AI Summary & Key Takeaways
AI InsightIn this context, the analysis proceeded with 13 clusters, which achieved the highest average Silhouette index (0.231), and the resulting clustering structure demonstrated an accuracy rate of 82.8% and an overall (average) F1-score of 81.5%, supporting the validity of the classification.
📌 3 Key Takeaways
Abstract
This study focuses on the spatially identified structural character of the landscape in the case of Bartın province in the Western Black Sea Region of Türkiye. In the first stage, Landscape Character Types (LCTs), which are the integrated expression of reclassified climate, geology, geomorphography and landscape pattern components, were created at two levels as L1-Regional and L2-Subregional (L1: 157 types, 4018 units; L2: 449 types, 7757 units, respectively). In the second stage, the raster data for the L1-LCTs in the study area were transformed into {1(present), 0(absent)} values and PCA scores. Clustering algorithms, including twostep and k-means methods, were applied to L1-LCTs data, and clustering structures were compared across different cluster numbers. The optimal number of clusters was identified based on Silhouette index values, and the optimal cluster-based performance metrics were calculated accordingly. In this context, the analysis proceeded with 13 clusters, which achieved the highest average Silhouette index (0.231), and the resulting clustering structure demonstrated an accuracy rate of 82.8% and an overall (average) F1-score of 81.5%, supporting the validity of the classification. In the third stage, landscape diversity, landscape density, naturalness ratio, cluster ratio, relative landscape richness, average nearest neighbor distance and Shannon diversity metrics were calculated using FRAGSTATS based on the L2-LCTs included in the 13 clusters. When the clusters are examined, clusters 6 and 9, which constitute 24.67% and 17.35% of the study area respectively, have high normalized landscape naturalness rate (0.99 and 1.00 respectively) and normalized landscape diversity values (0.67 and 0.65). This study provides a robust framework for identifying spatially distinct natural and agricultural systems through structural landscape analysis. By integrating spatial typology with quantitative metrics, the method enables a consistent and diagnostic understanding of landscape variation, which can guide planning, management, and protection decisions.