Urban green space classification using Object-Based Image Analysis (OBIA) and LiDAR fusion: Accuracy evaluation and landscape metrics assessment

Abstract

With over two-thirds of the global population projected to live in cities by 2050, accurately mapping urban green spaces is increasingly important for sustainable development. This study integrates Object-Based Image Analysis (OBIA) and LiDAR data fusion to improve green space classification in three urban catchments in Helsinki, representing high (Ita-Pasila), intermediate (Pihlajamaki), and low (Verajamaki) land-use intensities. Using highresolution color-infrared (CIR) aerial orthophotographs enhanced by LiDAR-derived vegetation height data, the method effectively identified vegetated areas. Results were validated against a reference dataset using standard accuracy metrics and landscape structure indices. The results show that the OBIA method yielded green space area estimates within 1-4 % of the reference, but tended to produce more fragmented landscape configurations in high land-use intensity urban areas, resulting in higher numbers of patches and lower aggregation indices. Conversely, results in less urbanized Verajamaki closely matched the reference data both spatially and structurally. These discrepancies underscore the inherent challenges in interpreting spatial patterns within complex urban morphologies, particularly where spectral information is limited by shading, like in Ita-Pasila. Nevertheless, the OBIA-LiDAR fusion approach demonstrated strong reliability in less structurally complex environments and provides valuable data for watershed-scale hydrological and ecological modeling.