Accuracy of Airborne Lidar-Derived Elevation: Empirical Assessment and Error Budget

As part of a countywide large-scale mapping effort for Richland County, South Carolina, an accuracy assessment of a recently acquired lidar-derived data set was conducted. Airborne lidar (2-m nominal posting) was collected at a flying height of 1207 meters above ground level (AGL) using an Optech ALTM (Airborne Laser Terrain Mapper) 1210 system. Unique to this study are the reference point elevations. Rather than using an interpolation approach for gathering observed elevations at reference points, the x-y coordinates of lidar points were located in the field and these elevations were surveyed. Using both total-station-based and rapid-static GPS techniques, observed vertical heights were measured at each reference lidar posting. The variability of vertical accuracy was evaluated for six land-cover categories. Root-meansquared error (RMSE) values ranged from a low of 17 to 19 cm (pavement, low grass, and evergreen forests) to a high of 26 cm (deciduous forests). The unique error assessment of lidar postings also allowed for the creation of an error budget model. The observed lidar elevation error was decomposed into errors from lidar system measurements, horizontal displacement, interpolation error, and surveyor error. A crossvalidation approach was used to assess the observed interpolated lidar elevation error for each field-verified reference point. In order of decreasing importance, the lidar system measurements were the dominant source of error followed by interpolation error, horizontal displacement error, and surveyor error. Observed elevation error in steeper slopes (e.g., 25°) was estimated to be twice as large as those on low slopes (e.g., 1.5°). Introduction The use of airborne lidar (LIght Detection And Ranging) sensors for topographic mapping is rapidly becoming a standard practice in the aeroservice community. Counties are collecting such data for a variety of management purposes—stormwater assessment, flood control, visualization, etc. Several efforts are underway to collect lidar data statewide, primarily for flood-plain mapping associated with the Federal Emergency and Management Agency’s (FEMA’s) flood mitigation efforts. Also, several federal agencies in the United States, Great Britain, and other countries are using airborne lidar for topographic mapping applications. While a general understanding of the relative accuracy of lidar is known, too few empirical studies exist for assessing the accuracy of digital elevation models (DEMs) created from these data. During the initial years of lidar mapping efforts Accuracy of Airborne Lidar-Derived Elevation: Empirical Assessment and Error Budget Michael E. Hodgson and Patrick Bresnahan (i.e., 1995 to 2000), most aeroservice companies would routinely quote accuracies of 15 cm RMSE. Most would now agree such accuracy is only achievable under the most ideal circumstances (e.g., low altitude collections, flat terrain, minimal or no surface vegetation or obstructions, much human analysis, etc.). A few empirical studies have been conducted to date and suggest accuracies of 26 cm to 153 cm root-mean-squared error (RMSE) for large-scale mapping applications (Adams and Chandler, 2002; Bowen and Waltermine, 2002; Hodgson et al., 2003). There is a need for numerous research efforts in quantifying the accuracy of lidar data incorporating various platform parameters and environmental conditions—collection, processing, geography. In this article, the elevation error was assessed for a spatially dense (2-m nominal postings) lidar dataset collected in Richland County, South Carolina. The analysis tested the hypothesis that mean elevation error would vary between landcover categories while holding terrain slope constant. The uniqueness of this lidar accuracy assessment is that the reference data were collected at actual lidar points rather than between lidar points, requiring spatial interpolation (Figure 1). This approach allowed for an assessment of observed lidar error without the additional error introduced by typical spatial interpolation. An error budget for observed lidar elevations was created, which included lidar system, horizontal, interpolation, and surveyor errors. Results from this analysis could be used to estimate observable elevation errors from similar lidar datasets in other environmental conditions.