Adriano Rolim da Paz

Extracting coarse-resolution river drainage networks from fine-resolution DEM or DTM is essential for large-scale hydrological and environmental modeling. Flow direction upscaling methods are preferred for preserving drainage patterns, but they face significant computational challenges when handling massive fine-resolution datasets. This study introduces eCOTAT + , an enhanced version of the COTAT + algorithm, designed to improve computational performance through two key strategies: (i) tile-based domain partitioning to reduce memory usage by processing localized subsets of data sequentially, and (ii) OpenMP-based parallelization to accelerate execution using multi-core processors. The eCOTAT + was applied to a 1-m LiDAR-derived DTM covering a 40× 40 km region in northeastern Brazil (1.6 billion pixels) and upscaling it to 100 m resolution. Twelve tiling configurations and eleven parallelization schemes (1–20 cores) were tested. Results showed that intermediate tile sizes (~ 20 × 20 coarse-resolution cells) offered the best trade-off between runtime and memory demands. Tiny tiles lead to high overhead from frequent memory and file operations, while large tiles diminish the advantages of parallelization. Parallelization significantly reduced runtime up to 8–12 cores, beyond which performance declined. A novel runtime-memory elasticity metric was proposed to quantify the balance between execution time and memory use across configurations. The upscaled drainage networks retained hydrological realism compared to the fine-resolution reference, confirming that computational optimizations did not compromise output quality. The eCOTAT + algorithm enables fast and scalable flow direction upscaling for both macro- and microscale applications, facilitating the use of massive elevation datasets even in standard computing environments.

Revista Cartográfica. Publicación del Instituto Panamericano de Geografía e Historia - IPGH

Digital Terrain Models (DTM) derived from LiDAR technology are increasingly available. However, working with these DTMs causes high computational costs and requires high-performance equipment, which may even make the use of these models unfeasible. The objective of this work is to evaluate the effects caused by changing spatial resolution in topographic representations and drainage networks extracted from LiDAR-DTM. For this, three resampling techniques were applied, mean aggregation, bilinear interpolation and nearest neighbor interpolation, to coarse a 1 m spatial resolution LiDAR-DTM in multiple resolutions (2, 10, 30 and 100 m). A sub-basin (550 km²) of the Sirinhaem River basin was taken as the case study area. The results show that there was no significant difference between the resampling techniques, but between the spatial resolutions, varying according to the applied metric. The spatial resolution of 2 m is more suitable in case of the need for a coarser resolution.