Laserfiche WebLink
<br />~D.SS.n <br /> <br />I. The low- and moderate-resolution L1DAR elevation data sets that we investigated on <br />relatively Ilat, bare-sand surfaces were offset above the true ground surfaces at our four <br />study areas (Figures 32-34). Without correction for these vertical offsets, almost all the <br />L1DAR data sets have RMSE values greater than our desired 20 cm and greater than <br />contractor specifocations (15 cm); many of the data sets have RMSE values in the range of <br />40-100 cm. Vertical offsets have also been reportcd in previously published evaluations <br />ofLlDAR data. which suggest that some L1DAR surveys are no more independent of <br />ground control than photogrammetric surveys. <br /> <br />2. A fter correction of thc low- and moderatc-resol ulion LI DA R data sets for thei r observed <br />vertical offsets, the resulting bare-sand elevation data for the moderate-resolution L1DAR <br />surveys produced better vertical accuracies (RMSE = 9-26 cm in August and RMSE = 13- <br />36 cm in September) than the lower resolution L1DAR survey (RMSE = 26-103 cm) and <br />the photogrammetric survey (RMSE = 32 cm; Figure 35). However, the higher <br />photogrammetric error was due to isolated editing errors by the photogrammerrist (Figure <br />36). <br /> <br />3. Photograrnmelry produced more accurate ground elevations on the cobble bars (RMSE = <br />16 cm) and on the vegetated sand surfaces (RMSE = 75 cm: Figure 37) than did either <br />L1DAR survey. <br /> <br />4. The low-resolution. March L1DAR survey produced better elevation accuracies within the <br />vegetated terrain (RMSE = 14-189 cm; Figure 37) than the moderate-resolution, August <br />L1DAR survey (RMSE = 58-279 cm; Figure 37). This is attributed to the March <br />collection period during leaf-off conditions and to the smaller scan anglc used during thc <br />lower resolution LlDAR ~;ur"ey. <br /> <br />5. In terms of reproducibility. which is an important consideration in a monitoring program, <br />our analyses of replicate collections of the moderate-resolution L1DAR data showed an <br />average RMSE value of29 cm for bare-ground surfaces in our four study areas (after <br />correction tor vertical offsets). In terms of vegetated terrain, the reproducibility of these <br />data averaged 95 cm (RMSE). Thus, the RAMS moderate-resolution L1DAR data are <br />both inaccurate and imprecise in CRE vegetated terrain, which is one reason we rank the <br />performance of moderate-resolution L1DAR be/ow that of photograrnmetry for our <br />requirements at this time. <br /> <br />6. Oll'-'ecent evaluations of high- (1.5 points/m') and very-high (10 points/m') resolution <br />L1DAR data on bare ground showed vertical accuracies of 17 cm and 8 em, respectively, <br />much higher accuracies than the photogrammetry data or lower resolution L1DAR dala. In <br />particular. the very high-resolution UDAR data was found to have very high precision <br />(reproducibility) of 4-5 cm and to have essentially no vertical offsets. Figure 38 shows the <br />correspondcnce between the very high-resolution L1DAR elevations and ground surveyed <br />elevations for poinls on bare ground, near vegetation, and within vegetation. Within the <br />densely vegetated areas, the accuracies ofbolh high-resolution L1DAR data sets <br />decreased. but this may be a result of our processing. Visual inspection of these data sets <br />within vegetation suggests that more rigorous (smarter) proccssing may be able to reduce <br />the observed 0.7-1.3 m errors found in thcse data. An advantage of the very high- <br />rcsolution LIDAR system is that it achicves very high accuracy without any ground <br />control and is therefore non-invasive, except for thc 100-m Ilight AGL required to obtain <br />the high density data. Although the cost tor the very high-resolution L1DAR data is <br /> <br />2S <br />