WSP11909 - Laserfiche WebLink

Home Browse Search

. " These issues include: (1) the precision and vertical accuracy of these techniques for producing "bare earth" elevation data. not only on bare ground surface, but also within riparian vegetated zones, and (2) the ability of these airborne elevation data to accurately estimate sediment volume within the riparian ecosystem. DATA SETS AND STUDY AREAS To directly address these issues, we collected and compared the following data sets: (I) high- resolution RAMS LIDAR and moderate-resolution ALMS LIDAR data acquired at a spot spacing of 1.5 m and 3.75 m. a spot diameter of 0.5 m and 1.0 m, during August and March. 2000. respectively; (2) a second high-resolution RAMS L1DAR data set acquired over the same river reaches just two weeks (September, 2000) after the first high-resolution LlDAR collection. using the same LlDAR system and acquisition parameters, (3) high-resolution (25-cm C. l.) photogrammetric elevation data produced from I :4,800-scale photography acquired between (September, 2000) the two high.resolution RAMS data collections; and (4) over 2,000 ground elevations obtained transect and stakeout surveys on bare and vegetated surfaces during these airborne data collections. For this study we selected four large river reaches used by GCMRC for long-term monitoring of sediment. The four study areas are located within the northern 100 km of the Colorado River that flows through Arizona. All study areas have at least one debris flow and various types of vegetation stands. Bare ground in the river reaches consists of fine- grained sand bars and river terraces, talus slopes and associated cobble (25 em) bars, and isolated 1-2 m boulders. Vegetation consists of patches of short (=25 cm) grass; low-lying (=2 m). scattered bushes and dense marsh vegetation; and individual stands and dense groves of high ~4-6 m) mesquite and Tamarisk trees. Elevations in our study region decrease downstream from 952 m AMSL (above mean sea level) to 818 m AMSL. DATA ASSESSMENT The photogrammetric DEM (25-cm cells) used in our analyses was produced from contour vectors using the GRID function in Arc/Info. LIDAR point elevations for each study area were converted to a raster image file of elevations having a 25-cm cell size. These uninterpolated elevation raster images were used in our assessments of vertical accuracy and precision; interpolated OEM versions were used in our assessments of the accuracy of the airborne elevation data for estimating sediment volume within each study area. Because there were very few spatial coincidences between the LIDAR points and ow ground-transect points, we perfonned a I-m radial search around each transect point to determine a corresponding LIDAR elevation. When more than one LlDAR point occurred within this radius for a particular transect point, we calculated a distance-weighted average LlDAR elevation for that transect point. High-resolution, orthorectified color.infrared il1llgery was used to create digital masks of bare ground and vegetated ground, which were used in our examinations of specific surfaces in each study area. Two measures of error were used to assess accuracy and precision: the mean error (which is the sum of the differences between the airborne elevations and the corresponding surveyed-ground elevations divided by the number of values) and the mean absolute error (which is the sum of the absolute differences between the airborne elevations and the corresponding surveyed- ground elevations divided by the number of values). Vertical accuracy on bare ground The lower resolution ALMS LIDAR data are less correlated with the true bare.sand surfaces than the two higher resolution RAMS LIDAR data sets, with standard deviations on the mean errors in the ALMS data being factors of2 to to higher than those in the RAMS data. This is attributed to the larger spot diameter and hence higher positional uncertainty of the ALMS first return, which may have hit nearby bushes scattered about the sand surfaces. Least- squares regression lines for most L1DAR data sets are parallel to, but offiet above, the true ground surfaces. Vertical offsets in the ALMS data show a wide range (-5 em to +55 em), whereas the August RAMS data have quite consistent offsets (+1 J cm to +21 em) with only one area's offset differing more than I em from a +20 em offset. This RAMS consistency is encouraging for large-area topographic mapping large areas of the river corridor. However, the replicate (September) RAMS data show a more random, wider range in vertical offset (+17 cm to +48 cm), which argues against a regional offset approach. The unadjusted mean absolute errors (vertical accuracies) of the ALMS (33-84 cm) and September RAMS (19-49 cm) data are consistency worse than the 15-20 cm accuracy generally purported by commercial vendors. Only the August RAMS data have vertical accuracies (7-22 em) dose to this advertised accuracy. Assuming offset adjustment is an operational necessity, we corrected the data sets for the observed offsets. The adjusted vertical accuracies for the ALMS data are still low (21-84 em). but the adjusted vertical accuracies orthe September RAMS data (12-26 em) and August RAMS data (7-15 cm) are close to or within our level of acceptability. Thus, acceptable LfDAR accuracies for GCMRC monitoring of bare-sand deposits can be achieved by high-resolution LlDAR, but use of the data will require EV ALUA nON OF LIDAR AND PHOTOGRAMMETRY FOR MONITORING VOLUME CHANGfS IN RIPARIAN RESOURCES WITHIN 1lIE GRAND CANYON, ARIWNA Pecora 15fLand Satellite Information IVIlSPRS Commission I/FIEOS 2002 Conference Proceedings