Diachronic mapping of invasive plants using airborne RGB imagery in a central Pyrenees landscape (South-West France)

The rapid spread of invasive plant species (IPS) over several decades has led to numerous impacts on biodiversity, landscape and human activities. Often, long-term invasions lead to a significant change in ecosystems due to significant competition with native species, but they also adversely affect diverse agro-ecosystems. Early detection and knowledge of the dynamics and spatio-temporal distribution of IPS is crucial to better understand invasion patterns and conduct appropriate activities for management. Therefore, high resolution remote sensing using aerial photos provides great potential for detecting and mapping the spatial spread of IPS.

To characterize the spatial dynamic of IPS, mapping on two study sites located along the Pique valley in the central Pyrenees in the southwest of France was performed. The targeted species included Reynoutria japonica and Impatiens glandulifera. The areas occupied by these species over the past decade were assessed from ortho-rectified RGB aerial photographs (2019, 2016, 2013, 2010, resolution 0.2 – 0.5 m).

A supervised classification based on the random forest algorithm was performed using pixel attributes. From previous aerial imagery obtained, the original spectral bands (R-G-B) was used, to which vegetation indices (CIVE, VDVI, NGRDI) and textures (Energy, Entropy) were added to improve target species detection. Ground truth data were also collected during field investigation and were randomly divided into two independent groups, one for learning (50%) and the other one for validation (50%).

The classification models yielded a mean prediction accuracy (F-score) of 0.90 with values ranging from 0.87 to 0.92 at site 1, and 0.87 at site 2 with values ranging from 0.81 to 0.91. The model’s ability to correctly detect IPS allowed further examination of the processes favoring their emergence. Classifications results show that the expansion of IPS in this region was closely related to the presence of roads and to environmental disturbance by human activity such as land clearing.