Feature Articles Spotted Towhee Population

FEATURE ARTICLES

The C:c~mior 109:72I-733 (L,The Cooper Orn~rhologicalSociety 2007

SPOTTED TOWHEE POPULATION DYNAMICS IN A RIPARIAN RESTORATION CONTEXT

STACYL. SMALL''^, FRANKR. THOMPSON1112, GEOFFREYR. GEUPEL~,AND JOHN FAABORG' 'Division of Biological Sciences. 105 Tttcker Hall, U~ziversit-vof Missouri-Columbia, Coluinbia, MO 65211 'USDA Forest Service, North Ceitrrul Research Sturion, 202 Anheuser-Busch Natural Resources Building, University of Missouri- Columbia, Cohmzbia, MO 652 11 3PRB0 Conservntion Science, Palomarin Field Station, P. 0.Box 1 157, Bolinas, CA 94924

Abstract. We investigated factors at multiple scales that might influence nest predation risk for Spotted Towhees (Pipilo maczzlatus) along the Sacramento River, California. within the context of large-scale riparian habitat restoration. We used the logistic-exposure method and Akaike's information criterion (AIC) for model selection to compare predator search, predator abundance, restoration, and temporal effects hypotheses. Our candidate models represented restoration, temporal, nest activity, nest concealment, agriculture, and flood effects. Restoration sites did function as breeding habitat, with nest survival comparable to mature forest sites and some young being produced. The best-supported models contained the covariates cowbird young in host nest and calendar date. All supported models contained the covariate cowbird young, the only variable with a strong effect. In contrast to our prediction, nest survival increased with cowbird young in the nest. implicating adult cowbirds in nest predation events. Nest survival declined throughout the breeding season and increased marginally with nest concealment. We calculated the finite rate of population increase (A), using a model that incorporated nest success estimates for parasitized and unparasitized nests, double brooding, the locally observed cowbird parasitism rate for Spotted Towhees. number of young fledged from successful parasitized and unparasitized nests, adult survival and a range of juvenile survival values, estimated as percentages of adult survival. Values of A did not approach the replacement level of one, indicating that this population is not self-sustaining. We recommend extensive habitat restoration in the floodplain landscape, ideally in conjunction with cowbird control. to reduce nest predation and parasitism pressures. Key words: bird population dytza~nics,cowbird parasitism, floodplains, nest predation, nest survival, restoration, ripririan forest. Dinamica de Poblaciones de Pipilo maculatus en el Contexto de Restauracibn de Ambientes Riberefios Hesumetz. En el context0 de un esfuerzo de restauracion de ambientes ribereiios de gran escala, investigamos 10s factores que pueden afectar el riesgo de la depredacibn de nidos de Pipilo rnaczrlutus a diferentes escalas a lo largo del Rio Sacramento, California. Usamos el mCtodo de exposicion logistics y el procedimiento para selecci6n de modelos basado en el criterio de informacibn de Akaike para comparar las hip6tesis de busqueda de depredadores, abundancia de depredadores, efectos de restauracibn y efectos temporales. Nuestros modelos candidatos representaron 10s efectos de restauracion, temporales, de la actividad de 10s nidos, de la ocultacion de 10s nidos, de la agricultura y de las inundaciones. Los sitios de restauracion si funcionaron como ambientes de

Manuscript received 16 January 2007; accepted 6 September 2007. Present address: River Partners, 806 14th Street, Modesto, CA 95354. E-mail: [email protected]

[72 I I 722 STACY L. SMALL ET AL.

reproduccion: la supervivencia de 10s nidos en estos fue comparable a la de sitios ubicados en bosque maduro y se produjeron algunos juveniles. Los modelos mas robustos incluyeron las covariables de presencia de polluelos de 19rlolothrusnter en el nido y la fecha. Todos 10s modelos apoyados incluyeron la covariable presencia de polluelos de M. uter, y Csta fue la hica covariable con un efecto grande. En contraste con nuestra predicci6n, la supervivencia de 10s nidos fue mayor en nidos con polluelos de :V. arer que en nidos sin Cstos, lo que involucra a 10s adultos de ,W. Lrter en 10s eventos de depredaci6n de nidos. La supervivencia de 10s nidos disminuy6 durante la temporada de nidificacibn y se increment0 marginalmente con la ocultaci6n de 10s nidos. Calculamos la tasa finita de increment0 de la poblacijn (A) usando un modelo que incorpora estimados del &xito reproductivo para nidos parasitados y no parasitados, Ias nidadas dobles, la tasa de parasitismo sobre nidos de P. rnacul~~rusobscrvada localmente. el numero de volantones criados en nidos parasitados y no parasitados, la supervivencia de 10s adultos y un rango de valores de supervivencia de 10s juveniles estimado como porcentajes de la supervivencia de 10s adultos. Los valores de ILno se aproximaron a1 nivel de reemplazo de uno, indicando que esta poblacibn no puede sostenerse a si misma. Recomendamos una restauracion extensiva de la llanura de inundation y un programa de control de la poblacion de &I. atel- para reducir las presiones de depredacion y de parasitismo.

INTRODUCTION models proposed to explain variation in nest conservation of riparian bird populations predation in a riparian restoration context over inhabiting North America's ] arge river flood- a ten-Yeal- period (19942003)- Based upon our plains may depend upon ma,or habitat resto- best approximating models, we then predicted ration efforts conducted at the landscape scale nest survival under cOndi- (Askins 2002). Birds serve as excellent bioindi- caters of riparian ecosystem function (~i~hNest predation is the primary source of nest 2002), but to evaluate restoration success in failure for most ~asserines (Ricklefs 1969, terms of habitat quality requires delnographic Martin 1992, 1993) and for Spotted Towhees studies at the appropriate scale for species that On Our sites 2005). Krementz and depend upon this habitat (Smallwood 2001). Powell (2000) attributed population declines of However, in a synthesis of 37 099 U.S. river the congeneric and ecologically similar Eastern restoration projects, Bernhardt et al. (2005) T~whee(Pipilo er~thro~thalmus)to high nest reported that few (510%) are monitored for Predation- Nest predation reduces not only biological function, although habitat improve- fledging success but also recruitment of Year- ment is a commonly stated objective of such lings, with consequences for long-term popula- projects. tion regulation (Sherry and Hollnes 1992). The Sacramento River, ~~lif~~~i~~~largest Predation has been identified as the primary river system, has undergone massive riparian nest mortality factor for Spotted Towhees on habitat loss over the past century, prinlalily our sites; nest predators for this species include a result of agricultural conversion and water rats (Raftus ~PP-).raccoons (Proc?'on lotor). regulation (Hunter et al. 19991, contributing to snakes, and birds (Small 2005)- However, regional bird population declines and local cowbird parasitism pressure is unusually high extirpations of open-cup nesting birds. Flood- there compared to Previous anecdotal reports plain habitat restoration has been conducted in for this species (Greenlaw 1996)- In 1994-2003, the Sacramento Valley for nearly 15 years, with 38% of Spotted Towhee nests observed on our the objective of increasing and improving sites were parasitized, and 60% of those cases habitat function for riparian and aquatic life were multiple parasitisms (Small 2005). (Golet et al. 2003). We sought to evaluate large- Nest predation dynamics may operate at scale riparian restoration efforts in the Sacra- multiple scales (Donovan et al. 1997, Thomp- mento Valley in terms of avian nest survival. To son et al. 2002). In developing our models of do so, we conducted a muItiscale analysis of nest predation risk, we thus incorporated risk factors for nest predation on Spotted habitat variables measured at several scales Towhees (Pbilo mncu/atus), taking an infornla- (Table 1): local (nest site, 5 m), territory (50 m), tion-theoretic approach (Burnham and Ander- and patch (500 m), measured as radii around son 2002) to evaluate a set of a priori candidate observed nests, as well as landscape-level flood TOWHEE POPULATION DYNAMICS IN RIPARIAN RESTORATION 723

TABLE I. Covariates included in nest survival models of Spotted Towhees along the Sacramento River, California. Effect directions were not predicted for interaction and random effects covariates.

Direction of Covariate Explanation predicted effect Nestling stage Nestling stage Calendar date Calendar date of nest observation Cowbird young Presence of cowbird young in nest Restoration site Habitat type Nest height Nest height from ground Nest concealment Nest concealment at 1 m radius. average of six ocular estimates Flood duration Mean duration of annual flood events Flood intensity Mean flow (cfs) of annual flood events Flood timing Median annual flood date Percent agriculture 50 m Amount of agriculture in 50 m radius around nest Agricultural edge density 50 m Meters of agricultural edge in 50 m radius around nest Percent agriculture 500 m Proportion of land cover in agriculture in 500 m radius around nest Nest stage*calendar date Interaction of nest stage and calendar date Plot-year Random effect of plot and year

effects and temporal effects. We proposed six METHODS working hypotheses to explain predation risk for Spotted Towhee nests in the Sacramento STUDY .AREA Valley, each represented by a set of candidate We 1ocat.ed and monitored active Spotted models. The hypotheses pertained to the Towhee nests on five nest plots-two restora- following effects: 1) restoration, 2) temporal, tion plots and three mature forest plots-from 3) nest concealment, 4) nest activity, 5) agricul- 1994 through 2003 on the Sacramento River ture, and 6) floods, represented by 44 total National Wildlife Refuge in Butte, Colusa, candidate models (Table 2). Glenn, and Tehama Counties, California. Plots Our hypotheses may be generally categorized were 16-20 ha in size and located within the 1- as representing either functional or numerical 2 year floodplain of the Sacramento River. predator responses. In general, functional (i-e., These sites were primarily mixed riparian forest behavioral) predator responses occur at local surrounded by agriculture. Mature forest and and patch scales, while numeric (i-e., predator riparian restoration plots were interspersed and abundance) responses occur at the landscape at least 1 km apart, arrayed along approxi- scale (Chalfoun et al. 2002). Four of our mately 97 river kilometers. Restoration sites hypotheses primarily address the functional had been planted with a mix of native riparian response of predator search effectiveness (i-e., shrubs and trees, using horticultural techniques restoration, temporal effects, nest concealment, described by Alpert et al. (1999). Forest gaps on and nest activity hypotheses). Two of our the restoration sites were occupied by native hypotheses (i-e., agriculture and flood effects mugwort (Artenzusia dozrglasiana) in some loca- hypotheses) pertain to predator abundance, tions and invasive herbaceous plants in others, a numeric response to large-scale phenomena. including Johnson grass (Sorgl~zrmhalepense), Our study species, the Spotted Towhee, is Bermuda grass (Cynodon dactylon), and peren- a ground- and shrub-nesting songbird that is nial pepperweed (Lepidium latifoliunz). Mature resident to California bat migratory in other forests were primarily a mix of Fremont parts of the western United States (Greenlaw cottonwood (Populus fremontii) and willow 1996). It is an excellent indicator species for (Salis spp.) mix, with a midstory of California restoration success on the Sacramento River, as wild grape ( Vitis cal$orniccr), blue elderberry it nests in low substrate on both restoration and (Sanzbucus mexicana). and box elder (Acer mature riparian forest sites, colonizing restora- negundo). Open gaps were primarily occupied - tion sites within several years after planting. by mugwort, Himalayan blackberry (Rubus 724 STACY L. SMALL ET AL.

TABLE 2. Working hypotheses and candidate models of Spotted Towhee nest survival along the Sacramen to River. California, 1994-2003.

Hypothesis Model Global NStg, Date, NStg*Date, CowY. NHt, NCon, HabR. Ag50, AgED5O. Ag500, FDur. FInt. FTim, p Restoration HabR. Ag500, NStg, Date, NStg*Date, 11 HabR, p HabR, Ag500, pHabR, Ag500. p HabR, NStg, p HabR, Date, p HabR, Ag500, NStg p HabR, Ag500, NStg, Date, p Temporal NStg, Date. NStg*Date, p Date, p Nest concealment NCon. NHt, NStg, Date, NStg*Date, p NCon, p NHt, p NCon, NHt, p NCon, NHt, NStg. p NCon, NHt, Date, p NCon, NHt, NStg, Date, p Nest activity NStg, CowY, NCon. Date, NStg*Date, p NStg, P CowY, p NStg, CowY. p NStg, Date, p CowY, Date, 11 NStg, CowY. Date, p CowY. Ncon, Date, p NStg. CowY, Date, NStg*Date, p Agriculture Ag50, AgED50, Ag500, NStg, Date. NStg*Date, p Ag50, P AgEDSO, p Ag500. p Ag50, Ag500, p Ag50, AgEDSO. Ag500. p Ag50. AgED50, AgSOO, NStg, 11 Flood effects FDur, p FInt. 11 FTim. 11 FDur, Flnt, FTim. p FDur, FInt, FTim, HabR, p FDur. FInt. FTim, HabR, NStg, Date, p Null P

"Abbreviationsof covariates: NStg = nestling stage, Date = calendar date of nest observation, NStg*Date = interaction of nestling stage and date, CowY = presence of cowbird young in nest, NHt = nest height, Ncon = nest concealment, HabR = habitat type, Ag50 = percent agriculture in 50 m radius around nest. AgEDSO = meters of agricultural edge in 50 m radius around nest, Ag500 = percent agriculture in 500 m radius around nest, Fdur = flood duration, Fint = flood interval, Ftim = flood timing, p = random effect of plot and year. discolor), or willow shrub-scrub. Scattered included a random effects variable as a study patches of woody river debris could be found design element to account for potential spatial throughout the forest floor. autocorrelation (plot-year). Restoration effect. We predicted that nest CANDlDATE MODELS survival for this species would be higher on All candidate models were run with and restoration sites. In general, predators may be without the following covariates: nest stage less familiar with territories on newer sites and, (i-e., eggs or young), calendar date, and an as habitat structure changes annually on an interaction term (stage*date). Also, all models early successional riparian site, predators must TOWHEE POPULATION DYNAMICS IN RIPARIAN RESTORATION 725 learn territories and potential nest sites anew cover, as greater nest concealment may hinder each year. Also, compared to mature riparian predators' visual search efforts and reduce the study sites, the understory habitat of the number of opportunistic nest predations. We restoration sites was shrubbier, with more also predicted that nest survival would decrease diversely structured, low, dense vegetation. This with nest height (Budnik et al. 2002, Burhans et offered more potential towhee nest sites and al. 2002, Patten and Bolger 2003), as the bulky presented a potentially greater nest searching Spotted Towhee nest typically becomes more challenge to predators, in accordance with the visible from below when suspended in vegeta- "many nest-site" hypothesis of Martin and tion rather than embedded in ground vegetation Roper (1988). The candidate models represent- or litter. The candidate models representing the ing the restoration effect hypothesis included nest concealment hypothesis included combina- combinations of the covariates type of habitat tions of the covariates percent nest conceal- (i.e., restoration or riparian forest), percent ment, nest height, as well as the set of covariates agriculture in surrounding landscape at 500 m that were used with all models. radii around nests, and the set of covariates that Nest activity. The nest activity hypothesis were used with all models. predicted that increased activity around the Teznporal effects. Temporal variation, within nest, associated with parental care and nestling and between breeding seasons, may produce begging, would attract the attention of nest demographic patterns akin to the spatial predators, resulting in higher nest mortality variation described by source-sink models (Martin et al. 2000, Tewksbury, Martin et al. (Burhans et al. 2002), and some studies have 2002). Both nestling begging (Haskell1994) and found that year (incorporated into our random parental activity (Budnik et al. 2002) have been effects variable, plot-year) is important in shown to increase nest predation risk. The nest explaining nest predation (Burhans et al. 2002, activity hypothesis also predicted that nests Morrison and Bolger 2002, Nur et al. 2004). with cowbird young, which may beg loudly and Calendar date may describe patterns of preda- persistently, would have higher predation rates tor activity 01- abundance, and year may (Dearborn 1999), as would more exposed nest capture effects not explicitly described by other sites that provided less concealment of parental models, as predator populations may vary with activities. The candidate models representing multiple interacting factors from year to year. the nest activity hypothesis included the fol- including agricultural harvests, flood regime, lowing covariates: nest stage, cowbird young in and precipitation. We predicted that nests later the nest, nest concealment, as well as the set of in the season and those in the nestling cycle covariates that were used for all models. would have lower nest survival, as predators Agriculture effects. We predicted that nest may cue into nest activities and learn bird predation would increase with more agriculture territories over the season (i.e., a functional in the surrounding landscape. Human-altered predator response). The candidate models riparian landscapes may support higher densi- representing the temporal effects hypothesis ties of nest predators, parasites, and exotic included combinations of the covariates calen- species (Andren 1992, Brown 1997, Bayne and dar date, nest stage, and a date*stage interac- Hobson 1998, Saab 1999, VanderWerf and tion term. Smith 2002, Schmidt 2003). Most nest preda- Nest concealnzent. The effects of nest con- tors videotaped on our sites are associated with cealment on predation rates are uncertain, with agriculture (Small 2005). The candidate models some studies finding a positive relationship representing the agriculture effects hypothesis between the two (Budnik et al. 2002, McLaren contained combinations of the covariates per- and Sealy 2003), others finding a negative cent agriculture at 50 and 500 m radii around relationship (Forstn~eierand Weiss 2004), and nests, agricultural edge density at 50 m radii still others finding no effect (Howlett and around nests, as well as the set of other Stutchbury 1996, Burhans and Thompson covariates that were used for all models. 1998). Brown (1997) found that predation by F/ood effects. Our flood hypotheses predicted birds but not by black rats (Rattus rattus) was that nest survival would increase with longer associated with nest exposure. We predicted and more intense floods annually, as well as that nest survival would increase with nest floods nearer to the breeding season. We 726 STACY L. SMALL ET AL.

TABLE 3. Hydrology data collection locations at bridge river gauge stations and nearest study sites, Sacramento River. California.

Study site distance to River gauge station Latitude Longitude Study site Habitat type station (km) Vina-Woodson 39.92"N 122.99'W Kopta Slough Restoration 3.2 Flynn Mature Forest 22.5 Hamilton City 39.75"N 121.99"W Phelan Island Restoration 8.1 Capay (Kaiser) Mature Forest 12.9 Colusa 39.21 "N 12 1.99"W Sill Norte Mature Forest 6.4 hypothesized that floods would reduce rodent care activities close to the nest. We counted populations by reducing their adult numbers nestlings upon each nest check, and the number and depressing their reproductive success (i-e., of young in the nest upon the final active nest numeric predator response). Schmidt and Ost- check was considered to be the number of feld (2003) found that Veery (Carhurus fusces- young fledged. cens) nest mortality increased with small After either fledging or nest failure, we mammal density. The candidate models repre- measured nest height from the ground and senting the flood effects hypothesis contained estimated percent concealment from 6 direc- combinations of the following hydrology co- tions-above, below, and side in the four variates, measured annually: 1) flood duration cardinal directions-from 1 m distance. We (maximum and median length, in hours, of high averaged concealment estimates to create a sin- flow pulses), 2) flood intensity (mean discharge gle nest concealment variable. We noted nest per flood event, measured as cubic feet per locations on paper nest plot maps in all years second [cfs]), and 3) flood timing (median and recorded locations with a GPS unit starting ordinal date of high flow pulses) recorded at in 2000. We checked GPS locations against field nearest bridge gauges (Table 3), as well as the maps and manually added nest locations for set of covariates that were included with all years previous to 2000 as point data, based on models. field maps.

NEST DATA LANDSCAPE DATA We obtained nest data by locating and moni- We obtained landscape data using Fragstats toring nests. following methods described by (McGarigal and Marks 1994), with a cell size of Martin and Geupel (1993). Nests were located 5 m, based upon California Department of using parental behavior cues, systematic Water Resources GIs coverage from 1998 and searches, and adult flushes. We minimized 1999. We classified cover type as riparian forest, disturbance to nest sites during nest checks by agriculture, or other. varying paths taken to nests, checking nests from a distance whenever possible, and avoid- HYDROLOGYDATA ing nest checks when corvids were present. Nest We calculated hydrology parameters using U.S. searching began in late April, early in the Geological Survey hourly discharge data taken Spotted Towhee breeding season, and lasted from three river gauges at bridge stations until breeding activity declined in late July. nearest our study sites (north to south): Vina- Nests were located in diverse substrates, in- Woodson Bridge, Hamilton City, and Colusa cluding on the ground in leaf litter and grass. (Table 3). Where hourly data points were and in shrubs, trees, tall forbs, and woody river missing, we substituted values by taking the debris. We determined nest fate by checking mean of the readings immediately before and nest contents every 1-4 days (rarely at intervals after the missing data points. greater than every four days). As we were We defined the threshold for a high pulse specifically interested in analyzing nest pre- event to be the rate of flow at which the river dation risk factors, we categorized a nest as connects with the floodplain on our study sites. successful if it was not depredated, based upon We estimated flood stage for our study sites, direct observations of fledglings or parental located within the 1-2 year floodplain, to be TOWHEE POPULATION DYNAMICS IN RIPARIAN RESTORATION 727

50 000 cfs, based on field notes of high water Donovan and Thompson (2001) to estimate events and corresponding discharge readings at fecundity and calculate the finite rate of nearest bridges, as well as hydx-ologic modeling population increase (A), based on productivity conducted as part of the Aimy Corps Hamilton and survival parameters. The model accounts City Comprehensive Study (G. Lemon, Cali- for double brooding and number of nest fornia Department of Water Resources. pers. attempts, and we modified it to incorporate comm.). the different fledging rates for parasitized and We calculated flood metrics for the winter- unparasitized nests. We used empirical data to spring flood season starting from 1 October and compute all model inputs except juvenile ending with the last floods of the season, survival, for which we calculated a range of typically ending by mid-June of the following values, based on percentages (0%100%) of calendar year. We related these flood metrics to adult survival. We used local adult survival the overlapping spring-summer breeding sea- estimates for the Spotted Towhee in the son. Sacramento Valley for the period 1995-2000, reported by Gardali and Nur (2006). We ANALYTICAL METHODS incorporated three nest attempts to account We employed Akaike's selection criteria cor- for double brooding or renesting attempts after rected for small sample sizes (AIC,) to assess nest failures, and we used model-based esti- support for our competing models, based on mates of nest survival for parasitized and differences in AIC, scores, AIC differences (Ai), unparasitzed nests, obtained following methods and Akaike model weights (w,; Burnham and described by Shaffer and Thompson (2007). Anderson 2002). To obtain our likelihood estimator, we used the logistic-exposure method RESULTS (Shaffer 2004). This method uses a modified We used a sample of 208 nests located and logi t-link function that accounts for exposure monitored between 1994 and 2003. Descriptive period or the number of days a nest was under statistics for model covariates follow (values are observation. One nest check interval was means + SE unless otherwise indicated): nest counted as an observation. We used PROC height = 36.4 + 3.8 cm, range = 0-500 cm; NLMIXED (Rotella et al. 2004), which allowed percent nest concealment = 69 + 2, range = us to incorporate the random effect of plot-year 8%-100%; median calendar date for nest as a study design element. Our binomial initiation = 7 June; mean flood duration = response variable (survive) indicated that the 133 2 12 hr; mean river discharge per flood nest was not depredated during a given nest event = 65 202 + 341 cfs; median high water check interval. Before proceeding to rank date = 10 May. models, we tested whether our global model, The top model of predation risk included the containing all covariates, differed from the null covariates cowbird young, date, and nest hypothesis. concealment (Table 4). Our estimate of daily

We estimated period nest survival rates nest survival was 0.94 +_ 0.02 (95% CI: 0.90- within a range of ecological conditions, based 0.98) based on this model. The model-based on our best-supported model. To do so, we estimate of period nest survival was 0.26 (95% employed the estimate statement using SAS CI: 0.09-0.71), using a total nest period length software (SAS Institute, Cary, NC) and used of 22 days. Based on the Hosmer-Lemeshow . the observed range of values for the covariate of goodness-of-fit test, our top model did fit the interest, while holding the other covariates in data (x' = 4.5, P = 0.81). The second-best the model constant at their sample mean or, in model contained these same covariates, plus the the case of calendar date, sample median covariate nest stage and the stage*date in- (Shaffer and Thompson 2007). Before predict- teraction term. All supported models contained ing nest survival based on our top model, we a combination of these five covariates, and the tested the fit of this model to the data, using the covariate cowbird young appeared in every Hosmer-Lemeshow goodness-of-fit test (Hos- supported model. mer and Lemeshow 1989). Comparison of model-based estimates of nest To estimate population growth rates, we used survival showed that nests with cowbird young a one stage, female-only model described by had a much higher period survival rate (0.73, TABLE 4. All supported models of Spotted Towhee nest survival, based on 11 18 nest observations, in the Sacramento Valley, California, 1994-2003. K is the number of parameters in the model, -2log(P) is the likelihood estimator, AIC, is Akaike's Information criteria adjusted for small sample size. AAIC, is the difference in AIC, values from that of the top model, and ,vi is the weight of evidence.

Model" K - 2logt4 AAIC, Wi CowY, Date. NConb CowY, Date, NCon, NStg, NStg*Date CowY? Date CowY, Date. NStg CowY, Date, NS~~,NStg*Date CowY CowY. NSte

" Abbreviations of covariates: CowY = presence of cowbird young in nest. Date = calendar date of nest observation, Ncon = nest concealment, NStg = nestling stage, NStg*Date = interaction of nestling stage and date. AIC, of the top model = 732.86

95% CI: 0.48-1.12) than those without (0.04, direction of cowbird young was positive- 95'Yo CI: 0.01-0.28). Model-based estimates of opposite the predicted direction. The effects of nest survival throughout- the breeding- season the other two covariates in the top model were indicated a within-season decline in nest sur- in the predicted direction, but minimal. The vival (Fig. 1). Nest survival increased only covariate calendar date had an odds ratio of marginally with nest concealment, but its post 0.99 (95% CI: 0.99--0.99), with the predicted hoc inclusion improved the top models. These negative effect. The covariate nest' concealment models were proposed under the nest activity had an odds ratio of 1.01 (95% CI: 1.00-1.02), hypothesis, but because of the unexpected with the predicted positive effect (Table 5). positive effect of cowbird young, this hypoth- Of non-depredated nests, parasitized nests esis was not supported. fledged far fewer host young overall (0.74 per Of the covariates in the top model, cowbird nest; 95% CI: 0.46-1.03; n = 73) than unparasit- young had the strongest effect, with an odds ized nests (2.60; 95% CI: 2.31-2.89; rz = 39). ratio of 11 .O1 (95% CI: 3.16-38.41). The effect The finite rate of population growth (A), ranged- from 0.25 to 0.33 based on period nest success rates of unparasitized (0.05) and parasitized nests (0.76), an observed cowbird parasitism rate of 38% during the period 1994-2003 .r (Small 2005), female young fledged per successful nest from unparasitized (1.30) and parasitized (0.37) nests (assuming half of fledglings are female), apparent adult survival (0.25 + 0.1 1 for the period of 1995-2000 in the Sacramento River Valley (Gardali and Nur 2006), and a range of juvenile survival values calculated as percentages (0%- 100%) of adult survival U, 0.002 - (Fig. 2). No agriculture or flood effects were detected for this species, and there was no ~~rnrn~~t- detectable difference between restoration and Q, CU.-OI h N a forest sites in nest predation rates. tn-- Date ~3 DISCUSSION FIGURE 1. Model-based estimates of Spotted Towhee period nest survival along the Sacramento 'potted Towhee populations are increasing River. California. 1994-2003. as related to calendar lo calf^ in some areas of western North America date of nest observation. ' and declining in others, but are currently stable TOWHEE POPULATION DYNAMICS IN RIPARIAN RESTORATION 729

TABLE 5. Effect sizes and odds ratios for covariates in the best approximating model of Spotted Towhee nest survival along the Sacramento River, California, 1994-2003. Confidence intervals are associated with odds ratios.

Covaria tea Effect size and direction Odds ratio 95% CI CowY Date NCon

" Abbreviations of covariates: CowY= cowbird nestlings in nest at time of observation, Date = calendar date of nest observation, NCon = nest concealment. range-wide and in California, while the conge- in values of h not approaching the replacement neric and ecologically similar Eastern Towhee is level of one. Porneluzi (2003) found in an showing population declines throughout the Ovenbird (Seiurus arrrocapi!la) population that eastern United States (Sauer et al. 2007). Our low reproductive success led to lower adult results indicate that the Sacramento Valley is return rates, which may be the case in this likely a population sink for Spotted Towhees. population. In addition to high predation rates, nests are Lack of a detectable restoration effect parasitized at a higher rate than historically indicates that restoration sites are functioning reported for this species (Greenlaw 1996), as well as forest sites as breeding habitat in the reducing annual productivity. This species' Sacramento Valley, in terms of nest predation. relatively large body size, compared to most Twedt et al. (2001) similarly found that Eastern of the locally extirpated host species, may be Towhee nest success did not differ between one important reason this species persists as bottomland forests and cottonwood plantations a breeder in the Sacramento Valley, as parasit- in the Mississippi Valley. Habitat restoration is ized towhee nests still can fledge some of their therefore likely to be an effective songbird own young. conservation measure in this western riparian In addition, the apparent adult survival rate system, but identification of limiting factors for for the Sacramento Valley Spotted Towhee this population will be an essential step toward population during the period 1995-2000 was in preventing long-term declines. the low range (0.25 t 0.1 1) for resident and Nests with cowbird young in them were I1 partial migrant species (20%58%; Greenberg times more likely to escape nest predation than 1980). Adult survival rates were not high were those without. Dearborn (1999) found enough to offset poor reproduction, resulting that cowbird beggng in Indigo Bunting (Pas- serina cyanea) nests increased nest predation risk; therefore, we expected higher nest predation rates in nests with cowbird young. In contrast to our prediction, nest survival increased when there were cowbird young in the nest. Differential body size between our study species and buntings may partially account for these contrasting results. Spotted Towhees are larger, weighing 33-49 g (Greenlaw 1996), compared to Indigo Buntings that weigh 12- 18 g (Payne 1992). Cowbird begging intensity 0.W 0.02 a05 0.07 0.10 a12 0.15 0.17 020 0.22 0s has been shown to represent honest signaling, EstlmW juvenile su&d being driven by hunger levels (Hauber and Ramsey 2003, Lichtenstein and Dearborn FlGURE 2. Finite rate of population increase (h) 2004). Cowbird nestlings may have been more values for Spotted Towhee population, Sacramento Valley, California, using a range of estimated juvenile satiated and therefore begged less intensely in survival values, based on percentages (0%100%) of a nest of the relatively large-bodied Spotted apparent adult survival values. Towhee, which can sometimes provision a 730 STACY L. SMALL ET AL. cowbird nestling and one of its own young. results, especially in the observed context of However, differential body size does not high cowbird pressure and few effective host explain the observed positive effect of cowbird species, lend support to the forced- re-nesting young in the nest. hypothesis and also corroborate those of We propose two alternate explanations for a Brown-headed Cowbird removal experiment this unexpected result: 1) a key nest predator that resulted in lower nest predation in the was discriminating between nests that contain Fraser River Delta (Smith et al. 2003). Sinall et cowbird young and those that did not, or al. (2005) had previously identified nest pre- 2) nests containing cowbird young were being dation as the primary nest mortality factor for preferentially defended. For either explanation Spotted Towhees. Our results suggest that to hold, the predator must have been able to cowbirds are playing a greater role in nest discern cowbird from host nestlings and have mortality than previously believed, possibly had an investment in the survival of cowbird through the dual mechanisms of nest parasitism offspring. In our study system, only Brown- and predation. headed Cowbirds (Moiothrus ater) fit this Cowbird pressures compound the most crit- description. While there are numerous pub- ical avian conservation threat on North Arner- lished records of adult cowbirds killing nest- ican floodplains, namely anthropogenic habitat lings of host species, reviewed by Granfors et al. change and loss (DeSante and George 1994, (2001) and captured in video footage from our Askins 2002, Rothstein and Peer 2005). The own study sites (Small 2005), we have not arrival of the cowbird in the Sacramento Valley observed cowbirds actively defending nests in in the 1930s (Neff 1931) coincided with the ten years of monitoring and are aware of no onset of agricultural conversion in the region. published evidence of adult cowbirds defending Agriculture-dominated landscapes tend to sus- host nests. tain predator and cowbird populations, where In the highly agricultural Sacramento Valley, host nests are available adjacent to prime where several smaller-bodied cowbird host foraging sites (Goguen and Matthews 1999). species including Warbling Vireo (Vireo gilvus), Such is the case in the Sacramento Valley, Yellow Warbler (Dendroica peteclria), and Song where riparian floodplain forests have been Sparrow had been locally extirpated as breeders larcgely converted to agricultural production, by the end of the 20th centuiy (Gardali et al. and extant breeding riparian birds have been 1998), host nests may be a limiting factor for exposed to elevated nest predation and cowbird Brown-headed Cowbirds. Many of the extant parasitism pressures in riparian forest rem- host species are not ideal hosts, in that they are: nants. 1) larger than cowbirds (e.g., Black-headed Long-term management for riparian song- Grosbeak [Pheucticus m~lunocep/~alz~s]),2) pri- bird populations in the Sacramento Valley. marily seed-eaters (e.g., House Finch [Carpo- especially open-cup nesters, should be focused dacus mexicanus] and American Goldfinch on restoration of the floodplain landscape, with [Carduelis tristis]), 3) cavity nesters (e.g., Oak the long-term goal of continually increasing the Titmouse [BaeoIophus inornu tt4s], 4) sparsely proportion of riparian to agricultural habitat. distributed (e-g., Common Yellowthroat We have shown that restoration sites do [Geotl~lypistrichas], Yellow-breasted Chat [Ic- function as breeding habitat, with nest survival terin virens]), or 5) otherwise ineffective hosts comparable to mature forest sites and some (e.g., Lazuli Bunting [Passerinu unloena]; Gar- young being produced. Landscape-scale solu- dali et al. 1998). tions to predator and cowbird impacts in the Arcese et al. (1996) proposed the "forced re- western United States have been proposed by nesting" hypothesis, suggesting that cowbirds DeSante and George (1994), Tewksbury, Black depredate unparasitized nests to create new et al. (2002), and Rothstein and Peer (2005). nesting opportunities. They similarly reported This approach provides the dual benefits of higher predation rates on unparasitized nests in increasing available nesting habitat while de- a population of Song Sparrows (Melospiza creasing foraging habitat for cowbirds and n?elodili) on Mandarte Island, British Columbia known nest predators. However, interim cow- and also observed decreased Song Sparrow nest bird control may also be necessary to reverse mortality in the absence of cowbirds. Our the sink status of Spotted Towhees and TOWHEE POPULATION DYNAMICS IN RIPARIAN RESTORATION 731 facilitate reestablishment of locally extirpated HASSETT. R. JENKINSON.S. KATZ, G. M. breeding songbird populations. The Spotted KONDOLF,P. S. LAKE,R. LAVE,J. L. MEYER, T. K. O'DONNELL,L. PAGANO,B. POWELL,AND Towhee's regional persistence and its ability to E. SUDDUTH.2005. Synthesizing U .S. river produce some young on both restoration and restoration efforts. Science 308:63&637. ]nature forest sites afford managers an oppor- BROWN,K. P. 1997. Predation at nests of two New tunity to improve the prognosis for a species Zealand endemic passerines: implications for bird comniunity restoration. Pacific Conserva- that has suffered the impacts of habitat loss, tion Biology 3:91-98. nest predation, and nest parasitism in the BUDNIK.J. M., F. R. THOMPSON111, AND M. R. Sacramento Valley over the past half century, RYAN.2002. Effect of habitat characteristics on with potential multispecies benefits. the probability of parasitism and predation of Bell's Vireo nests. Journal of Wildlife Manage- ACKNOWLEDGMENTS ment 66:232-239. BURHANS,D. E., D. DEARBORN,AND F. R. Funding for this project was provided by The Nature THOMPSON111. 2002. Factors affecting predation Conservancy California, U.S. Fish and Wildlife at songbird nests in old fieids. Journal of Wildlife Service, and the National Fish and Wildlife Foun- Management 66240-249. dation, as well as a GAANN Doctoral Fellowship BURHANS,D. E.. AND F. R. THOMPSON111. 1998. awarded to SLS through the Division of Biological Effects of time and nest-site characteristics on Sciences at University of Missouri-Columbia. B. concealment of songbird nests. Condor- 100: Dijak of the USDA Forest Service provided GIs 663-672. technical support. PRBO Conservation Science staff BURNHAM.K., AND D. ANDERSON.2003. Model and interns contributed field and technical support, selection and multimodel inference: a practical especially G. Ballard. J. DeStaebler. T. Gardali, J. information-theoretic approach. 2nd ed. Spring- Gilchrist, A. King, and N. Nur, as did S. Abbott, A. er, New York. Stokes, E. Elliot-Smith, J. F. Falcone. M. Gilbart, J. CHALFOUN,A. D., F. R. THOMPSON111, AND M. J. Hunt, J. Harley, S. Hinic. D. Lontoh, L. Makley, M. RATNASWAMY.2002. Nest predators and frag- Okada, P. Pintz, 0. Richmond, M. Ricketts, T. mentation: a review and meta-analysis. Conser- Maguire, M. Rogner, and K. Shipp. G. Golet, D. vation Biology 16:30&318. Peterson, M. Roberts, J. Silveira, and D. Zeleke DEARBORN,D. C. 1999. Brown-headed Cowbird provided scientific and logistical support. The Sacra- nestling vocalizations and the risk of nest mento River National Wildlife Refuge permitted predation. Auk 116:448-457. access to study sites. California DWR provided DESANTE,D. F., AND T. L. GEORGE. 1994. hydrology and GIs land cover data. Comments from Population trends in the landbirds of western R. Cocroft, H. He, B. Kus, R. Semlitsch, and J. North America. Studies in Avian Biology Tewksbury improved the manuscript. J. Garcia 15:173-190. provided Spanish translation. DONOVAN,T. M., P. W. JONES,E. M. ANNAND.AND F. R. THOMPSON111. 1997. Variation in local- LITERATURE CITED scale edge effects: mechanisms and landscape context. Ecology 78:2064-2075. ALPERT,P., F. T. GRIGGS.AND D. R. PETERSON. DONOVAN,T. M., AND F. R. THOMPSON111. 2001. 1999. Riparian forest restoration along large Modeling the ecological trap hypothesis: a hab- rivers: initial results from the Sacramento River. itat and demographic analysis for migrant song- Restoration Ecology 7360-368. birds. Ecological Applications 11 :871-882. ANDREN;H. 1992. Corvid density and nest predation FORSTMEIER,W., AND 1. WEISS. 2004. Adaptive in relation to forest fragmentation: a landscape plasticity in nest-site selection in response to perspective. Ecology 73:794-804. changing predation risk. Oikos 104:487499. ARCESE,P., J. N. M. SMITH,AND M. I. HATCH. GARDALI,T., A. M. KING,AND G. R. GEUPEL.1998. 1996. Nest predation by cowbirds and its Cowbird parasitism and nest success of the consequences for passerine demography. Pro- Lazuli Bunting in the Sacramento Valley. West- ceedings of the National Academy of Sciences em Birds 29:174-179. USA 93:4608-46 1I. GARDALI,T., AND N. NUR. 2006. Site-specific ASKINS,R. A. 2002. Restoring North America's survival of Black-headed Grosbeaks and Spotted birds: lessons from landscape ecology. Yale Towhees at four sites within the Sacramento University Press, New Haven, CT. Valley, California. Wilson Journal of Ornithol- BAYNE,E. M., AND K. A. HOBSON.1998. The effects ogy 118:178-186. of habitat fragmentation by forestry and agri- GOGUEN,C. B., AND N. E. MATTHEWS.1999. culture on the abundance of small mammals in Review of the causes and implications of the the southern boreal mixed wood forest. Canadi- association between cowbirds and livestock. an Journal of Zoology 76:62-69. Studies in Avian Biology 18:10-17. BERNHARDT,E. S., M. A. PALMER,J. D. ALLAN,G. GOLET,G., D. BROW, E. CRONE,G. GEUPEL,S. ALEXANDER,K. BARNAS,S. BROOKS.J. CARR, GRECO.K. HOLL, D. JUKKOLA,G. KONDOLF, S. CLAYTON,C. DAHM,J. FOLLSTAD-SHAH,D. E. LARSEN,F. LIGON,R. LUSTER,M. MARCH- GALAT,S. GLOSS,P. GOODWIN,D. HART, B. E~I,N. NUR, B. ORR, D. PETERSON,M. 732 STACY L. SMALL ET AL.

POWER,W. RAINEY,M. ROBERTS,J. SILVEIRA,MARTIN, T. E., AND G. R. GEUPEL. 1993. Nest- S. SMALL,J. VICK.D. WILSON,AND D. WOOD. monjtoring plots: methods for locating nests and 2003. Using science to evaf uate restoration monitoring success. Journal of Field Ornitholo- efforts and ecosystem health on the Sacramento gy 64507-519. River Project, California, p. 368-355. hz P. Faber MARTIN.T. E.. AND J. J. ROPER. 1988. Nest [ED.],California riparian systems: processes and predation and nest-site selection of a western floodplain management. ecology. and restora- population of the Hermit Thrush. Condor tion. Riparian Habitat Joint Venture, Sacra- 9051-57. mento. CA. MARTIN,T. E., J. SCOTT:AND C. MENGE.2000. Nest GRANFORS,D. A.. P. J. PEITZ,AND L. A. JOYAL. predation increases with parental activity: sepa- 2001. Frequency of egg and nestling destruction rating nest site and parental activity effects. by female Brown-headed Cowbirds at grassland Proceedings of the Royal Society of London nests. Auk 118:765-769. Series B 267:2287-2293. GREENBERG.R. 1980. Demographic aspects of long- MCGARIGAL,K., AND B. MARKS.1994. FRAG- distance migration, p. 493-516. lit A. Keast and STATS: A spatial pattern analysis program for E. S. Morton [EDS.], Migrant birds in the quantifying landscape structure. Version 2.0. Neotropics: ecology, behavior, distribution, and Forest Science Department, Oregon State Uni- conservation. Smithsonian Institution Press, versity, Corvalis, OR. Washington, DC. MCLAREN,C. M., AND S. G. SEALY.2003. Factors GREENLAW,J, S. 1996. Spotted Towhee (Pipilo influencing susceptibility of host nests to brood manlla/zls). III A. Poole and F. Gill [EDS.], The parasitism. Ethology, Ecology, & Evolution birds of North America No. 263. The Academy 15:343-353. of Natural Sciences, Philadelphia, PA. and The MORRISON.S. A., AND D. T. BOLGER.2002. American Ornithologists' Union, Washington, Variation in a sparrow's reproductive success DC. with rainfall: food and predator-mediated pro- HASKELL,D. 1994. Experimental evidence that cesses. Oecologia 133:315-324. nestling beg@ng behavior incurs a cost due to NEFF, J. A. 1931. Cowbirds in the Sacramento nest predation. Proceedings of the Royal Society Valley. Condor 33250-252. of London Series B 257:161-164. NUR,N., A. L. HOLMES.AND G. R. GEUPEL.2004. Use of survival time analysis to analyze nesting HAUBER,M. E., AND C. K. RAMSEY.2003. Honesty in host-parasite communication signals: the case success in birds: an example using Loggerhead for begging by fledgling Brown-headed Cow- Shrikes. Condor 106:457-471. birds (Molorlzrus ater). Journal of Avian Biology PATTEN,M. A., AND D. T. BOLGER.2003. Variation 34:339-344. in top-down control of avian reproductive success across a frawentation gradient. Oikos HOSMER,D. W.. AND S. LEMESHOW.1989. Applied 101:479488. logistic regression. John Wiley & Sons, New PAYNE. R. B. 1992. Indigo Bunting (Passeri~za York. c~anea).In A. Poole, P. Stettenheim, and F. Gill HOWLETT,J. S., AND B. J. STUTCHBURY.1996. Nest [EDS.],The birds of North America No. 4. The concealment and predation in Hooded Warblers: Academy of Natural Sciences, Philadelphia, PA, experimental removal of nest cover. Auk and The American Ornithologists' Union, Wash- 113:l-9. ington, DC. HUNTER,J. C., K. B. WILLETT,M. C. McCoy: J. F. PORNELUZI,P. A. 2003. Prior breeding success affects QUINN,4ND K. E. KELLER.1999. Prospects for return rates of territorial male Ovenbirds. preservation and restoration of riparian forests Condor 105:73-79. in the Sacramento Valley, California, USA. RICH,T. D. 2002. Using breeding land birds in the Environmental Management 24:65-75. assessment of western riparian systems. Wildlife KREMENTZ,D. G., AND L. A. POWELL.2000. Society Bulletin 30: 1128-1 139. Breeding season demography and movements RICKLEFS,R. E. 1969. Analysis of nesting mortality of Eastern Towhees at the Savannah River Site. in birds. Smithsonian Contributions to Zoology South Carolina. Wilson Bulletin 112:243-248. 9: 148. LICHTENSTEIN,G., AND D. C. DEARBORN.2004. ROTELLA,J. J., S. J. DINSMORE,AND T. L. SHAFFER. Begging and short-term need in cowbird nest- 2004. Modeling nest-survival data: a comparison lings: how different are brood parasites? Behav- of recently developed methods that can be ioral Ecology and Sociobiology 56:352-359. implemented in MARK and SAS. Animal Bio- MARTIN,T. E. 1992. Breeding productivity consid- diversity and Conservation 27: 187-205. erations: what are the appropriate habitat ROTHSTEIN,S. I., AND B. D. PEER.2005. Conserva- features for management?, p. 455-473. 111 J. M. tion solutions for threatened and endangered I. Hagan and D. W. Johnston [EDS.], Ecology cowbird (Molotlzrus spp.) hosts: separating fact and conservation of Neotropical migrant land- from fiction. Ornithological Monographs 57: birds. Smithsonian Institution Press, Washing- 98-1 14. ton, DC. SAAB,V. 1999. Importance of spatial scale to habitat MARTIN,T. E. 1993. Nest predation and nest sites: use by breeding birds in riparian forests: new perspectives on old patterns. BioScisnce a hierarchical analysis. Ecological Applications 43:523-532. 9:135-151. TOWHEE POPULATION DYNAMICS IN RIPARIAN RESTORATION 733

SAUER.J. R., J. E. SINES.AND J. FALLON.2007. The SMALLWOOD,K. S. 2001. Linking habitat restoration North American breeding bird survey, results to meaningful units of animal demography. and analysis 1966-2006. Version 7.23.2007. U .S. Restoration Ecology 9253-261. Geological Survey Patuxent Wildlife Research SMITH.J. N.M., M. J. TAITT: L. ZANETTE.AND 1. H. Center, Laurel. MD. MYERS-SMITH.2003. How much do Brown- SCHMIDT;K. A. 2003. Nest predation and population headed Cowbirds (Molothrus ater) cause nest declines in lllinois songbirds: a case for meso- failures in Song Sparrows (Melospiza melotiia)?: predator effects. Conservation Biology 17:1141- a removal experiment. Auk 120:772-783. 1150. TEWKSBURY,J. J., A. E. BLACK,N. NUR,V. A. SAAB. SCHMIDT,K. A., AND R. S. OSTFELD.2003. Songbird B. D. LOGAN,AND D. S. DOBKIN.2002. Effects of populations in fluctuating environments: preda- anthropogenic fragmentation and livestock graz- tor responses to pulsed resources. Ecology ing on riparian bird communities. Studies 84:406-4 15. in Avian Biology 25: 158-202. SHAFFER,T. L. 2004. A unified approach to TEWKSBURY,J. J., T. E. MARTIN,S. J. HEJL, M. J. analyzing nest success. Auk 121521-540. KUEHN,AND J. W. JENKINS.2002. Parental care SHAFFER.T. L.. AND F. R. THOMPSON111. 2007. of a cowbird host: caught between the costs of Making meaningful estimates of nest survival egg-removal and nest predation. Proceedings of with model-based methods. Studies in Avian the Royal Society of London, Series B Biology 34:84-95. 269~423--429. THOMPSON,F. R., 111, T. M. DONOVAN,R. M. SHERRY,T. W.. AND R. T. HOLMES.1992. Popula- DEGRAAF,J. FAABORG,AND S. K. ROBINSON. tion fluctuations in a long-distance Neotropical 2002. A multi-scale perspective of the effects of migrant: demographic evidence for the impor- forest fragmentation on birds in eastern forests. tance of breeding season events in the American Studies in Avian Biology 259-19. Redstart. p. 43142. In J. M. I. Hagan and D. TWEDT.D. J., R. R. WILSON,J. L. HENNE-KERR, W. Johnston [EDS.],Ecology and conservation of AND R. B. HAMILTON.2001. Nest survival of Neotropical migrant landbirds. Smithsonian In- forest birds in the Mississippi Alluvial Valley. stitution Press. Washington. DC. Journal of Wildlife Management 65:45&460. SMALL,S. L. 2005. Mortality factors and predators VANDERWERF, E. A., AND D. G. SMITH.2002. Effects of Spotted Towhee nests in the Sacramento of alien rodent control on demography of the Valley, California. Journal of Field Ornithology O'ahu 'Elepaio, an endangered Hawaiian forest 76:252-258. bird. Pacific Conservation Biology 8:7%81.