Cloud-optimized Earth observation (EO) platforms increasingly serve massive geospatial data directly from object storage, and the Cloud Optimized Point Cloud (COPC) format extends the widely used LAZ standard with a spatial octree index that enables HTTP range reads of a single cloud-hosted file. Yet many scientific workflows still require downloading the entire granule before analysis can begin, imposing heavy transfer, storage, and time costs. This presentation benchmarks LAS/LAZ versus COPC performance using R and the lasR package on NASA's Multi-Mission Algorithm and Analysis Platform (MAAP), drawing on the MAAP STAC catalog and its GEDI CalVal LiDAR COPC collection. We measure conversion cost, compare on-disk versus remote cloud read behavior, test sensitivity to spatial window size, and demonstrate end-to-end generation of Digital Terrain Models (DTM) and Canopy Height Models (CHM) leading to best practices for scalable 3D LiDAR data processing.

From this presentation, you’ll have the following takeaways-

COPC is not always faster, it depends on the scenario. It's suited to metadata checks and queries on large tiles that aren't local yet, but once files are cached or the query window is large, local LAS performs better.Remote COPC is scientifically usable.The same product built two ways, local COPC (download first) and remote COPC (read only what's needed), closely matched for both DTM and CHM.
Going cloud-ready is cheap. Converting LAS to COPC is practical: it takes a negligible amount of time and shrinks files by up to four times, reducing storage and transfer costs.
COPC complements LAS. Use COPC for browsing, selecting, and small first-time queries; use local LAS for dense, repeated work on cached files.
Over the past decade, airborne LiDAR has become central to forest structure, topography, and validating missions such as GEDI. LAS/LAZ were built for sequential processing on local archives, while rasters and tables moved to cloud-native formats like COG, Zarr, and GeoParquet. COPC is the LiDAR equivalent of COG: LAZ plus a spatial octree for HTTP range reads.
COPC lets users inspect metadata and pull small spatial windows remotely without downloading large ALS tiles. Yet many cloud workflows still stage the entire granule first, undercutting platforms meant to keep data and compute together.

The open question is not whether COPC is faster, but when. In this presentation we focus on addressing these problems:
Quantitatively benchmarking cloud-optimized point-cloud workflows against conventional LAS for real scientific tasks, rather than relying on general claims.
Determining whether converting LAS to COPC is practical in terms of conversion cost and storage footprint.
Establishing when remote cloud access genuinely outperforms downloading first, across local-versus-remote reads, query extent, and catalogue browsing.
Testing whether analysis-ready products derived directly from remote COPC are scientifically trustworthy, not just faster to produce.

All benchmarks were run on the NASA's MAAP R workspace, chosen because it supports data-proximate analysis next to the STAC-indexed cloud archive. We use the MAAP GEDI CalVal collection, published in paired LAS and COPC form, which allows a fair like-for-like comparison without building a custom archive. Three granules span a realistic size range, from roughly 69–72 MB (about 2.6 million points) to a 191 MB NEON Yellowstone tile of about 7.1 million points.
Conversion was affordable and cut storage/transfer by about four times. Locally, COPC was much faster for shallow header reads, but LAS was faster for dense subsets and full-tile summaries. Comparing download-LAS, download-COPC, and remote-COPC workflows showed a common trap: judging remote reads against cached files measures local disk speed, not first-time staging cost. Across 100 to 1000 m windows, remote COPC beat cold download for small queries—especially on the large tile—but that gain faded as windows grew, and cached local files stayed fastest. DTM/CHM from local and remote COPC agreed closely, with remote runs only modestly slower.

Attendees will leave with concrete, evidence-based guidance on when cloud-native point-cloud access pays off and when it does not. COPC excels for catalogue browsing, metadata checks, and small, first-time spatial queries against large remote tiles, while conventional local LAS remains the better choice for dense queries once files are cached. COPC does not retire LAS; it adds another efficient way to work with analysis-ready Earth observation data. The talk uses entirely open-source tooling (R, lasR, rstac, terra, GDAL) on an open platform, so the workflow is directly reproducible by the FOSS4G community.