You are here: Home > Activities > Meetings > 2002 > abstracts > Wednesday >
Society for Conservation Biology: 2002 Annual Meeting

Abstracts

< Go Back

Society for Conservation Biology: 2002 Annual Meeting

Society for Conservation Biology 16th Annual Meeting July 14-July 19 2002
co-hosted by DICE and the British Ecological Society


Abstracts for Spatial Ecology and Conservation
Session Four

Wednesday 17th July, 10.15 - 12.15, Grimond Lecture Theatre 3

Chair: Alexander Harcourt




(BLOCK CAPITALS indicate the presenting author)

10.15 - 10.30
SEOANE, JAVIER, Javier Bustamante and Ricardo Díaz-Delgado. Department of Applied Biology, Estación Biológica de Doñana, CSIC, Avda. Maria Luisa s/n, 41013 Sevilla, Spain, <seoane@ebd.csic.es>.

INCORPORATING EXPERT OPINION IN PREDICTIVE HABITAT MODELS? IS IT WORTH THE EFFORT?

Predictive habitat modelling for conservation and planning may be acomplished either with fully automatic procedures for the selection and transformation of predictive variables, or with some aid of expert opinion. On one hand, the first approach is easily integrated into a GIS framework but it is supposed to pick up spurious relationships between predictors and response. On the other hand, the second approach renders more interpretable models, but it is time-consuming, difficult to standardise, and may incur other subtler statistical problems. In this work we compare the predictive ability of bird presence/absence additive models that incorporate expert opinion in the selection and transformation of potential environmental predictors with others developed according to statistical criteria only. Estimates of discrimination ability (AUC and Kappa) are calculated for three different sets of evaluation data: the building data, a ten-fold crossvalidated set and data from other geographic area. Models built according to statistical criteria alone had greater discrimination ability than those incorporating expert opinion when evaluated with building data. There were not any clear differences among models in the rest of evaluation situations. The results suggest that unsupervised fitting procedures can generate habitat suitability maps in an adequate-cost effective way.

10.30 - 10.45
TRAVIS, JUSTIN M.J. Institute of Mathematics, University of St Andrews, St Andrews, Fife KY16 9SS, UK, <justin@mcs.st-and.ac.uk>.
NEIGHBOURHOOD SIZE, DISPERSAL DISTANCE AND THE COMPLEX DYNAMICS OF A SIMPLE POPULATION MODEL.
The use of population models is widespread in conservation biology. The reliability of their results depends upon both model formulation and paramaterisation. Spatial processes are important for many species yet most population models are spatially implicit. Spatially implicit models assume individuals interact equally with every other individual and that dispersal occurs with equal likelihood to any location. For many species of animals and for all plant species these assumptions are unlikely to be true. Here one much-studied population model — the Ricker model — is reformulated such that interactions only occur between individuals located within a certain distance of each other and dispersal is local. Results demonstrate that global population stability is greater when interaction neighbourhoods are small and dispersal distance low. With small interaction neighbourhoods population dynamics are stabilised due to stochastic variability in local population densities. Reduced dispersal distances increase global population stability through the spatial structuring of the population. Different areas of the population behave asynchronously of one another and thus local extinction does not automatically imply global extinction. The proper inclusion of space into population models will result in more accurate predictions of population dynamics and this should have considerable benefits for conservation biology.



10.45 - 11.00
FLEISHMAN, ERICA, and Ralph Mac Nally. Center for Conservation Biology, Department of Biological Sciences, Stanford University, Stanford, CA 94305-5020, USA, <efleish@stanford.edu> (EF), Australian Centre for Biodiversity: Analysis, Policy and Management, School of Biological Sciences, Monash University 3800, Australia (RMN).

VALIDATION TESTS OF PREDICTIVE MODELS OF BUTTERFLY OCCURRENCE

Statistical models increasingly are being used to predict the occurrence of species based on environmental information. We previously used 14 topographic variables from 49 locations in the Toquima Range (Nevada, USA) and species inventories conducted over four years to model logistically occurrence of resident butterfly species. To test the models, validation data were collected over two years in 39 locations in the nearby Shoshone Mountains. We used a series of "classification rules" based on conventional logistic criteria and Bayesian criteria to assess the success rates of predictions. The rules represented a gradient of stringency in the "certainty" with which predictions were made. More stringent rules reduced numbers of predictions but greatly increased their success. Success rates for absence predictions were uniformly higher than for presence predictions. When comparing rates for locations with one and two years of inventory data, success rates for presence predictions increased with two years of data while rates for absence predictions decreased. Neither overall success rates nor the success rate of absence predictions was correlated with the fit (explained deviance) of the initial model, although the success rate of presence predictions was correlated with fit. Classification rules for making presence and absence predictions may be decoupled.



11.00 - 11.15
UNDERWOOD, EMMA, Rob Klinger, Linda Mutch, and Peggy Moore. Dept. of Environmental Science and Policy, University of California - Davis, Davis, CA 95616, USA (eunderwoodrussell@ucdavis.edu) (EU, RK), National Park Service, Three Rivers, CA 93271, USA (LM) and U.S. Geological Survey, El Portal, CA 95318, USA (PM).

PREDICTING PATTERNS OF NON-NATIVE PLANT SPECIES INVASIONS IN YOSEMITE NATIONAL PARK, CALIFORNIA

Yosemite National Park is one of the most well known parks in the United States — harboring phenomenal scenic attractions and great ecological value. However, one of the immediate challenges confronting park management is the invasion of non-native species. Effective management of invasives would be greatly assisted by information on their potential distribution. We performed a series of community scale analyses on plot data to identify key environmental and disturbance-related data layers that correlate with the percent cover of non-native species. We then developed two predictive models. The first using elevation, slope, and vegetation alliance to predict the environmental niche of co-occurring species; the second using disturbances such as distance from roads, trails, and campgrounds. Verification of results was performed using plot data reserved from the model. Correct prediction of presences for the environmental and disturbance models was 76% and 65% respectively, while plots that failed to be predicted raise some interesting questions. The resulting maps of potential invasion were composited to provide the basis for stratifying fieldwork plots for monitoring. This research contributes to a nationwide National Park Service strategy for providing scientific data for managing natural resources, particularly for invasives which greatly threaten both biodiversity and ecosystem functioning.



11.15 - 11.30
SANDERS, SUZANNE and James B. McGraw. Department of Biology, West Virginia University, Morgantown, West Virginia 26505-6057, USA, <ssander4@wvu.edu>.

HABITAT SUITABILITY MODELING OF GOLDENSEAL WITH GEOGRAPHIC INFORMATION SYSTEMS

Goldenseal, Hydrastis canadensis, is a rare perennial herb of the eastern deciduous forest of North America. This species grows in rich woodlands, although its spatial pattern and ecological requirements are poorly understood. A traditional quadrat sampling approach failed to discern spatial pattern because this species is only sparsely distributed across the landscape. We developed a habitat suitability model based on characteristics at thirty known goldenseal patches in north central West Virginia, USA. We used geographic information systems (GIS) and Mahalanobis distance multivariate analysis to quantify habitat suitability across the landscape of the study area. Four percent of the study area was classified as highest suitability, while 36 percent was categorised into the next highest suitability class. We surveyed 27 one-hectare sites during summer, 2001 to determine model accuracy. Goldenseal was located at four surveyed sites, three of which were in two highest suitability classes. The surveyed site with the greatest abundance of goldenseal was classified only as moderate to poor goldenseal habitat. Characteristics such as stand age, canopy composition, and soil classification were not available digitally for inclusion in the GIS model. Inclusion of these may be necessary for improved model accuracy.



11.30 - 11.45
BUSTAMANTE, JAVIER and Javier Seoane. Department of Applied Biology, Estación Biológica de Doñana, CSIC, Avda María Luisa s/n, 41013 — Sevilla, Spain, <busta@ebd.csic.es>.

PREDICTIVE SUCCESS OF TOPOGRAPY AND VEGETATION IN HABITAT MODELS FOR FOREST RAPTORS

Resource managers need to predict the distribution and abundance of species and the suitability of available habitats. We test the effectiveness of topographic and vegetation variables estimated with a GIS to predict the distribution of the Buzzard (Buteo buteo), Short-toed eagle (Circaetus gallicus), Booted eagle (Hieraaetus pennatus) and Black kite (Milvus migrans) in Southern Spain. We used road census in 10x10 km squares to sample raptor distribution and adjusted Generalised Additive Models with a stepwise variable selection procedure. In most cases it was possible to build predictive models that improved significantly a classification by chance with only topographic or only vegetation variables, but models were not good. Models improved their predictive ability if variables from both sets were included, and in three out of four species the inclusion of spatial coordinates to account for neighbourhood effects improved these models. Models with high predictive accuracy were not necessarily easy to interpret according to the ecology of the species. Models implemented in a GIS allow us to build predictive maps of habitat suitability for these species.



11.45 - 12.00
BIEK, ROMAN and Mary Poss. Fish and Wildlife Biology Program (RB) and Division of Biological Sciences (MP), University of Montana, Missoula, MT 59812, USA, <rbiek@selway.umt.edu> (RB).

USING PHYLOGEOGRAPHY OF A MICROPARASITE TO ASSESS SPATIAL POPULATION STRUCTURE IN ITS MAMMALIAN CARNIVORE HOST

Molecular approaches are widely used to infer spatial population structure in conservation. However, the genetic population structure of a species may reflect processes on temporal scales much larger than those of specific conservation interest, for example if populations became fragmented relatively recently. In this study we demonstrate that phylogenetic data of an endemic, rapidly-evolving, and non-pathogenic retrovirus commonly found in Rocky Mountain populations of cougars, Puma concolor, can provide recent information on population subdivision and movement of its host. Based on sequence data from two viral genes, we show that most infected cats within an area carry closely related viruses and our data indicates that many such regional virus variants circulate in Rocky Mountain cougars. Further, using serial sampling of infected individuals we estimated that the virus genes examined evolve at rates of 0.1-0.5% per year, suggesting that virus transmission that occurred among cougar populations within the last few decades should be detectable. This molecular technique thus holds the promise to provide current information about population connectivity, an issue of much interest to conservation.



12.00 - 12.15
QUINTANA-ASCENCIO, PEDRO F., Eric S. Menges and Rebecca W. Dolan. Archbold Biological Station, 123 Main Drive, Venus, Florida 33960, USA, <pfquintana@archbold-station.org> (PFQA, ES) and Friesner Herbarium, Butler University, Indianapolis, Indiana 46208, USA (RWD).

SPATIAL PATTERN OF GENETIC VARIATION IN HYPERICUM CUMULICOLA: IMPLICATIONS FOR MANAGEMENT

We present an analysis of spatial distribution patterns and overall genetic constitution of H. cumulicola in gaps among dominant shrubs in Florida scrub. We measured the length and width of all gaps, mapped gap centers with GPS, and determined polar coordinates for individual plant locations. We used spatial autocorrelation to calculate f, the average kinship coefficient between individuals within distance intervals. Gaps (mean length and width in m & se; 5.0+0.2 x 3.7+0.4 se, n = 111 gaps; 6.8+0.5 x 4.7+0.4; n = 85 occupied gaps) supported from 1-115 plants. The numbers of plants increased exponentially with gap size: no gaps smaller than 23 m2 sheltered plants, while all gaps larger than 210 m2 did. We found five isozyme loci variable, and there was spatial patterning of multi-locus genotypes among gaps. Across all gaps, genetic similarity was higher than expected for plants growing within about 10 m of each other, and was highest within 3 m. Plants more distant than about 10 m were not particularly similar to each other genetically. However, there was no genetic spatial pattern within individual gaps. Our results show spatial scales relevant for the understanding of H. cumulicola genetic variation and contribute to evaluating population consequences of fire, which affects inter-gap distances, gap integrity, and gap sizes.