from (a) annual sample intervals and (b) actual netting intervals on the basis of data from standard-effort capture–recapture alone or with the addition of resighting of birds banded through standard-effort capture–recapture. See Table 2 for names of species

(Figure 2b). Resighting detected the majority (72.8%, n 5 18) of individ- 3.3 | Correlations with life histories uals recaptured with standard-effort netting, whereas standard-effort We were unable to measure egg temperature for three species (Tem- netting recaptured 11–14% of individuals not detected by resighting. minck’s sunbird, Sunda laughingthrush and chestnut-hooded laughing- thrush). Egg temperatures for the remaining 15 species were not 3.2.1 | Detection probabilities correlated with estimates of apparent survival based on standard-effort Detection probabilities when resighting was included were nearly dou- netting alone (Figure 4a). Mountain wren-babblers are a low- ble those obtained from standard-effort netting alone (i.e., Figure 3a) temperature outlier because they take 5–8 hr off-bouts and allow eggs when we used annual sample intervals (Table 3). When modelling to cool to the ambient temperature. If this low outlier is removed, the actual netting intervals, detection probabilities were three to four times correlations with apparent survival based on standard-effort netting greater with resighting added than when based on standard-effort net- alone are still not statistically significant for all four datasets (r 52.08 ting alone (Figure 3b and Table 3). Actual netting intervals yielded sub- to 2.48, p 5 .08–.78, n 5 14). stantially lower detection probabilities than annual intervals (Table 3). Apparent survival estimates when including resighting with The recapture probability estimated by Barker models was similar to standard-effort netting were correlated with egg temperature in all detection probabilities estimated on the basis of standard-effort netting cases (Figure 4b). Removal of mountain wren-babblers had little effect alone for the same dataset that used actual netting intervals (Table 3). on statistical significance in the four datasets (r 52.57 to 2.61, p This result is consistent with the fact that detection probability is 5.034–.020, n 5 14). Likewise, apparent survival estimates obtained recapture probability for standard-effort netting alone. At the same from Barker models for the dataset including resighting, and based on time, Barker models estimated recapture probabilities that were sub- actual netting intervals while excluding hatch-year birds, were strongly stantially lower than resighting probabilities (Table 3). correlated with average egg temperatures (Figure 4c). Removal of 1394 | MARTIN ET AL. mountain wren-babblers slightly increased statistical significance 36 r = -0.40, P = 0.14 (r 52.73, p 5 .003, n 5 14). (a) 35 Variation in apparent survival estimates was greater when esti- mates were based on standard-effort netting alone (Figure 4a) than 34 when resighting was added (Figure 4b,c). 33

32 4 | DISCUSSION annual intervals, no hatch-year: r = -0.15, P = 0.60 31 annual intervals, with hatch-year: r = -0.38, P = 0.16 netting intervals, with hatch-year: r = -0.47, P = 0.08 The overlap in apparent survival estimates of tropical and temperate 30 songbird species obtained from standard-effort netting is not consist- C) 29 o ent with core tenets of life history theory (i.e., Charnov & Krebs, 1974; 0.4 0.5 0.6 0.7 0.8 0.9 1.0 36 Pianka, 1970; Roff, 2002; Williams, 1966) given the well-documented (b) r = -0.59, P = 0.02* slower life history strategies of tropical species (e.g., Martin, 2015). 35

Standard-effort netting is a widely used technique in the tropics 34 (Table 1), but our results showed that it can greatly underestimate 33 apparent survival of tropical birds (Table 3). Addition of resighting dra- 32 matically reduced overlap in apparent survival estimates between tem- annual intervals, no hatch-year: r = -0.75, P < 0.001* perate and tropical songbirds (Table 1) and corresponded to latitudinal 31 annual intervals, with hatch-year: r = -0.63, P = 0.013* netting intervals, with hatch-year: r = -0.59, P = 0.02* differences in life histories as predicted by theory. 30 Life history traits should also covary with apparent survival among species within the tropics given that variation of some traits can be 29 0.4 0.5 0.6 0.7 0.8 0.9 1.0 quite extensive among tropical species (Martin, 2008; Martin et al., 36 rp = -0.70, P = 0.004* 2015). Yet, apparent survival estimates that were based on standard- Average egg temperature over 24 h ( (c) 35 effort netting alone did not covary with parental effort measured by 34 average egg temperature (Figure 4a). In contrast, apparent survival esti- mates that were based on standard-effort netting plus resighting yielded 33 correlations that were consistent with life history theory (Figure 4b,c). 32 Thus, apparent survival estimates from standard-effort netting alone are 31 not only negatively biased, but also appear to include spurious variation that can obscure biological relationships across species. 30 Our estimates from standard-effort netting were typical of other 29 tropical studies that used this field method. On average, we recaptured 0.4 0.5 0.6 0.7 0.8 0.9 1.0 21% of marked individuals in a subsequent year through standard- Apparent survival probability effort netting (Figure 2a). Such information is rarely reported, but Fran- cis et al. (1999) likewise found that 11–30% of marked individuals of FIGURE 4 Correlations of average 24 hr egg temperature and 14 species were recaptured at least once in a subsequent year in a apparent survival estimates from (a) standard-effort netting alone (Cormack–Jolly–Seber models), (b) standard-effort netting and 10-year study in Peru. Also, Parker, Becker, Sandercock, and Agreda resighting (Cormack–Jolly–Seber models) and (c) standard-effort – (2006) found that a mean of 17% (9 38%) of marked individuals were netting and resighting (Barker models). Points are based on actual recaptured in a subsequent year in a 7-year study in Ecuador. All but netting intervals and excluded hatch-year birds, whereas correla- one of the tropical studies used annual sample intervals (Table 1). Our tions with other apparent survival estimates are displayed as text. Phylogenetic least squares analyses of these correlations are pre- average detection probabilities (.22–.23) derived from standard-effort sented in Supporting Information Table S11 netting and annual sample intervals (Table 3) were similar to those reported in other tropical studies that used standard-effort netting and annual sampling intervals (range .15–.29; Faaborg & Arendt, 1995; survival estimates (Blake & Loiselle, 2013). Even for species with high Johnston et al., 1997; Parker et al., 2006; Ruiz-Gutierrez et al., 2012, as detection probabilities, apparent survival estimates from standard- reported by Sandercock et al., 2000). Likewise, our apparent survival effort netting were substantially lower than estimates from standard- estimates from standard-effort netting did not differ from estimates of effort netting and resighting. This point is emphasized by consideration other similar studies in the tropics (Tables 1 and 3). of the species (grey-throated babbler) with the highest detection prob- The low detection probabilities in our study and other tropical abilities on the basis of standard-effort netting alone (Figure 3) and the studies that were based on standard-effort netting alone yielded appa- largest sample sizes (Table 2). Apparent survival was .12–.14 lower rent survival estimates with wide confidence intervals. Greater preci- when based on standard-effort netting alone than when resighting sion, however, does not necessarily improve the accuracy of apparent was added, whether based on annual or on actual netting intervals MARTIN ET AL. | 1395

(Figure 1). This increase in annual apparent survival from .64 to .77 has Many tropical species are permanent residents and could better learn a major effect on expected lifespan (Snow & Lill, 1974). Thus, apparent which areas (i.e., net sites) to avoid. In addition, birds with higher sur- survival probabilities are substantially underestimated when based on vival generally have larger brains, which allow better memories and standard-effort netting alone regardless of detection rates. greater cognitive function (Sol, Szekely, Liker, & Lefebvre, 2007). The Higher apparent survival estimates from resighting studies were higher survival of tropical birds might reflect larger brains and greater hypothesized to reflect sampling of breeding, territorial birds that do ability to remember the locations and appearance of nets than temper- not include subordinate and young age classes that may disperse or die ate birds. The apparent survival estimates we used here for most tem- at a higher rate (Francis et al., 1999; Johnston et al., 1997; Sandercock perate species were based on inclusion of resighting and should et al., 2000). This possibility seems unlikely in our study for three rea- provide a robust comparison. Yet, careful evaluation of apparent sur- sons. First, apparent survival rates of birds marked via standard-effort vival estimates in temperate birds from standard-effort netting versus netting increased considerably when we added resighting, even though addition of resighting is lacking and needed to assess any relative dif- standard-effort netting did not restrict sampling to individuals in any ferences between regions fully. particular social or age class. Second, apparent survival estimates did Different analytical approaches yielded a few interesting results. not differ statistically when hatch-year birds were excluded or included First, exclusion of hatch-year birds did not lead to different apparent (Table 3). Third, an average of 21% of individuals of each species that survival or detection probabilities (Table 3). Second, use of actual net- were never recaptured were later observed breeding. These observa- ting intervals, as did Blake and Loiselle (2013), increased apparent sur- tions indicate that many birds that were never recaptured are not vival estimates for two species (Temminck’s sunbird and mountain leaf transients. warbler) that had fairly low estimates on the basis of annual intervals. The low proportion of individuals that were recaptured via This result is particularly interesting given that detection probabilities standard-effort netting in subsequent years (Figure 2a) and the result- for both species that were based on annual sample intervals were not ing low detection probabilities (Figure 3) indicate a problem in estimat- particularly low (Figure 3). The lack of difference in estimates for the ing apparent survival probability. Individuals may not be detected other 16 species, but increases in estimates for these two species, sug- because they avoid nets, largely move above net level or move out of gests that use of multiple sample periods and modelling these actual the study area. Individual heterogeneity models attempt to account for periods might be beneficial. Barker models yielded slightly, but consis- such issues by estimating heterogeneous detection probabilities among tently, lower apparent survival estimates that strongly covaried (i.e., r 5 individuals (e.g., Kendall & Nichols, 1995). However, even these models .95) with Cormack–Jolly–Seber estimates from the same data (Table 3). cannot accurately account for net avoidance if avoidance persists Nonetheless, Barker estimates slightly increased the correlation with throughout the life of individuals. The high number of individuals that average egg temperature (Figure 4). Finally, we recommend that future were never recaptured, but we knew to be alive, and even breeding in field studies of apparent survival probability add colour-banding and the netting area on the basis of resighting (e.g., Figure 2), suggests that resighting of captured birds. Simply adding resighting to standard- net avoidance may persist in tropical birds. Indeed, we frequently effort netting substantially increased detection and apparent survival observed banded birds approaching a net and then moving above or estimates for most species (Figure 1 and Table 3). However, evaluation around it. Second, the propensity to avoid nets seemed to differ among of the effect of different levels of resighting effort would be useful. species, which could cause spurious variation and explain why apparent Average apparent survival estimates for our 18 Malaysian species survival estimated from standard-effort netting alone was not corre- were similar to those of 18 Neotropical songbird species studied in lated with egg temperature (Figure 4a). For example, the mountain Venezuela with similar methods (Table 1; Martin et al., 2015). The wren-babbler is an understorey species that moves in conspecific social groups, primarily at net heights (within 2 m of the ground). Yet only higher apparent survival rates of tropical than temperate songbirds fit- 15% of individuals of this species captured through standard-effort ted predictions of theory that were based on the life history strategies netting were recaptured (Figure 2a), whereas 76% of marked individu- of tropical songbirds. Although we believe that resolution of the argu- als were resighted (Figure 2b), in subsequent years after initial capture. ment over whether tropical songbirds have higher apparent survival In contrast, grey-throated babblers move in conspecific social groups, rates than temperate species is important, appropriate estimation of but over a much greater height range. Despite common movements apparent survival is crucial for examining biological relationships among above net height, grey-throated babblers appeared to avoid nets less tropical species (e.g., Figure 4). We found that large numbers of individ- than mountain wren-babblers, as reflected by their high detection uals of many species appeared to avoid nets, resulting in underestima- probabilities that did not increase as much as those of other species tion of apparent survival. Yet our detection and apparent survival when resighting was added (Figure 3). Ultimately, differences among probabilities from standard-effort netting alone were similar to those species in net avoidance can cause spurious variation in apparent sur- from other tropical studies with similar methods, which may indicate vival estimates that can obscure relationships between apparent sur- that bias is pervasive in estimates of apparent adult survival from tropi- vival and life histories (Figure 4). cal netting studies. Most importantly, our analyses suggest that Tropical birds may avoid nets more than temperate birds, which standard-effort netting not only negatively biases estimates, but also complicates comparisons of apparent survival rates between tropical yields greater variation in estimates with lower confidence that can and temperate birds on the basis of standard-effort netting alone. obscure biological relationships (Figure 4). 1396 | MARTIN ET AL.

ACKNOWLEDGMENTS Charnov, E. L., & Krebs, J. R. (1974). On clutch-size and fitness. Ibis, 116, 217–219. We appreciate helpful comments from J. LaManna, S. Williams, B. Dobzhansky, T. (1950). Evolution in the tropics. American Scientist, 38, Sandercock and four anonymous referees on the manuscript. We 209–221. are grateful to Sabah Parks and the Sabah Biodiversity Centre in Faaborg, J., & Arendt, W. J. (1995). Survival rates of Puerto Rican birds: Malaysia for authorization and facilitation of this study and to a Are islands really that different? The Auk, 112, 503–507. number of field assistants for help in data collection, particularly C. Felsenstein, J. (1985). Phylogenies and the comparative method. The Armstad, B. Jurunin, R. Gobbo and E. Conrad. This work was sup- American Naturalist, 125,1–15. ported by the National Science Foundation (DEB-1241041, DEB- Fogden, M. P. L. (1972). The seasonality and population dynamics of 1651283, IOS-1656120) and conducted under University of Mon- equatorial forest birds in Sarawak. Ibis, 114, 307–343. tana IACUC #059-10TMMCWRU. Any use of trade, firm or product Francis, C. M., Terborgh, J. S., & Fitzpatrick, J. W. (1999). Survival rates of names is for descriptive purposes only and does not imply endorse- understorey forest birds in Peru. Proceedings of the 22nd International Ornithological Congress, Durban. BirdLife South Africa, ment by the U.S. Government. Johannesburg. Ghalambor, C. K., & Martin, T. E. (2001). Fecundity-survival trade-offs DATA ACCESSIBILITY and parental risk-taking in birds. Science, 292, 494–497. Embryonic temperature and development time data are presented Jetz, W., Sekercioglu, C. H., & Bohning-Gaese,€ K. (2008). The worldwide by Martin et al., (2015), and apparent survival and detection data variation in avian clutch size across species and space. PLoS Biology, 6, e303. presented here are provided in Dryad DOI: 10.5061/dryad.k1p41 Johnston, J. P., White, S. A., Peach, W. J., & Gregory, R. D. (1997). Sur- Data files: Martin et al 2017 GEB Tropical survival. vival rates of tropical and temperate passerines: A Trinidadian per- spective. The American Naturalist, 150, 771–789. AUTHOR CONTRIBUTIONS Karr, J. R., Nichols, J. D., Klimkiewicz, M. K., & Brawn, J. D. (1990). Sur- vival rates of birds of tropical and temperate forests: Will the dogma T.E.M. conceived and designed the study. T.E.M., W.M.B., R.R. and survive? The American Naturalist, 136, 277–291. J.C.M. contributed to data collection. W.M.B. established all netting Kendall, W. L., & Nichols, J. D. (1995). On the use of secondary capture- sites and subplots in collaboration with T.E.M. T.E.M. and M.M.R. recapture samples to estimate temporary emigration and breeding completed all data summaries and analyses. T.E.M. wrote the initial proportions. Journal of Applied Statistics, 22, 751–762. draft of the manuscript, but all authors contributed to conceptual Lack, D. (1954). The natural regulation of animal numbers. London, U.K.: and written presentation. Oxford Press. MacArthur, R. H. (1972). Geographical ecology: Patterns in the distribution ORCID of species. Princeton, NJ: Princeton University Press. Martin, T. E. (1995). Avian life history evolution in relation to nest sites, Thomas E. Martin http://orcid.org/0000-0002-4028-4867 nest predation and food. Ecological Monographs, 65, 101–127. Martin, T. E. (2002). A new view for avian life history evolution tested REFERENCES on an incubation paradox. Proceedings of the Royal Society B: Biologi- – Barker, R. J. (1997). Joint modeling of live-recapture, tag-resight, and cal Sciences, 269, 309 316. tag-recovery data. Biometrics, 53, 666–667. Martin, T. E. (2008). Egg size variation among tropical and temperate Blake, J. G., & Loiselle, B. A. (2002). Manakins (Pipridae) in second- songbirds: An embryonic temperature hypothesis. Proceedings of the – growth and old-growth forests: Patterns of habitat use, movement, National Academy of Sciences USA, 105, 9268 9271. and survival. The Auk, 119, 132–148. Martin, T. E. (2015). Age-related mortality explains life history strategies – Blake, J. G., & Loiselle, B. A. (2008). Estimates of apparent survival rates of tropical and temperate songbirds. Science, 349, 966 970. for forest birds in eastern Ecuador. Biotropica, 40, 485–493. Martin, T. E., Bassar, R. D., Bassar, S. K., Fontaine, J. J., Lloyd, P., Math- ... Blake, J. G., & Loiselle, B. A. (2013). Apparent survival rates of forest ewson, H., Chalfoun, A. (2006). Life-history and ecological corre- birds in eastern Ecuador revisited: Improvement in precision but no lates of geographic variation in egg and clutch mass among passerine – change in estimates. PLoS One, 8, e81028. species. Evolution, 60, 390 398. Brawn, J. D., Karr, J. R., Nichols, J. D., & Robinson, W. D. (1999). Demog- Martin, T. E., Martin, P. R., Olson, C. R., Heidinger, B. J., & Fontaine, J. J. raphy of forest birds in Panama: How do transients affect estimates of (2000). Parental care and clutch sizes in North and South American – survival rates? Proceedings of the 22nd International Ornithological birds. Science, 287, 1482 1485. Congress, Durban. BirdLife South Africa, Johannesburg. Martin, T. E., Oteyza, J. C., Boyce, A. J., Lloyd, P., & Ton, R. (2015). Adult Burnham, K. P., & Anderson, D. R. (2002). Model selection and multimodel mortality probability and nest predation rates explain parental effort inference: A practical information-theoretic approach. New York, NY: in warming eggs and embryo development time. The American Natu- – Springer-Verlag. ralist, 186, 223 236. Burnham, K. P., Anderson, D. R., White, G. C., Brownie, C., & Pollock, K. Michod, R. E. (1979). Evolution of life histories in response to age- – H. (1987). Design and analysis methods for fish survival experiments specific mortality factors. The American Naturalist, 113, 531 550. based on release–recapture. American Fisheries Society Monograph, 5, Moreau, R. E. (1944). Clutch size: A comparative study, with reference to 1–437. African birds. Ibis, 86, 286–347. Charlesworth, B. (1994). Evolution in age-structured populations. Cam- Pagel, M. D. (1992). A method for the analysis of comparative data. Jour- bridge, U.K.: Cambridge University Press. nal of Theoretical Biology, 156, 431–442. MARTIN ET AL. | 1397

Parker, T. H., Becker, C. D., Sandercock, B. K., & Agreda, A. E. (2006). White, G. C., & Burnham, K. P. (1999). Program MARK: Survival estima- Apparent survival estimates for five species of tropical birds in an tion from populations of marked animals. Bird Study, 46, 120–139. – endangered forest habitat in western Ecuador. Biotropica, 38,764 769. Williams, G. C. (1966). Natural selection, the costs of reproduction, and a Pianka, E. R. (1970). On r-and K-selection. The American Naturalist, 104, refinement of Lack’s principle. The American Naturalist, 100, 687– 592–597. 690. Pollock, K. H., Nichols, J. D., Brownie, C., & Hines, J. E. (1990). Statistical infer- Willis, E. O. (1974). Populations and local extinctions of birds on Barro ence for capture-recapture experiments. Wildlife Monographs, 107,3–97. Colorado Island, Panama. Ecological Monographs, 44, 153–169. Pradel, R., Hines, J. E., Lebreton, J.-D., & Nichols, J. D. (1997). Capture– Wolfe, J. D., Stouffer, P. C., & Seeholzer, G. F. (2014). Variation in tropi- recapture survival models taking account of transients. Biometrics, 53, cal bird survival across longitude and guilds: A case study from the 60–72. Amazon. Oikos, 123, 964–970. Roff, D. A. (2002). Life history evolution. Sunderland, MA: Sinauer Associates. Ruiz-Gutierrez, V., Doherty, P. F., Jr, Martínez, S. C., Schondube, J., Mun- BIOSKETCH guía, H. V., & Inigo-Elias,~ E. (2012). Survival of resident neotropical birds: Considerations for sampling and analysis based on 20 years of THOMAS E. MARTIN is interested in understanding causes of variation in bird-banding efforts in Mexico. The Auk, 129, 500–509. demography and life history strategies among songbirds across the Sandercock, B. K., Beissinger, S. R., Stoleson, S. H., Melland, R. R., & world. Hughes, C. R. (2000). Survival rates of a Neotropical parrot: Implica- tions for latitudinal comparisons of avian demography. Ecology, 81, SUPPORTING INFORMATION 1351–1370. Additional Supporting Information may be found online in the sup- Skutch, A. F. (1949). Do tropical birds rear as many young as they can porting information tab for this article. nourish? Ibis, 91, 430–458. Skutch, A. F. (1985). Clutch size, nesting success, and predation on nests of Neotropical birds, reviewed. Ornithological Monographs, 36,575–594. Snow, D. W. (1962). A field study of the Black and White Manakin, Man- How to cite this article: Martin TE, Riordan MM, Repin R, Mou- acus manacus, in Trinidad. Zoologica, 47,65–104. ton JC, Blake WM. Apparent annual survival estimates of tropi- Snow, D. W., & Lill, A. (1974). Longevity records for some neotropical cal songbirds better reflect life history variation when based on land birds. Condor, 76, 262–267. intensive field methods. Global Ecol Biogeogr. 2017;26:1386– Sol, D., Szekely, T., Liker, A., & Lefebvre, L. (2007). Big-brained birds 1397. https://doi.org/10.1111/geb.12661 survive better in nature. Proceedings of the Royal Society B: Biological Sciences, 274, 763–769.