DOT Project Number:  90-00-LRTF-404

Fiscal Year:  2004

Award:  $21,000.00

Principal Investigator:  Diane Debinski, Department of Ecology, Evolution and Organismal Biology, Iowa State University, debinski@iastate.edu

Other Project Participants:  Stephen Hendrix, Department of Biological Sciences, University of Iowa; Jessica Skibbe, Department of Ecology, Evolution and Organismal Biology, Iowa State University

Summary Report:

SUSTAINING POLLINATOR DIVERSITY IN A FRAGMENTED LANDSCAPE: THE ROLE OF ROADSIDES AND RAILROAD RIGHTS-OF-WAY

Introduction
The fragmented landscape of the Midwest USA creates prairie pieces embedded in an inhospitable matrix, potentially impermeable to dispersing individuals. Concern over the loss of pollinator diversity and its ecological and economic consequences has been growing in recent years as conservationists and others have begun to focus on previously undervalued ecological services. The major cause of the loss of pollinator diversity throughout the world is thought to be habitat fragmentation. Additional causes of pollinator diversity loss include increased application of insecticides, habitat loss due to agriculture, and urbanization.

We examined butterfly responses to linear prairies (roadsides, railroad rights-of-way, trails, etc.) and block prairies (state preserves and private prairies) in order to better understand the conservation value of these most common types of remnants. We studied the butterfly community in prairie remnants to answer four questions: 1) How do different guilds of species respond to local environmental variables? 2) How do configuration and composition of landscape elements of differing diversity of floral resources affect butterfly diversity at point locations in the landscape? 3) Which of five spatial extents, ranging from local to 0.5-2 km radius around a site (landscape) predict point-level butterfly diversity? and 4) How do local and landscape variables affect butterflies in linear compared to block habitats?

Methods
In the summer of 2004 we sampled the butterfly communities in ten block and 25 linear habitats in the northwest quarter of Iowa (Fig. 1). The block habitats sampled represent a wide range of sizes (26 –160 acres) and the landscape surrounding them varied considerably in the relative amount of landscape elements. Two 100m by 5m transects were established at each site.

We sampled butterflies three times during the growing season (June, July, and August) visiting each site once for butterflies. Butterfly surveys were conducted between by walking each 100 m transect at a pace of 10 m/min during ideal environmental conditions and observing butterflies within 2.5 m on either side and in front of the observer. During each visit, all butterflies (including skippers) were counted.

Floral resources at the site were surveyed by direct counting flowering ramets within the 100m transect during each round of butterfly sampling. A relevé of the transect was performed to approximate relative proportions of forbs, native grass, non-native grass, standing dead vegetation and bare ground.

We quantified resources in the landscape surrounding each block habitat and ten linear habitats within 2 km centered in the middle of the site. We analyzed the landscape data using nested spatial extents of 0.5, 1.0, 1.5 and 2.0 km because different guilds of pollinators (e.g., immobile vs. mobile; habitat-sensitive vs. disturbance-tolerant) may respond to landscape features at different spatial extents. We used digital color infrared orthophotographs taken in 2002 along with the program ArcMap 8.3 to divide each landscape into visually distinct, homogeneous polygons. The boundaries of the polygons were checked for any recent land use changes and each polygon was assigned a landscape element and given a resource value of 0-5 on the basis of diversity of floral sources in the polygon.

We used a number of multivariate analyses (canonical correspondence analysis, multivariate analysis of variance, partial least squares regression, and multiple linear regression) to assess differences in community composition between linear and block habitats and the effects of local and landscape variables on butterfly abundance. Butterfly habitat guilds were classified a priori based on host plant and nectar resource affinities. Disturbance-tolerant (DT) species survive well in agricultural settings and/or in urban areas and occur throughout Iowa. Habitat-sensitive (HS) species are associated with native prairie habitat and require conservative plants for either larval or adult resources.

Results
In 2004 we surveyed 1130 individual butterflies and 32 species. The most common species in order of decreasing abundance were Eastern-tailed Blue (Cupido comnyntas), Orange Sulfur (Colias eurytheme), Common Wood Nymph (Cercyonis pegala), Long Dash Skipper (Polites mystic) and Regal Fritillary (Speyeria idalia).

The total number of species was significantly greater in linear compared with block habitats (F1,33=14.01, p=0.001) as was the total number of DT species (F1,33=13.01, p=0.001). Additional response variables tested were not significantly different between linear and block habitats (total abundance: F1,33=1.30, p>0.05; DT abundance: F1,33=2.47, p>0.05; HS abundance: F1,33=0.32, p>0.05; and total number of HS species F1,33=1.65; p>0.05). Multivariate analyses clearly separated linear from block habitats based on butterfly community composition. Multivariate analysis of variance indicated no differences in environmental variables between linear and block habitats at the local scale, but significant differences in landscape environmental variables between linear and block habitats.

The local variable that was most highly correlated with butterfly community response was litter. HS abundance was positively correlated with litter while DT abundance was negatively correlated with litter. At the landscape scale, the amount of roads surrounding the habitat was the most important variable at all spatial extents.

Results from partial least squares regression suggest there are indeed effects of the landscape on butterflies at all extents investigated (local; 0.5, 1.0, 1.5, and 2.0 km). The 0.5 km landscape extent was the most highly correlated with butterfly community abundance. Variance partitioning using partial canonical correspondence analysis indicated that landscape variables at all spatial extents add additional explanatory power beyond local variables with little overlap in percentage of variation explained.

Discussion
From this we conclude that linear habitats harbor a different assemblage of butterflies than block habitats, and linear habitats provide important resources for habitat-sensitive species in Iowa, thus increasing the overall diversity of butterflies. Our results provide evidence for mass effects, where local populations are not self-sustaining on linear habitats due to a decrease in species richness with increasing habitat size. Additional information on birth and death rates in linear habitats is necessary to validate source-sink dynamics.

In terms of management decisions, varying the amount of litter will produce opposite effects on abundance of DT and HS species. These data suggest that by decreasing the frequency and intensity of prescribed fires, managers could increase the diversity of HS species on prairie fragments.

We know that butterflies immigrate to high-quality patches, which suggests they are either assessing patches from a distance or moving within the landscape and sampling. Our results support the hypothesis that butterflies are making decisions based both on the local and landscape environmental factors. Additionally, the sphere of environmental influence for a butterfly occurs at a relatively small spatial extent suggesting that butterflies are sampling the landscape immediately surrounding the prairie remnant.

Roads are highly and positively correlated with some species such as Anatrytone logan, Celastrina neglecta, and Pholisora catullus. A. logan is a habitat-sensitive species which uses grass species as host plants and may be responding positively to roads because of the associated increased connectivity grassy roadsides provide. Alternatively, C. neglecta and P. catullus, both DT species, larvae require host plants commonly found in roadside ditches (Melilotus officinalis and Chenopodium album respectively) and may be responding to a higher abundance of those species. Although grassland and minimum distance to the nearest polygon with high floral resources are associated with block habitat sites, only Speyeria idalia is positively correlated with the minimum distance to nearest polygon with high floral resources. This species is of special concern in Iowa and given these results, in order to increase the abundance of S. idalia, patches of habitat should be created with high floral diversity in close proximity to preserves.

This research allowed us to evaluate the role of small linear strip prairies in the larger context of preserving butterfly diversity. From such information we developed predictive models that are useful in guiding development of future roadside plantings and railroad right-of-way management. Additional research on butterfly movement and behavior is necessary to evaluate the safety of linear habitats for butterflies. Negative factors of roadsides such as mortality due to automobile traffic and pesticide drift and/or direct spraying on agricultural buffer strips, roadsides, and railroad rights-of-way may outweigh the benefits such as nectar and host plant resources prairie plantings and remnants provide. Critical questions therefore include: how are butterflies using the roadsides and do the benefits of roadside prairie fragments outweigh the ecological costs?