GEOG 883
Remote Sensing Image Analysis and Applications

Remote Sensing Workflow Tutorials

PrintPrint

In your first assignment, you will execute a remote sensing workflow to map water from Landsat satellite imagery. The four videos below show the four key components of this assignment: 1) downloading and preprocessing imagery, 2) feature extraction, 3) data conversion and attribution, and 4) data sharing.

Video: Downloading and Preprocessing Landsat Data (9:51)

Click here for transcript of Landsat ArcGIS video.

PRESENTER: In this video, we'll show you the access, download, and work with Landsat imagery using ArcGIS Pro. We're starting out in GloVis, which is the USGS Global Visualization Viewer. GloVis is a wonderful online tool that allows you to search different USGS data sets, apply search criteria, and easily download these data so you can bring them into your GIS for further analysis

GloVis makes it really easy to access the data you want. You simply pan or zoom to the location, select the data sets, apply any search criteria you want, and then you can download that data directly from GloVis to your computer.

I'm interested in obtaining a Landsat scene over the Presidential Range, New Hampshire. So I'm going to use the Jump tool to enter the latitude and longitude coordinates for Mount Washington, which is New Hampshire's highest peak, and then just zoom in a little bit.

Over on the left, for Choose Your Data Sets, I'm going to turn on the radio button for Landsat 8 OLI. As soon as I do this, GloVis begins displaying some Landsat scenes. I can browse through those scenes, but this would take a long time. And there's an awful lot of scenes with a tremendous amount of cloud cover. To improve my search, I'm going to go over to the metadata filter and constrain my search to only those scenes with 0% to 20% cloud cover and those acquired in the months of June, July, and August.

Now I'm able to browse through the collection of Landsat 8 scenes that meet my search criteria. I have the ability to exclude certain scenes from my Browse by clicking on the Hide Scene button. This is useful, because I clearly have some Landsat scenes in this case that have extensive clouds over the Presidential Range. The whole scene may not be cloudy, but the area around the Presidential Range is, so I'm going to hide those scenes. If at any time I want to show those scenes again, I can go back to the Choose Your Data Sets and click on Clear Hidden Scenes.

For any scene, I can move over and click on the View Metadata button to obtain all the information about that scene. This would include everything from the percent cloud cover, to the acquisition date, to even specific sensor parameters that I may want to use for calibration. Once I've found the Landsat scene that I'm interested in, I simply move over and click on Download. GloVis will package up the Landsat scene.

I want to make sure that I click the Download button for the Level-1 GeoTIFF product, as this is the full Landsat 8 data set. Once the download is ready, I'll save the tar.gz file to my computer, where I can uncompress it and make use of the Landsat data.

You may need a specialized ZIP utility to uncompress the tar.gz file that GloVis uses to package up your Landsat data. I recommend 7-Zip. Here, I'm going to first uncompress the GZ file, then unpack the TAR file. Once the data are uncompressed, you'll see that each separate Landsat band is stored as its own individual GeoTIFF file.

In addition to the individual GeoTIFF band files, I also have some metadata in calibration files that may be useful to me. The metadata is stored in the underscore MTL file. Opening it up will allow me to read through the metadata and access key information, particularly things like the date the scene was acquired on, the path row, cloud cover, and calibration parameters.

Although it's certainly possible to add each individual band and work with it separately within ArcGIS, what I really want to do is combine all these bands together into a single multispectral composite image. To stack the bands that I'm interested in working with into a single multiband raster data file, I'm going to use the Composite Bands Geoprocessing tool.

The Composite Bands Geoprocessing tool will produce a new raster file containing all of the bands that I enter into the Composite Bands tool. My input rasters are going to be all the bands that I'm interested in. In my case, I'd like to load in bands 1 through 7. These are the 30 major multispectral bands that are commonly used in a multiband raster data stack.

Once I have these rasters selected, it's imperative that I pay attention to their order in the Composite Bands tool. The order they're listed in the Composite Bands tool determines their band order within the multiband raster output. In this case, you can see that band 1 is accidentally at the bottom. So I'm going to need to reorder the bands by moving band 1 up to the top so that it becomes layer 1 in the resulting output raster data set.

I'm going to save my composite raster as a GeoTIFF file by navigating to the directory that I want to store it and typing .TIF at the end of the file extension. To store it as imagined format, I would replace the .TIF with .IMG. Alternatively, I could save it inside a geodatabase.

Running the Composite Bands tool stacks all of these rasters into a new single composer band raster file. And in my case, I've stored it in GeoTIFF format. Once the Composer Bands tool is finished running, it's important to go into the properties of my new layer to confirm that the tool ran as expected.

By going into the Source tab and Raster Information, I can access the key source information about this data set. Specifically, I want to check the number of bands, which is 7. I see that the cell size is 30 by 30, which is the resolution of these particular Landsat 8 bands. The format is GeoTIFF, and the pixel depth is 16-bit, just as I expected.

The best way to adjust your symbology for imagery is to move over to the Appearance tab and go to Band Combinations. You can see that ArcGIS already has two defaults-- Natural Color and Color Infrared. Unfortunately, these band combinations are for aerial imagery, not for Landsat 8 imagery.

Under the Appearance tab, I'm going to go to Custom, and I'm going to create Custom Band Combination settings for some common Landsat 8 band combinations-- 3-2-1, 4-3-2, et cetera, until I have all the band combinations set. This will allow me to quickly and easily symbolize my Landsat 8 multiband composite image.

Now that I have my band combinations set, I can make some other changes to the appearance of my imagery. Activating DRA, or Dynamic Range Adjust, automatically modifies the contrast stretch within my field of view as I pan around the image. You may find that certain stretch types help you pick out particular features of interest.

But it's important to keep in mind that there may not be one single stretch type that's optimized for all of the features you're interested in on the landscape. Remote sensing is an art and a science. Determining the best symbology takes a bit of trial and error. To make it easier for me to compare the different band combinations, I'm going to copy and paste the raster layer in my table of contents, and adjust the band combination for each one of these layers, and then give each layer a meaningful name for quick reference.

I'm interested in doing some vegetation analysis around the Presidential Range, so I'm going to use the Locate tool to find Mount Washington. Typing in Mount Washington will pull up a list of similar names, and then we'll display the list of those locations that match on the map.

Now I'm going to compute NDVI, or the Normalized Difference in Vegetation Index. This is very easy to do using the raster function NDVI. So I'm opening up the Raster Functions tab, searching for NDVI, and clicking on it to launch the function. Any one of my layers can be entered as the input raster.

It's imperative that I understand my bands. The visible band for this Landsat 8 composite is band 4, and the near infrared band is band 5. Entering the bands incorrectly here will produce invalid NDVI output. I want my NDVI values to have a range from negative 1 to 1. So I'm going to check the box for Scientific Output. And then when I'm ready to execute the NDVI tool, I'm going to click Create New Layer.

The Raster Function NDVI creates a virtual layer pointing back to my original imagery, so it executes almost instantaneously. After renaming my layer in the table of contents, I'm going to head over to the Symbology tab. I'm going to make sure I'm not displaying any background values of zero, and then I'm going to adjust the stretch type until I come up with something that I think displays the NDVI values the best.

Finally, I'm going to tinker with the color scheme, choosing a ramp that goes from red, meaning no vegetation, to green, which means higher concentrations of vegetation. As I explore my NDVI layer, I can see that areas with high amounts of vegetation, such as the forested areas, have high NDVI values, displayed as green, and features that are devoid of vegetation, such as roads, and water bodies, and mountaintops, have much lower NDVI values, symbolized by the red color. Clicking on any individual pixel will display the NDVI value in the Identify Window.

One of the advantages of NDVI is it minimizes the effect that lighting has on my data. As we pan around the Presidential Range, we see that forested areas and shadowed valleys are considerably darker than forested areas that are in unshadowed regions. However, when we move over to the NDVI layer, we see that the effect of shadowing has been greatly reduced. This is because lighting has a similar effect on both the near infrared and red bands. And thus, when we apply NDVI, which is a ratio between those bands, the effect is minimized.

In this video, we showed you how to use GloVis to access the Landsat scene that you're interested in, create a multiband raster composite within ArcGIS, display the multispectral imagery using different band combinations, and then finally compute NDVI.

Credit: Penn State College of Earth and Mineral Sciences

Video: Unsupervised Classification (4:56)

Click here for transcript of Unsupervised Classification video.

PRESENTER: This video will demonstrate a workflow in which a pixel-based unsupervised classification is used to map land cover from Landsat imagery. Unsupervised pixel-based classification approaches work by partitioning the input imagery into a set of output classes, based only on the spectral values. Because they're only making use of the spectral information, they are limited in their utility. They cannot make use of spatial information such as size, shape, or even texture.

One can use the Iso Cluster unsupervised classification geoprocessing tool to perform an unsupervised classification. But a more streamlined approach is to use the Image Classification Wizard. To do this, first select the image data set in the table of contents that you want to classify. And then from the Imagery tab, choose Classification Wizard.

The first step is to configure the classifier. Under Classification Method I'm choosing Unsupervised, and for the classification type I'm going to use a pixel-based. Now I already have an existing classification schema. So I'm going to load that here. If you don't have a classification schema, you can always select default and modify it later.

Once you've finished entering the configuration parameters, you can click Next. This will move to the next step in the Image Classification Wizard, which is called Train. For an unsupervised classification, the train phase is where you'll enter some of the key parameters. The most common parameter you'll want to modify is the maximum number of classes. This is the total maximum number of output classes that you can have in the resulting classification. You want this to be higher than the number of classes you need. But note that this is the maximum, and it may not be achieved.

I'm going to adjust the maximum number of classes to 10, and leave all the other defaults in place. Clicking Run will produce the initial classification. You'll want to review your output here, and you may want to go back to the previous stage and make some adjustment to the input parameters, such as adjusting the number of classes.

If you're happy with the output, simply click Run in the classified window, to move to the next phase. This will produce another raster data set with the same number of classes. And in the Classify window, you can now click Next to transition to the next phase.

In the assign class phase, I'm going to assign a land cover class to each one of the categories in my unsupervised classification. It's important to note that although classes may be spatially dissimilar, and then have unique values in the unsupervised classification, they could belong to the same land cover class.

Prior to assigning classes to the unsupervised classification, you may want to make some modifications to your classification schema. For example, you can right-click, and edit the properties of the schema, giving it a new name, and entering descriptive information.

You can make modifications to individual classes, adjusting their name, color, and default numerical value. No two classes should have the same numerical value. You can also add or remove classes from the existing schema. And then finally, if you're happy with your classification schema and are considering reusing it in the future, be sure to save it using one of the Save options.

Assigning the appropriate land cover class to each one of your unsupervised categories may take some time. You'll want to compare your unsupervised classification to at least your input imagery, and perhaps even to some other reference data that you may have access to. To associate each land cover class with its corresponding unsupervised category, you'll want to select the class in the assign class dialog, click on the Assign Class button, and then click on a pixel belonging to that class in the unsupervised classification.

The unsupervised class assignment information is updated in both the assign class dialog and in the table of contents. Clicking on a class in the assign class table, will highlight that particular class in the unclassified raster, helping you understand what pixels belong to that particular category. As was mentioned earlier, one or more unsupervised categories could have the same land cover class.

You'll want to continue this process of associating land cover classes with unsupervised categories, until all of your unsupervised categories have an associated land cover class. As I'm working with Landsat data which is at an angle, I even have a background category that I need to consider for my classification. Once you've completed the class association process, you can click Next to move to the reclassifier stage.

For an unsupervised classification, it's unlikely that you'll need to apply any reclassification routines. So you can click Run to finalize your classification. This produces the final classified raster, which is stored inside your project geo database. You have the option of removing any intermediate products from your table of contents to clean up your ArcGIS project.

Unsupervised classification is an easy and straightforward way to getting at feature extraction. However, it should generally only be applied to multispectral imagery, and in those cases where the land cover classes can be clearly separated by only spectral information.

Credit: Penn State College of Earth and Mineral Sciences

Video: Raster to Vector Conversion (4:36)

Click here for transcript of Raster to Vector Coversion video.

PRESENTER: This video will walk you through workflow in which raster land cover data are converted to vector for further attribution and analysis.

The raster land cover data set contains four categories and was produced by performing an unsupervised classification on landset eight multispectral imagery. Raster is a fine format for storing land cover data. But in this particular case, I want to know the area and attribute the extent of this particular lake. To reduce the amount of data I have to convert to vector, I'm first going to clip the raster down to my current view extent.

Using the raster Clip function, I can select my input raster, leave the clipping geometry set to the active view extent, and then run the process to clip the land cover data set down to the current zoom level. Zooming back out, we can see that the result of the raster Clip function is a new layer that is clipped down to my previous zoom extent. Because this is a raster function, I didn't actually write out any raster file, but created a virtual layer based on the clipping extent.

I can now use this clipped version of the land cover data set to convert to polygon format. From the analysis menu, I'm going to choose Tools to launch the geoprocessing window, navigate to Toolboxes, and, under Conversion Tools, go to the From Raster toolset. Obviously, I'm going to use the Raster to Polygon tool, which allow me to convert raster data to polygon format.

The input raster is the clipped land cover. The field is the numerical value associated with each pixel, and the output polygon features are going to be written to my project geodatabase. I'm leaving Simplify polygons checked, which will result in smoother vector features, and then clicking Run will execute the geoprocessing task. The resulting output is a vector polygon representation of the clipped land cover data set.

The original land cover raster data set had four possible values. These were stored in the Value column in the attribute table. This was the field used for attribution when we did the raster-to-polygon conversion. When we opened up the attribute field for the vectorized version of the land cover data set, we see a field called gridcode. Gridcode stores the land cover code and corresponds to the value field in the original raster land cover attribute table.

To symbolize the vector version of land cover data, we'll navigate to the Appearance tab and open up the symbology. We're going to choose to symbolize the vector data based on unique values. And the field we want to do that is gridcode. Recall this is the unique land cover code corresponding to the value field in the original land cover raster. If we want to take the time, we can go in and adjust the labels so that they display the actual class names, as opposed to just a numerical value.

In this example, I'm only interested in the vector polygon representation of a single lake so that I can know the area of the lake, and also attribute it with additional information. To select this individual polygon, I'm going to choose the Selection tool from the main menu, select the lake, and then convert the selection into a standalone layer. I haven't created a new feature class. The layer I've generated is simply a virtual subset that points back to the original feature class.

Opening the attribute table for this new layer, we see that it contains all the same attributes as the original feature class. The only difference is that it has a single record representing the individual lake that I selected and created the layer from. I'd like to make some modifications to the attribute table and, thus, it makes sense for me to export this layer to brand new vector feature class. To do this, I'll right-click on the layer in the Table of Contents, choose Data, and Export Features.

This will write the polygon feature out to a brand new feature class within my geodatabase. To avoid me making modifications to the wrong layer, I'm going to remove the original vector layers from my Table of Contents. I'm now going to make some modifications to the attribute table of my lake polygon feature class. I'm going to start by removing those attribute fields which are no longer relevant to the data set. Now I add a new field where I can store the name in the lake, then I'm going to populate that attribute with the lake name.

When I'm finished with my edits, I'll be sure to go to the Edit menu and save my edits to finalize my vector feature class. Because my vector data is stored within the geodatabase, the area for the polygon-- in this case, the lake area-- is stored within the Shape Area field. The units correspond to the coordinate system units for the polygon feature class squared-- in this case, meters squared. This video reviewed the steps necessary to take the output from an unsupervised classification and convert it to vector format for additional attribution analysis.

Credit: Penn State College of Earth and Mineral Sciences

Video: Creating a Web App (8:06)

Click for transcript of Creating a Web App video.

[MUSIC PLAYING] JARLATH O'NEIL-DUNNE: This video will show you the steps to create an ArcGIS online web app by publishing a map using ArcGIS Pro.

There are four steps to creating a web app. First, you're going to configure your map using ArcGIS Pro. You're then going to publish the map to ArcGIS Online. You're going to create the web app and then finally you're going to share your web app.

Prior to getting started, you're going to want to head over to ArcGIS Online and make sure you have the ability to both login and publish data. You may have to reach out to your organization's Esri administrator in order to get those permissions. Prior to publishing your map, you're going to want to devote the necessary time to set up a decent ArcGIS project. You can see here that I've got two layers, one displaying neighborhood tree canopy, and one displaying neighborhood surface temperature, both for the city of Pittsburgh. I've taken the time to give the layers meaningful names and also to configure the legend so that the data can be easily understood.

You can adjust the display of your data along with the legend information over in the Symbology tab. For this percent neighborhood tree canopy data, I'm using a meaningful color scheme. You'll also notice that I've adjusted the labels, so they actually show the percent values. Under the advanced tab, you can get in and adjust those labels, eliminating unnecessary decimal places and choosing to display percent data with the percent symbol.

I've also taken the time to configure my fields so that when I click on an individual neighborhood, only the requisite information comes up. You can see that each one of these fields has been carefully formatted to eliminate excess decimal places and show only the data I want to get across to the end user. This complete neighborhoods layer has an awful lot of information attached to it and I don't need to display all of this. So by going up to the Data tab and selecting the fields option, I can configure my attributes to only display what I want the end user to see. Checking the visible button controls what fields are displayed. You want to make sure that all the fields you're choosing to display have a meaningful alias and that if they are numeric data, that you've formatted them correctly.

Once you've looked everything over, don't forget to save your project. Now you're ready to publish this map and get it up on ArcGIS Online. The first step is to click on the Share tab. Under the options for Share As, click on the button for Web Map. This opens the Share As Web Map interface. You're going to fill out some key information required for ArcGIS Online and you're also going to configure your Web Map settings. The main field cannot have any spaces or unusual characters. Your summary should be a clear, concise description of your Web Map. Add any keywords in the Tags section that you think would help people find your map.

Give some thought as to who you want to grant permission to view your Web Map. Is it just your organization, everyone in the world, or a select group of people for a group that you've created in ArcGIS Online. Chances are the default configuration setting is just fine, but if you're using your map for other purposes such as data editing, you may want to change the configuration setting.

Prior to sharing your map, you'll want to click on the Analyze button and make sure that the map's been configured correctly. If there are any errors that pop up here, those ones with a red x, you'll need to address them. Warning signs you can ignore, but you might want to consider fixing. You're now ready to publish the map. You can confirm your configuration settings, check any messages. Then when that's set, click the Share button.

Once the share process is complete, you're set to move over to ArcGIS Online. Sharing a Web Map creates three new pieces of content within ArcGIS. We really care about the first one, which is the actual Web Map. So let's go in and explore the Web Map. This is a good place to check out and make sure that your settings, configurations, and everything in that Web Map turned out as you expected it. So opening in that Web Map Viewer, turning on and off the layers, and checking your legends and symbology is a good thing to do prior to going ahead and using this Web Map within a Web App.

Once you've confirmed everything looks good, navigate back the Content section of ArcGIS Online. We're now going to use that Web Map as the foundation for building our Web App. So from the Create menu, scroll down to the Apps section and choose Using the Web App Builder. So we're going to create a new Web App here. You're going to want to give it a title, tags, and some summary information, just like you did when you created your map in ArcGIS Pro.

This will put you in the Web App Designer interface. First, you want to decide on a theme. Once you've selected a theme, you can make some modifications to its properties, like the colors. Don't worry if you don't like the theme. You can always play around and preview your Web App and come back and adjust the theme later.

Moving over to the Map tab, this is where you're going to choose the Map, the one that you just published ArcGIS Online, that's going to participate in the Web App. Go ahead and select your Map, but keep in mind that you cannot make any modifications to the Map within the Web App Designer interface. You've got to go back to the Web Map in order to make any changes. Given that we just made some modifications to the web app, it's a good idea to save it.

We can further customize our Web App through the use of widgets. Each Web App by default has certain widgets included. We can choose to remove those widgets, add additional widgets, or reorder them. For this Web App, I'm going to add a Base Map widget and a Swipe widget. I'm now going to reorganize the widgets to change the order that they appear in the Web App. Under the attributes tab, you can make some other minor changes like adjusting the header text that will appear in the top header. Down below, you can click on the previews button. This is a really nice feature with the Web App builder because it allows you to see how your Web App will appear on various devices.

Once you're happy with the previews, the final check should be to go ahead and launch your Web App to see how it works within a browser. This is where you'll want to check out all your widgets, preview the legends, and just make sure everything, in general, is working and appears as you wish it would. Remember, you can't make any changes to a lot of the key components here. You would have to go back to the actual Web Map to do that or possibly republish your data. This is why is important to get it right early on.

We're nearly done. Our final step is to get back into the ArcGIS Online content, select our Web App, and go into the Share settings. Remember we set the share settings when we published our map, but this is the app and we have to set the app settings independently. Make sure that if you want everyone to see this, those not in your organization, that you check the box for everything. Once you have that set, scroll down to the bottom, copy the link and you're ready to share your app. [MUSIC PLAYING]

Credit: Penn State College of Earth and Mineral Sciences

Other Examples

Below is an example of a water extraction workflow using ArcGIS 10.x

Video: Introduction to Remote Sensing Workflows (10:43)

Click here for transcript of Introduction to Remote Sensing Workflows video.

Jarlath O'Neil-Dunne: This video will provide you with a basic introduction to remote sensing workflows using ArcGIS. The example we'll use is mapping lakes in Kenya using Landsat satellite imagery.

The desired end state of our project is to produce updated lake boundaries for a few lakes that lie in the Great Rift Valley region in Kenya.

We’ll begin by outlining our remote sensing workflow. We've already defined the task, which is to create Lake polygon boundaries. Next, we need to determine our data needs. Data needs, of course, are a balance between what you'd like to have, what's available, and what you can afford. In this particular case, we're going to use freely available Landsat satellite imagery. We're going obtain that data from the USGS Glovis site. We're going to assemble the separate Landsat bands into a single composite image, and overlay those bands in ArcGIS to take a look at our data to better understand it. We’re then going to carry out an unsupervised classification, using a nicer data classification algorithm. And then finally, we're going to do a quick evaluation of our data to understand its strengths and weaknesses.

There are two reasons for selecting Landsat imagery for this project. First, its acquired at regular intervals. Second, it's freely available from the USGS.

We need to obtain the appropriate Landsat satellite scene, but before we do that, let's get oriented in ArcGIS. We're going to load in an ArcGIS base map. In this case, the imagery with labels base map and use it to zoom into our area of interest. Here we are in the Great Rift Valley and you can see there are two major lakes here that are labeled, Lake Nakuru and then Lake Naivasha.

Landsat scenes are organized by their path and row numbers. So let's head over to our search engine to see if we can locate a Landsat path row shapefile. Sure enough, we can find one from the USGS. Now land areas are typically imaged when the satellite is in its descending orbit. So we can get the WRS2 (the World Reference System 2) descending shapefile. This is for the most recent Landsat satellites. We're gonna save that to our local drive and unzip it so we can bring it into ArcGIS.

Moving back into ArcMap, we've loaded in the WRS2 shapefile. Let's use the Identify tool to click on the Landsat scene corresponding to our area of interest, and then the identify tool you'll notice that we want the Landsat scene with path 169, row 60.

Landsat imagery can be obtained freely from the USGS Global Visualization Viewer. We’ll first want to specify the collection. We're going to go for Landsat 8, which was launched in February of 2013. Once we've specified the sensor, we're going to go and enter the path row. Remember this was 169 and 60. Once we've entered the path row, we’re to click on the Go button and Glovis is going to transfer us over to that portion of the globe. Scrolling down, you can see that we can view the previous and next scene. This will scroll through all the Landsat scenes available for that particular area. Once we've found a scene of interest, we can click on the Go button and add it to our cart. Once we've added it to our cart, we can download the data by sending it over to Earth Explorer. You're going to need to log in with your EarthExplorer ID, but once you've done that you can go ahead and click on the little download icon and you'll be able to download the full Geo TIFF product.

Once we've uncompressed the file we downloaded, let's head over to ArcCatalog. Here we see that each of the eleven Landsat bands is a separate raster file. In addition, we have a QA or quality assurance band and a metadata text file. Now we're not going to need all of these Landsat bands for our work. If we look at the Landsat spectral coverage for each band from the USGS website, it looks like bands 2 through 7 are probably going to be optimal for our work.

In order to more effectively work with our Landsat image, we're going to want to combine bands 2 through 7 into a single multispectral image raster file. To do this, we're going to use the Composite Bands tool. We're going to select bands 2 through 7 and drag them into the Composite Bands tool. The Composite Bands tool is going to produce a single output file that has all those bands. In this example, we're saving it with the dot TIF extension, meaning it's going to be a Geo TIFF file. To check our progress from the geoprocessing menu, we can choose results. And once the process is complete, we can preview the results in our catalog.

Now let's head over to ArcMap to do some data exploration. The first thing that we're going to do is switch up the ban combinations. We're going to create a color infrared composite by assigning bands 4, 3 & 2 to the red, green and blue color guns respectively. Band 4 corresponds to the near-infrared band, and this color infrared composite will really help distinguish water because water absorbs practically all named near-infrared energy. We can also make use of some of ArcGIS image analysis tools. By selecting the image in the image analysis window, we can activate the DRA or Dynamic Range Adjust, and also play around with the contrast and brightness sliders. Adjusting the digital imagery will help us identify certain features of interest.

All the work we've done up until this point in time is leading us to the data analysis phase, where we're going to use an unsupervised isodata classification in an attempt to map water. The isodata classification simply takes our multi-spectral image and groups it into a set number of classes, based on the digital values of the pixels. We're going to use 20 classes, in this case, and give the output a new file name. So we can expect a new raster file in which we have 20 classes, based on the similarity of the spectral values of the pixels. When we look at the output we see that we only have 13 classes. This means the algorithm could only identify 13 unique clusters of data.

In examining the results of our isodata classification, it looks like class four, the bright pink class, best corresponds to water and that all other classes are not of interest to us. To confirm this, we can double-click on our isodata classification layer, to access the layer symbology properties. Under the Symbology tab, we can remove all those classes except class 4. This doesn't remove those values from the raster, it simply hides them for display purposes.

Now we're going to want to create a vector layer that contains only those pixels corresponding to class 4. This is going to be a two-step process. In the first step, we're going to create a new raster that only contains those pixels with class 4. All other pixels are going to be no data values. We're going to do this using the Reclassify tool. Within the reclassify tool, we're going to load in our raster data, turn all pixels with a value of 4 to 1, and click on the checkbox that says change missing values to no data. We're going to save this as a new geo TIFF file, and then run the reclassify tool. Our new raster layer only contains pixels with ones and no data. However, further exploration of this data set yields some problems. You can see that we've got a lot of shadows that fell into our water class.

We're going to deal with these false positives by converting the raster data to a vector layer and then querying by size. Let’s create a new file geodatabase to store the vector output in. Then we're going to use our conversion tools to convert the raster data to a polygon. By storing the vector data in our geodatabase, we'll make sure that the area field, the shape area filled more specifically, is populated automatically.

The vectorized version of our classification really illustrates the problem that we have with false positives. As you can see, we have all these small, quote-unquote, water polygons, that are actually shadows. Because we stored our vector output in a shapefile, it contains the shape area field. The shape area field contains the area of each polygon and map units. Because our Landsat satellite was in UTM, this means it's going to be square meters. It looks like somewhere around 20 million square meters is a good cutoff.

Using Select by Attributes, we can select only those polygons that exceed 20 million square meters. Our query is going to be a shape area is greater than 20 million. Once we click OK, we'll have selected only those polygons that met that criteria. Right-clicking on the layer, we can go to selection and choose create layer from selected features. To make this layer permanent, we’re going to right click on it and go to Data, and export our data to a new feature class. This new feature class will only contain those polygons that exceeded 20 million square meters.

We're going to finish up by doing a very quick evaluation of our classification. Flipping back and forth between our Lake polygons and the imagery, you can see that we have some issues, both with respect to the lake edges and with some clouds and haze that were in the middle of the lakes that seem to throw off the classification.

In this video, we walked you through the entire remote sensing workflow everything from defining the task and obtaining the data to extracting features and doing a course evaluation

Credit: Penn State College of Earth and Mineral Sciences