I'd like to show you how to use a raster weighted overlay. This is a really useful tool to be able to compare multiple datasets, and use those as criteria in a kind of constraint analysis that's a term we might use or multi-criteria decision support kind of things. All this really means is you're taking multiple datasets, and you're comparing the same cell through each of these data sets to see do they match certain criteria, and if so give them a weighting, compare them, and give them a final score that can help us to understand the interactions between these different data sets and what might be the most meaningful result for whatever it is that we're trying to work on. So, let's work through a typical scenario where you're interested in buying a house, and you have certain criteria that are important to you in terms of where you might want to live. So, for example if you have kids you might want to know where their closest high schools are, you might also want to have libraries within a certain distance, the all-important golf course that might be important to you. So you can set distance criteria for each of these things when I say close to, that may depend on who you are, how far you're willing to travel, whether you have a car or not, and so we can work on these criterion and weight them in different ways as we'll see, and then lastly is lots of trees or other vegetation. So, right now I'm not saying that this has to be a certain distance, I'm just going to measure distances away from these things and then give them weightings. So for example, this is the Euclidean distance measure away from high schools in the city of Toronto. Now just to be able to see these a bit better, I use the symbology to classify these into groups. Now, this is not in the dataset, I haven't classified the data yet. So the original Euclidean distances are a continuous range of distance values away from all of these high schools. I've just made it look this way so it's a little easier to see that. Now, I want to simplify my Euclidean distances into a range of values from one to 10 because what I'm going to do here is give a score for each of the criteria that I'm interested in so that they're all sharing a common score value. So, this is something that you have to kind of figure out for yourself is to what's the most useful one? The very common ranges from one to 10. But even then you have to think about well, what does that mean? Is one the most important? Is 10 the most important? As a starting point, I would recommend that you use one to 10 where 10 is the most important thing. Because what we're going to do is add these different features together or different layers together, and so the higher the number the higher the score will be, and that usually means the more important it is or the more desirable or whatever. But it could also be the opposite. It could mean a higher cost, so it could be a social cost, or environmental cost, or cost for traveling a certain distance. So, part of the secret to me are the way of thinking about this in terms of weighted overlay is that what those weights are really depends on who you are, what your background is, and what you're bringing to this in terms of how you want to view this problem. Not to get too far ahead but often with weighted overlay, you can incorporate other people, you can have stakeholders involved, you can have a public meetings to say which weightings are most important for what and try different scenarios. So, that's a whole lot to pack into this little dialogue box. But really what it comes down to is that you have to come up with a scheme that works for you. So, what I'm doing here is I'm setting it up so that areas that are closer to the things I'm interested in have a higher score and areas that are farther away have a lower score. So, I used an equal intervals scheme here to divide up the distances. So, anything that's within 500 meters of a high school, will get a score of 10. Anything that's within 500 to 1,000 gets a score of nine and you work your way down to there. Once it gets down to a level of one, anything that's beyond that distance I'm just not going to consider at all, it's not going to get a score at all, and it basically will get us for change missing values to no data, it's going to assign anything beyond that to no data. So, that's a way of saying okay, anything that's within my scoring range, I'm interested in using from a weighted overlay, anything that's beyond that I'm not going to use. You could set that differently, you could have 20 values, you could set the ranges differently, but that's up to you. Okay, so here's my classified version. So now, instead of doing the classification with symbology, I've actually used the Reclassify tool. So, these cell values are one to 10 that are actually stored in the dataset, and so what I've done here is the areas that have a higher score or darker because those are the ones that have more value or more importance to me. So cartegoraphically, I wanted to use dark for close, and light for farther away. I did a similar thing with the distance to libraries, so here's my Euclidean distances to libraries. I'm going to do the same kind of reclassify for those, and so here's my reclassified distances to libraries. I'll do the same thing with golf courses, reclassify those, so there's my reclassified golf courses. So, this is a good example here where you can see that anything that's beyond a certain scoring distance I'm not going to and consider at all, I'm only going to consider places that are within a certain distance of my areas of interest or my points of interests. My last criterion is just lots of trees and other vegetation. This is sort of a way for me to just say okay, if if it has a high NDVI value, then it's going to get a higher score and if it has a lower one it's going to get a lower score. So, there's nothing here that I actually have to reclassify. I've got two, three, four, five, this can actually go directly into my weighted overlay function, and then I can change the score values right there as you'll see. Okay, so this is the weighted overlay dialog box. There's a couple of things here that are important, so I just want to make sure this is all clear to you. So what's happening here is that it's listing one of my input raster datasets here, so for example this is the reclassified distances to high schools, and then I'm setting the percent influence. Then I'm telling you what fields I'm using. So, here we have one to 10, so I've already reclassified this and given it a score. Sometimes depending on the dataset, you could just have skipped that step and done that right in here. I tend to like to do it separately just so I know what I'm getting and I know what it looks like before I put it into the overlay, so you can do that separately, sometimes you can do it right in here and then here we have the scale value, and so because I've already set those myself, they're going to be the same. So one is one, two is two and so on. So, then the second dataset is a library distance, this is also got a 25 percent influence. These have similar values as well. So, the 25 percent influence, what I have here is four input datasets. I have three distances to high schools, libraries, and golf courses and the last one is to the NDVI data set, and so those weighted influences have to add up to 100 percent. The whole idea is that you're going to take those scores and it's going to multiply them together, and then add them up, so that you end up with a final score that's still out of one to 10 based on those weightings. Now, I've used an equal influence here which is actually easy to set by the way. You can just click this button here and say set equal influence, and it'll just divide 100 by however many inputs you have. So, that's a good place to start, but then the whole idea of weighted overlays and you can adjust those weightings, you could run it again and say well, what if I made this 10 percent, and this one 30 percent, or 40 percent? This is a great way of being able to run multiple scenarios to see if we change the weightings, what happens to our result? If we scroll down to the bottom of the list, you can see here that the NDVI is on majority. This is a case where I'm actually going to assign the scores directly from the original dataset. As you can see here I'm saying not all the input data sets have to be integer. So, I couldn't just for example use the raw NDVI scores from negative one to positive one. I have to reclassify those first so that we have integer values to use for the weighted overlay. So, here we've got NDVI five that's going to get a score of 10. So, remember basically what I'm doing is I'm converting a system that's NDVI values from one to five into a score from one to 10. So, then I'm going to have an NDVI four is going to be a nine, and then really, since I said that I want lots of vegetation, anything that has a score of three is actually going to get a much lower ratting, I think it should only be a five, and anything with NDVI of two, it is going to only give a score of one. So, you don't have to have values for every single one from one to 10. For example, here I've only got four different values, and I could have made these whatever I want, but I feel like well, if it's four or five that is going to have a high score, anything lower than that I really wanted to drop off quickly. Then this will produce a new output raster which I've just used the imaginative name of weighted overlay, and so here's our result. I've just used a simple grayscale to begin with here, and you can see that we have a range of values from three to 10. Why not one to 10? Well, it may just turn out that that there's no place where we actually have a score low enough to produce a one. So, don't necessarily think of that as a mistake, it just may be that with the particular combination of scores and values and geographic datasets that you have, this is what you ended up with. So, I've done it here so that the lighter the value, the lower the score, the darker the value the higher the score. I see this as a little homage to Ian McHarg who used to do this with transparent overlays. So, it's kind of simulating that effect. Here's a different version where I've got the same color scheme except that I've color-coded these scores that are eight, nine, and 10 as red because maybe those are really the only areas that I really want to focus on. That's the beauty of weighted overlay is, now that you have those weights, you can then decide well, what's the cutoff for whatever it is that I'm interested in? If it's locating distances to a landfill or whatever, you can say, well, where are the areas that we really want to zero in on? So for me if I was buying a house based on these criteria, I would look at the areas that are darkest red first, and then the lighter red, and then the slightly lighter red again. The nice thing as you can see here is that there's really very few areas that have the dark red high score of 10. So, when I look at that I might think, okay, well, I have to be realistic about this, maybe I need to change my criteria, maybe I have to spend more time considering areas that have a score of nine, or a score of eight, or maybe I have to go back and re-weight things, maybe I need to adjust things a little bit. So, I hope that you see this this is a great way of being able to look at these different combinations.
