Forest structure analysis on very high resolution images

Forest structure information has been of interest for a long time, being linked to thepossibilities of maximizing forest production in terms of ecological stability. Forest structure parameterscan be used in assessing the degree of closeness to natural ecosystems for any type of forest stand. Theremote sensing methods have the advantage of large scale interpretation of results, facilitating theanalysis at different levels – from tree level to landscape level, all this in the context of satellite imagesprice decrease. The high resolution images have their advantages and disadvantages in forest structureassessments but, accurate information about the forest ecosystems can be obtained with the rightprocessing techniques. The evolution of satellite sensors parameters has lowered the analysis level fromthe landscape level somewhere close to tree level; the modern very high spatial resolution satellites havea spatial resolution similar to medium scale aerial photos, with the advantage of price and multispectralcharacteristics. The different spectral channels can be sensitive to several forest parameters and, in theright combination, can offer an “image” of forest state and structure. Before using the data extractedfrom satellite images, is important to make accuracy studies and establish the correct level of analysis.In this context, the paper’s objective is to analyze the possibilities of measuring forest structureparameters on IKONOS and aerial images and establish the optimum level for information processing.The parameters taken into account were crown diameter, canopy cover index and number of trees perarea unit; the forest stands are consisting of mixed forests of silver fir and Norway spruce, with a highhomogeneity. The results of the study showed that a better accuracy is obtained on IKONOS rather thanaerial photos, especially in the case of crown diameter and canopy cover index; in the case of number oftrees per area unit, the two types of images being similar. In both cases, remote sensing methods tendto reduce the variability of the data, in comparison to ground methods. The best comparability of data isachieved at the stand level, even if the information is originating from measurements of individual trees(as in the case of crown diameter). The conclusion that should be drawn from this study is that the levelof analysis can be lowered even to the tree level, but the data should be analyzed on a stand level forbetter comparability.