This study evaluates a Level of Detail 1 (LoD1) workflow for estimating building floor counts and gross floor area in Buntusu Subdistrict, Makassar. The study combines 2024 high-resolution aerial imagery with Digital Surface Model (DSM) and Digital Terrain Model (DTM) data. Building footprints were generated with the Segment Anything Model (SAM) and refined through manual digitization, while building heights were calculated from a normalized surface model (nDSM = DSM - DTM). Heights were converted to floor-count estimates using an operational floor-height assumption and contextual correction for atypical roofs and non-occupiable architectural elements. Accuracy was assessed using 200 buildings selected by stratified random sampling and verified in the field. An exact match was obtained for 173 buildings, producing an overall floor-count accuracy of 86.5%; 27 buildings differed from the observed count. The workflow delineated 6,612 buildings. Estimated gross floor area ranged from 5.63 to 3,786.06 m² per building, with a mean of 173.00 m², a median of 129.65 m², and a total of 1,118,107.23 m². Larger values were concentrated mainly along major roads and in highly accessible locations. The findings show that a LoD1 workflow can support neighborhood-scale mapping of vertical built intensity, while accuracy remains sensitive to roof geometry, vegetation, elevation-model quality, and the assumed floor height. Parcel-level floor area ratio analysis was not performed and requires cadastral parcel data and zoning information.
Copyrights © 2026