SWARS Step-by-Step Video Instructions

SWARS Spatial Analysis, The Book: Present the Result

When you produced something, you need to find a good way to present it.  Pretty maps are always helpful.  Conventional wisdom says a picuture is worth a thousand words.  That, is very much true.

So here, I present you a “slightly fine-tuned” map I made from the priority data we just produced in our Palau SWARS example.

This map shows you not only the overall distribution of the three priority land classes (high, medium, low) across the entire island.  It also illustrates the distribution of the three classes within each watershed.  Better yet, the size of each pie is in propotion to the land area of that particular watershed.  I wish I had another adminstrative polygon layer, something like a village boundary.  Since I don’t, I used the watershed layer to run the stats.

Now, how would you tell this story?

SWARS Spatial Analysis, The Book (*About RCV*)

(*)

The Story about RCV

The concept of Raster Class Value(RCV) might be a bit confusing, understandably so.  It’s therefore necessary to paint some more ink on the subject.  I surely do hope this extra effort would be helpful.

Let’s use an example of car buying.  Say you are looking for a new car and have eyes fixed on 3 models B, F, and T.  All 3 of them look rather attractive overall, but each has its distinctive pros and cons.  B is a gorgeous car with great reputation but the price tag is kind of intimidating.  F has its quite unique history and some cultural fame attached, also lots of horsepower.  But F is notorious for its reliability.  T looks rather dull, even boring, but you pretty much get everything from it — reliable, great gas mileage, plenty of accessories, etc.

Obviously, if there is that one criterion you emphasize so much and would not give in a single inch, things would be easier.  For example, if you are absolutely fixed on your budget, then B is likely to be the first one out.  On the other hand, if you definitively do not want to deal with broken car and all that issues for the years to come, you might as well just kick out F first.  If you, however, are determined not to have a vehicle that looks just like the other 90% cars sitting in the shopping mall’s parking lot, I guess a T wouldn’t be your first choice. So you keep looking and comparing and just can’t decided which one is the clear winner.

You look it up on some “expert” websites, and you notice they all use this “comparing chart” thingy.  Basically, they all drum up this numeric “scale” say 1 to 5,  then they break a vehicle down into individual components or features and measure each feature using that numeric scale.  So, you would see model B has a measurement of horsepower-4, gas millage-3, styling-5, etc.  Then by adding up those individual numbers, you get a overall “score” for each model.  Whichever comes up the best would therefore be the number one recommendation for your purchase.  Simple enough.

That numeric scale, is our RCV scheme!

Just like the vehicle evaluation scale, it really doesn’t matter what are the numbers you use.  Using 1 through 5 (step by 1) doesn’t make that much of a difference from using 100 through 500 (step by 100).  The point here is to assess the relative performance/value of each vehicle in that particular feature.  They all have engine horsepower, but 500hp is obviously stronger than 200hp thus deserve 5; getting 45 miles per gallon is definitely more desirable than 15 miles per gallon thus also should be given the top ranking of 5.

For our project, the “vehicles” we are comparing are different locations in our project area.  What we are assessing is their relative importance/significance to the issue we identified to study.

Apparently we don’t seemingly have such distinctive individual “cars” to look at and compare.  What we deal with here, are the individual land cells/grids, say a 10 meter by 10 meter cell!  The features we are measuring are things like distance to river, existence of invasive plants, proximity to fire risk, and being able to contribute to public drinking water system, etc.

The Raster Class Value Scheme is, therefore, just like the numeric scale used to evaluate vehicle features, a common measurement system we create to compare the relative importance of land at different locations in a particular aspect in terms of natural resources or threats.

Having such a relatively objective value system would enable use to conduct an assessment that involves many features/aspects, apply simple but scientific mathematical calculations, and produce clear, understandable, and explainable results.

What is the exact numeric scheme to use?  Really, it doesn’t matter that much.  You should be able to produce the same result using either 1 through 5 or 100 through 500, as I have demonstrated during the Hawaii Workshop.  My only recommendation here is to use a scheme that contains more numeric values than you think you would absolutely need.  If you need at least 5 values, instead of using 1 through 5 (step by 1: 1, 2, 3, 4, 5), you should start by using 1 through 10 (step by 2: 2, 4, 6, 8, 10).
Another important point, you should always make sure you use the full range of the scheme, by at least assigning the highest value in the scheme to the most important class within each layer.  Because if you don’t, you are implicitly applying a layer weighting!!  Leave that to the layer weights!!!

SWARS Step-by-Step Video Instruction — Start

SWARS Spatial Analysis, The Book (9)

(9)

Let’s Model

Mm… That sounds a bit funny —  “let’s model”?

I feel it’s necessary to demystify the concept of “modeling” first, for it sometimes has a repelling effect on people new to it.

Yes, modeling can be a huge and extremely complex project involving thousands and thousands of data inputs and mathematically equations, that will take truck load of Phds to accomplish.  But, modeling can also be as simple as adding two entries together!

Our SWARS spatial analysis is luckily one of the simpler models.  Essentially, we are just adding a bunch of raster layers together to see their aggregated impact on the land.

In this example of studying Palau’s water quality issue, we only have 6 input layers and one mathematical equation!

Again, I throw out this very same point which I have made perhaps a million times: there are always multiple solutions to accomplish a task in ArcGIS! Do you feel a little bit of nausea yet hearing this again?

Remember I showed at least three different ways to implement the final SWARS Weighted Overlay Model Analysis during the Honolulu Workshop?  Well, I do not plan to mess up your brain again in this writing.  Instead, I will only provide one method here.  It’s not the simplest or fastest.  Nor is it the most comprehensive one.  But it is one that is clear and straightforward, and probably matches SWARS spatial analysis methodology materials you find elsewhere.  So, sit tight and fasten your seat belt!

First, we need to create a blanket new model.

Go to the ArcToolbox window (on the left here), select the <Favorites> tab, then right-click in the blank area to open the menu list.  Select <New Toobox…>.  Name the new Toolbox “Palau SWARS Model”.

Then, create a new model by right-clicking on the Palau SWARS Model Toolbox and selecting <New> then <Model…> as shown below.

Open the new Model, click and drag the six prepared input layers into the canvas.  You can do this click-and-drag one layer at a time or select them all first and then click-and-drag all at once.  You will now have the six inputs in the model as seen here.

Next, we need to get this magic <Single Output Map Algebra> tool (SOMA) into the model.

In the ArcToolbox window, switch to <Index> tab.  In the blank area, start typing “single output map algebra“.  You should see the <Single Output Map Algebra> tool (SOMA) poping up in the list at the bottom very quickly.

Now, click-and-drag the tool into the model canvas.  It appears as below.

Use the <Add Connection> tool to link the six input layers into the SOMA tool.

Select the tool fist.

Click on one input layer, then click in the SOMA tool box.  A link is established when you see that arrowed line show up going from the layer to the tool.

Next, open the SOMA dialog by either double-clicking on the tool or right-clicking it and select Open.

When the dialog opens, type Very Carefully in the Map Algebra Expression box this following equation:

prio_watshed * 25 + landcover_rec * 25 + river_rec * 15 + strmbuf_rec * 15 + invasden_rec * 15 + slope_rec * 5

Make sure you have a space/blank between every single layer name and symbol!  This is very important!!

Name the output SWARS_SOMA and click OK.

When the dialog closes and the active window returns to the model canvas, you should now see the SOMA tool box turned into color (as shown below)!  If not, that means you didn’t set the parameters right and the model will not run.

OK, all set and ready, execute the model by clicking the <Run> button or right-clicking the tool box and select <Run>!

One thing here, when the model is done, ArcMap won’t bring in the output raster automatically into the map, kind of strange.

So, I will open a new ArcMap project and add in the result SWARS_SOMA grid.  As you can see, I also put the hillshade layer (created from the DEM) underneath for later visual effects.

Reclassify SWARS_SOMA into a three-class grid using the Natural Breaks method in the reclassify tool.  Name the output raster SOMA_Rec.

Assign SOMA_REC a good symbology (maybe as shown here) plus 35% Transpanrency, and put ontop it the Watershed polygon layer.  Wah-Lah! This is what you will see.

Finally, we just need to calculate the acreage numbers for each priority classes.

Open the attribute table from SOMA_Rec.

Add a new field called “Acres” using Double data type.

Calculate the acres for all three rows/classes using the <Field Calculator…> tool.

Still remember what’s the calculation equation to use?

If you forgot, this is the equation that’s based on cell count, known cell size (10 by 10 meters) and the square meters to acres conversion factor.

And here are the results.

Let me also show you a little map I made from the data.

SWARS Spatial Analysis, The Book (8.1)

Prepare Input Layers

— Priority Watershed —

Be creative!  That’s a key when you prepare the input layers.

In our Palau example, we have a Watershed polygon shapefile.  But we don’t yet know which are the higher priority watersheds, unless you can id those by the names.  To demonstrate how to prioritize the watersheds, I make up a WaterIntake points layer since we are looking at an issue of water quality.  What I will do, is to prioritize the watersheds based on the number of intake points within each watershed.

Here is an ESRI article on how to ** Count number of points within a polygon! The method here is a bit simpler.

Righ click the Watershed layer in TOC.

Select <Joins and Relates>, then click <Join…>.

In the Join Data dialog (as seen below), make sure you select “Join data from another layer based on spatial location” in the first drop-down option.

Then, choose WatershedIntake layer as the layer to join to the watershed polygons.

You can pretty much leave everything else as default but do try them and see what you get.

Name the output to watershed_intakes and hit OK to run the spatial join.

The output is a new watershed polygon layer with a new attribute named “count” showing the number of waterintake points within each watershed.  That count is what I will use to prioritize the watersheds.  The more intake points a watershed contains, the higher its priority shall be.

Next, I will convert this new Watershed polygon layer to a raster using the <Feature to Raster> convertion tool under <Spatial Analylst>, using the “Count” attribute as the field for convertion, and name the output raster as “watershedg” with the “g” for grid.

e

The last step is to Reclassify the watershedg grid to our RCV scheme (1~10).

Open the raster Reclassify dialog by going to <Spatial Analyst>, then <Reclassify…>.

In the dialog, make sure the Input Raster is “Watershedg”, and the Relcass Field is “Value”.

We will then type in the new values (1~10) in the reclassify table based on the original values.  Here is how I assigned the new RCVs.

Original Value	New RCV Value
7	10
4	8
3	6
2	4
1	2
0	1
NoData	0

Name the output as “Prio_Watshed” and click OK.

Waa-Lahh!  The Priority Watershed layer is now ready for our model!!

Ask me now why I assigned the new RCVs the way I did it, it’s completedly up to you.  But do make sure you assign the highest RVC, 10 in this case, to the most important class!

Another thing, you notice yet that I assigned value o to the “NoData” class?  Why did I do that?

a Step-by-Step Guide for SWARS Geospatial Analysis

As promised, I’m starting to put together this step-by-step guide for conducting the spatial analysis for SWARS. This will be a very detailed and technical guide designed for you GIS specialist. Because of that “high requirment”, it’s not gonna be ready in a day or two. Plus, I will have video clips to accompany some of the steps. So, read along as new writings become available here or on the page desiganated for the document here: http://docs.google.com/View?id=dg4763rp_42hf45563n.

I don’t think you would need this document to be completedly available to start your work on SWARS, especially after the intense training last week in Honolulu. You should have learned enough to get started.

As I said more than a million times, there is a lot of work needed to be done before you even start to prepare your first layer. Defining the issue(s) alone will take some lengthy discussions sand debates. Then the layer rankings and weights, the class values for your rasters and how to get your original layers, continuous raster or themaic vector, all these questions are to be answered first.

Hoptefuly our repeated practices in the workshop on all that tasks have prepared you for the coming battles. Don’t be shy! You need to and should be active in those meetings and discussions. You need to help to initiate, facilitate, and lead some of the discussions. Don’t forget, in the end, you are responsible to implement all the anlyses. Without a well defined path and target, you could run into all kinds of problems later on. If that happens, yourself is partially to be blamed.