Camels

Well, the final report on the Camel project has officially been released. Here is a link where you can download it. There is a separate report (54) for the stuff I worked on. THis was incorporated into chapter 11 of the main report. Essentially it is a multi-criteria evaluation using weighted linear combination. There is python code in the appendix for performing (modularly) a weighted linear combination, including calculating weights. To calculate the weights you need NumPy.

Flow Mapping

Getting a head start on a new project that is more cartographic. It will involve mapping migration/flows from Australia to the Northern Territory (probably smaller geographic units than states). I like making maps, and so I'm excited to do some cartography beyond standard ArcGIS layouts. There are different possibilities on how to map this. Initially, I think I will use something like this that creates a more trunk/branch flow map instead of the typical straight line between places (Tobler's Flowmapper). The project lead doesn't like this style too much, but thought the trunk/branch style might work. We might pursue other mapping techniques, which would be cool to try and apply different map techniques to this area...

The Flow Map Layout tool that comes with the article is actually pretty slick, at least when using the supplied examples. Instructions are lacking unfortunately, so I'm not sure how to use multiple root points. The output is nice, and allows for export to an eps file. That could be brought into Adobe Illustrator and edited. Ad ambitious as I can be, I'm thinking of creating something similar for ArcGIS but using a shapefile. This would give a lot of control over the final product. In their tool, you can move things around too...Anyway, there are limits and I'm a control freak.

Camel project finished up as much as that type of project can finish. It's one of those projects that you can keep adding to and making it more sophisticated. It was part of a larger part, but it sounds like this will also be produced as a separate GIS/Model report. See how that goes. It was an interesting and fun project to work on. Good people too.

Back

I'm back, which I'm sure is a relief to all 2 of you who read this. We were on a 5 day backpacking trip along the Jatbula trail near Katherine Northern Territory. If you're interested, here is the blog entry on it.

The camel project finished up shortly before the trip. I was a little nervous but it seemed to work out. In the least, it gives some base on which to develop a more robust decision support system for camel management. The difficulty that I ran into was because everything was tied to access (i.e. roads), most of the management plans overlapped. There were ways around it though, for example doing a cost distance surface from the abattoir limited camel regions to how close they were to the abattoir along the road. I actually used the cost distance surfaces most of the time. Instead of using point locations for say boreholes (necessary for transporting the camels because of the amount of water they require) I used those as starting locations for a cost-distance surface and the roads a friction surface. I also used the major roads as a starting place, with tracks and 4wd roads as a friction surface. This created an implicit cost (cost as in the cost of implementing a camel management action) where it was less money to implement the plan on a major road (easier access) than on a track and then of course off road was the most costly. I originally based aerial culling off of roads, but then switched that to community base, which gave a different range....It was an interesting project. Difficult due to the time constraints, though.

Finishing an article based on some school choice analysis I worked on in 2007. Hopefully it will be accepted for a special edition on mapping school choice. I used kernel density estimation to explore some of the changes in distribution across the school district. Also calculated network distances to see how distance from home to school has changed.

Working on another article using kernel density to calculate a segregation index. It is based on the methodology in this article by O'Sullivan and Wong, A Surface-Based Approach to Measuring Spatial Segregation. I'm not entirely sure if this will actually happen, but I'm experimenting a little bit.

Camel's 'R' Us

I am working on creating scripts to create weighted linear combination. Basically duplicating different features in Idrisi so that I can have two models, one in ArcGIS and Idrisi. This is for the "Camel Project." I have a script now that calculates the weights for a pairwise matrix. Once I found a little information on this topic it was fairly easy to create. It is mostly just multiplying the matrix until the differences between the weights calculated from each iteration equals zero. Unfortunately, the python script is dependent on NumPy (Numerical Python), so it isn't as self-contained as I would like. I didn't feel like writing a function to multiply matrices, although I did this in VB.NET. I also am able to calculate the consistency ratio. I would share the script, but I because I am getting paid to do it, it is up to the client.

I also have managed to use Raster Calculator to duplicate the Fuzzy tool in IDRISI. It is a duplicate of the Linear function, and there are two ways to calculate it. One for monotonically increasing, and the other for monotonically decreasing. The formula is for 0 to 255.
Con is a map algebra function that means conditional. Con(test, true, false).

Here is how it works for monotonically decreasing:

INT((Con([RasterToBeScaled] <> 900, 0,( ((900 - [RasterToBeScaled]) * 255) / 900) + 0))) + .5)

Con([RasterToBeScaled] <> 900, 0 if the cell value is greater than 900 give this cell a value of 0. If it is false then calculate the value
( ((900 - [RasterToBeScaled]) * 255) / 900) + 0)))

INT and +.5 rounds the values properly and removes the decimal places.
The range of this example is from 0 to 900, so you need to change those values depending on what your "maximum and minimum" values are going to be. You really divide 900 - [RasterToBeScaled] by the range (max - min raster values). Since my range was 0 to 900 I skipped this and put 900.

If you need this to increase then you have to change some of it around.
INT((Con([RasterToBeScaled] <> 900, 255,( (([RasterToBeScaled] - 0) * 255) / 900) + 0))) + .5)

Now, whatever is less than zero is zero and greater than 900 is 255. Everything in between is calculated using the formula:
( (([RasterToBeScaled] - 0) * 255) / 900) + 0))
So now we subtract the cell value by the lowest value in your range and everything else is the same.

This is similar, but not quite what I was looking for.

Hope that helps...

“Trust in Allah, but tie your camel”

I was doing some AJAX today, trying to dynamically load markers from points stored in a geodatabase. Basically, just an XML response formatted specifically for my data, and parsed using Javascript. It works, surprisingly given my programming skills :). Anyway, probably wouldn't have gotten finished if it weren't for FireBug. A handy little Javascript debugger that lets me know when I have errors in my code. I use Visual Web Developer express for ASP.NET programming, and the 2005 version doesn't really have anything like this for Javascript (I think I read version 2008 does). FireBug works with Mozilla Firefox, but I think there is an alternative version for Internet Explorer. Be sure to disable FireBug when not debugging because it will report errors on any website, and slows down Gmail...

So I've started work on another project. I am a little unsure of confidentiality, so I won't list all that are involved. I'm sure the project will be made public, or at least the publication will be. It probably wouldn't be too difficult to guess who this work is for if you are in Australia. I'll be modeling camel management plans...while it is tempting (and easier) to just turn in a set of photographs of myself shooting camels, I'm of course referring to GIS-based models. Since the turn around time on this is pretty quick (2 months!), the model is a fairly simple Multi-Criteria Evaluation. I'm planning on using IDRISI (Andes?) to do this, but in the back of my head I'm thinking of writing a plug in for MapWindow. Two months isn't long, and IDRISI has a number of built-in tools for performing a Multi-Criteria Evaluation, as well processing rasters (distance surfaces, friction surfaces, etc...). The reason I want an open source solution is to create a user interface that could be used by anybody. That way a land manager could come in, and given a set of criteria (most-likely predefined), they could spit out a map showing potential locations for different camel management plans.

As idyllic as camels look in the Australian desert, they aren't native species. They were brought in as pack animals, and in many cases the train routes and road routes actually follow the same route as the old camel trains. Now their population is approaching one million, and they can be quite destructive to infrastructure, and biodiversity (or vice versa). Probably the unique part of this model is it is trying to identify locations for management plans based on a perspective, and also including perspectives in the model. Obviously, when a question is asked it comes from a certain perspective, and to answer that question certain criteria will be relevant. On top of that, there will be a layer that explicitly shows where certain management plans cannot be implemented based on the community, land owner, etc... This will most like be a constraint factor...but also could be a distance surface I suppose...Have to think about that one.

I'll keep you up-to-date on it.