Development of national forest inventory model using artificial intelligence algorithms: A case study of Istanbul Regional Directorate of Forestry

This thesis emphasizes the necessity of aligning National Forest Inventory (NFI) systems with international standards to ensure data collection and analysis are comparable and consistent. The research aims to evaluate the current state of Turkey's NFI practices, compare these with international standards and practices in developed countries, and propose recommendations for improvement. Within the scope of this thesis, the historical development of Turkey's NFI practices was reviewed, and the country's compliance with international inventory standards was evaluated. It was determined that the initial NFI initiatives in Turkey began in 1937; however, systematic and continuous implementation has not yet been established. Currently, data on Turkey's forest resources are primarily collected through management plans and limited databases such as Inventory-Statistics (ENVANİS). It has been identified that these databases inadequately cover biodiversity and ecological services. The COST Action E43 initiative was comprehensively examined as part of an international comparative analysis. This initiative aimed to harmonize European countries' NFI systems through common definitions and reporting techniques. The study revealed that the common definitions and "bridging" methods developed under COST E43 play a critical role in standardizing different national definitions. Methods employed in the thesis include Geographic Information Systems (GIS), remote sensing technologies, and artificial intelligence-based analyses. Additionally, methodologies utilized in the Istanbul Regional Directorate of Forestry (OBM), ranging from determining sample point quantities to applying these sample points in the field, were examined and discussed. Land use maps, one of the multi-layer filters for detecting forested areas, their changes, and identifying sample points, were produced for the years 2019-2023. Using this method, land-use classes were trained on a 3×3 km sampling grid, and data derived from Sentinel-2 satellite imagery were analyzed using the U-Net deep learning model. For terrestrial (numerical) data analysis, Random Forest (RF) and Artificial Neural Networks (ANN) algorithms were utilized, integrating these algorithms with both numerical and GIS-based relational databases. Additionally, web-based reporting systems were developed using the RShiny application. A significant finding of this research indicates that Turkey's existing NFI system has substantial shortcomings in terms of international standards, necessitating a broader scope. Adopting a more holistic approach to ecosystem services—such as biodiversity, carbon stocks, ecological services, and water budgets—as practiced in European standards was emphasized. The application in Istanbul OBM demonstrated high accuracy in land-use classification through remote sensing and artificial intelligence techniques. Another key outcome was the comparative analysis of the effectiveness of different algorithms in estimating forest carbon stocks and biomass. The results showed that double-entry volume calculations and RF and ANN models provide high accuracy in diameter increment, tree height, and age estimations. Moreover, using Tier 2 level for carbon stock calculations indicated that the contribution of forest areas to climate change could be monitored at international standards. Key recommendations for improvement presented in the study include making Turkey's NFI practices more systematic and continuous, revising data collection methods according to international reference definitions, and enhancing the effective use of GIS and remote sensing technologies. It is suggested that the proposed methods for data collection and analysis will provide a scientific basis for formulating forest policies in Turkey and play a critical role in meeting international reporting obligations. In conclusion, this thesis provides a comprehensive roadmap for harmonizing Turkey's forest inventory systems with international standards, contributing significantly to the sustainable management and conservation of the country's forest resources. It explicitly underscores the vital role of international cooperation and standardization processes in ensuring the sustainability of forest ecosystems.