AN OPEN-SOURCE GEOCOMPUTATION PIPELINE FOR MUNICIPAL LANDSCAPE ASSESSMENT: FROM TERRITORIAL DELIMITATION TO GEOMORPHOLOGICAL CLASSIFICATION
Abstract
Automated landscape assessment is essential for regional land-use planning, yet traditional GIS workflows often rely on fragmented, manual graphical interfaces that lack scientific reproducibility. This study addresses this operational limitation by developing an integrated, open-source geocomputation pipeline that automates vector boundary isolation, digital elevation processing, and geomophological mapping within a single R environment. It was hypothesized that a native C++ implementation of local computer vision algorithms in R could accurately categorize municipal topographies and generate rapid spatial statistics without data conversion errors. Applied to Bom Sucesso County (Minas Gerais, Brazil), the pipeline utilized the geobr and sf libraries to isolate official boundaries, the geodata library to dynamically stream SRTM remote sensing elevation datasets, and the terra framework for mask-cropping and coordinate projection (EPSG:32723). Geomorphological characterization was performed via the rgeomorphon package using a 7-cell search radius and a 1-degree flatness threshold. Non-parametric descriptive statistics of the localized Digital Elevation Model revealed an elevation range spanning from 790.0 to 1232.0 meters above sea level, with a mean topographic height of 942.7 meters. Spatial metrics quantified through the geomorphons algorithm demonstrated a highly dissecated landscape, strongly dominated by Valley formations (17,239.03 ha) and Ridge systems (15,139.47 ha), followed by substantial Spur networks (10,546.40 ha) and Slope areas (12,365.46 ha), while strictly Flat terrains occupied only 453.95 ha. In conclusion, the successful validation of the pipeline confirms the hypothesis, proving that in-memory geocomputation optimizes processing speed, eliminates manual intermediate steps, and provides a highly reproducible architectural framework for municipal environmental modeling.