Detecting Spatial Outliers in Landscape Structure Using K-Means Clustering and Chernoff Face Analysis Across Temporal Scales

Environmental datasets are often characterized by complex spatial structures and the presence of atypical observations that may influence the interpretation of landscape patterns. This study proposes a comparative framework for identifying spatial outliers in landscape structure using two complementary approaches: K-means clustering and multivariate visual exploration based on Chernoff faces. The analysis is conducted on two temporal snapshots (1956 and 2019) representing long-term changes in land use and land cover in the Zemplínska Šírava region, Eastern Slovakia. Outlier detection results from both approaches are systematically compared to assess their consistency and robustness. The two methods show substantial correspondence in the identification of anomalous landscape units. The number of land-cover classes increases from 19 in 1956 to 25 in 2019, reflecting increased landscape heterogeneity over time. Persistent spatial outliers across both methods and time periods include road networks and associated land and broad-leaved forest with continuous canopy, indicating the structural stability of these landscape elements despite long-term transformation. The results demonstrate that combining clustering-based approaches with multivariate visual analytics can improve the interpretation of complex spatial patterns in environmental data. However, the study is exploratory in nature, and the interpretation of Chernoff faces involves inherent visual subjectivity, which should be considered when evaluating the results. The proposed framework should therefore be regarded as a complementary exploratory tool rather than standalone analytical evidence. Future research may extend this framework by integrating identified spatial outliers into environmental assessment models focused on biodiversity patterns, ecological connectivity, and sustainable landscape planning.