G. Nyarko, H. Bayor


Species distribution modelling is important in conservation planning and many other fields of study. It is however often fraught with bias in the location data used to develop the models. Spatial thinning is one of the bias correction methods. It has been reported to be superior to the background correction method in modelling experiments. However, the effect of spatial thinning on predicted areas and model assessment characteristics are unreported. We examined the effects of spatial thinning on the potential distribution of 10 African indigenous vegetables (AIV). The aims of our study were to investigate the effect of different spatial thinning distances on (1) the potential predicted areas (present and future 2070) of 10 species of AIV and (2) model evaluation statistics. We applied spatial thinning to the location data using the R package ‘spThin’ at distances of 0, 10, 20, 40, 60, 80 and 100km. For each species MaxEnt was used to run 10 replicate models with cross-validation and a threshold of 10% training presence. There were between 54 and 564 location data points a species after cleaning of GBIF data and 153-25 after thinning at a spatial resolution of 100km. The area under the curve (AUC) of the receiver operating characteristic, Boyce Index and the true skill statistic (TSS) decreased with increasing spatial thinning distance but sensitivity remained relatively constant. There was consistency in the direction of prediction for eight of the 10 species while spatial thinning influenced the direction of prediction for two species. Future 2070 suitable climatic envelope may be larger than the present for six species, remain the same as present for three species and become smaller for one species. We concluded that while spatial thinning may be useful in correcting for under-estimation caused by clustered data, it might also lead to incompleteness in environmental space leading to unexpected results if not done with caution. Although the differences in the extent of suitable climatic envelope may imply reduction of overall biodiversity, no species was under serious threat of complete loss of suitable environment in the future.

Keywords: Area under the curve (AUC), Boyce Index, Sensitivity, Spatial Thinning, Specificity, True Skill Statistic, Vegetables

Full Text:



  • There are currently no refbacks.