Object-based analysis of 8-bands Worldview2 imagery for assessing health condition of desert trees

High spatial resolution panchromatic and multispectral WorldView2 images were used to assess the health condition of Tamarugo (Prosopis tamarugo Phil.) trees in the hyperarid Atacama desert in Northern Chile. Tamarugo is a very valuable species for biodiversity conservation due to its endemic character and limited distribution range. An object-based analysis of a panchromatic WorldView2 image was carried out to identify single trees and to measure canopy area. The NIR/red-edge ratio index (NIR/Re) and Red-edge Normalized Difference Vegetation Index (ReNDVI) were calculated using the near-infrared band (772-890 nm) and the red-edge band (704-744 nm) of the 8-band multispectral WorldView2 image in order to assess health condition at the tree level. The widely used Normalized Difference Vegetation Index (NDVI) was also calculated as reference. Finally, the results were compared with field data taken over 48 trees. The developed object-based algorithm worked out satisfactory with WorldView2 panchromatic data, especially in defining canopy contours and splitting overlapping trees. Within the study area - Bellavista plantation in Pampa del Tamarugal National Reserve- a total of 90,350 trees were automatically identified with a total canopy area of 5.2 km2. The tree average canopy size was 57.6 m2. Single canopy identification allowed an accurate assignment of average spectral values from vegetation index images to the object trees. This way single values of the three vegetation indices were used to evaluate vegetation condition at tree level. Correlation coefficients between vegetation index values and green canopy percentage (GC%) measured for 48 trees in the field were significant in all cases, with the highest one for the traditional NDVI.