z-logo
Premium
Predicting habitat suitability for rare plants at local spatial scales using a species distribution model
Author(s) -
Gogol-Prokurat Melanie
Publication year - 2011
Publication title -
ecological applications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.864
H-Index - 213
eISSN - 1939-5582
pISSN - 1051-0761
DOI - 10.1890/09-1190.1
Subject(s) - habitat , edaphic , ecology , rare species , population , scale (ratio) , species distribution , spatial ecology , ranking (information retrieval) , environmental science , geography , biology , cartography , computer science , soil water , demography , machine learning , sociology
If species distribution models (SDMs) can rank habitat suitability at a local scale, they may be a valuable conservation planning tool for rare, patchily distributed species. This study assessed the ability of Maxent, an SDM reported to be appropriate for modeling rare species, to rank habitat suitability at a local scale for four edaphic endemic rare plants of gabbroic soils in El Dorado County, California, and examined the effects of grain size, spatial extent, and fine‐grain environmental predictors on local‐scale model accuracy. Models were developed using species occurrence data mapped on public lands and were evaluated using an independent data set of presence and absence locations on surrounding lands, mimicking a typical conservation‐planning scenario that prioritizes potential habitat on unsurveyed lands surrounding known occurrences. Maxent produced models that were successful at discriminating between suitable and unsuitable habitat at the local scale for all four species, and predicted habitat suitability values were proportional to likelihood of occurrence or population abundance for three of four species. Unfortunately, models with the best discrimination (i.e., AUC) were not always the most useful for ranking habitat suitability. The use of independent test data showed metrics that were valuable for evaluating which variables and model choices (e.g., grain, extent) to use in guiding habitat prioritization for conservation of these species. A goodness‐of‐fit test was used to determine whether habitat suitability values ranked habitat suitability on a continuous scale. If they did not, a minimum acceptable error predicted area criterion was used to determine the threshold for classifying habitat as suitable or unsuitable. I found a trade‐off between model extent and the use of fine‐grain environmental variables: goodness of fit was improved at larger extents, and fine‐grain environmental variables improved local‐scale accuracy, but fine‐grain variables were not available at large extents. No single model met all habitat prioritization criteria, and the best models were overlaid to identify consensus areas of high suitability. Although the four species modeled here co‐occur and are treated together for conservation planning, model accuracy and predicted suitable areas varied among species.

This content is not available in your region!

Continue researching here.

Having issues? You can contact us here