ES 775 Lab 4

RESAMPLING AND CHANGE DETECTION

Devils Lake, North Dakota

The Landsat system has now provided more than 40 years of data. These datasets represent the largest, longest, and most comprehensive body of information ever assembled for documenting land-surface conditions and environmental changes. Accessing older Landsat data from the 1970s may be difficult, however, because of the awkward data storage format (X-format) and variable quality of the archived data. In spite of these complications, it is often essential to use older data for studies of environmental change.

This exercise is based on the Devils Lake, North Dakota region—see city of Devils Lake. Devils Lake occupies a group of large, natural basins that were excavated during late Pleistocene glaciation, about 12,000 to 14,000 years ago. Individual basins of the lake vary considerably in their surface area, depth, and water quality. Two Landsat MSS datasets and one Landsat TM dataset are provided.

Date Bands File
14 May 1973 MSS 1, 2, 4 D73-1.*, etc.
19 June 1988 MSS 1, 2, 4 D88-1.*, etc.
8 Sept. 2000
TM 2, 3, 4
D00-2.*, etc.

Because of its topographic and geologic situation, Devils Lake is subject to large changes in its elevation, area, capacity, chemistry, and biomass. These changes take place mainly in response to climatic fluctuations. Recent dramatic rises in lake elevation have prompted much concern from local residents. Lake elevations above 1440 feet have led to significant flooding of adjacent cities, parks, roads, farms, sewage treatment plants, a military base, an Indian reservation, and other human structures. This elevation was surpassed during spring flooding in 1997. Devils Lake reached 1446 feet elevation during the summer of 1999, at which point the lake overflows and drains into Stump Lake. This level has been maintained for the past decade. For more information about the hydrology of Devils Lake, visit the U.S. Geological Survey.

Devils Lake historical record.
Devils Lake 10-year record.

Ground view of Devils Lake main bay (left) and Sullys Hill (right) as seen from the west. Image date Oct. 2003; © J.S. Aber.
Blimp aerial photograph looking toward the NNW over the eastern margin of East Devils Lake. Notice the old shore line that was flooded by rising water in the late 1990s. Image date Oct. 2003; © JSA with W. Jacobson and S. Salley.
Blimp aerial photograph looking toward the north over the eastern margin of East Devils Lake. Road through center of scene has been raised to keep above the rising water; road to right (*) is now under water. Image date Oct. 2003; © JSA with W. Jacobson and S. Salley.

Exercise

Begin by making a standard false-color composites of the 1988 and 2000 scenes, as in previous exercises. Also examine space-shuttle stereophotos (in GSA lab) from 1985 to gain a better impression of regional topography and drainage.

Left: 1988 MSS 124 composite.
Right: 2000 TM 234 composite.

Note: The MSS and TM datasets differ in their geometry two ways. First is a difference in spatial resolution (see metadata files). Secondly the TM dataset is georectified in the UTM grid, but the MSS datasets are not. This gives rise to slightly different size, shape and orientation of the scenes; nonetheless, they depict approximately the same portions of Devils Lake. Another difference involves digital data values; MSS is 7-bit (0-127) and TM is 8-bit (0-255).

1. What are the spatial resolutions for the MSS and TM datasets?

2. Describe the surface drainage that flows into and out of Devils Lake.

Now examine the older image set from 1973. Make a standard false-color composite, D73-124X, as you did above. This scene is closely equivalent in ground coverage to D88.

3. How many rows and columns does D73 have; how does this compare with D88?

4. Explain the different shape of scene D73 compared to D88.

In raw displays, X-format data experience a vertical foreshortening of about 30%, because of the rectangular pixel shape. The stated spatial resolution of older Landsat MSS data is approximately 79 m, but in reality X-format datasets were stored in pixels with rectangular shape. There is some confusion in documentation about the exact dimensions that were employed for X-format pixels. The x (column) resolution is given as 56 or 57 m; the y (row) resolution is stated as 79 or 80 m. For purposes of this exercise, we will use 57 by 80 m as the assumed pixel dimensions for X-format datasets.

USGS Landsat MSS for current information.

Idrisi has the ability to display rectangular pixels correctly. Use Idrisi Explorer to examine metadata for the grid setup of D73 files. Compare the min/max x and y values to resolution and to the number of rows and columns.

5. What is the relationship among, min/max x and y values, resolution, and the number of rows and columns?

In order to create a proper display, it is necessary to recalculate the min/max x and y values based on different resolutions in the x and y dimension. In this case, the number of rows and the min/max y values may be left "as is," but a change is necessary for the max x value. Before changing metadata files, copy the original D73 files to another name. Then update the max x value in the new files. Calculate the new value for max x as follows.

Max x = [number of columns] • 0.057 (km)

6. What value did you obtain for max x? How do the new min/max x and y values for D73 compare with those for D88?

Note: To "save" the new metadata click button in lower left corner of the Metadata window. Now make a new false-color composite for May 1973, as above. Compare the new D73 image with D88.

Sample images of Devils Lake, ND.

Left: 1973 MSS 124 composite. Right: 1988 MSS 124 composite with identification of main sections of Devils Lake. East D.L. = East Devils Lake. Your images should appear similar. Click on the small images to see full-sized versions.

7. How does the shape of the May 1973 image compare to the June 1988 image? Explain how you produced this change in shape.

8. Describe the differences that are apparent between May 1973, June 1988, and Sept. 2000 paying particular attention to the appearance of Devils Lake and its shoreline.

It is obvious that Devils Lake covered a much greater surface area in 1988 compared to 1973, and the lake spread over a still greater surface area in 2000. You can make a quantitative determination of lake surface area using the MSS 4/2 band ratio (roughly equivalent to TM 4/3 ratio). With this ratio, open water usually has quite low values (< 1); whereas land areas have greater values (> 1).

Use OVERLAY function to make MSS 4/2 ratio images for 1988 and 1973, and TM 4/3 ratio for 2000. Remember to use haze corrected bands (min = 1) for inputs to the ratio overlay—see Lab 1.

Note: D00 4/3 ratio and D88 4/2 ratio have a couple anomalously high values. For each, reset metadata so that display max = 15. Save metadata, and then display the images using default options.

Examine the real-number files using default palette and autoscaling. Determine the ratio values that represent land and water. You will now create a boolean image, that is an image in which pixels have values of either 0 or 1. Land pixels = 0, and water pixels = 1. Read about and use RECLASS (under GIS Analysis—Database Query; also 5th icon from right) to create boolean images of land/water from 2000, 1988 and 1973 ratio images. Name the output images D00-WAT, D88-WAT and D73-WAT; choose user-defined classification and enter the following values. Click yes for any warning messages.

TM/MSS Reclass Values
New Value Old Values
1
0 to < 1.1
0
1.1 to 99

Tip: to expedite multiple reclass operations, you can save a reclass
file (*.RCL) for the first one, then apply it for the others.

The images will display automatically with the Qualitative palette. After you have completed the reclass operations, delete the real-number ratio images (input files), which take up considerable disk space.

Next read about and run the GROUP module (under GIS Analysis—Context Operators) on each boolean image. Group will join together adjacent like pixels and renumber the groups consecutively. Begin with MSS water images for 1973 and 1988. Run GROUP; choose diagonal links and ignore background. Name the resulting files D73-GRP and D88-GRP. The grouped files will display automatically with the qualitative 256 palette and autoscaling. There are many hundreds of groups that represent lakes of various sizes throughout the scene.

In this case, you are interested only in the groups that make up Devils Lake. Use the feature properties function (box on right) to determine the group numbers for Devils Lake in the 1973 and 1988 images.

Note: roads and bridges divide the lake into several contiguous groups; include East Devils Lake in your analysis.

9. What groups make up Devils Lake in D88-GRP?

10. What groups make up Devils Lake in D73-GRP?

To continue processing the 1973 and 1988 images, you will use another technique for reassigning pixel values. Read about ASSIGN (under GIS Analysis—Database Query) and attribute values files. Use EDIT (icon with red arrow in blue box) to make attribute values files similar to the following examples (separate file for each image). Save as integer data type.

D88-LAKE D73-LAKE
12 1 74 1
519 1 297 1
887 1 447 1
888 1 517 1
1063 1 586 1
599 1

Now run ASSIGN using the appropriate attribute values files. Name the lake images D88-LAKE and D73-LAKE. Your images will display automatically with the Qualitative palette. You should now have images that show Devils Lake only (bright red, value = 1) in a black background. These images give a strong visual impact for the dramatic increase in surface area for Devils Lake from the early 1970s to late 1980s.

Now turn your attention to the TM water image from 2000. Perform the GROUP function as before. The resulting image has several thousand groups, and the initial display does not portray the Devils Lake group clearly. Go to "Layer Properites" (box on right) and change the display as follows: equal intervals, 256 classes, display min = 0 display max = 255, palette = qual. Click "Apply" to see a new display. This forces groups 0 to 255 to display as separate colors, and all other groups are lumped together.

At this point, it should be clear that all portions of Devils Lake and East Devils Lake are included in only two groups (1 & 2). Use reclass or assign operations to create a D00-LAKE image, in which Devils Lake = 1 and other other portions = zero.

The next procedure is to run AREA (under GIS Analysis—Database Query) on the lake images. Indicate tabular output and calculate area as km². Round your answers to whole km² values.

11. What are the areas (in km²) for Devils Lake in 1973, 1988 and 2000? Round your answers to whole km².

12. How much (%) did Devils Lake expand in surface area between 1973 and 1988, between 1988 and 2000, and from 1973 to 2000?

As your final task, make image compositions for D00-LAKE, D88-LAKE and D73-LAKE. The compositions should include: title, subtitle (your name and date), scale bar, north arrow, and legend (with captions). Create a custom palette with a pale gray-brown background (0) color and bright blue-green (cyan) for Devils Lake (1). Save digital files to turn in.

Turn in


Return to course schedule.
ES 775 © J.S. Aber (2017).