[MUSIC] Hello, everyone, and welcome back. In this lesson, we're going to go over the raster data model in depth. We'll discuss the core building blocks of raster data, how transformations and overlay operations work, and then briefly discuss imagery and surfaces with rasters. So first, let's think again about what a raster is. Given an area of interest, you can divide a raster up into a number of square cells packed in next to each other in a grid. In most cases, these cells are all the same length on each side, and we call this length the cell size. So, if I say that a raster has a cell size of 10 meters, that means each cell is a 10 meter by 10 meter square, for a total of 100 square meters in area. Now we know the structure, but to make it more useful, they need to store information about the landscape, right. So, each cell can hold one, and only one value, for now, anyway, and that value represents some theme on the landscape. A really common scenario is elevation rasters, where the value of the cell is some measure of the elevation for that cell. Maybe it's the average or the value in the center, or a spot measurement. When each cell represents its own value, we quickly get a complete landscape picture, because you have multiple cells next to each other with different values. Now these values are always numeric, but you can associate textual information with them. Think of a raster representing land cover and having some sort of key to indicate that one means a paved surface and two means water and three means vegetation and so on. Where vector data can pick and choose the locations it represents by either putting features on the landscape or not, raster data must cover an entire area even if the values are null. But this is a useful property because it allows us to represent a full area with continuous data, or at least as best as computers can. If we want higher resolution data, and have a data source available, we just use a smaller cell size, more accurately representing features on the landscape. So, while vector data often represents discrete observations, raster data is more suited to data that is continuous across some area. A common raster format you might be used to, is the images coming out of your camera, or even information being displayed to you on your computer screen right now. Cameras capture light in the same sort of grid that we use for rasters and save images as raster formats like TIF and JPEG. Conceptually, rasters in GIS aren't any different, except they have spatial information that we use to georeference these rasters. On your computer screen, if you look really close you'll see the pixels. These are the cells of the raster. Within the pixels, the color is homogenous. It's the smallest unit of information possible in the raster. But together they provide us with more information. Computers like rasters because they are easy to work with and to do math on, and they make efficient data structures. To understand rasters, it might help to think of the most basic way to store raster data. First, just like vector data, we need to specify a coordinate system, otherwise our coordinates don't mean anything. Then we need the coordinates of the origin of the raster. sort of the (0,0) point on a typical Cartesian graph that you're used to drawing lines and mathematical functions on. This point is usually in the top left of a raster. From there, we can just start specifying a raw list of values. And since we also tell the GIS the size of the cells, It can start to construct the raster, plunking one cell down after another. Okay, so a raster represents a location with a grid of values, and these values represent a generalized bit of information for an entire cell. Where rasters start to get difficult is when we start trying to understand their locations in relation to each other. It's easy when two rasters align completely, cell for cell, but what happens if two rasters are slightly misaligned? When one has 10 meter cells and the other has 15 meter cells, and we want to compare the values between them. Cells could have as many as nine values from another raster that they touch. Here, we have to make decisions about how to choose which cells overlay each other. Is it by where the center is located or which one is the most dominant or some other metric? In many cases, ArcGIS assists us in this, but sometimes we have to be aware of the limitations imposed on our analysis. Similarly, re-projecting rosters or trying change their cell size can introduce error into our data for the same reasons as we were just talking about. While we're still learning about projections in these courses, know that re-projecting can create the same multiple cell situation that we just talked about, and regardless, your data will be modified in some way. So re-projecting rasters in contrast to vector data is a lossy or destructive operation, in that we lose information when we do it and care should be taken. These are the types of situations where even though rasters can have higher data density than vector data, they can be less precise than vector data in many operations. The bottom line is you need to take care with your data alignment. One really great tool we get when working with rasters is map algebra. It's less complicated than it sounds, but it allows us to create really basic functions that operate on our rasters without any complex programming. Need to multiply one raster by another? You can, just as if they are two numbers, and ArcGIS will go through the rosters and multiply them together. And we'll show you some similar use of map algebra in the next lecture. This makes them great for all kinds of work, because it's easier for newcomers to modify and explore them. At the same time, we don't have the same ability to create selections on rasters. Each pixel isn't the same as a feature, so the selection tools don't work. With map algebra, however, we can still do a sort of select by attributes by writing expressions against the value of the raster. Want a new raster where all the elevation values are above some height? You can do that easily and get a new data set back. It's not the same as selections, but there are analogous tools. Now, the last thing we'll discuss as far as raster concepts go, is this assumption I've given you that rasters have only one value per cell. This is true, but if we stack rasters in the same projection, and the same cell sizes into a single file, we can work with them as if they're a single raster with multiple values per cell. We call these mutli-band rasters, and each band is effectively a single raster. We use these most commonly, to represent data from different sensors captured at the same time. So that when the cells align, we can access whichever sensor we need. The most basic example of this comes back to images, again. For aerial imagery we have three or more bands. For those of you familiar with how humans and cameras see light, this makes sense. We have red light, green light, and blue light, and each of those are captured into separate bands and stored in the same raster. The values of each cell or pixel are a measure of the intensity of the light. To display them out to the screen as aerial images, ArcGIS interprets the red band as red light, the green band as green light, and the blue band as blue light, and displays it to you accordingly. This gets much more complicated from here, and we'll save this idea for a future lecture, but know that our hard and fast rule of one value per cell, isn't so hard and fast after all. In most cases, it's true. But these special multi-band rasters don't follow the rule. Okay, that's it for this lecture. In this lecture, we learned how rasters store information and display it back to us, the limits of that data when it comes to transforming data, a bit about the different tools available for working with rasters, and about multi-band rasters and imagery. Coming up next we'll look at these concepts within ArcMap. See you there.