The bare earth digital elevation models (DEM) are a representation of the earth's surface where all man-made structures and vegetation have been removed. The bare earth DEMs were created from a subset of LiDAR returns that were classified as ground. They are regularly gridded at a six-foot post-spacing and were derived using TIN processing of the ground point returns. The elevation values are in feet. The DEM data format is ArcInfo interchange (.e00). TerraPoint surveyed and created this data for the Puget Sound LiDAR Consortium under contract.
The data is broken down into USGS quarter quads. Data from adjacent projects has been merged and is reflected in quarter quads where project boundaries meet. A quarter quad index is available at www.pugetsoundlidar.org.
In the bare earth DEMs where there are few survey points (i.e. bare-earth surfaces in heavy timber, where there are few ground reflections), TINing the points produces large triangular facets where the surface has significant curvature. Similar, though finer, textures are evident where vegetation reflections are incompletely filtered. Elevations are likely to be less accurate in these areas.
Top surface DEMs where project areas meet may have different vegetation heights. Survey projects are flown during winter leaf-off season; therefore adjacent project areas may be 1 or more years apart. Since vegetation is in a state of constant change it is expected to have differing vegetation heights in these areas.
LiDAR data values for water surfaces are not valid elevation values. Lidar surveying produces few survey points on water. Mirror-like surfaces fail to scatter the laser beam and unless the beam is perpendicular to the surface, no light is reflected back to the detector. Or intense reflections may lead to negative blunders, points that are too low. Interpolation between the nearest on-land points and sparse water points produces large triangular facets that may not accurately reflect the water-surface elevation. Where the water surface is surveyed adequately, adjacent swaths may be flown at different tide stages, producing swath-parallel cliffs. Ideally, lidar topography would be clipped to eliminate all open-water areas, but at present this is very labor-intensive.
User should carefully determine the place-to-place accuracy and fitness of these data for your particular purposes. For many purposes a site- and use-specific field survey will be necessary.
Applicable to data collected in 2000 and earlier: When merging adjacent DEM quarter quads there are instances where no-data cells exist at the boundary of the quads. This can be corrected by running a focalmean neighborhood command on the merged data. You can use the following ArcInfo command:
outgrid = con(isnull(ingrid), focalmean(ingrid, rectangle, 4, 4), ingrid)
Ingrid is the merged quads. This command finds all null values (no-data cells) and fills them in with the mean of the values within a specified neighborhood, which is a 4x4 cell area in this example. This area can be increased or decreased.
File names, formats, and values: All file naming convention and file formats are check for consistency.
Internal Consistency Analysis This analysis calculates and displays the internal consistency of tiled multi-swath (many-epoch) LiDAR data. The input for this analysis is the All-return ASCII data, but it only uses the first returns. The data is divided into swaths, or flightlines, and they are compared with each other. Since the contract specifications require 50% sidelaps, it means that all areas should have been flown twice. The results of this analysis is to verify that the data was generally flown to obtain the 50% sidelaps, that there are no gap between flightlines and also that overlapping flightlines are consistent in elevation values.
Visual inspection of shaded-relief images: During the visual inspection, hillshades are derived from the bare earth DEMs. The hillshades are examined for any obvious data errors such as blunders, border artifacts, gaps between data quads, no-data gaps between flight lines, hillscarps, land shifting due to GPS time errors, etc. The data is examined a scale range of 1:4000 to 1:6000. During this process we also compare the data to existing natural features such as lakes and rivers and also to existing infrastructure such as roads. Orthophotos area also used during this phase to confirm data errors. If any of these data errors are found, they are reported to TerraPoint for correction.
Internal Consistency: Data are split into swaths (separate flightlines), a separate surface is constructed for each flightline, and where surfaces overlap one is subtracted from another. Where both surfaces are planar, this produces a robust measure of the repeatability, or internal consistency, of the survey. The average error calculated by this means, robustly determined from a very large sample, should be a lower bound on the true error of the survey as it doesn't include errors deriving from a number of sources including: 1) inaccurately located base station(s), 2) long-period GPS error, 3) errors in classification of points as ground and not-ground (post-processing), 4) some errors related to interpolation from scattered points to a continous surface (surface generation).
Conformance with independent ground control points: Bare-earth surface models are compared to independently-surveyed ground control points (GCPs) where such GCPs are available. The purpose of the ground control evaluation is to assess that the bare earth DEMs meet the vertical accuracy specification in the PSLC contract with TerraPoint:
"The accuracy specification in the contract between the Puget Sound LiDAR Consortium and TerraPoint is based on a required Root Mean Square Error (RMSE) 'Bare Earth' vertical accuracy of 30 cm for flat areas in the complete data set. This is the required result if all data points in flat areas were evaluated. Because only a small sample of points is evaluated, the required RMSE for the sample set is adjusted downward per the following equation from the FEMA LiDAR specification (adjusted from the 15 cm RMSE in the FEMA specification to 30 cm to accommodate the dense vegetation cover in the Pacific Northwest)."
During this step, the bare earth DEMs were compared with existing survey benchmarks. The differences between the LiDAR bare earth DEMs and the survey points are calculated and the final results are first summarized in a graph that illustrates how the dataset behaves as whole. The graph illustrates how close the DEM elevation values were to the ground control points. The individual results were aggregated and used in the RMSE calculations. The results of the RMSE calculations are the measure that makes the data acceptable for this particular specification in the contract.
Lidar data were collected in leaf-off conditions (approximately 1 November - 1 April) from a fixed-wing aircraft flying at a nominal height of 1,000 meters above ground surface. Aircraft position was monitored by differential GPS, using a ground station tied into the local geodetic framework. Aircraft orientation was monitored by an inertial measurement unit. Scan angle and distance to target were measured with a scanning laser rangefinder. Scanning was via a rotating 12-facet pyramidal mirror; the laser was pulsed at 30+ KHz, and for most missions the laser was defocussed to illuminate a 0.9m-diameter spot on the ground. The rangefinder recorded up to 4 returns per pulse. Flying height and airspeed were chosen to result in on-ground pulse spacing of about 1.5 m in the along-swath and across-swath directions. Most areas were covered by two swaths, resulting in a nominal pulse density of about 1 per square meter.
GPS, IMU, and rangefinder data were processed to obtain XYZ coordinates of surveyed points.
For data acquired after January, 2003 (NW Snohomish, Mt Rainier, Darrington, and central Pierce projects), survey data from areas of swath overlap were analysed to obtain best-fit in-situ calibration parameters that minimize misfit between overlapping swaths. This reduces vertical inconsistency between overlappoing swaths by about one-half.
Heights were translated from ellipsoidal to orthometric (NAVD88) datums via GEOID99
Return points were then classified semi-automatically as ground (and water), not-ground (vegetation and structures) and blunder. For 2000 and 2001 data, the despike virtual deforestation algorithm described by Haugerud and Harding (2001) was used. After 2001, TerraPoint shifted to Terrascan software, which includes additional classification algorithms, allows for greater intervention by a human operator, and generally produces better bare-earth surface models.
Ground returns were used to construct a triangulated irregular network (TIN) which was then sampled at 6 ft intervals to produce the elevation raster.
TerraPoint shipped data in quarter-quad (3.25 minute by 3.25 minute) tiles. These tiles were checked for registration to a common lattice, shifted (up to on-half pixel--3 feet--in X and Y) if necessary, and merged to a continuous surface.
The PSLC has 4 different products available:
1. Bare earth DEM - these are in ArcInfo interchange format (.e00). These files are a representation of the ground surface. All vegetation and man-made structures have been removed. These files are about 35 MB compressed and about 110 Mb uncompressed.
2. Top surface DEM - these are in ArcInfo interchange format (.e00). These files are a representation of the top surface when the area was flown. You can see vegetation, buildings, bridges, etc. These files are about 35 MB compressed and about 110 Mb uncompressed.
3. Bare earth ASCII data - these files are plain text files with X,Y,Z values. The points in this file are all the returns classified as a ground return. The bare earth DEMs are derived from these ASCII files. These files are about 35 MB compressed and about 110 Mb uncompressed.
4. All-returns ASCII data - these files are plain text files with X,Y,Z values and also additional values such as GPS time, return number, etc. These files are very large, about 2 GB per USGS quarter quad.
The All-returns ASCII and bare earth ASCII files are available upon request to the Puget Sound LiDAR Consortium. This data is too large to put online, but it is still in the public domain and therefore interested users may obtain it free of charge. Depending on the amount of data requested, the user would receive a CD-ROM or a DVD-ROM. Other arrangements are also possible and will be evaluated on an individual basis.