Search

Search Results

• 1. [Article] Where's the ground surface? Elevation bias in LIDAR-derived digital elevation models due to dense vegetation in Oregon tidal marshes

  Light Detection and Ranging (LIDAR) is a powerful resource for coastal and wetland managers and its use is increasing. Vegetation density and other land cover characteristics influence the accuracy of ...

  Citation × Citation Citation Abstract Copy Title: Where's the ground surface? Elevation bias in LIDAR-derived digital elevation models due to dense vegetation in Oregon tidal marshes Author: Ewald, Michael J. Light Detection and Ranging (LIDAR) is a powerful resource for coastal and wetland managers and its use is increasing. Vegetation density and other land cover characteristics influence the accuracy of LIDAR-derived ground surface digital elevation models; however, the degree to which wetland land cover biases LIDAR estimates of the ground surface is largely unknown. The minimum-bin LIDAR gridding technique has been proposed as a way to mitigate dense vegetation interference to generate LIDAR-derived digital elevation models (DEM). Past research has focused on the ability to resolve the marsh plain elevation and only limited research has investigated the overall DEM accuracy across the landscape using a wide range of cell sizes and land cover classes. I compared LIDAR-derived DEM accuracy across a 174 ha tidal wetland restoration site with a mix of native wetland and non-native agricultural pasture species. I found an optimum cell size of 1.4 m (1.96 m²) with a mean positive bias of 4.5 cm and a mean absolute error of 24.3 cm. At cell sizes smaller than the optimum, vegetation interferes with the LIDAR sensor and positively biases DEM models. At cell sizes larger than 1.4 m, the DEM captures and favors low features within the landscape, such as channels and ditches, which thereby degrade overall DEM performance. In addition, I investigated LIDAR interference by twelve common tidal wetland vegetation associations across six Oregon estuaries, using survey-grade Global Positioning System (GPS) measurements of the wetland surface and quantitative vegetation data (percent cover by species) for each measurement location. The fundamental vertical accuracy (FVA) of the LIDAR datasets was 4.5 cm root mean square error (RMSE) and had no consistent positive or negative bias in open landcover. Within wetland vegetation communities, my results suggest that LIDAR estimates of the ground surface in tidal wetlands are typically 10 cm to 30 cm above GPS measurements. Plant associations dominated by Carex obnupta and Carex lyngbyei exhibited the largest discrepancy between LIDAR and GPS measurements (mean discrepancies 36.6 cm and 48.8 cm respectively). The smallest errors observed in the study were about 10 cm to 11 cm and occurred in several different plant associations, including two low tidal marsh associations dominated by a mixture of Deschampsia cespitosa, Distichlis spicata, Sarcocornia perennis and Jaumea carnosa. These results suggest that the minimum-bin gridding technique for LIDAR may mitigate vegetation interference by densely vegetated land covers within LIDAR-derived DEM. However, care should be taken to select an appropriate cell size and validate the results before relying on the DEM for analysis. My research yields new information for coastal LIDAR users and increases our understanding of uncertainty in LIDAR-derived datasets to improve the ability to accurately evaluate and manage coastal environments. Title: Where's the ground surface? Elevation bias in LIDAR-derived digital elevation models due to dense vegetation in Oregon tidal marshes Author: Ewald, Michael J. Light Detection and Ranging (LIDAR) is a powerful resource for coastal and wetland managers and its use is increasing. Vegetation density and other land cover characteristics influence the accuracy of LIDAR-derived ground surface digital elevation models; however, the degree to which wetland land cover biases LIDAR estimates of the ground surface is largely unknown. The minimum-bin LIDAR gridding technique has been proposed as a way to mitigate dense vegetation interference to generate LIDAR-derived digital elevation models (DEM). Past research has focused on the ability to resolve the marsh plain elevation and only limited research has investigated the overall DEM accuracy across the landscape using a wide range of cell sizes and land cover classes. I compared LIDAR-derived DEM accuracy across a 174 ha tidal wetland restoration site with a mix of native wetland and non-native agricultural pasture species. I found an optimum cell size of 1.4 m (1.96 m²) with a mean positive bias of 4.5 cm and a mean absolute error of 24.3 cm. At cell sizes smaller than the optimum, vegetation interferes with the LIDAR sensor and positively biases DEM models. At cell sizes larger than 1.4 m, the DEM captures and favors low features within the landscape, such as channels and ditches, which thereby degrade overall DEM performance. In addition, I investigated LIDAR interference by twelve common tidal wetland vegetation associations across six Oregon estuaries, using survey-grade Global Positioning System (GPS) measurements of the wetland surface and quantitative vegetation data (percent cover by species) for each measurement location. The fundamental vertical accuracy (FVA) of the LIDAR datasets was 4.5 cm root mean square error (RMSE) and had no consistent positive or negative bias in open landcover. Within wetland vegetation communities, my results suggest that LIDAR estimates of the ground surface in tidal wetlands are typically 10 cm to 30 cm above GPS measurements. Plant associations dominated by Carex obnupta and Carex lyngbyei exhibited the largest discrepancy between LIDAR and GPS measurements (mean discrepancies 36.6 cm and 48.8 cm respectively). The smallest errors observed in the study were about 10 cm to 11 cm and occurred in several different plant associations, including two low tidal marsh associations dominated by a mixture of Deschampsia cespitosa, Distichlis spicata, Sarcocornia perennis and Jaumea carnosa. These results suggest that the minimum-bin gridding technique for LIDAR may mitigate vegetation interference by densely vegetated land covers within LIDAR-derived DEM. However, care should be taken to select an appropriate cell size and validate the results before relying on the DEM for analysis. My research yields new information for coastal LIDAR users and increases our understanding of uncertainty in LIDAR-derived datasets to improve the ability to accurately evaluate and manage coastal environments. Close