Uso de imágenes hiperespectrales para la predicción del marchitamiento de Pinus halepensis (Mill.) en el bosque mediterráneo

Abstract

The increasing negative effects of climate change and the emergence of invasive species in forests around the world require the development of innovative methods to monitor and quantitatively measure the health status of woodlands. These effects are especially notable in the Mediterranean area, where the decline of stands due to recurrent droughts has increased the damage caused by secondary pests whose populations would otherwise be in balance. Remote sensing technologies allow us to work on large surfaces with reasonable precision. In particular, new spectral indices obtained from high-resolution hyperspectral and thermal images have been shown to be good predictors for the early detection of physiological changes related to diseases. In this pilot study developed in a stand of Pinus halepensis in the Comunitat Valenciana, a controlled simulation of a decay is carried out by means of sequential girdling of trees, making a subsequent field monitoring of the caused decay. Through a hyperspectral camera, the spectral information of each of these trees is analyzed in relation to their discoloration and state of observed decay. The proposed methodology allows the detection of affected trees three months before the appearance of visual symptoms, obtaining a precision higher than 0.9 with Random Forest and Support Vector Machine classifiers. The vegetation indices with better results were PRI, VGO1, VGO2, GM1 and OSAVI. This pilot study allows us to think that some of these indices can be used in the early detection of general pine wilt and, therefore, have application in the monitoring of the main threats to European forests, borer pests or quarantine organisms such as Bursaphelenchus xylophilus.