Introduction:

The focus of this example was to compare TIN derived DEM sources of various resolutions, with LiDaR based DEMs. Also, TIN derived DEMs were used to calculate stream depth DEMs. Finally, a comparison of two DEMs from different years (2010-2011) were used to conduct a change detection analysis, using a 0.2m disturbance threshold. The change detection analysis was conducted using ArcGIS ModelBuilder, and the corresponding flow chart is included to clarify workflow. The study area for this exercise is located in Central Oregon, in the Pat's Creek Reach, in the Bridge Creek watershed (roughly depicted below).

Methods and Data Sources:

DEM, survey point data, water surface, breaklines and all other data were provided by gis.joewheaton.org unless otherwise noted.

Using a refined TIN, DEMs were created at various cell sizes. DEMs were derived from the TIN at 1m, 0.25m, 0.20m, 0.15m, 0.10m, 0.05m resolution. At 1m resolution, the stream channel is slightly noticable but definition is fuzzy. As cell cize decreases, resolution increases. However, as cell size reaches 0.05m, dimpling on the surface decreases the smoothness of the surface. This dimpling is likely the result of errors from surveying during the original point based surveying. This is confirmed by the cone shape appearance that matches survey point data coordinates. Dimpling is greatest at 0.05 m and hardly noticable at 0.25m. The DEM created at 0.15m provides enough detail while limiting the dimpling effect to best represent further DEM based analyses.

Using the 0.15m DEM, bathymetric depths were determined by calculating the difference between water surface and minimum DEM values.

TIN based DEM vs LiDaR based DEM:

Next TIN based DEM data was compared against LiDaR based DEM provided by opentopography.org. LiDaR data for Pat's Cabin Reach is not nicely clipped based on survey data, and provides a larger extent of elevation data. TIN DEM data was compared to LiDaR DEM data at 1m resolution, to keep comparisons consistent.

Several differences immediately stand out about the LiDaR DEM. First, vegetation is detected. Because fixed-wing aerial surveying collected LiDaR data, vegetation was included in the sampling. In contract, during site surveying, leading to creation of the TIN based DEM, surveys avoided vegetation to only detect surface changes. As a result, LiDaR data is not a good source at identifying stream channels and corridors. Interference from vegetation make channel identification more challenging than identification using the TIN based DEM. Also, though the 1m TIN DEM was rough in comparison to 0.25m TIN DEM, it still detects stream channels better than the 1m LiDaR based DEM. Further, near IR, or IR sensors used to colled LiDaR data cannot penetrate water surfaces, further complicating how stream channels can be identified by air. Meanwhile, field samplers can easily survey wadeable stream reaches on foot. The trade-offs between these DEM types include time investments (LiDaR is fast, surveying takes time), cost investments (LiDaR is expensive, surveying is cheap), and manpower investments(LiDaR is doing quickly via aircraft, surveying requires field visits). Each should be considered to determine future DEM sampling methods.

Change Detection Analysis:

Finally, change detection analysis was used to identify changes to Pat's Cabin reach stream channel between 2010 and 2011. Between these two years, flooding scoured and eroded stream features and redeposited sediments to reshape the Bridge Creek channel in the Pat's Cabin Reach. To detect these changes, DEM data for 2010 and 2011 were compared to identify areas of scouring and areas of deposition. Using ArcGIS Modelbuilder a workflow was established and diagrammed below.

The 2010 DEM was subtracted from the 2011 DEM to establish the initial change in surface elevation, deemed the "DEM of Difference" (DoD). Next, a 0.2m threshold eliminated any changes within +/-0.2m from the change surface. This removed slight changes and possible sampling errors. Next, the DoD was reclassified to remove the data within the +/-0.2m threshold. The reclassification valued all remaining cells to a value of 1. By multiplying the remaining cells by the elevation changes from the DoD, the final change detection DEM was established.

The final change surface depicts scouring area (red where values are negative) and deposition areas (blue where values are positive) ranging from -2.5 to +2.5. Scouring was most common in the thalweg and cut bank of the river channel and deposition was common along point bars. The large western point bar was the location of large deposition between 2010 and 2011. The backdrop of this map is the 1m LiDaR DEM. Original 2010 and 2011 DEMs were included for comparative purposes.

Conclusions:

This workflow again reinforced the uses and differences in DEM types, and conceptions. Trade offs exist for different DEM derivation sources, and TIN and LiDaR DEMs were compared directly in this lab. Raster calculations were further used to determine bathymetry, and scouring and deposition in river channels. Though data requirements are specific and harder to derive, raster data have great quantitative and comparative value, which were highlighted during the change detection analysis.