Accuracy assessment and correction of a LIDAR-derived salt marsh digital elevation model

详细信息	查看全文 | 推荐本文 |

• 作者：Christine Hladik ; Merryl Alber ; ^{malber@uga.edu}
• 关键词：LIDAR ; DEM ; Salt marsh ; Georgia ; Sapelo ; RTK GPS ; Ground truth
• 刊名：Remote Sensing of Environment
• 出版年：2012
• 期刊代码：79_00344257
• 类别：ear
• 出版时间：June, 2012
• 卷：121
• 期：Complete
• 页码：224-235
• 文件大小：1814 K

摘要

Accurate habitat mapping in salt marshes is critical for both management and conservation goals. Information on marsh elevation is important to coastal managers, particularly for flood inundation mapping, coastal hazard assessments and modeling sea level rise. Elevation is also an important determinant of the frequency and duration of tidal flooding, which in turn affects species patterns in marshes: elevation differences of less than 10 cm can affect plant distributions and productivity. Light Detection and Ranging (LIDAR) can provide synoptic elevation information in many environments, but its accuracy in salt marshes is limited by a combination of sensor resolution (scan angle and frequency, pulse width, footprint size), instrument errors (GPS and inertial measurement unit errors), and poor laser penetration in dense vegetation. This means that uncorrected digital elevation models (DEM) are generally not accurate enough to distinguish elevation changes in salt marsh environments at the resolution that is used to determine tidal flooding or vegetation patterns. In this study, we used a LIDAR-derived DEM for the salt marshes surrounding Sapelo Island, GA obtained with a state-of-the-art Optech Gemini ALTM LIDAR system with a high laser pulse rate frequency of 125 kHz and advanced IMU/GPS technology, and evaluated its accuracy with elevations collected using real time kinematic (RTK) GPS. We found that DEM mean vertical errors for different cover classes ranged from 0.03 to 0.25 m in comparison to the RTK ground truth data, with the larger offsets for taller vegetation. We developed species-specific correction factors for ten cover classes and used these correction factors to modify the LIDAR-derived DEM in four areas of the study domain where vegetation boundaries were mapped directly in the field. Application of the derived correction factors greatly improved the accuracy of the LIDAR-derived DEM within these areas, reducing the overall mean DEM error from 0.10 卤 0.12 (SD) to 鈭?#xA0;0.01 卤 0.09 m (SD), and the Root Mean Square Error from 0.16 m to 0.10 m. In the corrected DEM, the ground elevations of all vegetation classes were no longer significantly different than the true RTK ground elevations. Our results suggest that these types of corrections can greatly improve the accuracy of LIDAR-derived DEMs in salt marshes and further emphasize the importance of accuracy assessments before DEM data are used, especially in environments such as salt marshes where small differences in elevation can have significant effects on inundation patterns and plant distributions.