Collaborative Research: Spatial inference and prediction with biogeographical data

点击翻译按钮获取中文摘要

Maps of actual or potential species distributions are required for many aspects of resource management and conservation planning including biodiversity assessment, habitat management and restoration, single- and multiple species and habitat conservation plans, population viability analysis, modeling community and ecosystem dynamics, and predicting the effects of climate change on species and ecosystems. A growing number of quantitative methods are being used both inferentially, to identify the parameters that determine habitat suitability, and predictively, to assign habitat value to locations where biological survey data are lacking (most of the earth's surface). There are three impediments to the effective use of these modeling tools by both researchers and conservation and resource managers: a) too few of the existing applications explicitly incorporate the spatial dependence inherent in biospatial data into the modeling methods b) the statistical and GIS modeling tools are not always well integrated, and, c) the proliferation of potential methods and conflicting results regarding their efficacy is daunting to users. The investigators will 1) synthesize existing information on spatial prediction using biogeographical data, 2) strategically plan and execute a set of modeling experiments, and, based on these, 3) develop a framework to guide the operational use of these methods for biodiversity assessment and landscape management. Comparative modeling experiments will be executed using species distribution and abundance data spanning the three major ecological regions in southern California (desert, mountain, coastal), for plants from vegetation surveys and reptiles and amphibians (herptiles) surveyed in a multi-year monitoring program. The methods tested will include parametric and non-parametric statistical (generalized) models, machine learning approaches, and those incorporating spatial dependence (regression kriging, spatial autoregressive models).The proposed research is innovative because it will provide a broad comparison of modeling methods for real biological datasets that vary in their sample design, measurement scale, and spatial dependence, but were collected in the same bioregion, and will focus on biogeographical modeling of spatial dependence in plant and animal species distribution and abundance. It will result in a framework that can be used by researchers and resource managers to select an approach to modeling that is best suited to their biogeographical data and questions. The project will directly benefit society because it is collaborative with the Biological Resources Division of the US Geological Survey, the federal agency with a leadership role in spatial data archiving and analysis and biological information infrastructure. Thus, the framework and recommendations will be directly conveyed to resource and data managers.