Modeling and Predicting Species Occurrence Using Broad‐Scale Environmental Variables: an Example with Butterflies of the Great Basin

摘要: : If occurrence of individual species can be modeled as a function easily quantified environmental variables (e.g., derived from geographic information system [GIS]) and the predictions these models are demonstrably successful, then scientific foundation for management planning will strengthened. We used Bayesian logistic regression to develop predictive resident butterflies in central Great Basin western North America. Species inventory data values 14 49 locations (segments canyons) Toquima Range ( Nevada, U.S.A.) were build models. Squares also accommodate possibly nonmonotonic responses. obtained statistically significant 36 56 (64%) butterflies. The explained 8–72% deviance those species. Each independent was at least one model, squared versions five contributed Elevation included more than half Models four variables; only variable about conducted preliminary tests two our by using an existing set on neighboring Toiyabe Range. compared conventional classification with posterior probability distributions modeling. For latter, we restricted high ( 70%) predicted presence or absence. perform further after conducting inventories new nearby Shoshone Mountains, which have computed remotely acquired topographic data, digital-terrain microclimatic models, GIS computation. Resumen: En la medida en que se pueda modelar presencia de especies enfuncion ambientales faulmente cuantificables (por ejempto, derivadas SIG) y las predicciones estos modelos puedan mostrar con exito, el fundamento cientifico para planeacion del manejo sera reforzada. Empleamos regresion logistica Bayesiana desarrollar predictivos mariposas residentes region Gran Depresion oeste Norteamerica. Usamos datos inventario valores localidades (segmentos canones) cordillera USA) construir modelos. Los cuadrados tambien fueron usados acomodar posibles respuestas no monotonicas. Obtuvimos estadisticamente significativos mariposas. Estos explican 8-72% variabilidad estas especies. Cada una independientes fue significativa al menos un modelo versiones cuadradas cinco contribuyeron los La elevacion incluyo mas mitad incluyeron 1-4 casi solo significativa. Realizamos pruebas preliminares dos cuatro usando juego existente sobre vecina Toiyabe. Comparamos clasificaciones logisticas convencionales contra distribuciones probabilidad posteriores bayesianos. Para ultimas, restringimos nuestras alta predecir o ausencia. Futuras seran realizadas despues llevar cabo inventarios nuevas Shoshone, cuales hemos calculado topograficos adquiridos por percepcion remota, terreno digital microclimaticos, calculos SIG.