Mapping Lepidoptera and plant alpha-diversity across a heterogeneous tropical dry forest using field and remotely sensed data with spatial interpolation

Abstract

We advocate that to secure ecological integrity, conservation should focus on the spatial overlap of biodiversity occurring for multiple taxonomic groups. However, there is a lack of practical and accurate measures for estimating the continuous distribution of insect diversity across spatially complex landscapes to enforce such conservation strategy. Based on 216 sampling data, we predicted plant and caterpillar species density across a 200 km2 spatially complex tropical dry forest landscape, using reflectance values of a SPOT 5 satellite imagery, altitude values from a digital elevation model, and improved model performances by incorporating spatial autocorrelation of residuals with kriging interpolation. Spectral information and altitude accounted for 21–25 % of the variation in woody plant species density whereas the residuals did not demonstrate spatial autocorrelation. In contrast, spatial dependence accounted for most of the 35 % explained variation in caterpillar species density; the spatial range of autocorrelation of caterpillars however, is broader than that of host plants. Although local species density of caterpillars and woody plants showed no relationship, specific areas demonstrated an overlap of species diversity for both taxonomic groups. Spatially explicit models with spectral information as predictors have the potential to estimate woody plant species density across complex landscapes, while the prediction of caterpillar diversity could be done analyzing the spatially autocorrelated structure of species density, implicating labor intensive, but not necessarily costly sampling schemes.