[GRASSLIST:2704] Re: stitching geotiffs

[Morten Hulden]

On Fri, 20 Feb 2004, Richard Greenwood wrote:

> Martin du Saire wrote:
> 
> > Outstanding!  Thank you Richard and  Glynn.  I think I may eventually 
> > have to r.patch the individual maps.  I was playing around with two maps 
> > within a single zone, and the r.stats -a -n map1,map2 output did not 
> > reflect what I expected to be the composite of the two (Actually, the 
> > numbers made no sense at all).  One last question:
> > 
> > Without having specific datum in my adjacent zones (14, 16), is there 
> > some way to make upper-bound estimates on the amount of distortion 
> > resulting from the r.proj method=nearest (the maps are land-use/cover)?
> > 
> > Thanks again.
> > 
> > Martin
> 
> That's a hard question that I will not try to answer directly. Map 
> distortion can involve linear, directional, and/or area distortion. So 
> first you have to define which of those are relevant to your 
> application. I think in your case the most significant distortion will 
> be angular. And I would guess that areas would be most important for 
> land use applications. Within a UTM zone I am thinking that area 
> distortion is limited to about 0.04% because the central meridian has a 
> scale factor of 0.9996 (anyone is welcome to correct me on that). But as 
> you move outside the zone, the increase in distortion would not be linear.
> 
> I am kind of a trial-and-error guy, so if it was me, I would simply 
> reproject the extreme cases into the target location and see if they 
> meet your requirements. Extending UTM zones is not generally considered 
> a good practice (although I am guilty of having done it a couple times). 
> A better solution might be to create a projection for your project.


For visual checking you can use r.mapcalc to create a raster grid
in one location, then project it into an other location and check 
the distorion. e.g. 

	r.mapcalc testgrid='if((row()%5) && (col()%5))' 

will create a grid where every 5th row and column have a different color. 
I bet someone has a more sofisticated recipe.