Local Climate Zone Classification Using YOLOV8 Modeling in Instance Segmentation Method
Melike Nicancı Sinanoğlu, Şinasi KayaLocal climate zones play an important role in understanding microclimates in urban areas, contributing to urban planning, environmental sustainability and human comfort. Istanbul, a city connecting the European and Asian continents, creates microclimate diversity in city areas with the influence of different land use models. The concept of Urban Head Island (UHI) arises when urban areas have different temperatures with neighboring rural areas. The lack of a universally accepted definition of the concept of urban and rural areas has created difficulties in the evaluation of this concept. In response to this situation, a standardized Local Climate Zone (LCZ) classification system for urban temperature observations was created. This study performs LCZ classification with YOLOV8, one of the deep learning-based image segmentation models, using high-resolution Istanbul Google Earth images. Labeled data was created from WUDAPT's Google Earth images according to the post "Things to consider when creating LCZ training areas". Model training was carried out by creating a dataset by labeling high-resolution, bird's-eye view images of Istanbul obtained from Google Earth, paying attention to the diversity of LCZ categories. Box P 0.263, R 0.341, mAP50 0.317, mAP50-95 0.219 and Mask P 0.254, R 0.318, mAP50 0.404, mAP50-95 0.305 model metric values obtained after training were calculated. Although these values are below 50 percent, the LCZ class predictions appear to be largely accurate in the labeled result images. Metric results are important for improving the model and detecting weak points. This research contributes to the field of urban climate studies by providing a robust and scalable approach to LCZ classification using advanced deep learning techniques. The results can form the basis for urban planning, environmental sustainability and informed decision-making processes in the context of Istanbul's urban environment.
