[GRASSLIST:5813] Re: Calculate fetch or rotate raster

[David Finlayson]

Thanks for your help.  I am interested in hearing more about your
GRASS/SWAN workflow.  We are looking into SWAN for future wave modeling.

I have written a Python script to dump the raster to ascii (r.out.arc),
and calculate fetches according to the Shore Protection Manual.
Unfortunately, there is a cost to dumping the grid and then loading it
again into Python, and an even greater cost in my rotation algorithm.  My
code runs very slowly on larger grids.  (Warning: geologist programming!)

I wonder if it would be difficult to bind Python to the C library of
Grass?  If that were successful, we could build a simplified object model
for efficient scripting directly on the grass databases.

-- 
David Finlayson
University of Washington
School of Oceanography
Marine Science Building (MSB) 112
Seattle, WA  98195
(206) 616-9407

http://students.washington.edu/dfinlays




On Fri, 14 Mar 2003, H Bowman wrote:

> > Is anyone aware of a script to calculate (wind) fetch from a raster?
>
> not per se, but I am currently working on wave exposure models and might
> offer something useful.
>
> - I've adapted r.los to return compass bearing instead of vertical
> angle. By running it from a series of regularly spaced viewpoints around
> the edge of the map and extracting the answers that coincide with the
> coastline it gives a pretty good qualitative view of what areas are
> directly exposed to the ocean. r.los is severly limited to the number of
> cells it can handle, as the calulation time goes up by cells^4 or something.
> 12km x 12km x 50m res was about the best I could get. It's a pity that
> MASK doesn't work with it, as the coastline only represents a handful of
> cells to analyze..
>
> - r.cva would be a better solution, but I keep on making it segfault and
> haven't played with it much. This would also solve a lot of the
> computational time issues, and would let you add weighting to swell from
> different compass bearings, etc.. r.cva is "cumulative viewshed analysis" and
> is a more powerful adaptation of r.los available at:
> http://www.ucl.ac.uk/~tcrnmar/GIS/r.cva.html
>
> - the main thing I use is the SWAN near-shore wave model with input
> bathymetry files generated by grass, and output fed back into GRASS,
> with the help of some glue programs. For non-commerical use, you can run
> it with the free Intel fortran compiler.
> http://swan.ct.tudelft.nl/
> http://www.intel.com/software/products/compilers/flin/noncom.htm
>
>
> So nothing to do with fetch, but maybe what you want to do with the
> answers from the fetch question.
>
>
> regards,
> Hamish
>