If we zoom in here again to our digital elevation model, these cells have a particular size that is representing that size on the ground. In other words, here we have a cell with a side of 30 meters by 30 meters and in Esri software the cells have to be square. What this is referred to as is, the Spatial resolution. So, we would say that there is a spatial resolution of 30 meters, and because they're square you can use that shorthand that it's going to be the same for both sides. So, that represents an area on the ground that's exactly 30 meters by 30 meters. So, let's just compare different spatial resolutions for a second so you can get a sense of what they're like. So, this is the original resolution of the data, 30 meters. I'll just put in the shoreline here for reference. Then I change the spatial resolution, I actually downgraded it to 100 meters and then to 500 meters and then to 1000 meters. And So, we would consider this a pretty low spatial resolution. There's no hard and fast rule as to what number exactly is the transition between low resolution and high resolution. I think it's something you get a feel for as you go along. So, certainly anything above 100 meters might be considered a fairly low resolution but it all depends on the type of data that you're working with. So, in terms of resolution and values, you have one value per cell and remember that, because there's only one value per cell and what that really implies is that there's no variability within one cell. So, what do I mean by that? Is that if we have a dataset here, where remember, this is a 1000 meter spatial resolution, so that's in other words one kilometer. What this is implying is, if I only have one value here, let's say this is 75 meters above sea level, what I mean by that there's no variability is that it implies that if you were actually to go to that location if this was realistic, there would be a perfectly flat square that's one kilometer by one kilometer that's exactly 75 meters above sea level and then if you went to the next square over on the ground in reality, there would be this sort of cliff-face between the two. Let's say this was 100 meters above sea level, you would have to climb this cliff-face to get to the next perfectly flat one kilometer by one kilometer square, that's 100 meters above sea level. Of course that's not realistic but it's just something that you need to be aware is inherent in the nature of raster data. It doesn't matter even if we make those cells smaller you still have the same thing. There's always going to be this trade-off between the size of the cells and the amount of variability that you can capture and the amount of data that that's going to take in order to store that. So, in other words just to summarise this part, you might say," Well, of course it's ridiculous. You're not going to have varies in the ground that are that huge like a kilometer across that are perfectly flat. What if you made them say ten meters across? That would be better. Right?" Well sure, it would but even then within a 10 meter cell you might have a little hill. Let's just say if I draw it here, if that's a 10 meter cell, if there was a little hill in there, it wouldn't actually capture all of the variability inside that 10 meter cell and you say," Okay. Well then what about one meter?" Same problem again. So, the cells are getting smaller but there's always going to be a bit of variability inside that cell. It's just a matter of how important is it to you to have that level of resolution in order to be able to capture that amount of variability. So, let me show you how this works in terms of data volume. So, here we have a pond, a body of water and we're going to try and capture that pond as a raster dataset. So, you'll notice that what happens here is that we have four cells that capture that pond. If I compare that to the original outline, you'll notice that we're missing bits of the pond that have disappeared from our Raster dataset. Why is that? Well, typically, there's different rules that can be used for this but if part of the cell when it's been converted to raster, part of the pond is less than one cell. So, if that's less than one half, then it gets counted as not pond. Let's say this is just called land or something. So, that cell will be coded as land and not as water. So, you can see how that works, that any of the cells that are here they're going to be coded as whatever the majority of that cell is, which in this case will be land. Okay. We got that part. So, if we increase the spatial resolution, we have smaller cells to try and capture this same geographic feature, the same pond. Now we're getting a bit more detail to it because we're having these smaller cells with a higher resolution. So, what happens to the number of cells that we're using to capture this geographic feature? Well, if we double the resolution, in other words, if we go from a cell let's say 10 meters across to a cell that's five meters across, so that's doubling the resolution. I know it sounds confusing because we're going from ten to five but that means we actually have twice the number of cells in this direction in a linear way. What that's actually doing is it's doubling it in both the X and the Y and so, it's also doubling it in this direction and what that ends up with is we have one, two, three, four cells where we only used have one. So, we've quadrupled the number of cells in order to double the spatial resolution. I hope that makes sense. So, why am I telling you all this? Well, the fact is that with roster data there's this temptation to always want smaller and smaller cells to increase the spatial resolution more and more. This can really have an impact on the amount of data that we have to store because remember, every cell has one value that's stored for that location. If we chart this out, if you look at the number of cells that are being used to capture that location, this is where I started with where I showed you the 1000 meter spatial resolution and we had 1333 cells that were used to capture all of the elevation data for that location and then 500 meters we have 5,418, at 100 meter spatial resolution we have over 133,000 and then when we get to 30 meters look at this we have almost 1.5 million cells that are being used to store that data. So, all the point am trying to get across here is that, there is this trade-off that you can, if the data allows or the sensory you are using or whoever's being captured, you may be able to increase the spatial resolution but that's going to have a huge effect on the amount of data you have. Why is that important? Because you have to store that data, you have to process that data. So, you need to spend more money on storage, you have to spend more money either on faster processors or you have to spend more time waiting for that processor to go through all of those extra data values that you've created by increasing the spatial resolution. You may not think that that's not important you think," Well computers are fast. What's the big deal?" Well there's this kind of never-ending race that's going on where, yes computers keep getting faster and faster but the datasets keep getting more and more detailed as well. And So, you can have a raster dataset that when you want to do something to it in terms of analysis you might have to wait an hour for it to just do that one thing or maybe it might be five hours or 10 hours or more. So, that has a big effect on the speed at which you can complete your work. So, that's something you want to be able to keep in mind when you're doing this kind of work. There's a reading that you can look at that Esri has on their website that relates to this called The cell size of raster data. They do a nice job of summarising the same kinds of concepts that I'm talking about here. So, if you want to have a little more background reading to go through this and I would highly recommend that you have a look at that site. For spatial resolution, the resolution that it's created at and displayed out are two different things, and it can be better or worse depending on what it is you're using this for. So, for example, here we have an image with a spatial resolution of 150 meters that's being shown quite zoomed in, and this is at a scale of one to 50000. This is actually part of downtown Toronto and this is Tron island and here this is the downtown area, and of course it's hard to really see that here, it's too pixelated or grainy. Two course too low of a spatial resolution they all mean the same thing for you to be able to pick out any kind of detail and you might look at that and say," That's lousy data". It's not lousy data, it's just not meant to be used at that scale. If we zoom out to a scale of one to 46,000,000, in other words you can see most of North America here, it's the same spatial resolution but if I showed you that image first you'd say,"Yeah, that's great. That looks really good" And So, it's not that the the spatial resolution is good or bad, it's what the scale is displayed at is and whether that's useful or representative. If raster data ends up having a high data volume a lot of the time, if that's an issue that comes up often, then we have to look at ways to be able to try and compress that data. In other words, can we find a way to make the file sizes smaller but still have data that we can work with that represent the areas correctly? And there's lots of different ways to do this. Some are incredibly advanced. I just want to show you a quick simple one just to give you an idea of how these compression routines can work. So, this one's called Run-length encoding. I've just got a sample dataset here in raster and you'll see that if I wanted to store information about how to recreate this dataset, if I was trying to compress this into a file you'll notice that I can take this whole first row. So, we have one, two, three, four, five, six, seven, eight zeros across from left to right. So, I can store that simply by saying, "I have a value of zero that goes for eight cells across". So, I've stored those eight values using only two values. All I have to store is a zero and an eight and from that I could recreate those eight raster cell values across. Then I could say, "For the next row there's a zero and a two", which means there's a value of zero for a run of two cells, then we have a run of three cells for a value of one, a value of zero for a run of three and so on. You get the idea. And So, we can actually compress the data in order, like we have eight across and eight down here. So, eight times eight is 64 numbers that we're trying to store. We can actually store that in much less than 64 numbers by using this thing called run-length encoding. So, like I said this is actually a simple method. There's much more advanced methods available to store their stuff if you've ever used a Jpeg file, that's a different kind of image compression that's used. There's some that when you compress it you lose a little bit of information and some you don't and I'm not gonna go through all of them here. Really all I wanted to do like I said, is just to give you a sense of well, "Yes we have these issues with big raster files and there are ways that we can compress that down if that may help us in terms of making the storage space required for these smaller".
