tag:blogger.com,1999:blog-7729093380675162051.post7251180491331087943..comments2024-03-28T13:56:47.604+11:00Comments on moyhu: A success for open coding!Nick Stokeshttp://www.blogger.com/profile/06377413236983002873noreply@blogger.comBlogger5125tag:blogger.com,1999:blog-7729093380675162051.post-48469445872167526402010-06-21T02:40:15.285+10:002010-06-21T02:40:15.285+10:00Thanks, I had read the 'dual" explanation...Thanks, I had read the 'dual" explanation without fully comprehending what it meant. There appear to be some quick ways to regrid if a single point is dropped. i'm wondering if methods that use CAM could regrid on a annual or decadal basis..stevenhttps://www.blogger.com/profile/06920897530071011399noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-52308121881933315992010-06-20T16:18:37.400+10:002010-06-20T16:18:37.400+10:00Voronoi has the same issue as Delaunay - you'd...Voronoi has the same issue as Delaunay - you'd have to redo when cells become empty. In fact, Voronoi tessellation is just a dual of Delaunay triangualtion, and has basically the same algorithm. You'll notice that many pairs of points can be connected by drawing a line between them that crosses just one edge of the Voronoi tess. And when that happens, it crosses at right angles. If you make all such connections that you can, you have a triangulation.Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-53422426902100054852010-06-20T15:58:35.194+10:002010-06-20T15:58:35.194+10:00hmm. ya regriding the whole world every month or e...hmm. ya regriding the whole world every month or even year would be problematic. like your approach better. for now I'll stick with a grid, also thought about this:<br /><br />http://www.bostongis.com/images/snippets/bostonschools_voronoi.png<br /><br />http://www.sgsi.com/MIUserGroup/Tech_FindNearest_SGSI.htm<br /><br />http://www.voronoi.com/wiki/index.php?title=Main_Pagestevenhttps://www.blogger.com/profile/06920897530071011399noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-30526785580920164072010-06-20T07:39:55.843+10:002010-06-20T07:39:55.843+10:00Steve,
The trouble with Delaunay is that the stati...Steve,<br />The trouble with Delaunay is that the stations reporting changes every month. If there's an empty cell, you need to re-grid. And Delaunay is striving for equilateral triangles, which you don't really need. The requirements for just dividing the area are much less strict than Delaunay, and the algorithm has to take advantage of that.<br /><br />The algorithm I'm currently looking at goes like this. You start with just a few triangles covering the Earth. You then start bisecting, with a requirement that all new triangles have a minimum number of stations (maybe 1). You have a priority based on size of triangle and number of stations within. Form a binary tree of the bisection process. Stop when you can't make any new qualifying triangles.<br /><br />Then as you proceed month-to-month, when a triangle gets empty, consult the tree and move up a level, reversing the bisection. That's quick to do. At any time you have a tesselation with all triangles qualifying.<br /><br />For a rectangular mesh, this is the idea of the <a href="http://en.wikipedia.org/wiki/Quadtree" rel="nofollow">quadtree</a>. But triangles are more flexible.Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-52690118117622277662010-06-20T02:26:58.962+10:002010-06-20T02:26:58.962+10:00Thanks Nick,
I've been puzzling over the weig...Thanks Nick,<br /><br />I've been puzzling over the weighting issue myself, i was going to look at delaunay triangulation or Voronoi (just to be different)<br /><br />Here is the puzzle. Take a grid with 4 cells. 3 cells have 1 station and the<br />4 cell has 10 stations. so at that gridding we would average the 10 to<br />get one value, and then average all 4.<br /><br />But, at some distance R, we can draw a grid where each station will be in its own grid square. and then the average will be the same as the simple average.. with a bunch of missing grids...<br /><br /><br /><br />so it was in puzzling over this I wanted to parameterize the gridding<br />and ask "what does grid size do to the average"<br /><br />Another way to think about this..You have 5*5 grid with 10 stations<br />and another grid with 1 station. which has better information.. which lead me to wonder.. what do the trends in sparse grids look like as opposed to highly sampled grids?stevenhttps://www.blogger.com/profile/06920897530071011399noreply@blogger.com