Improved methods of classification of multispectral aerial photographs: evaluation of floodplain forests in the inundation area of the Danube

The Gabčíkovo hydroelectric power plant has significantly influenced Danube water regime, thus the condition of floodplain forests in the region. Forest condition has been regularly monitored since 1995 using aerial photos. The subject of this study was to improve the procedure of floodplain forest health evaluation based on digital multispectral aerial images. Firstly, the forest mask was created with overall accuracy 89%, and next, tree health was evaluated using defoliation as health indicator. We applied orthogonal transformation of 4 original bands of multispectral imagery into two-dimensional space. Marginal values of digital numbers (DN) of the first component (New Synthetic Channel - NSC1) were defined by fully foliated willow and poplar. The second component (NSC2) was optimised for damage estimation. Calculated DN values of NSC2 represented a perpendicular distance from the line of DN values of the first component. The distance from the line was proportionate to tree damage extent in a given pixel. We generated linear regression model between pair values of NSC2 and defoliation evaluated for 38 trees in the field, respectively, from aerial photos. A decline prediction resulted in r-square equal 0.86. Finally, we used the model to predict defoliation for each picture element (pixel) of the component NSC2.