I am a grad student involved with hydrological modeling. In the model I work on, we are calculating the Topographic Index (TI; without transmissivity - or set to 1) using Quinn's MFD algorithm. We do not use the exponent factor p, which is the same as having it set to 1. I am interested in determining how the TI calculation in SAGA handles large flat areas that might exist on a watershed DEM that has been delineated in ArcMap. The watershed delineation process in ArcMap involves filling depressions according to Jensen and Dominique, which creates these large flat areas.
Our model uses a TI calculation algorithm that handles the flat areas reasonable well, but results in some striping in the direction of the processing (left -> right, then down, etc.). This results from the fact that we sort the elevation values from highest to lowest, moving from the top left corner to the bottom right during the sort, and then proceed with calculating the TI values from highest to lowest. When a large depression or flat area is reached, we temporally elevate the current cell to force flow into a neighbor or neighbors. This, due to the sorting having occurred left to right, then down, results in striping. The TI values are all close, and high (near the extreme of the distribution) as the should be, but not equal.
But the SAGA TI algorithm seems to be able to give them a uniform value. I was wondering if anyone could explain what steps the SAGA TI algorithm takes when handling such large flat areas.
See the two embedded pics below for more info - pay particular attention to the large flat area in the middle of the watershed. The SAGA output is on the top, my algorithm is on the bottom. I'm thinking that some sort of recursive tree search/dependence graph algorithm is needed. Any comments or ideas are greatly appreciated! I haven't been able to determine this from the help text.