[GRASSLIST:10557] Re: TOPIDX flow direction algorithm

[twiens]

I just found the answer and was about to post the answer
when I found your reply. 

My initial reading of the comments in the AML header was
wrong. The null values for flat areas is not related to the
D8 algorithm. This is a function of the topidx formula. The
D8 algorithm issue is that its calculation of slope is not
very accurate and thus when calculating the contributing
area for topidx the calculation is not as accurate as is
possible.

To address this problem there is an interesting algorithm
from (complete with GPL software) by Dr. David Tarboton
(http://www.engineering.usu.edu/cee/faculty/dtarb/).

The D8 algorithm slope angle calulation can only have 8
possible values. Dr. Tarboton has a different algorithm he
calls DInf which calculates slope direction using a
continuous value (in radians). The result is anything based
on this more accurate slope angle is of course more
accurate.

As per the null values in flat areas, Dr. Tarboton has a
suggested different algorithm that is comparable (documented
in his ArcInfo plugin which is also GPL) which doesn't give
null values for flat areas.

I apologize for not fully researching my question before
posting to the list. Thanks for your reply.

T

----- Original Message Follows -----
From: Huidae Cho <grass4u at gmail.com>
To: twiens <twiens at interbaun.com>
Cc: GRASSLIST at baylor.edu
Subject: [GRASSLIST:10556] Re: TOPIDX flow direction
algorithm
Date: Fri, 24 Feb 2006 11:32:19 -0600

> r.topidx has the same issue that flat areas have NULL
> topidx values since contributing areas cannot be
> calculated with the D8 algorithm when area is float.  Is
> there any way to work around this flaw or is this really a
> bug or a feature of the algorithm?
> 
> Huidae Cho
> 
> 
> On Fri, Feb 24, 2006 at 08:31:19AM -0700, twiens wrote:
> > I'm working with some ArcGIS users and in a note in the
> > AML script to calculate the topographic index, the
> > author notes that ArcInfo uses a D8 algorithm to
> > calculate the contributing area. The concern is that for
> > relatively flat areas, the index value will be
> > inaccurate or inappropriate. I've noted in areas that I
> > have generated a topographic index, that flat areas seem
> > to be set to NULL by topidx. 
> > I've not been able to find any reference to this issue
> > in the topidx documentation. Can someone tell me if this
> > concern is valid for r.topidx?
> > 
> > Thanks
> > 
> > T
> > --
> > Trevor Wiens
> > twiens at interbaun.com
>