Repository of DUTH

Parameter estimation in artificial acacia clusters using Sentinel-2 data (Master thesis)

Ζαφειρόπουλος, Βασίλειος/ Zafeiropoulos, Vasileios

From 1993 and onwards, the EU regulation began to apply in the European Union, including our country the regulation (EC) No 2080/92 providing for the afforestation of agricultural land and forest land and the development of forestry activities on agricultural holdings in order to improve the country's forest resources and to provide alternative or supplementary income to producers. This regulation marked a change in the EU's forestry strategy. In particular, it provided for the distinction between conifers and broadleaves, financial aid for the costs of installation, maintenance and loss of income. The shelf life of each plantation is 20 - 25 years. According to data from the Ministry of Agriculture, Robinia Pseudoacacia L occupies the largest area of forest tree plantations of the regulation 2080/92 as they account for 43% of their total area, namely 9,819 hectares. The science of remote sensing is an effective tool for studying the environment and vegetation. Much of modern remote sensing is done with the help of sensors mounted on satellites and manned or unmanned aircraft, and it is possible to obtain measurements related to the interaction of forest surfaces and electromagnetic radiation. The purpose of this thesis is to estimate the parameters of the Robinia Pseudoacacia L forestry species in forested plantations in northern Evros, using linear regression models, artificial neural networks and geographic information systems. The parameters of the forest species that were measured on the ground are the number of individuals, the diameter at the base of the tree, the diameter of breast height, the diameter of the crown and the height. Subsequently, prediction models of these parameters were developed using satellite images from the European Union satellite, Sentinel-2. Non-atmospheric corrected images and atmospherically corrected images were used respectively. Finally, using the neural networks, the accuracy of the prediction model was compared with that of linear regression