Digital elevation models (DEMs) of the lower Elwha River, Washington, water year 2013 to 2016

Image collection: Images collected before Nov 2014 were collected with a single Canon D10 using firmware hacked with CHDK and running an intervalometer script written in Lua. Until Mar 2016 they were flown with a D10 and a Ricoh GR, and after March they were done with two Ricoh GR cameras. Cameras were mounted in a customized wing mount and flown in four flight lines with a Cessna 172 at elevations between 1500 and 2000 feet above ground level. Image locations were established by matching features in some images with ground control points (GCPs). GCP locations were measured using either a RTK base station or a WSRN network fix using a TopCon GRS-1, Trimble R8 or Trimble R10 receiver. Occupation times were generally 180 seconds and estimated accuracy is 5 cm vertical, 2.5 cm horizontal.

PlaneCam structure-from-motion processing: Images were aligned using PhotoScan software (Agisoft, LLC). PhotoScan aligns images based on tie points shared between multiple images and scales the resultant 3D surface based on perspective differences between individual images. Geographic location and elevation are introduced to the surface model using the GCP locations in the model. The result of this structure-from-motion processing was the PlaneCam digital surface model (DSM).

DEM creation: Each DEM was created from the most recent lidar before 30 September of the given year, supplemented with an error corrected structure-from-motion DSM from a low flow summer Elwha PlaneCam flight as close to 30 September as possible. One surface (2014) also contains data from a November 7 2014 lidar flight because the previous flight in 2012 did not capture the 2013 overbank flows. Because higher flows were reached in 2013 than in 2014 before 7 November, this flight was still deemed worthwhile - most of the in-channel data and all of the erosion occurring after the summer 2014 planecam flight were specifically maintained as either summer low flow 2014 PlaneCam data or 2012 lidar data for this DEM.
DEMs were produced with the following process steps:
1) most recent lidar data used as a base data layer 2) identify points on hard, flat surfaces where both lidar and PlaneCam had good surface reconstruction resolution and no visible X/Y offset, and where change between multiple lidar flights was +/- less than 0.1 meters (lidar uncertainty) 3) develop error map for PlaneCam surface based on those points using empirical bayesian kriging and a cell size of 20 meters with a linear semivariogram composed of local models with less than or equal to 75 points, a local model overlap factor of 2, and 100 simulated semivariograms per model 4) apply error map to PlaneCam surface model and visually inspect to confirm reduced error (typically expressed as a reduction in large-scale warping in areas of low GCP point density - specifically the middle and lower river away from either reservoir or the river mouth where higher control point densities exist. 5) delineate area where erosion/deposition was active since the last lidar (scale of 1:500 +/- 250) and clip PlaneCam model to that area 6) evaluate surface difference areas for replacement with (1) lowest surface or (2) most recent surface 7) clip out areas of most recent surface as a separate polygon 8) merge rasters selecting the lowest surface between lidar and PlaneCam 9) overwrite merged raster with planecam data where higher but more recent surface accurately reflects channel conditions.
Several artifacts exist from this merging process:
1) where lidar data captured high water surfaces and subsequent low flow planecam flight didn't capture dewatered channel features or lower water surface, the higher water surface is an artifact. This is most pronounced in the 2014 DEM because the 2014 Nov 7 DEM was used with a WSE corresponding to flows of 3300-3560 cfs. In some areas, overhanging vegetation prevented PlaneCam surface reconstruction from the corresponding planeCam dem. This artifact is also visible in the 2016 DEM, most notably in the Elwha Canyon between the lower Elwha dam site and the Elwha surface water intake
2) the 2015 DEM used a topobathy lidar flight that extended to only the Highway 101 bridge. Because these data are more accurate than either the PlaneCam DEM or the previous (Feb 22 2015) DEM, they were used. This may result in an exaggerated estimate of sediment evacuation/recovery from 2014-2015 in the lower river because of (A) the previous issue and (B) the extensive subaqueous coverage of the 2015 topobathy data. Note that topobathy data were only used in-channel and the area outside of the active channel was comprised of the Feb-22-2015 lidar because leaf-off conditions were better for resolving bare earth.
Data sources are as follows:
2013_PlaneCamLidar_Final.tif -2012 10-17 October USGS lidar -2013 0919 PlaneCam -river profile surveys (East et al., 2015)
2014_PlaneCamLidar_Final.tif -2012 10-17 October USGS lidar -2014 30 September PlaneCam -2014 07 November USGS lidar
2015_PlaneCamLidar_Final.tif -2015 22 February USGS lidar -2015 23 September PlaneCam -2015 28-29 September USGS topobathy lidar
2016_PlaneCamLidar_Final.tif -2016 29 March USGS lidar -2016 30 September PlaneCam

Corrected name of theme keyword thesaurus from ISO 19115 Topic Categories to Category.

Edited metadata to add keywords section with USGS persistent identifier as theme keyword. No data were changed.