Mapping of Acacia dealbata cover using winter and summer multiespectral images captured by drone in Verín County (Ourense, Galicia, Spain)

Abstract

Acacia dealbata Link is an invasive plant that spreads favoured by disturbances. The objective of this work is to mapping the coverage of A. dealbata using multispectral and RGB images and to evaluate the behaviour of winter, during flowering, and summer images. These images were captured using a DJI Matrice 210 drone in three sites in Verín County.

We used a Maximum Entropy (MaxEnt) programme to estimate the probability of A. dealbata presence at a pixel level. Ground truth (points of known presence) were based on a visual interpretation of RGB winter orthoimages, while the set of predictor variables were the five reflectance bands and a vegetation index.

The specificity and sensitivity of the MaxEnt models for the three study sites were acceptable in terms of ROC (Receiver Operating Characteristic) analyses AUC (Area under the Curve). The AUC values ranged from 0.80 to 0.86 for the winter images and from 0.83 to 0.90 for the summer ones.

These AUC values, and also the confusion matrices´ parameters obtained, have proven that our approach is a valid procedure to identify patterns of species distributions at local scale. There is also a better performance of models using summer than winter images.