Estimation of crown area of wild pistachio single trees using DSM of UAV aerial images in Baneh Research Forest, Fars province

Crown cover of trees in arid and semi-arid regions is of great importance as the parts of ground under tree canopies are favourable environments for regeneration establishment and survival of other organisms. Therefore, it is essential to be aware of crown cover status of trees and shrubs, monitor their changes and assess their health. Remote sensing data obtained by most of satellites do not make unbiased measurement of crown area of single trees possible as their spatial resolution is not suitable for this purpose. On the other hand, these data are not available at the time researchers need, and if available, they are expensive. Continuous progress in remote sensing results in access of researchers to unmanned aerial vehicle (UAV) that imagery taken by this device have not only very high spatial resolution for precise study of biometric characteristics of single trees, but also availability of images at favourable time for researchers. Considering this issue, this study was aimed to evaluate UAV imagery and corresponding digital surface model (DSM) resulted from stereo images to estimate crown area of wild pistachio single trees in Zagros vegetation zone. Moreover, it was also aimed to investigate the impact of spatial resolution of DSM on accuracy and precision of estimating crown area of the trees.
A part of Baneh Research Site with area of 45 ha was selected that was purely cover by wild pistachio. In October 2016, the study area was covered by 1076 aerial images with 3 cm spatial resolution taken by a UAV flown at 70 m above the study area. The images were geo-referenced using 12 ground control points collected by Leica Viva GS15 three-frequency global positioning system. The DSM extracted using bundle adjustment method was then resampled to three different spatial resolutions of 3, 50, and 100 cm. An object-based processing method was proposed in order to automatically extract the crown boundaries from DSM. Number of 100 wild pistachio trees were randomly selected that their spatial position were registered and their crown areas were measured before. The mean crown area obtained by DSM of UAV imagery and the observed mean were compared by paired sample t-test. In addition, three indices of root mean squared error (RMSE), model efficiency (ME), and bias score (BS) were applied to assess the precision of results.
The orhtophoto mosaic of the study area was obtained with RMSE of 8 cm. Among 100 wild pistachio trees randomly selected for this study, 100, 89, and 80 trees were recognized on DSMs with spatial resolution of 3, 50, and 100 cm, respectively. Although there was no significant difference between observed mean crown area of 80 trees (51.3 m2) recognized on all DSMs and mean crown area estimated on DSMs with 3 (42.6 m2), 50 (44.5 m2), and 100 cm (39.2 m2) (α=0.05), but their correlation decreased. Moreover, RMSE increased and ME and BS decreased with decreasing spatial resolution of DSM.
In general, it was concluded that DSM of UAV imagery is an appropriate means to recognize and measure crown area of wild pistachio single trees in the study area that obviously separated tree crowns from their shadow and other objects. Moreover, it was revealed that with decreasing spatial resolution of DSM, data processing became easier and there was no significant difference between observations and measurements but the precision of results decreased.