Object-based assessment of tree attributes of Acacia tortilis in Bou-Hedma, Tunisia

Acacia tortilis subsp. raddiana represents the most important woody species in the pre-Saharan zone. It is the only forest tree persisting on the edge of the desert. Due to tree/environment interactions, canopy sub-habitats arise, enabling an increased storage of soil water, soil nutrients and soil oxygen. Depending on their density, they can also reduce erosion and reverse desertification. Soil erosion and desertification are the main problems faced by the UNESCO Biosphere Reserve in South-Tunisia (Bou-Hedma National Park). The restoration of the original woodland cover to combat desertification (particularly) by afforestation and reforestation of Acacia tortilis goes hand in hand with a climate change in the Biosphere Reserve, also influencing rural population outside the Biosphere Reserve. In order to study the different effects of woodland restoration in Bou-Hedma, the number of Acacia trees and their attributes have to be known. High resolution satellite imagery (GeoEye-1), was used with a GEOBIA approach. Field measurement of bole diameter, crown diameter and tree height were collected at > 400 locations. After segmentation, correlations with > 200 object features and tree attributes were calculated. For crown diameter and tree height, high correlations were observed with the features area and GLCM Entropy Layer 4 (90 degrees). Relations between these features and measured tree attributes were modeled, resulting in RMSE values of resp. 1.47 m and 1.62 m for crown diameter estimation and 0.92 m for tree height. The results show that a GEOBIA working strategy is suitable for estimating tree attributes in open forests in semi-arid regions.