Appendix F

Programs Used For Flow Direction

SAGA – http://geosun1.uni-geog.gwdg.de/saga/
ArcGIS – http://www.esri.com/
TauDEM – http://hydrology.neng.usu.edu/taudem/
TAS – http://publish.uwo.ca/~icreed/tas.htm

SAGA preformed the D8, Dinf, MFD, and DEMON flow direction algorithms that were tested of this study. The main reasoning in using SAGA was that most programs could not handle a large LiDAR dataset. TAS, TauDEM and at times ArcGIS would fail when attempting to use a 2-m LiDAR DEM. SAGA used a memory cache to process data. In doing this, process time increased significantly in calculating DEMON, Dinf, and MFD flow algorithms. Demon at 2-m grid size took at around a day to compute while MFD and Dinf was half that. D8 only took a few hours. Using 6-m grid cell size, process time for all flow direction algorithms decreased to a few hours.

SAGA was verified using ArcGIS for the D8, TauDEM for Dinf and TAS for MFD on a 10-m DEM. DEMON was verified by Dr. Steven Burges. Regression analysis was performed on the flow accumulation outputs from the 10-m DEM by the flow direction types. The only differences were in how the algorithms handled flat terrain. Appendix Figure F1 and Appendix Table F1 is an example of the regression analysis that was done to verify SAGA performed the flow direct algorithms accurately.

Appendix Figure F1. Correlation between Dinf algorithms by TauDEM and SAGA.

 

Appendix Table F1. Regression outputs from Appendix Figure E1

Regression Analysis
N = 1966041	Linear Regression:	Y = 0.861927 * X +1.228089
SAGA dinf:		(r=0.8654, r²=0.7489)
Min. = 7.043411 Max. = 20.423664
Arithmetic Mean = 9.711222
Variance = 1.725083
Standard Deviation = 1.313424
TauDEM dinf:
Min. = 7.043411 Max. = 19.798889
Arithmetic Mean = 9.584567
Variance = 1.711332
Standard Deviation = 1.308179