Analysis of Terrain Evolution using Lidar data and

Analysis of Terrain Evolution using Lidar data and open source GIS H. Mitasova, M. Overton, J. J. Recalde North Carolina State University Raleigh, USA R. S. Harmon Army Research Lab. , Army Research Office Research Triangle Park, Durham, USA GRASS GIS

Workflow for Terrain Modeling and Change Analysis Multitemporal set of LIDAR Points Density, noise and accuracy analysis using binning: • selection of resolution and approximation method, • systematic error removal Spatial approximation and analysis using RST: • smoothing of random noise, • computation of elevation surface and its parameters Multitemporal DEM, slope, aspect, curvatures rasters Map algebra, contouring, flowrouting: • natural feature extraction Multitemporal ridge, crest, shoreline, stream maps Map algebra, spatial query: • quantification of change Change maps, tables, graphs

GRASS 6: open source GIS http: //grass. itc. it • 1984 started at CERL as land management system • 1999 GPL, all common OS, 32/64 bit, code in C • Web-based infrastructure for managing the code • fully integrated 350+ modules • add-ons managed on wiki • raster: map algebra, DEM analysis, flow, buffers, solar, • image: rectification, multispectral, reclassification • vector: buffers, overlays, networks, • DBMS: attribute management, SQL • transformations: projections, raster-vector, interpolation • visualization: 2 D, 3 D visualization, ps maps • WMS support, Google earth through v. out. ogr

North Carolina Lidar Surveys 2005: USACE Topo/Bathy Mapping after hurricane Ophelia SHOALS 1000 T hydrographic (1 pt/5 m) and topographic lidar 2004: USACE Topo/Bathy Mapping SHOALS-1000 T 2003, June, Sept. 18, 21: NASA/USGS EAARL, acquired just before and after the hurricane Isabel landfall 2001: NC Floodplain Mapping survey, Leica Geoscan 1996 -2000: USGS/NASA/NOAA ATM II published vertical accuracy is between 15 cm - 30 cm

Evolving coastal landscape: Jockey's Ridge sand dunes ft 158 Nags Head Woods Jockey's Ridge Outer Banks Nags Head 2001 DEM (20 ft resolution) based on NC Floodplain lidar survey

Jockey’s Ridge 1999 sand pavement vegetation ocean 0 Photogr. 1974, 95, 98 500 m Lidar 1999 Lidar 2001 RTK-GPS 2004 N

Terrain Change Analysis Define site specific terrain change measures - vertical change: peaks, spatial pattern of el. change - horizontal migration: dune crests and peaks, lateral movement of ridges how fast it moves, is there acceleration? - area change: is the dune area shrinking – overcome by vegetation? - volume change: is dune gaining or losing sand? Methods for extraction of dune features: differential geometry and map algebra Measure and compute change: standard GIS tools

Tuning the level of detail for feature extraction RST tension and smoothing is used to create surface at a desired level of detail while keeping the 1 m res. tension 700 tension 100 profile curvature slope road

Feature extraction and change analysis

Dune Migration N a b c d sand lost sand gained stable c e 1974: 108 ft. 2001: 72 ft. 0 d 500 m rate of horizontal migration South and East dunes: 6 m/y Main West: 3 m/y e The main dune rotates clockwise while its peak moves southeast. Volume and area are relatively stable

Changes in elevation and shape 1974: 110 ft 1949: 138 ft 2001: 83 ft 2001 1974 View from the ocean (east)

Dune field evolution: 1974 -2001 N Relocation of leading south dune that was moving out of park boundaries more surveys are needed to understand whether the management is working

Working with lidar data from a diverse set of surveys Oregon inlet avg. no. of points per 2 m res. grid cell 1996 0. 2 ATM 1997 0. 9 ATMII 1998 0. 4 ATMII 1999 1. 4 ATMII 2001 0. 2 NCflood 2003 2. 0 EAARL 2004 15. 0 SHOALS 2005 6. 0 SHOALS Major overwash area used for testing 0. 3 m res. 2004 DEM

Increasing LIDAR point density 1998 2004 substantially improved representation of structures but much larger data sets 1 m resolution DEM computed by RST binned computed by RST 2004 lidar 0. 5 m resolution DEM

Mapping LIDAR point density 2001: NC Flood 2004: SHOALS pt / 2 m grid cell Point density maps created by binning draped over 2 m res DEM (2001) are used to select common resolution

Shifts in Lidar Surveys elevation difference [m] Boxplot elevation differences between the most accurate 2004 DEM and other surveys computed along 1 m wide road centerline strip: most surveys are shifted 10 -20 cm above or below road 1996 1997 1998 1999 a 1999 b 1999 c 2001 2003 a 2003 b 2005

Impact of shifts in Lidar data Do we have high erosion rate? A Is the road sinking? original: blue: 1999 black: 2001 A erosion 12 m B accretion 2 m elevation difference [m] corrected: red: 1999 violet: 2001 A erosion 4 m (!) B accretion 8 m B 1996 1997 1998 1999 a 99 b 99 c 2001 2003 a 03 b 2005

Analysis of systematic error Elevation difference between RTK-GPS survey (0. 03 m RMSE) and lidar data along centerline of a stable road. elevation [m] RTK-GPS 2001 lidar mean diff = -0. 23 m 0 0. 7 1. 3 2. 0 2. 6 3. 2 distance [km] RTK-GPS 2004 lidar mean diff = -0. 06 m Elevation difference between 2001 and 2004

Conclusions - analysis of points density and systematic error is essential when using lidar for assessment of topographic change from multiple sources - automated tracking of extracted features is needed for more efficient measurement of change - open source geospatial software provides powerful, customizable tools for analysis of terrain change - OSGEO foundation supports the development of open source geospatial software and promotes its use. First Joint Meeting 9/2006, Lausanne, CH => FOSS 4 G 2006 Conference http: //www. foss 4 g 2006. org

Acknowledgment Funding by US Army Research Office, NC WRRI and North Carolina Sediment Control Commission is gratefully acknowledged We also thank Geodynamics for performing the accurate road survey.

Communities growing together (General) statistical computing environment: http: //www. r-project. org/ Postgre. SQL Most advanced open source relational database http: //www. postgresql. org/ Rgeo: spatial data analysis in R, unified classes and interfaces (e. g, RGRASS) http: //r-spatial. sourceforge. net/ GRASS GIS Spatial Computing http: //grass. itc. it GDAL - Geospatial Data Abstraction Library http: //www. gdal. org Post. GIS: support for QGIS: user friendly Open Source GIS http: //www. qgis. org geographic objects to the Postgre. SQL object-relational database http: //postgis. refractions. net Spatially-enabled Internet applications http: //mapserver. gis. umn. edu/ . . . AND MANY OTHERS! http: //www. osgeo. org

Accuracy of approximated DEMs RMSE of interpolated DEMs, based on 50 pts measured on pavements using RTK-GPS: 0. 25 m (1995) and 0. 03 m (lidar 1999, 2001) Lidar 1999, first return data: spatial distribution of deviations Surface deviation from the given points as function of tension vegetation spatial distribution of vegetation all sand