Ground Reference Data Collection

To test the geometric accuracy of lidar data, a 9-day, one-field survey campaign was planned and designed (Irwin and others, 2024). Field data were collected by the USGS between October 25 and 31, 2020. The locations surveyed in Iowa are shown in figure 2. Real-time kinematic (RTK) Global Navigational Satellite System (GNSS), total station, and ground-based lidar were acquired to compare to the 3DEP data collected. USGS collected these data using survey-grade GNSS and total station instruments along with a terrestrial laser scanner (TLS). For this study, three Trimble R10-2 GNSS antennas (one base and two rovers), a Trimble SX10 robotic total station, and a RIEGL VZ-400i TLS were used. With proper RTK surveying practices, the Trimble R10-2 GNSS antennas can achieve ± 8 millimeters (mm) + 1 ppm horizontal accuracy, and ±15 mm + 1 ppm vertical accuracy relative to the base station. The Trimble SX10 used in this work was a 1-inch instrument with a distance measurement accuracy of ±2 mm + 1.5 ppm. The RIEGL VZ-400i has an accuracy of 5 mm and a precision of 3 mm. In the RTK method, a fixed base station with a known location (base station) transmits corrections to a moving GNSS receiver (rover). The rover uses these corrections to greatly enhance its positional accuracy, achieving centimeter-level precision (Van Sickle, 2008). The base station data were postprocessed through the National Geodetic Survey Online Positioning User Service (OPUS) at least 16 days after collection, when precise ephemeris of the GNSS satellites were available (Soler and others, 2011). OPUS static processing typically provides solutions that are within 1–2 centimeters (cm) horizontally and 2–4 cm vertically within the National Spatial Reference System. The postprocessed base station data were used to update coordinates for the rover points. Those rover points included control points that were established for the total station to occupy and backsight. The ground-based lidar is operated by strategically positioning the lidar scanner to ensure full coverage of the target area. These scans from multiple locations are necessary for complete coverage of most targets due to obstructions. The lidar systematically emits laser pulses and records the return time and intensity of the reflected light, capturing millions of data points. Reflective targets are placed in the region of data collection, and their positions are surveyed using the total station. These targets are used to georeference the lidar data scans and improve the accuracy of merging multiple scans (Bethel and others, 2025). The GNSS and total station data include points collected on ground elevations, infrastructure features, and TLS georeferencing targets. The TLS data include scans of buildings, parking lots, and groups of trees. All survey data were published (Irwin and others, 2024).

Ground Reference Data Processing

GNSS and total station data were collected using Trimble R10-2 GNSS antennas and a Trimble SX10 robotic total station. The data are processed to be in the Universal Transverse Mercator Zone 15 North projection and are referenced to the North American Datum 1983 National Adjustment of 2011 (epoch 2010) and the North American Vertical Datum 1988 GEOID12B for orthometric heights. Raw GNSS base station data were downloaded from of the base antenna in the manufacturer’s proprietary format. The base files were converted to the Receiver Independent Exchange (RINEX) format using the Convert to RINEX version 3.1.4.0 tool (Trimble Inc.). RINEX files were submitted to OPUS for correction in the National Spatial Reference System. Postprocessed base coordinates, returned from OPUS, were used to update all GNSS RTK points in Trimble Business Center 5.60 (TBC). Updated control point coordinates from the rover were then updated for the total station, thus updating the coordinates of all total station points in TBC. Point shapefiles were exported from TBC.

TLS data were collected with a RIEGL VZ-400i laser scanner. Scan data were downloaded from the scanner and processed using RIEGL’s RiScan Pro 2.14.1 software (http://www.riegl.com/products/software-packages/riscan-pro/). Raw scan data were imported using the download and convert tool. The coordinate reference systems were set via the GeoSysManager. Coarse scan registration was accomplished using the Automatic Registration 2 tool. Fine registration and georeferencing were then determined using the Multi Station Adjustment 2 tool. The point clouds were then colored from images taken by the integrated camera. The scan data were filtered to remove data with a pulse shape deviation of greater than 15. Points with a reflectance of less than −20 decibels were also filtered. Isolated points were removed when less than 5 points were within a 0.300-meter (m) search radius. Finally, all scanner positions were exported as merged files.

Interswath Accuracy Assessment

Interswath accuracy (Stensaas and others, 2018) describes the degree of agreement between overlapping strips of lidar data and measures the horizontal and vertical errors that may exist between conjugate features in the overlapping regions of the data. The errors are estimated by first identifying two overlapping swaths and calculating the distance between a point in one swath and the corresponding best-fit plane in a neighboring patch of points extracted from the overlapping swath. These measurements are aggregated to estimate the relative horizontal and vertical errors between swaths. The interswath accuracy is described in terms of the following metrics:

Same-Surface Precision Assessment

Same-surface precision of the data describes the consistency of the lidar measurements and is measured by extracting points inside a planar surface. Artificial surfaces, such as a flat roof plane or other built-up area, are preferred. Once the lidar points are extracted, a plane is fit to the points and the root mean square error (RMSE) of the planar fit is calculated. Figure 3 shows a synthetic computer-generated data with same-surface precision exaggerated for clarity. Figure 4 shows a real example of a polygon placed over an impervious planar surface.

Vertical Accuracy Assessment

The vertical accuracy of lidar data is measured using checkpoints collected using GNSS or total station survey on flat hard surfaces. 3DEP lidar points within 3 m of the checkpoints are extracted, and a two-dimensional surface is fit to the extracted points. Although a planar fit is normally used, a second order surface more accurately represents local topology (Dey and Wang, 2022); therefore, we decided to use a two-dimensional surface fit z=ax²+bx²+cxy+dx+ey+f. The parameters (a, b, c, d, e, and f) are obtained by extracting the (x, y, z) coordinates of lidar points in the spatial proximity of the checkpoint and fitting the second order equation to the points. The reason for a second order surface fit instead of first order planar fit is that many of these vertical checkpoints have been gathered on road or parking lot surfaces, which can have a mild curvature. The two-dimensional surface fit is used to account for this curvature. If the curvature of the underlying surface is very small, the coefficients of the higher order terms will be estimated as zero or to a very small quantity, which in turn yields the equation of a plane. Figure 5 demonstrates this phenomenon using synthetic computer-generated data and the error shown is exaggerated for clarity.

At each checkpoint, the error was estimated as follows:

The vertical error estimates at each checkpoint are combined and reported as RMSE.

Horizontal Accuracy Assessment

The horizontal component of accuracy for 3DEP lidar data is not explicitly verified in most 3DEP projects collected under Geospatial Products and Services Contracts. The Lidar Base Specification (LBS) (USGS, 2024) does not explicitly provide quality thresholds or methods for verification of horizontal accuracy of the data; therefore, the methods presented in this section are forward-looking research efforts to develop operational methods for verifying 3D accuracy (horizontal and vertical) in future versions of the LBS.

In this report, we present three methods to verify the horizontal component of the airborne lidar data accuracy. The first two methods involve using TLS scanned roof planes of houses and other structures and comparing the point cloud data with the 3DEP lidar data. Two methods were used to quantify the comparisons. In the first method, conjugate roof plane points (in TLS and 3DEP lidar data) are manually selected, and the coordinates of the roof are compared (fig. 6). The ellipsoids in figure 6 represent the location uncertainty (due to noise and other factors) for the coordinates determined by TLS and 3DEP lidar data. The TLS derived point (blue ellipsoid) is more precise than the 3DEP derived point (red ellipsoid) due to the higher precision of the TLS data and can be used as control points. This process is described in Kim and others (2020). In the second method, an iterative closest plane algorithm (Chen and Medioni, 1992) is used to determine the offset between TLS point cloud of the roof and the 3DEP point cloud.

The third method (fig. 7) for observing horizontal errors in the data used checkpoints collected on visible parking lot or road markings on freshly paved surfaces. In figure 7, the coordinates of blue and the pink dots (surveyed check points) are compared to the manually extracted coordinates of the intersection of the painted lines and the location of the end parking lanes (not shown). This method is used in the industry (Bethel and others, 2006) and is analogous to the data quality assessment methods used for aerial and satellite imagery. Because the LBS do not require horizontal accuracy reported with the 3DEP data delivery, these measurements are usually made by the data vendors for their internal processes and not externally reported.

Point Density Assessment

Point density refers to the concentration or frequency of point data within a defined area. In the context of spatial analysis, it represents the number of points per unit area. According to the American Society for Photogrammetry and Remote Sensing’s (ASPRS) recommendations on point density (Bethel and others, 2025), the recommended methodology is a three-step process:

Interproject Light Detection and Ranging Data Consistency Assessment

The 3DEP is envisaged to support national and panregional scientific and engineering goals; however, many of the lidar acquisition projects (including the current one under study) have scientific goals that are regional or subregional in scope. Their contractual requirements on geometrical data quality are also limited to their spatial extent (Dewberry Inc., 2022b). Lidar data for scientific projects such as river systems analysis, carbon sink assessments, interstate road networks, and watersheds require data assessments that are similar in scope. Many features such as streams, roads, and forests will span multiple and diverse collections or projects, using a varied set of sensors, data processing algorithms, and software. Data may be acquired by different vendors using different sensors at different times; therefore, features must be represented in a consistent manner by the lidar data to ensure accuracy and reproducibility in scientific studies, and the consistency between adjacent datasets must be quantified and documented in the metadata.

The interproject consistency is evaluated by first determining all the relevant projects that the current project under study overlaps. Then, the evaluation follows the same process as the interswath assessment; however, instead of two swaths of the same project being compared, the two overlapping datasets are assessed. Figure 9 demonstrates this process using synthetic points. The two boundaries are determined using alpha shapes and the analysis is performed using the “Intersection Points” in the overlapping region of the alpha shapes.

Because these are large datasets (for example, the IA_Eastern_1 dataset has 154,630,110,433 points and IA_Eastern_2 has 90,121,793,878 points), this analysis is not trivial. Large computing resources and data handling tactics are needed. Accordingly, this workflow uses lidar data stored in the Entwine Point Tile (EPT; Butler and others, 2021) format on the Amazon Web Services cloud computing platform. The breakdown of the steps involved are as follows: