Using Indicator Species to Predict Species Richness of Multiple Taxonomic Groups

:  Values of species richness are used widely to establish conservation and management priorities. Because inventory data, money, and time are limited, use of surrogates such as “indicator” species to estimate species richness has become common. Identifying sets of indicator species that might reliably predict species richness, especially across taxonomic groups, remains a considerable challenge. We used genetic algorithms and a Bayesian approach to explain individual and combined species richness of two taxonomic groups as a function of occurrence patterns of indicator species drawn from either both groups or one group. Genetic algorithms iteratively screen large numbers of potential models and predictor variables in a process that emulates natural selection. The best-fitting models of bird species richness and butterfly species richness explained approximately 80% of deviances and included only indicator species from the same taxonomic group. Using species from both taxonomic groups as potential predictors did not improve model fit but slightly improved the parsimony (fewer predictors) of the model of bird species richness. The best model of combined species richness included five butterflies and one bird and explained 83% of deviance, whereas a model of combined species richness based on six butterflies as indicators explained 82% of deviance. A model of combined species richness based on birds alone explained 72% of deviance. We found that a small, common set of species could be used to predict separately the species richness of multiple taxonomic groups. We built models explaining approximately 70% of the deviance in species richness of birds and butterflies based on a common set of three bird species and three butterfly species. We also identified a set of six species of butterflies that predicted ≥66% of both bird species richness and butterfly species richness. Our approach is applicable to any assemblage or ecosystem, and may be useful both for estimating species richness and for gaining insight into mechanisms that influence diversity patterns. Resumen:  Los valores de riqueza de especies son ampliamente utilizados para definir prioridades de conservacion y manejo. Debido a que los datos de inventarios, el dinero y el tiempo son limitados, se ha vuelto comun el uso de sustitutos, como las especies “indicadoras,” para estimar la riqueza de especies. La identificacion de conjuntos de especies indicadoras que pronostiquen la riqueza de especies confiablemente, especialmente en varios grupos taxonomicos, es un reto importante. Utilizamos algoritmos geneticos y un metodo Bayesiano para explicar las riquezas de especies individuales y combinadas de dos grupos taxonomico como una funcion de patrones de ocurrencia de especies indicadoras extraidas de ambos grupos o de uno. Los algoritmos geneticos reiterativamente filtran grandes numeros de modelos potenciales y variables predictoras en un proceso que emula a la seleccion natural. Los modelos que mejor se ajustaron a la riqueza de especies de aves y de mariposas explicaron aproximadamente 80% de las anormalidades e incluyeron solo a especies indicadoras del mismo grupo taxonomico. Utilizando a especies de ambos grupos taxonomicos como predictores potenciales no mejoro el ajuste del modelo pero mejoro ligeramente la parsimonia (menos predictores) del modelo de riqueza de especies de aves. El mejor modelo de la riqueza de especies combinada incluyo a cinco especies de mariposas y una de ave y explico 83% de la anormalidad, mientras que un modelo de riqueza de especies combinadas basada en seis especies de mariposas explico 82% de la anormalidad. Un modelo de riqueza de especies combinadas basado solo en aves explico 72% de la anormalidad. Encontramos que un conjunto pequeno, comun, podria ser utilizado para pronosticar, por separado, la riqueza de especies de multiples grupos taxonomicos. Construimos modelos que explicaron aproximadamente 70% de la anormalidad en la riqueza de especies de aves y mariposas con base en un conjunto comun de tres especies de aves y tres de mariposas. Tambien identificamos un conjunto de seis especies de mariposas que predijeron ≥ 66% de la riqueza de especies tanto de aves como de mariposas. Nuestro metodo es aplicable a cualquier ensamble o ecosistema, y puede ser util tanto para estimar la riqueza de especies como para incrementar el entendimiento de los mecanismos que influyen sobre los patrones de diversidad.

[1]  E. Fleishman,et al.  Comparative influence of spatial scale on beta diversity within regional assemblages of birds and butterflies , 2004 .

[2]  A. Pullin.  Ecology and Conservation of Butterflies , 1995, Springer Netherlands.

[3]  R. M. Nally.  Regression and model-building in conservation biology, biogeography and ecology: The distinction between – and reconciliation of – ‘predictive’ and ‘explanatory’ models , 2000, Biodiversity & Conservation.

[4]  Jorgen B. Thomsen,et al.  Biodiversity Hotspots and Major Tropical Wilderness Areas: Approaches to Setting Conservation Priorities , 1998 .

[5]  D. White,et al.  Rare Species and the Use of Indicator Groups for Conservation Planning , 2003 .

[6]  Charles T. Scott,et al.  SAMPLING METHODS FOR ESTIMATING CHANGE IN FOREST RESOURCES , 1998 .

[7]  D. Murphy,et al.  Zoogeography of Great Basin butterflies: patterns of distribution and differentiation , 1987 .

[8]  R. M. Nally,et al.  Issues and Perspectives in Landscape Ecology: Scale and an organism-centric focus for studying interspecific interactions in landscapes , 2005 .

[9]  J. Sekhon,et al.  Genetic Optimization Using Derivatives , 1998, Political Analysis.

[10]  John E. Estes,et al.  Species RichnessA geographic approach to protecting future biological diversity , 1987 .

[11]  Lucila Ohno-Machado,et al.  A genetic algorithm to select variables in logistic regression: example in the domain of myocardial infarction , 1999, AMIA.

[12]  L. Gustafsson.  Red-listed species and indicators: vascular plants in woodland key habitats and surrounding production forests in Sweden , 2000 .

[13]  Dean L. Urban,et al.  Focal Species, Community Representation, and Physical Proxies as Conservation Strategies: a Case Study in the Amphibolite Mountains, North Carolina, U.S.A. , 2002 .

[14]  S. Andelman,et al.  Umbrellas and flagships: efficient conservation surrogates or expensive mistakes? , 2000, Proceedings of the National Academy of Sciences of the United States of America.

[15]  E. Fleishman,et al.  Modelling butterfly species richness using mesoscale environmental variables: model construction and validation for mountain ranges in the Great Basin of western North America , 2003 .

[16]  E. Fleishman,et al.  Upsides and downsides: contrasting topographic gradients in species richness and associated scenarios for climate change , 2000 .

[17]  C. Hobohm.  Characterization and ranking of biodiversity hotspots: centres of species richness and endemism , 2003, Biodiversity & Conservation.

[18]  E. Fleishman,et al.  AN EMPIRICAL TEST OF RAPOPORT'S RULE: ELEVATIONAL GRADIENTS IN MONTANE BUTTERFLY COMMUNITIES , 1998 .

[19]  E. Fleishman,et al.  Natural History and Biogeography of the Butterflies of the Toiyabe Range, Nevada (Lepidoptera: Papilionoidea). , 1997 .

[20]  N. B. Kotliar,et al.  Multiple scales of patchiness and patch structure: a hierarchical framework for the study of heterogeneity , 1990 .

[21]  Paul L. Angermeier,et al.  Factors influencing behavior and transferability of habitat models for a benthic stream fish , 1997 .

[22]  G. Schwarz.  Estimating the Dimension of a Model , 1978 .

[23]  R. Mittermeier,et al.  Biodiversity hotspots for conservation priorities , 2000, Nature.

[24]  D. Pearson,et al.  Selecting indicator taxa for the quantitative assessment of biodiversity. , 1994, Philosophical transactions of the Royal Society of London. Series B, Biological sciences.

[25]  Jon Christensen,et al.  Auditing Conservation in an Age of Accountability , 2003 .

[26]  Ralph Charles Mac Nally,et al.  A Successful Predictive Model of Species Richness Based on Indicator Species , 2004 .

[27]  A. Bennett,et al.  Forecasting the impacts of habitat fragmentation. Evaluation of species-specific predictions of the impact of habitat fragmentation on birds in the box–ironbark forests of central Victoria, Australia , 2000 .

[28]  Ralph Mac Nally,et al.  Validation Tests of Predictive Models of Butterfly Occurrence Based on Environmental Variables , 2003 .

[29]  P. Kareiva,et al.  Conserving biodiversity coldspots: recent calls to direct conservation funding to the world's biodiversity hotspots may be bad investment advice. , 2003 .

[30]  E. Fleishman,et al.  Modeling and Predicting Species Occurrence Using Broad‐Scale Environmental Variables: an Example with Butterflies of the Great Basin , 2001 .

[31]  R. Siegel,et al.  USING POINT COUNTS TO ESTABLISH CONSERVATION PRIORITIES: HOW MANY VISITS ARE OPTIMAL? , 2001 .

[32]  Ralph Mac Nally,et al.  USING “INDICATOR” SPECIES TO MODEL SPECIES RICHNESS: MODEL DEVELOPMENT AND PREDICTIONS , 2002 .

[33]  M F Jefferson,et al.  Comparison of a genetic algorithm neural network with logistic regression for predicting outcome after surgery for patients with nonsmall cell lung carcinoma , 1997, Cancer.

[34]  M. Boyce,et al.  Relating populations to habitats using resource selection functions. , 1999, Trends in ecology & evolution.

[35]  J. W. Thomas,et al.  Ecological Uses of Vertebrate Indicator Species: A Critique , 1988 .

[36]  J. Reed,et al.  Effects of habitat type and degradation on avian species richness in Great Basin riparian habitats , 1999 .