Context and objectives
The forests in Bosnia-Herzegovina are suffering from important sanitary problems. In particular, bark beetles (Scolytidae), are killing large numbers of conifers. Fungi transported by the beetles stain the wood of the attacked trees and alter its technological properties, resulting in important reductions of the technological qualities of the timber. Control of these harmful pests can be achieved by sanitation thinning and clearfelling, and by using specific lures (pheromones) to attract the beetles to traps or trap trees.
Over half of the area suffering outbreaks are mined, therefore making direct control impossible. Mined areas attacked by bark beetles represent a threat for the neighbouring areas, as they act as " reservoirs".
Total treatment of these mined areas is not possible because of the high costs involved and because of higher priorities for de-mining. There is, however, a possibility to determine priorities among the mined outbreak areas, in order to direct local de-mining to limited high-priority zones.
This study contributes to this project by analysing the potential of high resolution satellite imagery to detect, characterise and rank infestation foci.
Expected scientific results
For the Kruskal-Wallis test and the multiple linear regression analysis, two analyses were performed. The first one considered all field observations. The second one addressed only the observations made in the pure Norway spruces stands, in order to decrease environmental heterogeneity.
The Kruskal-Wallis test detected only significant difference between infestation levels and reflectance values of the near infrared channel and middle infrared channel and this for all field observations; this is possibly a consequence of the small number of field observations.
The multiple linear regression however, showed a significant relationship between infestation levels and the near infrared channel data and middle infrared channel data. The relationship is stronger (R² = 0.313) in the pure Norway spruce stands than in the whole set of stands (R² = 0.183).
The NDVI analysis permitted to create a mask that only takes the forest zones and to eliminate the cloudy zones. We made several PCA with different combinations of spectral bands. Some of these neo - channels created could put in evidence some typical zones our images (road, spruces…).
For the supervised and a non supervised classification, we made several classification tests, considering different spectral band associations. Each of these classifications was made either inside a mask taking the whole set of stands (conifers + broad-leafs), or inside a mask taking only the pure Norway spruce stands. Whatever the association of spectral bands and the mask used, we could not accurately locate the infested zones.