00:01My name's Steve Kopp, this is Liz Graham, and we work on the Spatial Analyst team and today we're going to...
00:06...talk about an introduction to Spatial Analyst.
00:10Just curious, how many of you guys never used Spatial Analyst but want to?
00:17Great, so like almost half.
00:20How many of you use Spatial Analyst a lot but just specific pieces of it?
00:26And ninjas? Ninjas? People who know as much as we do. Alright. Or more.
00:37Not to imply that we are ninjas; we're definitely not.
00:40So, we'll leave you guys to the ninja status.
00:45Okay. So, for those of you don't use Spatial Analyst yet, we'll give you a little bit of an overview here...
00:51...at the beginning and we're going to talk a little bit about sort of an overview of Spatial Analyst, and then...
01:04So, we're not going to have a lot of time to go into depth on very many of those topics, and there's going to be...
01:08...a few times I'm going to refer to other workshops that are available on some of those topics.
01:13So, for some of the areas, it will talk about in these slides, there's entire workshops dedicated to those...
01:20...and we can point you to those, as well.
01:23There's also some demo theaters downstairs dedicated to some of these topics, as well.
01:29So first off, what is Spatial Analyst?
01:31Spatial Analyst is integrated raster and vector spatial analysis tools.
01:37And, it seems kind of obvious to include that slide and that bullet point, but I really want to reinforce this point.
01:44It's raster and vector.
01:45Spatial Analyst is not really just about raster analysis.
01:49So, we've actually gone to a fair bit of effort to allow what would have been traditional raster GIS capabilities...
01:55...spend most of our time going through the variety of analytic capabilities available in Spatial Analyst.
01:56...to accept feature data as input and to create feature data as output and to do that intelligently.
02:05So, it is an integrated system.
02:07And it's an extension product, as you know, that adds functionality to Desktop, where most of you use it...
02:13...as well as Engine and Server.
02:17Now, the graphic in the bottom right is actually Spatial Analyst running in a web page.
02:23This is actually a solar radiation calculation from a website from the City of Seattle.
02:29There were a few demos that you saw on Monday that were using Spatial Analyst and other geoprocessing...
02:34...functionality through web services.
02:37Yes, you can serve analysis capabilities. Okay.
02:41Not only can you serve them, but if you have geoprocessing functionality that's served, you can...
02:46...consume that directly as a tool in Desktop.
02:49That's not really what today's about; if you want to know more about that, I can point you...
02:52...to a workshop specific to that topic.
02:55Eight-thirty tomorrow morning.
02:59So, some key features of Spatial Analyst.
03:02There's about 170 geoprocessing tools for the Spatial Analyst extension that work on all the...
03:08...raster formats that you can use in ArcGIS.
03:10So, if you can read a raster dataset with ArcGIS, you can use it as input to a Spatial Analyst function.
03:18Similarly, it works on all vector formats, as well.
03:22Now, we include these points because this wasn't always true, and this isn't true in all GIS software.
03:28And also because the way that we do that changed a few years ago.
03:32We used to make scratch files in the background, and we don't anymore.
03:36So the reason that that's important for those of you who have been using it for a while is there were some...
03:40...limitations for those scratch file formats, and when we got rid of that, we changed the architecture...
03:46...how we read and write data through these analytic functions, it actually made it faster and it reduced...
03:52...some of the problems that you were running into that you maybe didn't quite understand.
04:00Spatial Analyst also includes a calculator with map algebra syntax, and I point that out because we think...
04:05...that's a pretty important key feature of doing spatial analysis.
04:10How many map algebra people do we have in the room today?
04:14Not very many here? Okay.
04:15We'll talk a little bit about map algebra; I won't dwell on it too much, though.
04:18And finally, Spatial Analyst has a really great developer experience, and there is a workshop dedicated to raster analysis...
04:25...with Python that specifically covers that, and a lot of that is about working with Spatial Analyst through Python.
04:34So what does Spatial Analyst look like in the desktop?
04:36So, Spatial Analyst is a toolbox in the ArcToolbox window with a set of functional categories.
04:45We can also access this functionality from the geoprocessing pull-down menu in version 10...
04:51...where I can also lock the toolbox.
04:53I can use these tools in ModelBuilder, which Liz will show us in a few minutes, and also use them in the Python window.
05:02What about Spatial Analyst toolbar? Anybody curious about that?
05:069.x people who haven't moved to 10 yet, there's a little surprise there for you if you used to use the toolbar.
05:11We got rid of the Spatial Analyst toolbar in many of the ways that you think of it.
05:15It's still there, but it only has a few tools on it now.
05:19That pull-down menu that was on the Spatial Analyst toolbar has gone away.
05:24The reason that that went away is it was causing a lot of confusion and it was basically duplicate functionality...
05:29...the things you could do in the toolbox and things you can do from the pull-down menu.
05:33They didn't share the same environment, they behaved a little bit differently, and what happened was...
05:39...when we made it possible to add geoprocessing tools to a dialog, to a pull-down menu using...
05:47...standard Windows customize, now you can make your own Spatial Analyst toolbar.
05:52So, I've got a little sample toolbar that I made here; I do a lot of hydrology work, and I work...
05:57...with the hydro tools in Spatial Analyst a lot.
05:59So here I created a little toolbar, it has a hydrology pull-down, and I put some of the commonly used tools on there.
06:05Sink, fill, flow direction, watershed, et cetera, as well as some custom tools...
06:10...a Python tool for calculating soil wetness, ModelBuilder models for doing condition numbers, things like this.
06:18So, I did all of this with no code; it's just standard Windows customize, drag-and-drop little things. It's really easy.
06:26The other thing that we added in version 10 was the ability to search; we improved the search capabilities...
06:32...to be able to search for tools.
06:35We did a lot of work on the names and words that we index to go with those to make it easier for you to find tools...
06:43...because the toolbox I showed you, you know, 170-plus tools just for Spatial Analyst...
06:48...the toolbox list has become too big to navigate that tree control.
06:52There's like 700-and-something tools in that list.
06:55So, using the search window is a much quicker way to do that now.
06:59We tried to put a lot of words in there that you might think of, you know, like, if you wanted to search for "clip" but...
07:04...you couldn't think of "clip," you can put in, like, "cookie cutter," and it'll actually find the Clip tool for you, okay?
07:11And, actually, if you search for tools and you're not finding the thing you're looking for...
07:15...and you have recommendations on new keywords we could add for things, just let us know, you know...
07:21...through tech support or through the forums or whatever, please add these keywords to this thing...
07:25...'cause this is what people in my industry always call this tool.
07:31Spatial Analyst includes a few additional pieces for the geoprocessing environment...
07:38...and these are things we want to set up at the beginning of a project before we start.
07:42Things like cell-size extent, map projection, and possibly a mask.
07:48Cell size is probably the most important one for Spatial Analyst users because it actually controls the results...
07:55...of your analysis and we default the cell size to the largest of the inputs because we don't want to fabricate data.
08:03What I mean by that is, in the example here, if I have two datasets, a 25-meter resolution and 10-meter resolution...
08:09...dataset and I want to create an output from that, the default behavior is it's going to create a 25-meter resolution output.
08:17The reason for that is I don't actually have enough information in my 25-meter resolution input...
08:24...to create a 10-meter resolution output.
08:27I can do it; I can change that environment setting to be the minimum of the inputs, and there might be reasons why...
08:33...you want to do that.
08:34I just caution you that when you change the default for the output cell size, do it for a reason and understand what that is.
08:44It's usually going to be best to set it to an explicit number or same as some other dataset...
08:49...that you want all the pieces of your project to conform to.
08:53The same with the extent.
08:55You want to set a processing extent to constrain the extent to a particular area.
09:00The extent for Spatial Analyst also includes another parameter called the snap raster, and what the snap raster...
09:06...does, it allows you to co-register the origin of the grid cells to be the same.
09:13The reason that that's important and you want to check the extent, the cell size, and the snap raster...
09:18...is because you want all the datasets that you create to be perfectly co-registered.
09:23Because if they're not, if they're offset by a quarter of a cell size or if they're different cell sizes...
09:29...every time you do an analysis, one of those inputs is going to get resampled to match one of the other ones.
09:36So if you set this at the beginning of your project, you're going to minimize the resampling that occurs...
09:41...and resampling of raster data is not something that you want to do repeatedly; it will degrade your data, right?
09:48So, it's better to leave the data like it is as much as you can.
09:53There's also a mask option, sort of to constrain your analysis to nonrectangular area, like a watershed or a...
10:01...census block or something like that, or a ZIP code.
10:05And then it's also important to think about and set a map projection.
10:10I know that most of the data that you find shows up in lat-long and you don't know what datum it is...
10:16...and you just do it in lat-long, it's fine, right?
10:20It's really not fine to do your analysis in lat-long very often, okay?
10:24Most of the time it's not.
10:26Because a lot of times you're going to do things like measuring distance and measuring areas...
10:30...and doing that in lat-long is just not going to work, okay?
10:34So think about the map projection that you need to use for the type of work that you're doing.
10:39There have been some changes in things like the buffer tool to use geodesic distances...
10:45...so you can actually input points, lines, and polygons in lat-long and get buffers out that are actually...
10:52...true-distance buffers, but that's only in the buffer command; that's not throughout the system.
10:57We're not using geodesic distance in area everywhere, so.
11:01Okay. Moving on.
11:04So one of the first things you want to do in your analysis is look at your data.
11:08You want to explore your data; you want to understand your data.
11:11So, the way that we do that, it starts with things like Identify, you know.
11:15Using the Identify tool and poking around and looking at my data.
11:19Changing the display of the data; using different stretches on my data.
11:23I can also do selections on the data, and the way that we do selections on raster data is through the attribute table.
11:29So, an integer raster dataset - a dataset that's integer values, not floating point values...
11:38Now, you don't have to have additional fields in that, connected to that to be able to do...
11:43...selections on it; you can search on the values.
11:46So if I have a dataset that was, say, land cover, and there was a land cover type coded...
11:54...with the value of 2, and that was Forest.
11:57If I wanted to, you know, highlight all of the, or select all of the forested areas, I could click...
12:02...on that in the attribute table, and it would select all the cells of value 2, or the forest cells.
12:08And then if I wanted to run an analysis, I could actually use only those selected cells if I wanted to.
12:13So that selection that I do on the attribute table is actually honored during the analysis...
12:18...just like if you did a selection on features and used that as input to your analysis.
12:24We have an interactive tool for doing histograms of the data distribution values.
12:28That's histograms of the entire dataset or of selected areas you can select through graphic.
12:35And also a zonal histogram tool, which I'll talk about a little more, which is in a little while...
12:40...which allows us to look at the distribution of values within zones.
12:46So with that, I'll switch over to Liz, and she'll show us a little bit of this.
12:54Throughout my demonstrations today I'm going to be walking us through a suitability analysis...
12:58...to find the best location for a new ski hill.
13:01But before I get started with that, I'd like to show you around the application that Steve was just talking about.
13:06So my data is in the Lake Tahoe area.
13:08I have some elevation data, some power lines, and some roads.
13:12The triangles represent peaks in the area.
13:15But first, let's enable the Spatial Analyst extension.
13:19To do that, you go to Customize > Extensions and check on Spatial Analyst.
13:23After doing this, you can access the Spatial Analyst functionality.
13:27The first thing I'll show you is the toolbar.
13:30Go to Customize > Toolbar and down to Spatial Analyst.
13:35As Steve mentioned, there's a drop-down list that has layers.
13:38I've selected Elevation.
13:40There's a Create Contours tool and a Create Histogram tool.
13:44You can see that this is the distribution of values in the elevation layer because that's the layer that I have selected...
13:49...in the drop-down list.
13:51If on the Drawing tool, I would use this Graphic Circle tool and draw a circle around Middle Mountain...
13:56...and then do the histogram, you can see I get just the distribution of values underneath the graphic.
14:03So that's a good way to explore your data initially.
14:06To access the Spatial Analyst tools, you can click this Toolbox button, which opens up the toolbox...
14:13...Spatial Analyst toolbox - made up of several toolsets that contain different functionality.
14:18But if you're unsure of the tool you're looking for, you can always use the Search tab.
14:22If you click Tools, you'll limit your search just to tools.
14:25And here, you can type in the tool name you're looking for...
14:28...something that reminds you of the functionality that the tool does that you're looking for.
14:32Another way to access the Spatial Analyst functionality is through the Python window.
14:37You can click this button to open up the Python window and type and execute Python expressions that use...
14:43...Spatial Analyst functionality, as well.
14:46Most often, Spatial Analyst tools are written in algebraic format, so the left-hand side is equals the tool...
14:53...or the function name or operators and that's the expression.
14:57We're not going to go into that a lot here today, but there's other workshops on that.
15:02So now we know how to enable and where to find the functionality, but as Steve pointed out...
15:06...before we do any actual analysis, it's important to set environment settings.
15:11If you go to Geoprocessing > Environments, this is where you can enable your different environment settings.
15:17Your workspace, your output coordinate, your processing extent - including your snap raster to limit...
15:23...the amount of resampling, and under the Raster Analysis tab is where you can set your cell size and your mask.
15:29I've set my cell size to 30; that's the coarsest resolution of any of my inputs for today's analysis.
15:36Now let's go back and search for a tool called Hillshade.
15:41It comes up there on a drop-down.
15:43If I cursor down to it and hold Ctrl Enter, the tool's just going to open.
15:47If I didn't do that, if I just hit Enter, it would just do a search and populate here and then...
15:51...I'd have to click and open the tool, so that's just a little shortcut.
15:55I'm going to select the input raster elevation.
15:58I'll supply a useful output name, such as Hillshade Oak.
16:05I'm going to leave the azimuth and the altitude of the sun as a default values.
16:11The z factor can remain 1 because all of my units, my x,y units and my z units, are all in the same measurement.
16:18They're all meters; I don't need to convert anything.
16:21This just tells me the dataset exists on my computer and I'm going to overwrite it and I'm okay with that.
16:29You can see what's added to the display here is a really nice illumination of the study area...
16:33...but what would make that even a better picture of what's going on is if I drag this hillshade result underneath...
16:38...the elevation dataset, and then if I add some transparency to the elevation dataset.
16:43To do that, I'm going to use the Effects toolbar.
16:47I also access that through the Customize toolbar's expression.
16:51Again, elevation is chosen in the drop-down list, and this is the transparency.
16:56I'm going to adjust the transparency on the elevation dataset to about 30.
17:00Now you get a clear picture of what's going on in the study area, where the ridges are...
17:04...and where the other slot areas are within the study area.
17:08And we'll pick up here after Steve gives us more insight on some functionality.
17:12Okay. Thanks, Liz.
17:18So there's a lot of tools in the Spatial Analyst toolbox, like I said, about 170 or so tools.
17:23These are some of the functional groupings of those tools, and you don't need to read all these because basically...
17:28...we're going to go through these, each of these a little bit, one at a time.
17:33We're going to start with the simplest stuff that there is, the mathematical operators.
17:37So, we can take two raster datasets and combine them together using sort of traditional mathematical operators.
17:44And what we do is combine spatial data with mathematical operators to create new spatial data.
17:50So, these are sort of the nouns, sorry, the verbs where the spatial data are the nouns of the language.
17:57So if we have two datasets we need to add together - we want to do a units conversion...
18:01...or something like that, this is the way that we do that.
18:04And if you look at this, you can understand now, when we talk about the cell size and the extent...
18:09...and the snap raster being the same, if you look at this graphic, where I'm trying to add two datasets together...
18:15...you can begin to understand why if they were different cell sizes, or if they offset a little bit...
18:20...I would need to resample one of them to match the other one to create a new output.
18:27We use these mathematical operators as well to do things like map queries.
18:31So we have a set of Boolean and logical operators that we can use on our spatial data.
18:37So in this case, I'm looking for places where the soil is sandy and dry.
18:42And I just use a Boolean "and" to do that.
18:44So, dataset and dataset, and it creates an output for me of true and false, or a Boolean output.
18:54We also have a set of mathematical functions that we can use to create new analytics capabilities.
19:00So, traditional arithmetic functions; the example at the bottom is about turning floating point data into integer data...
19:07...which might be something you need to do - trig functions and exponential functions.
19:12I might want to make a soil wetness map, where I know that the formula for that is the natural log of the area...
19:20...divided by the tangent of the slope.
19:22So, I can actually combine these things together in a single expression to make that result.
19:29The way that I do that is using this language we were referring to called map algebra.
19:35Now map algebra is a tool in the toolbox, but it allows you to access a lot of other functionality that is...
19:42...also available as geoprocessing tools.
19:45Now, the nice thing about the map algebra is it's very readable; it's a nice sort of natural-language way...
19:51...to look at spatial analysis functions.
19:55The example in here, smooth hill equals hillshade, focal statistics, elevation times .3048; does that sound like...
20:04...it might be anything you might understand what that is?
20:07What would it mean if I had elevation data and I'm trying to multiple it by a number like 3048?
20:13Does that ring a bell to anybody?
20:14That might be a units conversion, yeah?
20:17Okay. So that's really all I do.
20:19Elevation dataset, this is my raster dataset, the times sign, and the number I want to multiply it by.
20:26Now, just like in any other mathematical expression, I want to do that thing first and...
20:32...use the output of that as something else.
20:34So I put parentheses around that and then run the focal statistics function.
20:39The focal statistics function's default behavior is to calculate the mean value within a 3-by-3 neighborhood.
20:46So what I'm trying to do is smooth the data a little bit.
20:49I'm trying to create a dataset I want to use for cartographic effect.
20:53And then I'm going to create a hillshade of that, which was the tool that Liz just showed.
20:58So I create an expression with an output named "smooth hill" equals the hillshade of the result...
21:04...of the focal statistics function, which is taking the result of elevation times .3048.
21:10Okay? Seems pretty simple once you break it down, doesn't it?
21:14So there are some changes that we made to the map algebra, for those of you who usually use it.
21:18We changed the map algebra a little bit in version 10 to make it work better with Python...
21:23...which is the native scripting language of ArcGIS.
21:26So now we have a really tight integration between map algebra and Python, which allows us within...
21:31...the Python language to mix map algebra and all the other geoprocessing tools, as well as to...
21:37...import and use other Python libraries with our ArcGIS functionality.
21:47So moving on sort of in complexity a little bit, we have a set of distance and proximity tools in Spatial Analyst.
21:55So we can calculate straight-line distance - you know, Euclidian distance from point to point.
22:00We can calculate the distance; we can calculate the direction of movement; we can allocate space to a collection...
22:07...of points; and we can also do what's called cost weighted distance, or cost allocation where we use...
22:14...another raster dataset that provides what we should think of as an impedance surface...
22:18...or a cost of travel through an area.
22:21And in both of these cases, we can also do things like shortest path.
22:25What is the shortest distance between two points, if I'm using this cost distance method?
22:32So this is an example of some of the outputs that we might create.
22:35In the center, I've got a little dataset, point A and point B, and I want to find the distance between them.
22:42So I can do a map of straight-line distance, and this looks like buffers, doesn't it?
22:46Well, it's a little different than doing things like the buffer tool would do.
22:51What the buffer tools does is allows you to create circles of specific distance away from features.
22:59What we do with Spatial Analyst is we actually create a raster dataset where each cell is coded with the value...
23:07...representing the distance to the nearest point.
23:09So it's actually a continuous distance map.
23:12And then we can threshold that through the renderer and through reclassify to do other things with it.
23:18But just remember; it's a continuous map where every cell has a different distance...
23:23...and it's the distance to the nearest point.
23:27Another output of these distance functions, as I mentioned, is the direction grid, which is the...
23:31...direction of travel to the nearest point.
23:33So in this case, this would be a Euclidean direction output.
23:37We can do an allocation of space, so if I do an allocation - this is sort of like Thiessen polygons or Voronoi polygons.
23:45And we can also do a cost-weighted distance, and this is where I'm bringing in an impedance surface or a cost surface...
23:51...another raster dataset whose values are multiplied by the map distance to make a map that represents...
23:58...a difficulty of traveling between two places.
24:03So, if we look back again up at the top, obviously if I was doing Euclidian distance between two points, A and B...
24:09...it would be a straight line between them.
24:12But, maybe for some reason I can't travel through here.
24:15I want to consider some other things because what I can see in the map is, that's actually...
24:19...traveling over the top of a mountain.
24:22That might not be a place that I can go; maybe the vehicle that I'm traveling in won't go there...
24:27...or maybe I just don't want to walk up that hill.
24:30So what we can do is use the Spatial Analyst to combine multiple datasets to make a map...
24:35...that represents the cost of travel.
24:38That might be the cost of traveling uphill versus downhill.
24:41It might be, how difficult is it to travel through this land cover type?
24:45Is it easier to travel over barren ground or forest or rock or through agricultural fields or through urban areas?
24:54Or, is it a soil type that I can't travel over? Or is it too wet? Or is there a barrier in the way?
25:00Okay. So I can combine those things together and use those in the cost distance function to weight the result.
25:06And what I end up with is a path that goes down along the shoreline of the lake, right?
25:13Now, what if I don't want to know just the path?
25:16Maybe there's some reason I need to know other options.
25:19So what we can do is create a corridor of travel between those two points, okay?
25:25So what we're actually doing here is, from point A, calculating the cost distance to every point in the dataset...
25:32...and then the same thing from point B, the cost distance to every point in the dataset.
25:38So this is what the cost distance map actually looks like.
25:41And see, it's not concentric rings or buffers; it looks much different than that.
25:46Then we combine together those two cost maps, and what we see is, if we don't want to travel down along the lake...
25:52...there's actually an alternate corridor here over the top of the mountain.
25:56It's higher cost, but for some reason, I might be interested in that.
25:59Did somebody have a question?
26:01[Inaudible audience question]
26:20So the question was, what if I have a vector grid - so you mean like a regularly-spaced set of points...
26:28...that I wanted to use in my model.
26:30So in order to do that in these particular functions, you would actually convert those to a raster dataset.
26:38But if they're already gridded, you wouldn't actually probably do an interpolation; you just use like the Points...
26:43...To Raster tool and just convert it directly.
26:51Okay. Moving on. Density functions.
26:54So, we have variety of ways in Spatial Analyst to create surfaces, and one of those is to create density maps.
27:01And a density map might be what you guys think of when you think of the term heat map.
27:06How many people are familiar with the term heat map?
27:11Okay. And if I ask each of you around the room, What is a heat map, I bet we'd get at least...
27:17...three different answers; maybe more.
27:21And I would assume, also, that some people have made things with the density functions that they call heat maps...
27:27...and some people have used other tools in ArcGIS and called them heat maps.
27:31Heat map is kind of a generic term, but most commonly what people are thinking of...
27:35...when they think of a heat map is an amount of occurrence of something...
27:39...a frequency map of how common is something in a particular place.
27:44So normally, we would do that with the density functions.
27:47We have two ways to do that, the simple density and kernel density.
27:50And the difference is just that the kernel density function creates a smoother output.
27:54It fits a distance-weighted kernel over the data so you don't get these concentric rings...
28:00...where you have sparsely sampled data.
28:02But what we use density for is to make maps that represent frequency or magnitude over some area.
28:10So, the number of people per square mile, number of trees per hectare, you know...
28:15...number of roads per, you know, whatever.
28:19So we can use feature data, point data, or line data as input to these density functions.
28:26So just keep in mind - density functions are about magnitude per unit area.
28:30They're something you use on count data.
28:33I have a number of something.
28:36And I just make that point to contrast it from later on when we talk about interpolation.
28:43We have a collection of tools in Spatial Analyst for doing neighborhood and block statistics.
28:48You know, like, what does that mean?
28:50Well, neighborhood and block statistics are about looking at individual raster cells and their neighborhoods...
28:57...and calculating some new value.
28:59So we might use these for, a classic application would be like image processing functions...
29:06...where you do want to do high-pass and low-pass filters and things like that.
29:09Those are neighborhood operations.
29:11We use them for filtering; we use them for data smoothing and the block functions as well for data aggregation.
29:18But really all we're saying is, for a single processing cell, like the one in the middle here that has the value of 1...
29:25...we want to look at a 3 x 3 neighborhood and calculate some new value; and, in this case, we're going to...
29:31...calculate the mean value of that 3 x 3 neighborhood.
29:34That value is output at 3.22; I write that to that cell; and this is a scanning function that just...
29:40...goes across the entire dataset and does the same calculation for each cell.
29:45So every cell in the output on the left, on your right, is going to have a different value potentially.
29:53Contrast that with the block-statistics functions where, with that same 3 x 3 neighborhood, I'm actually...
30:02And then I don't move over one cell.
30:04I actually move over an entire block and make a new calculation.
30:09So we use this primarily for data aggregation purposes.
30:14And one of the more common utility functions that people use in Spatial Analyst is the zonal overlay tools.
30:21So first we need to understand what a zone is.
30:23That's all the cells that have the same value.
30:26So I talked previously about, I had a dataset that was land cover, and the value of 2 represented forest.
30:32So that 2, that forest, would all be one zone, okay?
30:37And what I can do is calculate, say, for each land cover type.
30:41What is the average elevation per land cover type? Okay.
30:47Or I can calculate, what is the average slope per watershed?
30:52So the zones that I'm using as input, these areas, they can be rasters or they can be polygons.
30:59...calculating the mean value and applying to the same block.
31:02So I can use a set of polygons as my watershed boundaries and input a continuous dataset like slope...
31:08...and then calculate statistics on those to drive, like, my hydrologic model.
31:15This is an example of slope at the top, a very continuous dataset.
31:21At the bottom, I have watersheds as polygons, and the output is computed as average slope within each watershed.
31:32Now similar to that, we have a data exploration tool called Zonal Histogram, and what Zonal Histogram...
31:38...allows us to do is to create graphs.
31:42Maybe we don't want to know what the average slope is in each watershed.
31:46We want to know, say, a breakdown of the distribution of slope within each watershed.
31:52So the graph at the bottom, I've got five watersheds that I was looking at and I have five slope classes in each of those.
31:58This is getting to that issue of raster/vector integration.
32:05...whereas in watershed 2, 3, and 4, there's a lot of higher slopes.
32:10And I can tell from just that bar graph that watersheds 4 and 5 - sorry, 5 and 6 - are actually much smaller in area...
32:19...because I can look at the cell counts, how large of an area they are.
32:23So, it's an interesting data exploration tool and really can help you in understanding...
32:27...your data and the phenomenon that you're modeling.
32:31So when I was talking about density a moment ago, I said that it's about working with count data.
32:37Contrast that with interpolation tools, which are about working with measurement data.
32:42So this is another thing that people might think of as a heat map.
32:46What we're doing here is taking measurements in space and trying to estimate values at unmeasured locations.
32:55I went out into the landscape and I sampled soil chemistry at a bunch of places and I want to create a map...
33:01...that represents, you know, soil pH.
33:04Now, I didn't go out into that field and sample every 30 meters across the entire landscape, you know.
33:11Off of, you know, a very large area, I might only collect a hundred samples, but I might create 5,000 cells.
33:19So we use interpolation tools to estimate values at unmeasured locations.
33:24Now there's a lot of ways that we can do that in Spatial Analyst, in 3D Analyst, in Geostatistical Analyst.
33:30We're not going to go into depth on that; we did a workshop on that earlier today, and those...
33:35...slides will be available online if you want them.
33:39But there's a variety of ways that we do that - I think I'll just flip over to this one and help you out...
33:45...because we get a lot of questions about interpolators.
33:48Because there's many ways to interpolate data, people have a lot of questions about it...
33:52...because how you chose an interpolator has a lot to do with what the output surface might be.
33:59And the thought that goes into choosing that interpolator has a lot to do with how that data was sampled...
34:04...and what the phenomena is that you're trying to model.
34:07So, a few quick pointers.
34:10You don't know anything at all about your data but you need to make a surface and see what your data looks like.
34:15Use the Natural Neighbors interpolator.
34:18So, why Natural Neighbors?
34:20Because interpolators have some characteristics that allow them to exaggerate data.
34:26They estimate values that haven't been sampled.
34:29So they might actually create values that are higher and lower than what was actually in the sample dataset.
34:36Sometimes these are good things, but if there's too much of it, they create what we think of as artifacts in the surface.
34:42They're exaggerations of things that might not really be there.
34:46So the good thing about Natural Neighbors is it's fast, it's simple, and it doesn't create any artifacts.
34:52It's very clean in that regard, very conservative.
34:56So if your input data is contours, there's really only one answer, and that's to use the TopoToRaster tool.
35:02The TopoToRaster tool is designed to work well with contour data input.
35:07I'm not going to go into why contours are difficult for interpolators today, but I'd be happy to...
35:12...explain it to you and it's in the other presentation.
35:15If you're not trying to create an elevation dataset from your contours - let's say your contours represent...
35:21...soil chemistry - then there's a parameter in the TopoToRaster tool called Drainage Enforcement...
35:26...just click that little checkbox off, okay?
35:30If you're making elevation data, particularly for hydrologic purposes, leave the drainage enforcement on.
35:37If you're not creating an elevation surface, turn the drainage enforcement off.
35:42TopoToRaster also allows you to input not only contour lines but also things like spot heights...
35:48...points, streams, lakes, cliffs, things like this, so.
35:53It's a very specialized tool and it's very useful if that's the type of surface that you need to create.
35:59So what if you know that the high and low values in your phenomena have not been sampled...
36:05...but are actually really important to you?
36:08Then, if you're just using Spatial Analyst, the recommendation is to use spline.
36:12But you need to be careful when you use spline because if you have points that are very close together...
36:18...but very different in measured value, that's when you're going to start seeing artifacts in the surface...
36:23...because spline is going to try to fit a smooth curve to those points, and you're going to end up...
36:28...with these big exaggerations in your surface.
36:30So you just need to be a little careful when you use spline.
36:34Now if you happen to have the Geostatistical Analyst extension, our recommendation...
36:39...instead would be to use local polynomial interpolation.
36:43A lot of people actually have the Geostatistical Analyst extension but don't actually know it, so...
36:48...if you're not sure if you're on a ELA or a university site license or some, you know, government contract...
36:54... or whatever, there's a really good chance you actually have Geostatistical Analyst and just don't know it.
36:59So it's worth looking into; there are some really nice interpolation methods in Geostatistical Analyst.
37:06So what if your surface is not continuous?
37:09And that may seem a bit like an oxymoron, that I have a discontinuous surface, but it's a real thing.
37:14So I'm trying to create, for example, a surface that represents a geologic formation, a buried sand lens.
37:24And, I want to model the top of that sand formation, but it's old geology, right?
37:29There's faults and things that happen, and those faults have offsets.
37:33So when I use an interpolation algorithm, if I know that there's a surface with a distinct offset...
37:40...I can include that as a feature, as a line feature, in my interpolation so that I don't look across...
37:46...that barrier for sample points on the other side, okay.
37:51So if that's the case, there's a tool called Spline With Barriers for you to use.
37:55So, if for some reason you need to use a geostatistical technique or want to explore that...
38:03...then I would recommend you look at the Geostatistical Analyst and use the empirical Bayesian kriging technique.
38:08This is a new one in 10.1; it's also known sometimes as easy-button kriging or black box kriging.
38:16It actually does a lot of the hard work of geostatistics for you because it does parameter estimations.
38:22So you don't actually have to do all the variogram modelings that sort of intimidates...
38:27...people about using kriging techniques.
38:30The other nice thing about empirical Bayesian kriging is it doesn't assume stationarity.
38:35So there's an assumption about stationarity in structure in data that's a requirement for traditional kriging.
38:42If you don't understand that, that's okay.
38:44You don't need to understand that in order to use EBK.
38:47You only need to know that that's no longer a requirement, because what it does is it fits the model...
38:53...it changes the fit of the model as you move across the dataset.
38:57If you're interested in geostatistical techniques, I would encourage you to look into the empirical Bayesian kriging...
39:02...and I believe Steve is doing - not me, Steve, but the other Steve - is doing a demo theater...
39:08...on surface interpolators this afternoon.
39:14So once we have - oh, sorry, this one is about interpolation with barriers - this is the result from the...
39:21...Spline With Barriers tool, and what you can see here in the top right is this is a geologic surface...
39:27...with a series of offsets and gaps in it.
39:30So, at the bottom left, this surface is actually a salt dome formation where there's actually gaps...
39:37...okay, or holes, in this geologic formation.
39:40It's not continuous.
39:43If you look at the shape of the contours you can see, obviously, you know, we've got 4,000 here...
39:48...and 5,000 over here - there's a big offset between the two sides of this surface.
39:56So once we've created those surfaces, we have a bunch of analytic tools we can use to analyze those surfaces.
40:03Liz showed us the Hillshade tool, which we commonly use for display purposes and we merge that...
40:08...with the elevation data, but also things like slope.
40:11We need to make maps that represent the steepness of the surface.
40:15You know, if I'm doing house siting suitability, building suitability, there's certain slopes that I can build on...
40:22...and certain slopes that I cannot build on.
40:24I need to be able to make a map of that.
40:28You need to calculate the aspect on the surface.
40:30I want to know which way the sun shines, okay?
40:33I want to know if, at high latitudes, am I on a south-facing slope or am I on a north-facing slope?
40:39This has a lot to do with what kinds of plants will grow there, how much solar energy they receive.
40:48And we also have a variety of viewshed tools, viewshed observer points and a few others.
40:52And these allow us to specify locations in the landscape and understand what can be seen from this place, okay?
40:59I'm putting in a fire tower and I want to see what I can see.
41:03I'm putting in wind turbines and I want to know where they can be seen from.
41:08We use the viewshed tools for this.
41:13We also have some solar radiation tools, and what these allow us to do is calculate...
41:18...the amount of solar radiation hitting the earth's surface.
41:22Now, this is not a generic tool that just looks at, you know, your latitude and some simple equation.
41:28It actually uses the morphology of the terrain and looks at shadowing effects at different times of day and times of year.
41:36So it actually has a calendar in it that allows you to specify, you know, give me the solar radiation budget sample...
41:43...you know, every three hours from June through July for this particular area, okay?
41:52The output of this tool is measured in watt hours per meters squared.
41:57Originally when we put this in, we did it for sort of natural resource application regions...
42:03...because this is information that's used, for example, in fire modeling to estimate plant moisture.
42:10It's used in a variety of things like snow hydrology and other things.
42:14But the place that it's actually gotten used the most is solar panels.
42:19People want to figure out, If I put solar panels on the roof of my house...
42:23...if I put solar panels on the roof of this building, how much energy can I actually create?
42:29Now, in an urban landscape, it's actually important that you understand the shadowing effect...
42:34...from adjacent buildings, and that's what this tool will do for you, okay?
42:38So if you look here in the middle, you would think in downtown LA, the roof of every building...
42:42...would have the same solar potential as every other building, wouldn't you?
42:45But it's not actually true, because if you look here in the middle, there's a couple of blue ones.
42:50They have really low solar potential because that's a shorter building.
42:54And on the south side of the building are some taller buildings that, at certain times of the day, shade it, okay?
43:01If you're shaded early in the morning or late in the day, it's really not a big deal; you're not missing much solar radiation.
43:07But if you're shaded in the middle of the day, it really makes a much bigger difference.
43:11So the way that this was done, there's a few, actually, websites, Solar Boston and Salt Lake City...
43:18...are two of the ones that are pretty nice.
43:29...is actually all available on the Resource Center.
43:31If you wanted to build one of these for your own city, it's a pretty clearly understood model of how to do that now.
43:37I think Ryan's actually doing a demo theater on this, this week, as well.
43:41So with that, I'm going to turn it back to Liz, and she's going to start telling us about some of the...
43:46...pieces of this ski suitability model we're going to build.
43:50So previously I showed you how to enable the Spatial Analyst functionality and how to access the tools.
43:57Let's go ahead and see some of the tools, or toolsets, Steve was just speaking about.
44:01He mentioned the density tools, such as kernel, line, and point density.
44:05The distance tools, such as corridor, cost distance, cost path, and path distance; as well as the interpolation tools...
44:13...IDW, kriging, natural neighbors, TopoToRaster.
44:18This is where you find your raster calculator to execute your map algebra syntax.
44:24The Math toolset contains these math functions, as well bitwise, logical, and trig functions.
44:31The Neighborhood toolset is where you find the block statistics and the focal statistics.
44:36Solar radiation, which was what Steve just finished speaking about.
44:39There's Area Solar Radiation, Point Solar Radiation, and this diagnostics tool - the Solar Radiation Graphics tool.
44:46In the Surface toolset, you've already seen the Execute Hillshade, but you also find aspect, slope, and viewshed.
44:54And down here in the Zonal toolset is where you'll find Tabulate Area, Zonal Histogram, and Zonal Statistics.
45:00So now that we know some of the functionality, what Spatial Analyst can do, and where to find the tools to do it...
45:07...let's think of some criteria that are going to be important when determining the best location for a new ski hill.
45:11Some of the criteria I came up with would be snow depth.
45:14We want a place that has a lot of snow so that we don't have to make as much snow...
45:18...so it's a little cheaper to keep our ski hill open during the season.
45:22So snow depth is going to be important.
45:25I thought maybe the distance to the center of town.
45:27There's a town down here, and I think if we built the ski hill closer to the center of town...
45:31...maybe we could get more people to show up in the evenings and maybe have some night skiing...
45:35...and maybe generate some more money for our ski hill that way, so that might be an important criteria.
45:40And gradient's an important criteria; we don't want to put it anywhere too flat.
45:43Even if it's a high elevation, the gradient's important for the slope, right?
45:47We need a good slope for a ski hill.
45:49So let's look at the tools and the data that I have and see if I can generate that information...
45:53...and use it in my suitability model.
45:56So I have elevation data, and Steve taught us about the Slope tool, so if I go here to the Surface toolset...
46:02...and open Slope tool, I can input the elevation data, give it a useful output name, such as Slope Out.
46:12I'm going to leave it in degrees, although I could also output in percent rise and again the z factor can remain 1.
46:19I go ahead and execute that tool and add it to the display with this dataset, and you can see that green...
46:25...represents relatively flat areas, small degrees of slope, and the red represents...
46:30...much cheaper areas relative to my study area.
46:33So here is relatively flat and over here it's relatively steep.
46:37So that's going to be important when we go to combine our criteria for our ski hill.
46:42Another thing I mentioned was snow depth.
46:43I think that would be important.
46:45Now I don't have a raster that represents snow depth, yet; all I have are these points.
46:49And at each of these locations, there's a weather station, and I know the average annual amount of snow.
46:55Using those points, I can use one of the interpolators and generate a continuous surface.
47:01The interpolation tool I'm going to use is Natural Neighbors, which is the most conservative of the interpolators.
47:08I'm going to input my snow-depth points, select the snow-depth field - that holds the value of the snow depth.
47:15The cell size is inherited from my environment settings.
47:19I may as well as give it a name such as, I'll put Snow Depth.
47:27And go ahead and execute that tool.
47:31The result is added to display - it's not very intuitive because of the color, so let's fix that.
47:35We'll go to Properties > Symbology and find a different renderer.
47:39How about light blue to dark blue.
47:42Areas where it's light blue are relatively, I'd say relatively shallow, and areas that are dark blue are relatively steep.
47:51So now we have a continuous surface representing snow depth.
47:54That will be an important criteria.
47:56The last one I mentioned is the distance in the center of town.
47:59I have a little line segment down here - you can't really see it - it's the center of town, and I'm going to generate...
48:05...using one of the distance tools - the distance from every pixel in my input study area to that location.
48:12And to do that, I'm going to use the Euclidian distance tool.
48:17My input feature, in this case, is the center of town.
48:23Again, I'll provide an output name, EucDistTown.
48:29The maximum distance I'm going to leave empty, and this is going to allow the tool to execute...
48:33...to the full extent of my study area.
48:34If I wanted just to cut it off after 5 kilometers or 10 kilometers, I could enter that here, but it's not necessary in our case.
48:41And again, the cell size is inherited from the environment.
48:44I'll run the tool.
48:48So now you can see that the yellow areas, or orange - I don't know what color that is - is relatively close...
48:52...to the center of town and blue is relatively far.
48:55Every pixel in this output has its own value.
48:58If I click here using the Identify tool, it's about 9 kilometers to the center of town.
49:04This location is about 15 kilometers to the center of town.
49:09And this one's about 21 kilometers to the center of town.
49:14So in the next demo, I'll combine these in a meaningful way to find the best location for a new ski hill.
49:18But right now, Steve will teach us a few more things.
49:23So, so we've got things like slope measured in degrees, distance in kilometers, snow depth in meters.
49:34And we want to combine those together to create a map that represents how good of a place is this to build a ski resort.
49:43So how do we combine things like that together?
49:47Well, the first thing that we want to do is that we want to change some of those variables...
49:53...to what we think of as a common scale.
49:56We want to remap these values where each dataset will represent its suitability for a ski resort.
50:04What does that really mean?
50:05It means, I want to take the slope map and turn it into something that represents how good...
50:11...are certain slopes for building a ski resort?
50:15How good is a certain snow depth for a ski resort?
50:18How good is a certain distance to town for a ski resort?
50:23The way that we do that is with the reclassification tools, and what this allows us to do is reclassify...
50:29...individual values or ranges of values into some new value that has a different meaning for us.
50:35So we can take slope ranges - I want to build a ski resort, so I want to take things, you know...
50:41...0 to 20 degrees and make them one class, and 20 to 30, and make them another class.
50:47I use the reclassify tools to do that.
50:52And then when I want to combine those things together, what I can do is assign weight values to them and say...
50:59...you know, slope is really important, 'cause it's about downhill skiing.
51:04And distance to town is kind of important.
51:07Or, having good snow is really important.
51:10So I can add what we call a percent influence - you can't quite read it in the dialogue...
51:14...but you'll be able to read it in the demo - a percent influence of how important each of those variables is.
51:22And each of those datasets, those variables I have reclassified into classes that make sense for my suitability model.
51:30And I can use the Weighted Overlay tool to assign individual values to those.
51:36Maybe a little better example of that.
51:38So we use the Weighted Overlay tool to do sort of the classic GIS problem of finding "where is the best place."
51:45And "where is the place" is a very, sort of generic concept.
51:49But the way that we're doing this for finding the best place for a ski resort is the same way that you would...
51:56...find the best place for a coffee shop, or a likely place to find an endangered species, or the best place to...
52:00Change the slope values, okay?
52:04...you know, grow a certain kind of plant, okay?
52:08It's really all the same pattern.
52:10So, think about what kind of problems you're trying to solve in terms of "where is the best place"...
52:17...and what kinds of criteria you might use from the GIS to do that.
52:21What criteria would you use, what kinds of classes would you use to describe those?
52:27And then you can use the Weighted Overlay tool to combine those things together...
52:31...to assign those weight values, the importance of pieces of this, and you run the Weighted Overlay tool...
52:38...and you get a result, and you're like, wow, nice map!
52:41And then what do you do?
52:43You're not done.
52:45The real power of the Weighted Overlay tool is that you can just go back to the same dialog, it's already populated with...
52:51...everything you set, and you say, "Well, you know, the snow depth is much more important than I thought it was."
52:57So change the snow depth percent influence.
53:03Play with it.
53:04It's all a subjective model.
53:06People use the Weighted Overlay tool, for example, the people from Trust for Public Land...
53:11...who won the award on Monday - the TPL people, Breece and those guys - they used the Weighted Overlay tool...
53:18...and technique in stakeholder meetings with citizens.
53:22They're trying to figure out the best place to build parks.
53:25They invite the community in and then actually have a documented workflow, these exercises that...
53:30...they take the community through to have them determine which criteria they think are important...
53:36...and to assign those weight values.
53:38And then they compare each other's maps, right?
53:41So, it's not something for one person to do as an expert and make a decision by themselves.
53:46This is actually a collaborative tool where you can actually engage other people and get expert opinions...
53:51...on different pieces of it.
53:53And what the Weighted Overlay tool itself allows you to do is just have an easy place to modify...
53:58...those weights and rerun the tool without having to go redo all your work again.
54:03You just change a few values, hit OK, and do it.
54:08So we'll let Liz show us how we actually do that.
54:13As promised, it's time for the modeling part.
54:17So to create a model, you right-click, say New Toolbox; that's the first step.
54:24As you create a new toolbox, you right-click New Model.
54:29I already have a model started here; I'm going to open it up in an edit session.
54:33I didn't only think of those three criteria that we looked at last time.
54:36I thought of a few more criteria that we should include, and then I categorized them into three submodels...
54:42...issues that had to do with the development the ski hill, the terrain of the ski hill, and the accessibility of the ski hill.
54:49So let's go ahead and zoom in and see part of this model.
54:52In terms of development, I thought things that might be important were the soil type.
54:56What soil's best to build on, what's difficult to build on, especially with slopes involved.
55:01Land use. Who owns the land?
55:02Will I ever be allowed to put a ski hill on there?
55:04Is it protected land or private land or public land?
55:07Oh, that's landowner; sorry - land use.
55:10Is it swamps? Is it forest? Are there already buildings there?
55:14I mean, that's important stuff to consider when building, when looking for a new location.
55:20Terrain. We already looked at snow depths and we generate our continuous surface that represents snow cover.
55:25Elevation. You saw me, I already used slope to calculate gradient across my study area, but elevation...
55:31...is also the input to the Solar Radiation tool, where I can get an output of how much sun this slope is receiving.
55:39Maybe too much sun would affect the snow depth, but maybe not enough sun would be really cold...
55:44...in winter months to ski there, so that might be important to consider.
55:48In terms of accessibility, it's expensive to build roads and power lines, so we want to consider how far our new...
55:54...ski hill is from existing infrastructures, or how close it is to the center of town, 'cause we need lots of people there.
56:00So there's all kinds of criteria to consider.
56:03But the first thing you need to do after you identify the criteria is derive some data from them - some...
56:09...continuous data - like we did in our last presentation, using Natural Neighbors or slope or what have you.
56:15And now you have a whole bunch of rasters that are in different units of measurement...
56:19...whether it's degrees or - what were some of the - the distance to the center of town.
56:26So now as Steve pointed out, we need to reclassify these datasets.
56:30So down here, I have a section of my model that's not complete.
56:32I have the output and the Slope tool, but now let's use the Reclassify tool to reclassify the data...
56:38...so that I can use it with the rest of my model.
56:40So if I go to the ArcToolbox, reclass toolset, and drag Reclassify onto my model...
56:47...and use this connecting one, the output of slope can become the input of Reclassify.
56:52If I double click Reclassify, you can see that the input is slope elevation.
56:58I'm reclassing the value field and these are the degree values.
57:02This class represents zero to two degrees, and it's being assigned a value of 1.
57:07Let's go ahead and set up our own classification.
57:09If I click this Classify button, you can see here that there are several different classification methods...
57:14...I can select, such as equal interval or natural breaks.
57:18And I can also choose the number of classes that I wanted.
57:22Now since we're trying to do something specific with this in analysis, let's consider how these values interact...
57:28...with the notion of building a new ski hill.
57:30I spent some time thinking about this and I think we need five classes.
57:33We need a class that represents relatively flat areas for parking lots and lodges; something that represents...
57:39...a relatively little bit of steepness, or a small degree of steepness, for bunny hills and greens...
57:46...something a little bit more for the intermediate skiers, more for black diamond...
57:50...advanced skiers; and then probably one category that's just too steep to ski, like it's not useful for a ski hill.
57:57I can create these classes by dragging these bars and maybe setting it to where I want...
58:02...but I think it's a little easier just to come over here and type.
58:06I'm going to have my first class, from 0 degrees to 10 degrees.
58:10My second class is from 10 degrees to 20 degrees.
58:14So 0 to 10 can be good for parking lots; 10 to 20 is good for beginner slopes; 20 to 30...
58:20...we'll call those intermediate slopes; 30 to 45 can be our advanced, our black diamond slopes; and then...
58:27...everything else, 45 to 60, too steep to ski.
58:30I'm going to say, Okay, and I'm going to go ahead and run the Reclassify tool.
58:39I want to add this to display, but first, I'm going to set layer symbology so it's something meaningful to us.
58:46So Slope Reclass is the layer of symbology I set up.
58:49Now, if I add this to display, you can see that the gray areas represents good for parking lots...
58:55...green, the 2, is relatively not steep, good for beginners; blue is for the intermediate; black would be...
59:04...black diamond slopes; and the red is too steep to ski.
59:08So we've done that, but how important are they in our model?
59:10That's the next question we need to ask ourselves.
59:12And how important is that information in relation to the other criteria in our model?
59:18So the output from Reclassify is using the Weighted Overlay tool.
59:23Going into this Weighted Overlay tool, I have three datasets.
59:26I have the reclassified snow depth values, the reclassified radiation values from the solar...
59:32...and now the reclassified gradient values.
59:39So I have my five classes here on the left; now let's, how important are they?
59:44Well, all reclassifies in this model are all weighted values on a scale of 1 to 9...
59:52...where 1 represents not important at all or undesirable or unsuitable for a ski hill, and 9 is the value...
59:58...I selected to be highly valuable for a new ski hill.
1:00:01So we need to maintain that same scale across all of our datasets.
1:00:07We know that 5 is highly unsuitable; it's not good for a ski hill, so I'm going to give that a 1.
1:00:13Parking lots are important, but I feel as though we could probably flatten some land if we need, so let's give that a 3.
1:00:19Now what kind of a ski hill do we want to focus on?
1:00:21Do we want it to be for beginners, intermediate, or advanced skiers?
1:00:25What's important to us?
1:00:27Well, I'm going to design it for advanced skiers, the adventurous people.
1:00:30So I'm going to go ahead and give this a 5; the intermediate's even more desirable, and the black diamonds...
1:00:35...are the most desirable slopes in our suitability analysis.
1:00:40Next over here, you can see I have a red X, and that's telling me that my percent influences do not yet equal 100.
1:00:46They only equal 70.
1:00:48So we need to set the percent influences across these three rasters to equal 100.
1:00:53So what's the most suitable, or what's the, has the most percent influence within our model?
1:00:59What's the most important criteria out of these three, and how do they relate to one another?
1:01:03Snow suitability, I think, is extremely important because snow is expensive to make and it's necessary for a ski hill.
1:01:10The slope, I think, is probably second to that.
1:01:12And sun, well, people are going to have to dress warmer if there's not enough sun.
1:01:15I mean, it's not quite as important, alright?
1:01:17So now I've sent my present influences, and the results of this weighted overlay...
1:01:24...and the results of my other two submodels all go into this weighted overlay and I do the same process.
1:01:31I'm going to run both of these weighted overlay tools and the final result will be added to display.
1:01:37It's running the two tools now.
1:01:41And let's take a look at our final suitability.
1:01:43So we know we set it up on a scale of 1 to 9.
1:01:46Not all the values are there; that's okay.
1:01:50Two is the least desirable area in our study area.
1:01:54Let's not build a ski hill over here.
1:01:56Eight, the darker green, is the most highly desirable location for a new ski hill, and that looks to be...
1:02:01...around Sourdough Hill in Lost Corner Mountain.
1:02:03So that would be the best place to build a new ski hill for advanced skiers.
1:02:07But if we opened up the model and started changing our weights around...
1:02:10...if we change the idea or if we got new development information, then we go back to the model...
1:02:15...modify some of the numbers in the Weighted Overlay tool, and run it again.
1:02:18Like Steve was saying, it's an iterative process.
1:02:25So everybody kind of get that?
1:02:26You understand how you would apply this to your own types of problems in trying to do...
1:02:31...suitability modeling and site selection?
1:02:34So there's a couple of other ways to do this.
1:02:36One of the ways we added a few years ago is known as fuzzy overlay.
1:02:40So in the same way that she was talking about using the Reclassify tool and the Weighted Overlay tool...
1:02:45...there's another sort of parallel way to do that, which uses fuzzy reclassify and fuzzy overlay.
1:02:51And the fuzzy part of this is not about fuzzy boundaries, it's about fuzzy sets.
1:02:55So, fuzzy in this case is about set theory and how much something belongs to one group or another.
1:03:02The difference between fuzzy overlay and weighted overlay, in this case, is I don't have to...
1:03:04...calculating the accumulated flow across the elevation surface.
1:03:06...specify discrete class boundaries.
1:03:09I don't have to say that, you know, 10 to 20 is this value and 20 to 30 is another value.
1:03:15I can use continuous functions, okay, so I don't have to specify discrete boundaries for the classes.
1:03:21The other thing that's different about it is, I don't have to have everything be additive.
1:03:26In the case of weighted overlay, it's an additive overlay.
1:03:29It's this and this and this, all added together and then normalized.
1:03:33In fuzzy overlay, it can be this and this, or some criteria or another criteria.
1:03:39So you can mix "and"s and "or"s in the same model.
1:03:44We have a few sort of specialized domain-specific functions in Spatial Analyst, as well...
1:03:49...some hydrologic tools, which we presented yesterday, and these slides will be available.
1:03:55Tools for creating watersheds and stream networks...
1:03:59...that basically work by calculating the direction of flow across an elevation surface and...
1:04:08And then we use this for creating things like watersheds and stream networks and doing stream ordering...
1:04:13...and calculating flow length and things like this.
1:04:18We also have a few groundwater modeling tools for doing simple, two-dimensional advection...
1:04:22...dispersion modeling, so if you happen to have a head grid and a hydraulic convectivity...
1:04:30...and porosity [inaudible], then you can use these tools to calculate...
1:04:34...two-dimensional vector flow fields of flow magnitude and direction and drop particles on them...
1:04:39...and see where they go, and do Gaussian dispersion from those.
1:04:43Sorry if I'm going a little quick through that; we're running a little short on time.
1:04:46I've got something else I want Liz to show you.
1:04:48And we have some multivariate statistics tools.
1:04:50These are things that you might think of more classically as image processing tools...
1:04:55...so if you come from an image processing background, you may see things like principal component...
1:05:00...analysis and national lightweighted classifiers.
1:05:03The better way to show this is for Liz to show us the Image Classification toolbar, which allows us...
1:05:08...to do interactive training sample selection and interactive classification of multiband imagery.
1:05:16And we'll just show that one instead of talking about it.
1:05:21So, I have a satellite image here of the Lake Tahoe area and I want to do a supervised...
1:05:25...classification of that using training samples.
1:05:28To access the Image Classification toolbar, I go to Customize > Image Classification, and here's the toolbar.
1:05:35Here, you can see I have the imagery selected in the drop-down layer, and this is how I draw my training samples.
1:05:41So the first class I'm going to make represents water in this image.
1:05:45So you can go ahead and make some areas that represent water.
1:05:49If I open up the Training Sample Manager, I can then select those training samples and merge them into one class.
1:05:56I'm going to call this Water and, conveniently, it's already blue.
1:06:01The next class I'm going to make is for the clouds in the study area.
1:06:04I need to make sure that they remain clouds in my output classified raster.
1:06:09Again, I'm going to go ahead and select them and merge them and call them Cloud.
1:06:15Oh, and that's white; we're close enough.
1:06:16Okay, our next one is the rock, so the bedrock over here.
1:06:20Now if I was doing this for real, not just demonstration purposes, I'd want to zoom right in and...
1:06:25...make sure I'm getting really homogenous samples of the rock, but obviously this is how it's done in a demo.
1:06:32So I'm selecting some bedrock here.
1:06:35One thing you might want to do, though, is if you select these, then there's some diagnostic tools up here.
1:06:39The Histogram tool shows you the distribution of the cell that I've selected.
1:06:45And as you can see, because they're all bedrock, these histograms all overlap pretty well.
1:06:50There's also the scatterplot.
1:06:52This might look dispersed until you look at the axes, and actually these clouds all overlap quite well.
1:06:59And if I look at the matrix of the statistics, then that can also help determine whether or not...
1:07:04...I've done a good job of selecting my training samples.
1:07:08So I'm going to go ahead and merge these, call them Rock, and color them gray.
1:07:14The last class I'm going to create in the study area has to do with the vegetation, which is the red.
1:07:19Again, I encourage you to zoom in and be much more meticulous than I'm being.
1:07:23 But there we go, now we have some vegetation.
1:07:26I can merge these again.
1:07:27If I mistakenly had merged the rocks in with that class, you can always use the Split button...
1:07:33...to break it back apart and go back to what you had previously.
1:07:36Or if one of those training samples doesn't fit after you look at the statistics, you can split them...
1:07:41...delete them, create new ones, all that.
1:07:44So I'll merge them; I'm going to call it Veg for vegetation, and it's green.
1:07:49From the classification - oh, before I do that, let's make sure that we have really distinct classes.
1:07:54We don't want a lot of overlap; we want to be very clear with how we do our classification.
1:08:00Here I've selected all the classes; now when I do the histogram, it's across all those classes...
1:08:05...and you can see that they're pretty distinct.
1:08:07Likewise with my scatterplots; they're pretty separate.
1:08:10All my blues fall together - all the greens, the grays.
1:08:14Oh, the whites are up here. They're hard to see, right?
1:08:16They're also pretty distinct classes, so that's a good sign in terms of classification.
1:08:21So I go ahead and use the Interactive Supervised Classification tool, and it produces this classification.
1:08:28If in this Effects toolbar I select that Output Layer I just generated and use this Swipe tool...
1:08:34...I can swipe back and forth and see how well I did.
1:08:37Look, I'm not too happy right here, if you guys can see where those black arrows are.
1:08:41It seems like I have more clouds in my image than result in my classification.
1:08:45So what I can do is I can go back and I can keep adding; just draw a few more training samples until, you know what?
1:08:53These are also clouds, and I can merge those in, and now when I do the classification - let's make sure...
1:09:00...I select the right one - did I get any more of those clouds? I'm not sure.
1:09:06So you can go back and you can keep reiterating the process and continuing on and updating...
1:09:11...deleting, or moving, et cetera, until you get a classification in which you're happy with.
1:09:19Was that the fastest image classification you've ever seen? That's pretty awesome. Yeah.
1:09:26For all of us who used to do it the hard way in the past, before there was like actually a way...
1:09:31...to look at your image on a computer display, that's pretty shocking.
1:09:36So after we do a classification like that, we end up with these kind of speckly-looking things, right?
1:09:44And that really doesn't look much like good GIS data.
1:09:46So there's a collection of tools in Spatial Analyst that allow us to do data generalization and...
1:09:52...cleanup of raster data, and these are tools like Majority Filter, Expand, Shrink, Nibble, things like this.
1:10:01So without going into too much detail on those, if you just watch the little graphic in the corner...
1:10:04...we run, for example, Nibble on that and remove all of the individual or two- or three-cell areas and then...
1:10:11...run Majority Filter on that, you can see that looks much more like what you would think of as GIS data, right?
1:10:17Now the way that those tools work, they actually try to preserve the morphology of the data.
1:10:23It's actually replacing cell values with, for example, the majority of the neighbors sharing those boundaries, so.
1:10:34With that, one of the things, if you haven't seen this slide yet in one of the geoprocessing...
1:10:40...or Spatial Analyst presentations, there is a geoprocessing resource center now.
1:10:44When we talked about resource centers in ArcGIS Online on Monday.
1:10:49There is a community for analysis, and if you go there, there are a lot of good things there.
1:10:57Things like these presentations, some of the demos that you see, a lot of sample tools and sample scripts.
1:11:03So there's an education gallery where we have a lot of articles and videos and things like that...
1:11:09...as well as a tool gallery.
1:11:10Now the important thing about the tool gallery, it's not just the things that you see the little bitmap pictures of.
1:11:17There's a button here called More Gallery Posts.
1:11:20Click on More Gallery Posts; there's a lot more stuff there, okay?
1:11:24Things that we built, things that you guys built; there's lots of good things there.
1:11:28So, we just moved over to this new gallery approach at 10.1; we're still migrating over some of the...
1:11:34...old materials and indexing those, but expect to see that growing a lot more in the coming year.
1:11:42Please fill out a session evaluation.
1:11:44If you haven't been doing that, the session ID for this particular workshop is 811.
1:11:50We do actually read all of your session evaluations and all those comments that you write, and we do change our...
1:11:57...presentations and create new presentations and drop old ones based on the type of feedback that you guys give us.
1:12:03So, if you need to run to another session, it's two forty-five now, and we'll hang out and answer any questions you guys have.