Influence of Trees in the Landscape on Parasitism Rates of Grassland Passerine Nests in Southeastern North Dakota

Abstract.

Woody vegetation has been linked to increased rates of Brown-headed Cowbird (Molothrus ater) parasitism for some grassland hosts. In northern North Dakota, however, studies reported that parasitism of grassland passerine nests was lower in landscapes with trees than in those without trees. We looked for evidence of this pattern elsewhere, using data from two studies conducted on the Sheyenne National Grassland in southeastern North Dakota. Specifically, we examined the probability of parasitism relative to percent tree cover within 2 km of a nest. We found a negative relationship for grassland passerine nests of all species tested. Our results support the suggestion that cowbirds are less likely to parasitize nests of grassland passerines where tree cover on the landscape is greater. This pattern could be explained by cowbirds switching to alternative hosts in woodlands, but this hypothesis needs further testing.

Resumen.

La vegetación leñosa ha sido vinculada al aumento de las tasas de parasitismo por parte de Molothrus ater para algunas aves hospederas de pradera. En el norte de Dakota del Norte, sin embargo, los estudios han mostrado que el parasitismo de los nidos de paserinos de pradera fue menor en los paisajes con árboles que en aquellos sin ???rboles. Buscamos evidencia de este patrón en otros sitios, usando datos de dos estudios realizados en Sheyenne National Grassland en el sureste de Dakota del Norte. Específicamente, examinamos la probabilidad de parasitismo con relación al porcentaje de cobertura de árboles dentro de un radio de 2 km desde el nido. Encontramos una relación negativa para los nidos de paserinos de pradera para todas las especies evaluadas. Nuestros resultados apoyan la sugerencia de que Molothrus tiene una menor probabilidad de parasitar nidos de paserinos de pradera en los lugares en que la cobertura de árboles en el paisaje es mayor. Este patrón podría explicarse por el cambio de Molothrus hacia hospederos alternativos en los bosques, pero esta hipótesis aún debe ser evaluada.

Introduction

Populations of many grassland passerine species in North America have declined in recent decades (Ig1 and Johnson 1997, Peterjohn and Sauer 1999, Sauer et al. 2008). These declines have been linked to habitat changes (loss, fragmentation, degradation) on the breeding grounds (Igl and Johnson 1997, Igl et al. 2008). Tree cover, for example, has increased markedly within many grasslands of the northern Great Plains (Grant and Murphy 2005). Presence of woody vegetation has been associated with increased rates of Brown-headed Cowbird (Molothrus ater; hereafter cowbird) parasitism of grassland passerines (Johnson and Temple 1990, Jensen and Finck 2004, Patten et al. 2006). Parasitism, in turn, has been linked to reduced productivity of some host species (Ortega 1998). Land managers need to understand the consequences of varying amounts of tree cover on parasitism rates of bird species occupying these grasslands (Grant et al. 2004).

On its breeding grounds, the Brown-headed Cowbird reaches its highest abundance in the northern Great Plains (Price et al. 1995), but much of our understanding of cowbird parasitism and host-parasite dynamics comes from outside this region (Igl and Johnson 2007). In northern North Dakota, T Grant (J. Clark Salyer National Wildlife Refuge [NWR], unpubl. data) found parasitism rates of grassland passerines averaging 14% in treeless landscapes (n = 1961 nests; and, contrary to expectations in this region (Robinson and Smith 2000), near zero in grassland patches within an aspen-parkland landscape (n = 595 nests; Grant et al. 2006). Grant (pers. comm.) suggested that cowbirds with home ranges containing primarily grassland need to target grassland hosts, whereas cowbirds with home ranges containing both aspen (Populus tremuloides) woodland and grassland might preferentially target woodland hosts. Following this line of reasoning, Pietz et al. (2006) speculated that parasitism of grassland hosts may be lower where tree cover is sufficient to provide cowbirds with woodland hosts as alternatives. Cowbirds might actually prefer woodland hosts, or parasitism of grassland hosts might simply be diluted if the number of potential hosts in the landscape increases with increased woodland. In either case, we would expect lower parasitism rates for grassland passerine nests in landscapes with more tree cover. Here, we evaluate this hypothesis with data from studies in grasslands of southeastern North Dakota.

Methods

We collected data during two different studies conducted on the Sheyenne National Grassland (SNG) in Richland and Ransom counties of southeastern North Dakota. The SNG is managed by the U.S. Forest Service and encompasses more than 28 000 ha of tallgrass prairie, wetland, oak savanna, and deciduous forest habitats. Its vegetation, soils, topography, and land use (e.g., grazing) were described by Svedarsky and Van Amburg (1996). In 1999, 2000, and 2001, we searched for nests within 11 grassland units of the SNG as part of a study evaluating the Bird Conservation Area concept (Fitzgerald et al. 1998; hereafter BCA study). These search areas were selected to fit a study design described by Winter et al. (2006) that included large and small patches of grassland embedded in wooded and treeless landscapes. In 2001, as part of a study that focused on cowbird activity (hereafter cowbird study), we searched for passerine nests within an additional 11 grassland units scattered throughout the SNG (Pietz, unpubl. data). In the cowbird study, search areas were selected to maximize the range of the distance from grassland nests to woodland.

Nest-searching and monitoring techniques in the two studies were similar. In the cowbird study, we found most of the nests while systematically walking through the habitat and sweeping the vegetation with long poles to flush nesting birds. We found additional nests by observing adult birds' behavior, by flushing nesting birds fortuitously, and by flushing birds with a rope dragged between two people. In the BCA study, we used all of these methods but found most nests through behavioral observations (Winter et al. 2003). We marked each nest with a small survey flag placed 4–5 m from the nest. We revisited nests at intervals of about 3 or 4 days until they failed or fledged young. At each visit, we recorded numbers of cowbird eggs or nestlings, if present. For analyses, we used only active nests, defined as those attended by an adult and containing at least one host or cowbird egg or young. A nest was considered parasitized if it contained one or more cowbird eggs or young on any visit.

To evaluate the relationship between parasitism and tree cover, we first created a GIS data layer for the SNG by digitizing tree cover on 1997 digital orthophoto quadrangles by using Map and Image Processing System, versions 6.6 and 6.7 (Microimages 2001, 2002). Tree cover was further refined with 2003 images from the National Agriculture Imagery Program. We used ArcGIS software (ESRI 2002) to calculate the percentage of tree cover within a 2-km radius (both studies) and a 4-km radius (cowbird study only) of each nest. We present percentage of tree cover only within 2 km of nests, however, because earlier analyses from the cowbird study showed that the percentages of trees within 2 km and within 4 km were highly correlated (r = 0.93), and our telemetry data (Pietz, unpubl. data) suggested that most female cowbirds moved <2 km between their breeding and feeding areas.

We used logistic-regression models (Allison 1999) to examine the relationship between percent tree cover within a 2-km radius of a nest and the probability of that nest being parasitized, and how this relationship varied among four study and year combinations (i.e., BCA study 1999, 2000, 2001; cowbird study 2001). We considered four models: (1) no relationship between percent tree cover and the probability of parasitism; (2) a relationship between percent tree cover and probability of parasitism, with the slope and intercept being the same for all study and year combinations; (3) a relationship between percent tree cover and probability of parasitism, with the slope being the same and the intercepts varying among study and year combinations; and (4) a relationship between percent tree cover and probability of parasitism, with the slopes and intercepts varying among study and year combinations. We conducted all analyses by using the logistic-regression procedure of SAS (SAS Institute 2004).

Akaike's information criterion, corrected for small sample sizes (AIC_c), was used to evaluate the four models (Burn-ham and Anderson 2002). The model with the lowest AICc value was considered the best model and was used to graph the relationship between percent tree cover and probability of parasitism. We also used the best model to calculate odds ratios (Hosmer and Lemeshow 2000) and the associated 95% confidence intervals to assess the relationship between the probability of parasitism and a 1% decrease in tree cover within 2 km of nests.

The species of primary interest in our studies were the Clay-colored Sparrow (Spizella pallida), Vesper Sparrow (Pooectes gramineus), Savannah Sparrow (Passerculus sandwichensis), Grasshopper Sparrow (Ammodramus savannarum), Bobolink (Dolichonyx oryzivorus), and Western Meadowlark (Sturnella neglecta). Because sample sizes for some species were small, we pooled nest data for these six species for analysis (hereafter selected species). We also analyzed four species with adequate samples of nests separately: the Clay-colored Sparrow, Savannah Sparrow, Bobolink, and Red-winged Blackbird (Agelaius phoeniceus). Although Red-winged Blackbirds are more typically associated with wetlands (Igl and Johnson 1997), we included them here because the nests in our sample were found in grasslands and this species appears to be a preferred host of cowbirds in this region (Igl and Johnson 2007). The Bobolink and Red-winged Blackbird each had one study and year combination in which none of the nests found were parasitized (BCA study in 2000 for the Bobolink; BCA study in 2001 for the Red-winged Blackbird). Including these data in models that have different slopes and/ or intercepts for each study and year combination (i.e., models 3 and 4 above) would result in quasi-complete separation of the data and model failure. Therefore, for these species, the study and year combination without any nest parasitism was excluded from analyses.

Results

For the nests of selected species, overall parasitism rates in the two studies were similar: 11% in the cowbird study and an average of 12% during the BCA study (10, 8, and 17% in 1999, 2000, and 2001, respectively). Percent tree cover within 2 km of these nests ranged from 2 to 26% in the cowbird study and from 1 to 41% in the BCA study.

The best model for the selected species, Savannah Sparrow, and Red-winged Blackbird was the model with different intercepts for each study and year combination but the same slope for all combinations (Table 1). For the selected species, the second best model (Table 1), with the same intercept and slope for all study and year combinations, was equally supported by the data (Burnham and Anderson 2002). The best model for both the Clay-colored Sparrow and Bobolink was the model with the same intercept and slope for all study and year combinations (Table 1).

The relationship between percent tree cover within 2 km of the nests and the probability of parasitism was negative for all species (Table 2, Fig. 1). Thus, the probability of parasitism was lower where tree cover within 2 km of a nest was greater, and greater where tree cover within 2 km was lower.

For every 1% decrease in percent tree cover within 2 km of a nest, the odds that the nest had been parasitized increased by 12% for the selected species, 15% for the Clay-colored Sparrow, 5% for the Bobolink, and 6% for the Red-winged Blackbird (Table 2). For the Savannah Sparrow the odds ratio was large; probability of nest parasitism increased by a factor of 992 for every 1% decrease in tree cover. All Savannah Sparrow nests were found in areas with 2-15% tree cover, and 92% of the parasitized nests were in areas with 2-3% tree cover. This resulted in a large slope parameter (-6.9); hence the odds ratio, which increases exponentially with the slope parameter, becomes very large.

Discussion

Our data support the hypothesis that parasitism rates are lower for grassland passerine nests in landscapes with more tree cover. For all species tested the relationship among models was similar.

Studies showing increased parasitism rates for grassland bird nests near wooded edges (Johnson and Temple 1990 [reanalyzed by Johnson 2001], Jensen and Finck 2004, Patten et al. 2006) probably have led to the common perception that grassland nests are at greater risk of parasitism where there are more trees on the landscape. Some authors, however, have recognized that availability of alternative hosts in woodland habitats might reduce the risk of parasitism for grassland-nesting birds (Robinson et al. 1999, 2000, Peer et al. 2000, Igl and Johnson 2007). Bird densities typically are higher in more structurally complex habitats, such as woodlands (Cody 1985, L. D. Igl, Northern Prairie Wildlife Research Center, pers. comm.); thus densities of potential cowbird hosts should be higher, resulting in a dilution of parasitism rates in grasslands with more trees. Cowbirds also might prefer woodland hosts if densities of suitable hosts are higher in woodlands or if woodland nests are easier to find (Robinson et al. 1999, 2000, Peer et al. 2000).

A threshold percentage of trees may be required on the landscape before cowbirds can efficiently switch from grassland to woodland hosts. If woodlands provide alternative hosts, our data suggest that this threshold is relatively low. In our data, 76% of nests occurred where tree cover within 2 km was <15%. Our results suggest that small changes in the amount of tree cover on the landscape (even 1%) may influence rates of parasitism on grassland hosts substantially.

Our results support the suggestion that cowbirds are less likely to target grassland nests in landscapes with tree cover sufficient to provide alternative hosts in woodlands. This suggestion offers a potential explanation for the lower parasitism rates noted by Grant et al. (2006) on grassland sites amid trees in northern North Dakota. We must be cautious, however, about extrapolating our results. We did not measure nest density or parasitism rates within wooded habitats of the SNG. Furthermore, by the categories of Ortega (1998:184), parasitism rates in our study areas were low to moderate (10-20%); our results may be less relevant to landscapes with higher levels of parasitism.

An alternative explanation for lower parasitism rates in grasslands with more tree cover could be lower densities of cowbirds in these areas. However, there is no evidence that cowbirds are less abundant in grasslands with more tree cover. To the contrary, for grassland plots at J. Clark Salyer NWR, Grant et al. (2006:698) reported that, despite parasitism being nearly absent in grasslands within aspen parkland, “cowbirds occurred ubiquitously with respect to percent woodland cover.” Grant et al. (2004) also reported that cowbirds were more likely to be detected in grasslands near woodlands. In addition, at Des Lacs NWR in northwestern North Dakota, Murphy and Sondreal (2003) found that cowbirds were more abundant in woodland-prairie edge than in prairie. Furthermore, Johnson and Igl (2001) reported a tendency in the northern Great Plains for cowbirds to favor smaller grassland patches rather than large ones, and Igl and Johnson (1997:81) characterized the breeding habitat of cowbirds in North Dakota as “open or semi-open deciduous woodland and edge.”

Another potential explanation for lower parasitism rates in grasslands with more trees could be lower densities of grassland host nests in these areas. We do not have data on community-wide host-nest densities for grasslands with differing amounts of tree cover on the landscape. The composition of the grassland bird community, however, has been shown to change with percent tree cover. For example, at J. Clark Salyer NWR, as woodland cover within a 500-m radius increased, the probability of occurrence decreased for the Savannah Sparrow and increased for the Clay-colored Sparrow (Grant et al. 2004). It is unclear whether such changes in species composition lead to changes in the overall density of host nests. Interestingly, T Grant (J. Clark Salyer NWR, unpubl. data) found that Savannah Sparrows were parasitized most heavily (resulting in lowest nest survival) on sites with the lowest overall nest densities. Thus, factors other than host-nest density may influence parasitism rates.

Patterns of cowbird parasitism can vary among habitats, across spatial and temporal scales, and among and within species (Hahn and Hatfield 1995, Shaffer et al. 2003, Jensen and Cully 2005, Igl and Johnson 2007). In the data from our two SNG studies, for example, we noted an apparent difference in parasitism rates for the same species in the same year. The parasitism rate for Red-winged Blackbird nests in our cowbird study sample (10 of 19 nests) was typical of what has been documented for Red-winged Blackbirds in this region (Linz and Bolin 1982 [42%], Koford et al. 2000 [43%], Igl and Johnson 2007 [43%]). Parasitism rates of Red-winged Blackbird nests found on our BCA study plots (0 of 28 nests in 2001; 4 of 64 nests total) were more typical of those found in eastern North America (e.g., Hahn and Hatfield [1995]: 0 of 35 Red-winged Blackbird nests parasitized).

In a New York landscape with 55% forest cover, Hahn and Hatfield (1995) found that cowbirds preferred hosts from the forest interior over those in adjacent old fields and edges. Their results support the suggestion that where adequate woodland hosts are available, hosts in other habitats are less likely to be parasitized. There remains a need for research that evaluates parasitism in both grassland and woodland habitats simultaneously, especially at the core of the cowbird's range.

Finally, our results should not be viewed as a rationale for enhancing tree cover in grassland landscapes. Many grassland bird species exhibit area sensitivity (sensuJohnson and Igl 2001) and may avoid nesting in areas with too much tree cover (Grant et al. 2004, Quamen 2007). Also, the levels of cowbird parasitism that we observed probably have less influence on nest survival than does nest predation. In the cowbird study, for example, 48% of the nests we found were destroyed by predators (Pietz and Buhl, unpubl. data). Tree cover may increase nest predation in grasslands by some mammals and birds, such as raccoons (Procyon lotor) and raptors (Sargeant et al. 1993, Renfrew and Ribic 2003), offsetting any benefits of reduced parasitism. Studies are needed that evaluate survival and overall productivity of parasitized and unparasitized passerine nests in grasslands and woodlands concurrently.

Acknowledgments

Thanks to U.S. Forest Service district ranger B. R. Stotts for permission to work on the SNG and to members of all the field crews for finding and monitoring nests. H. T. Sklebar, B. R. Euliss, M. A. Cunningham, and R. M. Bush provided the digital data needed to assess the effects of tree cover. L. D. Igl provided useful discussions and suggestions throughout the preparation of this paper. T. A. Grant, L. D. Igl, J. E. Austin, and anonymous reviewers provided helpful comments on various drafts. The studies were funded by the U.S. Geological Survey and the U.S. Fish and Wildlife Service, regions 3 and 6. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Literature Cited