ADVANCING FOREST LAND MONITORING IN ISTANBUL REGIONAL DIRECTORATE OF FORESTRY: INTEGRATING U-NET DEEP LEARNING
Abstract
This study presents a comprehensive analysis of land use and land cover change within the Istanbul Regional Directorate of Forestry (RDF) utilizing semantic segmentation referred to as pixel-based classification. Focusing particularly on forest land dynamics, Sentinel-2 satellite imagery spanning five years from 2019 to 2023 was processed using a U-Net architecture. The study area encompasses diverse forest ecosystems, urban/built-up areas, water bodies, rangelands, wetlands, and agricultural lands. Through the application of advanced remote sensing techniques, significant changes in forest and rangeland were identified and quantified, 15.250 and 13.226 hectares of area decreased in five years, shedding light on the drivers and implications of land use transformations in this critical region. Controversially, built area and agricultural lands were increased by 13.878 and 15.953 hectares over 5 years. The findings contribute to a deeper understanding of forest dynamics and inform sustainable management strategies for preserving the ecological integrity and socio-economic value of forested landscapes within the Istanbul RDF. Additionally, the results reveal the average F-1 Score for each land cover class is approximately 90% for each year, with forested areas achieving an average F-1 score of about 92%, demonstrating the robustness and accuracy of the classification approach.
Keywords
Forest Management, Remote Sensing, Deep Learning, Istanbul, U-Net
References
- Adepoju, M. O., Millington, A. C., & Tansey, K. T. (2006). Land use/land cover change detection in metropolitan Lagos (Nigeria): 1984–2002. P 1-5 in ASPRS 2006 Annual Conference Reno, Nevada May.
- Atalay, I., Efe, R., & Öztürk, M. (2014). Ecology and classification of forests in Türkiye. Procedia-Social and Behavioral Sciences, 120, 788-805. https://doi.org/10.1016/j.sbspro.2014.02.163
- Atmis, E., Günsen, B. H., Yücedağ, C., & Lise, W. (2012). Status, use and management of urban forests in Türkiye. South-east European Forestry: SEEFOR, 3(2), 69-78. https://doi.org/10.15177/seefor.12-08
- Atmis, E., & Cil, A. (2013). Sustainable forestry in Türkiye. Journal of Sustainable Forestry, 32(4), 354-364. https://doi.org/10.1080/10549811.2013.767210
- Bagan, H., & Yamagata, Y. (2012). Landsat analysis of urban growth: How Tokyo became the world's largest megacity during the last 40 years. Remote Sensing of Environment, 127, 210-222. https://doi.org/10.1016/j.rse.2012.09.011
- Belgiu, M., & Drăguţ, L. (2016). Random forest in remote sensing: A review of applications and future directions. ISPRS journal of photogrammetry and remote sensing, 114, 24-31.
- Bozkurt, S. G., Kuşak, L., & Akkemik, Ü. (2023). Investigation of land cover (LC)/land use (LU) change affecting forest and seminatural ecosystems in Istanbul (Türkiye) metropolitan area between 1990 and 2018. Environmental Monitoring and Assessment, 195(1), 196. https://doi.org/10.1007/s10661-023-10975-7
- Cui, L., & Shi, J. (2012). Urbanization and its environmental effects in Shanghai, China. Urban Climate, 2, 1-15. https://doi.org/10.1016/j.uclim.2012.10.008
- Diakogiannis, F. I., Waldner, F., Caccetta, P., & Wu, C. (2020). ResUNet-a: A deep learning framework for semantic segmentation of remotely sensed data. ISPRS Journal of Photogrammetry and Remote Sensing, 162, 94-114. https://doi.org/10.1016/j.isprsjprs.2020.01.013
- ESA. (2023, June). Copernicus Open Access Hub. https://scihub.copernicus.eu/dhus/#/home Foody, G. M. (2002). Status of land cover classification accuracy assessment. Remote sensing of environment, 80(1), 185-201.