Digital Elevation Models (DEM) Digital representation of the terrain surface enable 2.5 / 3D views Rule #1: they are models, not reality Rule #2: they always include some errors (subject to scale and data collection process)
Digital Elevation Models (DEM) All DEMs are created from remote sensing except maybe GPS data ASTER sensor on Terra
Geowall and Terrain Bender (TB) assignment (10%) Assemble image / DEM dataset for a location of your choice Create 2.5D (TB) two best views Create 3D (Arcscene) for Geowall Recommend ~ 2000 x 1500 pixels subscene Pick area of interest / project area? / other course overlap Landsat image: from earthexplorer.usgs.gov - Anywhere on the globe DEM data BC: TRIM DEM CANADA: DTED (NTDB) World: SRTM DEM
This week s lab and 10% assignment: Terrain Bender http://www.terraincartography.com/terrainbender/
The UNBC Geowall True 3D
UNBC example: Chilcotin
DEM applications in Geomatics (remote sensing image analysis applications in bold) Extracting terrain parameters Modeling water flow or mass movement Creation of relief maps and models Terrain analyses in geomorphology Rendering of 3D visualizations Rectification of aerial / satellite imagery Image reduction (terrain correction) Classification layers in mountain areas http://www.satimagingcorp.com/svc/dem.html
Classification needs to include the topography
Avalanche slopes : 25-45
Classifications and channel inputs Avalanche slopes : 25-45 DEM: DTED 1:250,000 = 100m pixels
Evolution of DEM creation 1950s Generation of contours from stereo photos 1980s Mass points from stereo photos 1990s Automated generation of masspoints 2000s Direct generation of grids from stereo-imagery e.g. high-res sensors, ASTER, RADAR 2010s LiDAR / digital photography / UAVs cloud of millions of points -> high-res grid
Early DEM generation: pick regular points or digitise contours
TRIM DEM masspoints ~70m spacing captured onscreen from stereo-photography soft copy (prior to fully automated image matching) LiDAR / digital photography have multiple points per square metre
DEM data types A. Discrete elevation data Contour lines from maps or digital files Mass points. and break lines These are interpolated into GRIDS (PCI etc..) B. Continuous DEM data Raster grids for remote sensing ideally ~same pixel size also Triangulated Irregular Networks (TIN) not useful in remote sensing
DEM creation by interpolation Inverse distance weighted - simple Nearest neighbour honours raw values Spline minimizes curvature -> smooth surface Kriging uses spatial correlation of points
TRIM DEM 25m raster grid Interpolated to 25m grid By 1:250,000 map sheets (100 tiles appended) 25m pixel size to match Landsat 5 and 7 rectified imagery but now imagery is 30m pixels Header file includes NAD83, UTM zone..
Resampling pixel size In the early days during resampling stage, pixels were rounded to match UTM grid and DEMs: Landsat MSS 80m raw pixels -> 50m corrected pixels Landsat TM 30 (28.5) m -> 25m BC TRIM DEM was built to 25m to match Landsat TM data software now can handle different resolutions Rapideye 6.5 m -> 5 m
TRIM DEM 25m raster grid Vertical accuracy to 10metres UTM pixel coordinates to 25m Data are 16-bit metre values 32 bit interpolated decimals -> dead-fart value 8-bit data previously 0-255 (pre-trim) Software step value e.g. 10
Generate and examine hillshade for errors / quality after acquiring or generating DEM
DEM data acquisition often stored in geographic pixel size may be e.g. 1 arc second (GDEM) or 30 arc seconds (GTOPO30) Rule #3 Display, examine check elevation values in histogram; use hillshade to view Rule #4 Reproject DEMs with caution (or don t) - if raster and vector data have different projections, reproject vector to match raster, not vice versa - Raster reprojection involves resampling every pixel and may introduce artefacts - If DEM download offers geographic or UTM, pick UTM
Reprojecting vectors simply reassigns coordinates to points using specified ellipsoid and projection
Reprojecting raster DEMs Reprojecting rasters involves resampling every pixel A. nearest neighbour B. bilinear (?) - compromise C. cubic convolution.. no longer a processing issue
Raster reprojection not so simple: Every pixel has to be reassigned / resampled ArcGIS also has a majority method
Resampling methods New DN values are assigned in 3 ways a.nearest Neighbour Pixel in new grid gets the value of closest pixel from old grid retains original DNs b. Bilinear Interpolation New pixel gets a value from the weighted average of 4 (2 x 2) nearest pixels; smoother but synthetic c. Cubic Convolution (smoothest) New pixel DNs are computed from weighting 16 (4 x 4) surrounding DNs http://www.geo-informatie.nl/courses/grs20306/course/schedule/geometric-correction-rs-new.pdf
Resampling http://www.geo-informatie.nl/courses/grs20306/course/schedule/geometric-correction-rs-new.pdf
Possible results of reprojection This can cause striping Avoid reprojecting rasters.. -Reproject vectors first - e.g. topo to raster -Cubic convolution best I hate this almost as much as light beer ->
DEMs: Sources of error and variability Resolution pixel size Data processing and interpolation Missing data clouds, steep slopes Change landslides, glacier ablation forest clearance
DEM error: BC provincial DEM 1985 (1981-1989) Homathko Icefield
DEM error in Google Earth Mt. Robson (AB/BC) AB BC +30m -100
#5: Generate and examine hillshade for errors / quality Stripes often caused by WGS84 - NAD83 Holes: due to clouds
Datums & Coordinate Systems GPS Datum: WGS 84 Origin is at the Earth s center of mass (geocentric) Geodetic Referencing System (GRS) 1980 earlier model For all practical applications WGS84 ellipsoid and GRS80 ellipsoid are identical
Figure of the Earth Best-fit ellipsoid (e.g., GRS-80, WGS-84)
WGS84 v GRS80 Landsat imagery: WGS84 Global DEMs: WGS84 Canada DEM data: NAD83 - based on GRS80 - Slight difference in semi-minor ellipse axis
Differences between Datums NAD27 NAD83 : 33 to 700 feet NAD83 WGS84: ~ 4.3 feet The Revenge of the Shift
Matching Datums I m using WGS84 I use GPS for all my mapping. I m using NAD83 Revenge of the Shift 4.3 feet 40