Python-Powered LiDAR Workflow Spots Sinking Ground Along Texas Oil Pipelines

Beneath the coastal plains of Southeast Texas lies a dense web of crude oil pipelines that quietly carries the lifeblood of the American energy economy. Above those pipelines, the ground itself is on the move. Decades of groundwater extraction, oil and gas production, and the natural compaction of soft deltaic sediments have made the region around Beaumont and Port Arthur one of the most subsidence-prone corners of North America. Detecting where the land surface is sinking, and by how much, has long been a slow, labor-intensive exercise in geospatial data wrangling. Now a team of researchers at Lamar University has shown that the job can be handed largely to a computer script, and that doing so slashes processing time by roughly three-quarters while producing maps of elevation change that closely match those made by hand.

The study, published in the journal Natural Hazards, was led by Ademola Ibironke and Xing Wu of the Department of Civil and Environmental Engineering, together with Fayez Albalawi, Aawaz Gautam, and earth scientist Joseph Kruger. Their focus was Jefferson County, a low-lying expanse of Gulf Coast wetlands, refineries, and pipeline rights-of-way where even modest vertical ground motion can threaten buried infrastructure, alter drainage patterns, and amplify flood risk. Rather than relying on a single snapshot of the terrain, the team exploited a powerful but underused resource: two independent airborne LiDAR surveys of the county, one acquired in 2006 by the Texas Water Development Board and another in 2017, eleven years apart.

Airborne LiDAR works by firing hundreds of thousands of laser pulses per second from an aircraft toward the ground and timing the returning echoes to reconstruct the three-dimensional shape of the landscape as a dense point cloud. When two such surveys are captured years apart, subtracting one from the other yields a map of surface-elevation change at decimeter-scale vertical resolution, far finer than most satellite or surveying alternatives. The catch is the data volume. A single LiDAR tile covering a fraction of a county can contain tens of millions of points, each tagged with coordinates, classification codes, and metadata. Aligning two point clouds acquired with different sensors, flight paths, and classification conventions, then computing reliable elevation differences, is exactly the kind of repetitive computational choreography that humans perform slowly and inconsistently.

The researchers set up a head-to-head comparison between two workflows. The first was a conventional approach built in ArcGIS Pro, the industry-standard geographic information system, in which an analyst manually loads the LAS-format point clouds, filters them, interpolates them into digital elevation models, and runs the differencing tools step by step. The second was a Python-enhanced workflow that wrapped the same core geoprocessing operations into an automated script, chaining together point-cloud processing, spatial matching between the two survey epochs, and the final elevation-change calculations without continuous human supervision. The goal was not to invent new science so much as to industrialize an existing one, turning a bespoke analysis into something a monitoring agency could run repeatedly across hundreds of tiles.

The headline result is strikingly simple: for a representative single LiDAR tile processed through both pipelines, the Python-enhanced workflow cut processing time by approximately 77 percent. That figure matters because the bottleneck in LiDAR-based change detection has never been the physics of the laser; it has been the hours of analyst time consumed by clicking through menus, waiting for tools to finish, and re-running steps when parameters change. By automating the chain, the team effectively converted a multi-hour manual session into a largely hands-off computation, while producing spatial patterns of elevation loss and gain that were broadly comparable between the two methods. The authors are careful to note that localized differences appeared at the level of individual points, a reminder that automation changes the numerical details even when the big picture holds steady.

Those spatial patterns themselves tell an interesting story. Across the pipeline corridors of Jefferson County, the differenced LiDAR surfaces revealed a patchwork of elevation loss and gain rather than a uniform regional signal. Some stretches of ground dropped measurably between 2006 and 2017, while adjacent areas rose, reflecting the competing influences of sediment deposition, land development, excavation, vegetation change, and possibly deeper-seated compaction. This spatial variability is precisely why a screening tool is valuable: instead of assuming the entire corridor behaves the same way, operators can use automated change maps to flag specific hotspots where the surface has moved enough to warrant a closer look with field instruments.

The team did not stop at the LiDAR comparison. They also qualitatively cross-checked their results against two entirely independent sources of vertical-motion information: repeated RTK and GNSS survey measurements on geodetic benchmarks in the county, and interferometric synthetic aperture radar, or InSAR, which measures ground deformation from satellites. The authors are refreshingly candid about the limits of this comparison. Because the GNSS, InSAR, and LiDAR datasets differ in acquisition periods, spatial resolution, measurement characteristics, and the locations of their observations, they cannot serve as direct validation of the LiDAR-derived rates. Instead, they function as independent sanity checks, and the broad agreement in the general pattern of motion lends confidence without overstating precision.

Equally notable is what the researchers refuse to claim. Observed elevation losses along the corridors were not interpreted as confirmed land subsidence, because land development, erosion, deposition, and other localized surface modifications can all imprint themselves on a differenced terrain model. A bulldozed pad site, a dredged canal, or a freshly deposited layer of sediment can look, in a LiDAR difference map, remarkably like genuine ground sinking. By framing their output as a screening product rather than a subsidence diagnosis, the team has drawn a scientifically defensible line that many remote-sensing studies blur. The maps identify where to look; they do not, by themselves, explain why the ground moved.

The broader context makes the work timely. Southeast Texas sits on sediments deposited by ancient and modern river deltas, and the northern Gulf of Mexico coast has documented some of the fastest relative subsidence rates in the United States, driven by fluid withdrawal, sediment compaction, and sea-level rise. Previous work by co-author Joseph Kruger and colleagues has used repeated GPS measurements of National Geodetic Survey benchmarks to chart subsidence rates for flood-risk planning in the region, and recent satellite-based vertical land motion studies in Greater Houston have highlighted how infrastructure itself can be monitored from orbit. The Lamar study adds a complementary, high-resolution aerial perspective that can zoom in on the narrow corridors where pipelines actually run, a scale at which regional satellite products and sparse benchmark networks leave gaps.

For pipeline operators, the practical appeal is scalability. A corridor network spanning hundreds of kilometers crosses many LiDAR tiles, and a workflow that requires an analyst to babysit every tile simply does not scale to routine monitoring. A scripted workflow that can be re-run whenever a new survey flies, with consistent parameters and reproducible outputs, turns multi-temporal LiDAR from a one-off research exercise into an operational screening capability. The authors suggest that the automated approach can help prioritize locations for further field investigation and for complementary ground-deformation monitoring, whether that means leveling surveys, GNSS campaigns, or targeted InSAR time-series analysis at flagged hotspots. In an era when coastal infrastructure faces rising seas, intensifying storms, and slow but relentless ground motion, teaching a script to find the sinking spots first may prove one of the quiet but consequential wins of modern geospatial automation.

Subject of Research: Automated multi-temporal LiDAR analysis of surface-elevation changes along pipeline corridors

Article Title: Automated multi-temporal LiDAR analysis of surface-elevation changes along pipeline corridors using ArcGIS Pro and Python

Article References: Ibironke, A., Albalawi, F., Gautam, A., Wu, X., & Kruger, J. (2026). Automated multi-temporal LiDAR analysis of surface-elevation changes along pipeline corridors using ArcGIS Pro and Python. Natural Hazards, 122(21), Article 668. https://doi.org/10.1007/s11069-026-08429-7

Image Credits: AI Generated

DOI: 10.1007/s11069-026-08429-7

Keywords: LiDAR, land subsidence, pipeline monitoring, Python automation, ArcGIS Pro, point-cloud processing, remote sensing, Jefferson County Texas, infrastructure monitoring, InSAR, ground deformation, geospatial analysis