Does plot size affect the performance of GIS-based species distribution models?
Journal of Geographical Systems
ITC, International Institute for Geo-Information Science and Earth Observation, P. O. Box 6, 7500AA Enschede, Netherlands; Department of Biology, McMaster University, 1280 Main St West, Hamilton, ON L8S4K1, Canada; Department of Biology, Dalhousie University, 1350 Oxford St, Halifax, NS B3H 4J1, Canada; Department of Biology, University of Florida, P. O. Box 118525, Gainesville, FL, United States; International Livestock Research Institute (ILRI), P. O. Box 30709, 00100 Nairobi, Kenya; Department of Biological Sciences, University of Windsor, 401 Sunset Ave, Windsor, ON N9B 3P4, Canada
Species distribution models are used extensively in predicting the distribution of vegetation across a landscape. Accuracy of the species distribution maps produced by these models deserves attention, since low accuracy maps may lead to erroneous conservation decisions. While plot size is known to influence measures of species richness, its effect on our ability to predict species distribution ranges has not been tested. Our aim is to test whether the accuracy of the distribution maps produced depend on the size of the plot (quadrat) used to collect biological data in the field. In this study, the presences of four plant species were recorded in five sizes of circular plots, with radii ranging from 8 to 100 m. Logistic regression-based models were used to predict the distributions of the four plant species based on empirical evidence of their relationship with eight environmental predictors: distance to river, slope, aspect, altitude, and four principle component axes derived using reflectance values from Aster images. We found that plot size affected the probability of recording the four species, with reductions in plot size generally increasing the frequency of recorded absences. Plot size also significantly affected the likelihood of correctly predicting the distribution of species whenever plot size was below the minimum size required to consistently record species' presence. Furthermore, the optimal plot size for fitting species distribution models varied among species. Finally, plot size had little impact on overall accuracy, but a strong, positive impact on Kappa accuracy (which provides a stronger measure of model accuracy by accounting for the effects of chance agreements between predictions and observations). Our results suggest that optimal plot size must be considered explicitly in the creation of species distribution models if they are to be successfully adopted into conservation efforts. © 2010 Springer-Verlag.
environmental factor; GIS; mapping; numerical model; regression analysis; spatial distribution; species richness; vegetation; Namibia