Analyzing Recent Landscape Change: Salt Creek Mudslide (Mesa County, CO, May 25, 2014)
Links for an early news item about the slide and a set of aerial photos taken shortly after the slide:
Note to Instructor: This exercise is developed for introductory GIS students who have recently learned the basic processes of creating geodatabases, new feature classes, heads up digitizing, and vector spatial analysis (including clipping, intersection, buffering etc.)— and to provide a context for practicing these skills while also learning the basics of using Landsat 8 imagery in order to analyze a recent event. You will likely want to delete some instructions and/or explanatory notes your students will not need, or add others that they will, and modify references to internal network pathways and server names etc.
Further, this version of the exercise is intended for completion in one two-hour class period and therefore requires the instructor to obtain,compile, and preprocess much of the data (listed below) in advance—mostly to avoid lost class time due to file copying and processing. If a two-class version of the exercise is desired, it can easily be accomplished by devoting the first session to:
1) Searching for and downloading the Landsat images and the Mesa County vector GIS data (guided by the table on the first page of the exercise);
2) “Unzipping” each set of Landsat files;
3) Moving Part A of the exercise to Day 1and modifying it to instruct students to create composite band rasters for each of the three Landsat scenes and then reduce the size of each by clipping (with the Extract by Mask tool) to the Mesa County boundary, or the Plateau Valley School District boundary, or the HUC subbasin unit boundary (as suggested in this version of the exercise), and then beginning Day 2 at Part B (and making slight edits to reflect whatever file naming scheme was developed in the revised Part A).
As written, the exercise employs a June 2013 image that has already been compiled into a composite raster (7 bands) and clipped to the regional study area, a June 2014 image in which 6 of the bands have been clipped to that same study area but with Band 5 still at ‘full size’ so that students can see what the data originally looked like, thengo through the process of extracting a raster by mask, and then create the composite band image. It also employs an October 2014 image that has already been compiled into a compositeraster and clipped. Bands 8-11 have not been included in the exercise (primarily to reduce file size and complexity of the instructions); Band 1 is included specifically, and only, to keep ArcMap from renumbering Band 2 as Band 1 during the composite band process, and then having to explain that/and deal with resulting errorsduring the exercise. Since the images are only used as a basemap for digitizing, and to keep the exercise feasible within the allotted time, no radiometric corrections were done…With an additional class period, students could probably perform all of these tasks—though much of that time would be spent waiting for processing to complete.
The vector data used in this exercise were obtained (as .shp files) from the sources noted below. Note that the Mesa County GIS department has created a lot of its data as countywide dataset using UTM Zone 12 coordinates—but the country is actually in two different UTM zones (12 and 13). Consequently, the eastern parts of the county (where this analysis focuses)thus have easting values so high that the ArcGIS recognizes they should be designated as Zone 13. This appears to cause problems when attempting to reproject the data into WGS 84 to match the satellite imagery, as I would normally do before providing data to students. So be aware that the data used in the exercise do NOT all have the same spatial metadata and will generate the standard ArcMap warnings about combining multiple spatial references.
Developed 2015 by the Integrated Geospatial Education and Technology Training (iGETT) project, with funding from the National Science Foundation (DUE-1205069) to the National Council for Geographic Education. Opinions expressed are those of the author and are not endorsed by NSF. Available for educational use only. See for additional remote sensing exercises and other teaching materials educational use only. See for additional remote sensing exercises and other teaching materials.
STUDENT HANDOUT
Overview
One important type of data that we now have access to are satellite images—available for some regions since the early 1970s and for most since the early 1980s. Formerly expensive and hard to access, the Landsat images are now available for free and on-line—a generous and unusually forward-thinking decision and use of your taxes. These images, or scenes, enable us to compare ‘before/after’ a particular event, or to study decades-long processes of change in the same area. Further, the frequency with which these scenes are obtained (every 16-18 days), and the large area captured in each scene, allow us to ‘revisit’ the same area much more frequently than do aerial photographs—even very shortly after major events--and often by viewing a single image rather than hundreds of adjacent air photos.
In this exercise you’ll learn how to use a satellite image within ArcMap to see evidence of recent landscape changes, in this case a recent geomorphological event which became a disaster when considerable private property was destroyed and 3 men killed—the Salt Creek Mudslide near Collbran, CO on May 25, 2014.
After opening and processing the images for use in ArcMap, you’ll apply your recently acquired skills to digitize the extent of the slide, and the lake that has formed at the top of the slide (and presents considerable risk for future problems), and then use vector analysis to determine the amount of private property and stretches of road that have been damaged or destroyed by the slide.
Data and Sources
The following data were obtainedin advance from the sources indicated and, to simplify your work, compiled into the saltcreek.mdb geodatabase. which you will find in the saltcreek_mudslide folder in the \\PickUp folder:
Original name: mesabndy.shp,roads.shp / Mesa County boundary
Mesa County roads / Mesa County (CO) GIS Dept.
Original name: parcel.shp
Name in geodatabase: pvSD_parcels / Mesa County parcels
clipped version of original / Mesa County (CO) GIS Dept.
Original name: schldist.shp
Name in ex: : pvSD50 / Mesa County School Districts
(PV50 district only) / Mesa County (CO) GDept.
Original name: GAS_PIPE.shp
Name in ex: gas_pipelines / Mesa County Gas Pipelines / Mesa County (CO) GIS Dept.
Original name: nforest.shp
Name in ex: same / Grand Mesa Uncompaghre Gunnison National Forest / Mesa County (CO) GIS Dept.
nhd24kst_l_extract
Name in ex: streams / (extracted from HUC 5 with HUC 51201, 51202, 51203, 51204 boundary polygon) / National Hydrography Dataset (NHD)
HUC51201to04 / Boundaries of HUC 51201, 51202, 51203, 51204 / National Hydrography Dataset (NHD)
Landsat 8 Images
LC80350332013171LGN00.tar.gz / June 20 , 2013 (Path 35/Row 33) /
LC80350332014158LGN00.tar.gz / June 7, 2014 (Path 35/Row 33) /
LC80350332014302LGN00.tar.gz / Oct. 29, 2014 (Path 35/Row 33) /
Note: before satellite images downloaded from the USGS can be used in ArcMap, they have to be ‘unzipped’ (twice—which has already been done for you), and then combined into a format ArcMap can display correctly (this has been done for 2 of the images to save time, but not the third). To reduce file size, the images have then been clipped to a regional study area boundary—except for one of the files that make up the June 7, 2014 image which you will clip yourself before you then combine it with the others into a “composite band image”.
Part A.Preparing the satellite image files for use in ArcMap
1.Copy the \\saltcreek_ mudslidefolder from the Pick Up folder to your \\giswork folder. Read through the instructions for Part A while you wait for the copying to finish.
2.Start a new, blank map and set the Workspace to your \\saltcreek_mudslide directory. Note that this directory contains a saltcreek geodatabase, a folder called composite_images, and another called June 2014_3533_scenes.
3.Add thecompjune2013 image and the compoct2014satellite images to your project from the composite images folder. Arrange them so the 2013 image is on top, and note the shape of these images (they’ve been clipped to the PV50 school district boundary), then turn them off until Part B.
Note:these are the two scenes for which a lot of processing has already been done in order to reduce file size/copying and processing time. You’ll now add the other images that have not been so completely processed already and learn how to do two of those processes.
4. Now, from the June 2014_3533_scenes folder, add the whole set of images with the code LC80350332014158LGN00 (yes, one file name is a little different than the others). Highlight all 7 of them (that is, the files for Bands 1-7) and add them to the map.
Before accepting the prompt to make pyramids, click the “Accept my choice and don’t show this again” option (it will then create all the pyramids at once instead of prompting you 6 more times). This will take a few minutes…. So keep reading while you wait!
Note the numbers 035 and 033 in the file names—these refer to Path 35, Row 33 in the World Reference System. These numbers are followed by the year, followed by 3 digits, which refer to the day of the year, and endwith a band number abbreviation B1, B2, B7 etc. (Note: earlier Landsat images use the numbers B10, B20, B70 in the file names).
Review. Right now, your map consists of 7 different images (plus the two you added earlier and then turned off), all of exactly the same place, each with a name that is quite similar to all the others, each one created from a different wavelength ‘band’, all displayed in gray scale—but you can only see one at a time, and not very clearly at that! You’ll have noted that 6 of the bands are clipped to the same irregular shape of something way up in the northeast corner of the other, much larger, parallelogram-shaped image. The large, B5 image is a standard, full scene Landsat image.
But one of the major features of satellite imagery that we want to take advantage of is the fact that we can use software to combine several of these different views into a single image by assigning a different color to each wavelength, and then we can change which combinations ofwavelength bands we’re looking at—including views that are created from invisible radiation and which may reveal aspects of the landscape that are not normally apparent!
Table 1belowsummarizes the wavelength band numbering system used by Landsat 8 (earlier satellites used different numbering) and the wavelength range that corresponds to each band. In a minute you’re going to see that ArcMap renumbers the ‘bands’ according to whatever order they’re in, and if you’re not careful you can lose track of what the band #s really mean….
Table 1. Landsat 8 Bands
Band # / Wavelength range (µm) / Color/name/intended purpose1 / 0.435 - 0.451 / blue (coastal margin)
2 / 0.452 – 0.512 / blue
3 / 0.533 – 0.590 / green
4 / 0.636 – 0.673 / red
5 / 0.851 – 0.879 / NIR
6 / 1.566 – 1.651 / SWIR-1
7 / 2.107 – 2.294 / SWIR-2
8 / 0.503 – 0.676 / Panchromatic (15 m)
9 / 1.363 – 1.384 / Cirrus cloud detection
10 / 10.60 – 11.19 / Thermal IR-1**
11 / 11.50 – 12.51 / Thermal IR-2**
**not currently functioning/calibrated correctly
Note:for accurate, automatedanalysis of satellite images, and especially to compare any landscape in one year to another, it is necessary to first ‘correct’ the data by factoring in the angle of the sun on the day of the year the image was gathered. BUT, since we’re only going to use the images as a base map from which to digitize, and we can do this just as well with uncalibrated data, we’re not going to take time to perform the radiometric correction.
5.But before we can combine the B5 image with the others, and then experiment with different band combinations, we need to clip it to the study area boundary that was used to reduce the size of the other images.
a. First, we need to turn on a set of tools you’ve never used before: the Spatial Analyst extension. Select Customize-Extensions, and then turn on the Spatial Analyst extension.
(Note: the reason for doing this is simply to allow the next tool to work instead of giving you cryptic error messages.)
b. Now, from ArcToolbox, expand the Spatial Analyst Tools, and select “Extract by Mask”
c. Set up the tool to clip the LC80350332014158LGN00_B5.tif image with the school district boundary (pvSD50). Save the resulting file into your June 2014_3533_scenes folder, and name it clip_LC80350332014158LGN00_B5.tif
[Note to instructor: or another name that matches the scheme used for the clipped B1-4, 6 images].
d. Click OK to run the process. When it finishes, remove the original LC80350332014158LGN00_B5.tif image from the TOC, and move the new, clipped version into its place (between B4 and B6).
6.Okay, make sure that these LC80350332014158LGN00 images are in sequential order from Band 1 at the top, to Band 7 at the bottom. IF YOU DON”T DO THIS, THE NEXT STEP WIL SCRAMBLE THE IMAGE BANDS in a way that will be hard to make sense of later…
7. Now, Create a “composite band” raster(this 3 min video demonstrates the process: )
a. From the “Windows” menu, select Image
Analysis.
b. In the Image Analysis window, highlight the Band 1-7 files for the LC80350332014158LGN00 scene (if you notice they’re out of order, go back and reorder them!).
c. and click on the “Composite Bands” icon.
(Note: Alternatively, you can access the Composite Bands tool directly from ArcToolbox-Data Management-Raster tools as shown to the right. This method will save you one step but it may be harder to understand what you’re doing the first time…..)
8.You’ll now see a new file called Composite_LC8xxx158xxxat the top of your TOC, and you’ll notice that the image is now in color though not yet particularly easy to make sense of. BUT, it’s only a temporary image as you may have noted when you first clicked on the composite bands button
9. So, let’s make this Composite_LC8xxx158xxxfile permanent by exporting it…
a. Yep, with the old right click, Data-Export Data method.
b. You’ll see an “Export Raster Data” window open, with a lot of options you could change, but don’t want to! Leave all the settings as they are and focus on where you’re saving to, and what to name the file.
c. You have the option of exporting this image INTO your geodatabase, or not. Since we can, and it simplifies some decisions, I suggest doing so. So, specify your saltcreek.mdb geodatabase as the Location, and notice that the file format is automatically selected for you.
NOTE: If you prefer to save the image in the compositeimages folder with the others (a reasonable choice!) instead of inside the geodatabase, you must then specify TIFF as the file format, and when prompted to promote the pixel depth to include the no data values, decline to do so!
d. Give the file a name like compjune2014, and click Save.
e. Wait patientlyfor this process to complete (this is a very complex, memory intensive operation!)—and read ahead in the instructions while you wait.
10. Okay, once the new composite image has been added,remove all of the original clip_LC80350332014158LGN00.TIF scenes, and the temporaryComposite_LC8xxx158xxxfrom your TOC.
You may want to create a Group Layer called Landsat 8 35-33 Scenes, or something similar, and move all three of the composite images into that group layer.
You should be left then with just the 3 Composite scenes, each of which now appears to be in color, and your TOC should look quite a bit like this (though only two scenes are shown here, and the display names have been changed to better keep track of the date of each image).
Part B. Viewing the satellite images in a way that takes advantage of their multispectral nature……
Overview.A few minutes ago you combined 7 different images into one—but in a way that kept track of the data in each of the 7 different files. ArcMap has now assigned the three primary colors each to a different radiation band, but by default to bands 1, 2, 3 respectively. Unfortunately, this makes no sense at all—since bands 1 and 2 consist only of blue light but are rendered in red and green, and band 3 is really green light but is shown as blue—so we need to change the way the data is being symbolized!
Now you’ll experiment with different combinations of wavelengths to see which is most helpful for distinguishing the evidence of the mudslide (oh yeah, that’s what we’re actually going to look at!)
1. Change the symbology of the June 20, 2013 image to a more natural representation
a. Open the layer properties, and open the Symbology tab as you are used to doing.