Detección de árboles dañados por plaga en bosques de Abies religiosa en la Reserva de la Biosfera Mariposa Monarca, mediante fotografías aéreas infrarroja

Abstract

Forest pests are pathogens that cause mechanical or physiological damage to trees, such as deformations, disrupted growth, weakening, or even death, leading to important ecological, economic and social impacts. This study focused on the development of a technique for the detection of forest pests using infrared aerial photography. The general reflectance characteristics of healthy and damaged leaves are currently well known; Reid (1987) already described these features, with a shift toward blue and a reduced infrared reflectance as the dominant effects. As the plant disease progresses, the above effects become more apparent. The use of infrared digital aerial photographs allowed to obtain VIR (visible + infrared) images with four bands and a resolution of approximately one meter per pixel. Trees with some degree of deterioration and even dead individuals were identified and located through visual interpretation.

Color and infrared digital aerial photographs captured in March 2009 were used; two cameras were used: a Nikon D2X camera for the acquisition of images in the visible range (EV), and a Canon EOS Digital Rebel camera for infrared (IR) images. Once individual photographs were processed and organized, V and IR images were superimposed using the Photoshop editing program (Adobe™) Once composite V+IR (VIR) images were obtained, those covering the sampling area were selected and georeferenced. Rectified images were required to elaborate a mosaic encompassing the sampling area. The rectified images and the final mosaic had a spatial resolution of 90 centimeters per pixel.

The detection technique was designed using three methodological approaches: automatic, semi-automatic and manual processes. The semi-automatic and automatic modalities correspond to an assisted and unassisted spectral classification, respectively, while the manual method consisted in the direct observation of the photographs processed. The technique developed used as basis the photographic mosaic of the sampling area.

The unassisted and assisted spectral classification technique was carried out in the ERDAS Imagine image-processing software package. For the unassisted classification, tests were carried out considering various numbers of categories: 5, 10 and 15; the assisted classification included the spectral properties of each category used for the partition to group images into five categories: healthy forest, diseased forest, Juniperus scrubland, bare soil and shaded areas.

The accuracy of the technique for the detection of damaged trees was verified through field work, visiting different checkpoints where the health status of the tree was corroborated by direct observation and infrared photography at ground level.

A representative sampling area of the A. religiosa forest was established in the Monarch Butterfly Biosphere Reserve (RBMM), sufficient to encompass the largest number of damaged trees, but not so large as to excessively prolong the information-processing phases and make field sampling unattainable.

The analysis comprised an area of 1907 ha in Sierra Chincua, where the greatest affectation was observed in a core zone including 97 points (62%) with more than twice the density of individuals (11 trees/km²), relative to the buffer zone (4 trees/km²). This greater damage is the result of forest management policies, which have set no management (including sanitation) in the core zone.

At the end of this research work, we concluded that digital aerial photographs proved useful for the detection of damaged trees in Abies religiosa forests of RBMM. It is possible to obtain multispectral images using a low-cost photographic technology that is relatively simple and widely available. Our study showed that the best method to detect damage in A. religiosa forests in RBMM is the visual interpretation of aerial photographs, yielding a detection efficiency of over 98%. The method used has a greater cost- effectiveness compared to helicopter overflight and field work. Likewise, the method developed in this research work is a contribution to the detection of forest pests.