Potentials of Geographic Information Systems and Remote Sensing for an Examination of Environmental Variability Through Analysis of Historical Spectral and Spatial Data.

Among the wetlands are saturated meadows, shallow marshes, and open-water lakes. The wetlands range in size from less than one acre to two thousand acres (Wolfe, 1984). This study is very important because a reduction in the size of wetlands can be of great economic lose to the area ranchers as these wetlands, especially the wet meadows provide abundant and nutritious forage which is used as winter cattle feed. These wetlands also used as grazing sites as well as a source of water to livestock.

The data used in this study were: 1) Landsat Thematic Mapper (TM) images (1986, 1991, and 1992); 2) black-and-white (B&W) aerial photographs (1939, 1954, 1961, and 1968); 3) National Aerial Photography Program, both color-infrared (CIR) and B&W photographs (1989 and 1993); and 4) National Wetland Inventory (NWI) data-sets.

Several standard, commonly used image-analysis techniques were considered. The Tasseled-Cap Transformation (TCT ) was used to identify open surficial water and wetland areas from the TM images and scanned CIR photographs. The transformed images were classified using unsupervised classification. In order to calculate areas of surficial water and wetlands, the classified image was multiplied with another raster image created by rasterizing a vector layer of polygons for each lake.

The study shows that remote sensing data with GIS can be used to successfully identify and measure variations in the extent of open surficial water and wetland areas of (even) very small lakes. The study also shows that wetlands within the study area are dynamic and change frequently. The change in open surficial water has a direct effect on the areal extent of the wetland. For example, in 1993, the areal extent of open surficial water showed a negative change for all the lakes. In the same year the areal extent of wetlands also decreased (regardless of a difference in procedures between digital classification and manual photo interpretation).

Finally, this study also shows that the extent of open surficial water has generally increased over the years (especially in bigger lakes like Enders and Willow). In other smaller lakes, the spatial extent of open water has either increased in most years or stayed the same. Wetland areas, too, either increased or remained the same in most cases. But, wetland areas around the big lakes like Moon lake and Ender lake, have decreased considerably.

I. BACKGROUND:

This document is the final report on the results of a study aimed at monitoring and mapping changes in the spatial extent of standing surficial waters and wetlands at Enders and surrounding Lakes in Brown County, Nebraska. The project was conducted by the Center for Advanced Land Management Information Technology (CALMIT), University of Nebraska-Lincoln for the Sandhills Task Force and the U.S. Fish and Wildlife Service. The purpose of the research was to identify, record, and map changes in the areal extent of surface waters and wetlands within a specific area including Enders Lake and to do so using historical remotely sensed data in conjunction with a Geographic Information System (GIS).

II. METHODOLOGY:

A. Study Area:

The study site is the "Enders marsh" and surrounding area, approximately 100 sq. miles in size, in Brown County, NE (Figure 1). The study site is located in the North-Central portion of the state in the physiographic region known as the Nebraska Sandhills. Specifically, the study area is located between 42⁰ 13' 00" and 42⁰ 22' 30" N latitude and 100⁰ 00' 00" and 100⁰ 13' 00" W longitude.

Figure 1: Study Area

B. Data Acquisition, Processing, and Analysis

The flow chart, shown as Figure 2, summarizes the steps used in executing the project.

Figure 2: Project Flow Chart 

(1) Data Acquisition:

The first phase of the project, to obtain the relevant spectral and spatial data, was begun with an in- house search of the image archives maintained by both CALMIT and the Conservation & Survey Division (CSD), UNL. Three satellite images over the study area, acquired by the Landsat Thematic Mapper (TM) sensor at 30-meters ground resolution, were identified in the CALMIT data archive. These images were acquired by the satellite sensor in 1986, 1991, and 1992 (see Figure 3 for specific dates). A number of black-and-white (B&W) aerial photographs, for the years 1939, 1954, 1961, and 1968 for Brown County, NE, were located in the CSD image archive, although only the 1954 photographs covered most of the study area.

The second step in identifying image resources was to search the holdings of the United States Geological Survey (USGS) EROS Data Center. Two sets of National Aerial Photography Program (both color-infrared (CIR) and B&W) photographs covering the study area for the years 1989 and 1993 were located in that archive and purchased (see Figure 4) . National Wetland Inventory (NWI) data-sets, in vector format, for the study area were also obtained to use as a reference.

(2) Hardware / Software

ERDAS Imagine (ver. 8.3.1) and ESRI's ARC-INFO software on Sun Sparc-20 workstations were used to perform all data analyses.

(3) Pre-Processing of TM Scenes:

Each TM image covers a ground area of approximately 13225 square miles (185 KM X 185 KM or 115 miles X 115 miles) , which is a digital data-set of 350 megabytes. To reduce the processing time and storage space, the TM images were down loaded from compact disk and the study area was "cut" out of

Figure 3: Landsat Thematic Mapper (TM) Images (False Color Composite Show TM Bands 4,3,2).

Figure 4: Image Mosaic of Aerial Photographs (Color Infrared and Black-and-white).

the overall coverage. All TM images were then geo-referenced to Universal Transverse Mercator (UTM) map coordinates.

(4) Pre-Processing of Aerial Color-Infrared and Black-and-White Photographs:

The color-infrared photographs were obtained as 9x9-inch positive film transparencies while the black-and-white photographs were the same size but stored on paper. A total of 25 photographs (either CIR or B&W) was required to cover the entire study area.

The first step, converting the photographs to a digital format, was accomplished by means of a SHARP JX 610 high resolution flatbed color scanner. Each photo was scanned at a resolution of 300 dots per inch (dpi) to get a spatial resolution (on the ground) of 4 meters. Next, all individual photographs were "stitched" together to create a digital mosaic of the whole study area (Figure 4). In order to mosaic photos together, each one had to be geographically referenced to known ground locations and geometrically rectified to a common coordinate system (UTM). Each photo was rectified using 10 to 15 Ground Control Points (GCP's), comprised of man-made features (e.g., road intersections, road junctions, bridges, railroad crossings, etc.) which could clearly be identified on both the image and a reference map (USGS topographic). Creating digital mosaics from individual photographs is labor-intensive and requires large amounts of time. Each photograph was joined to the adjacent photograph in such a way as to preserve the correct positional relationships of the features on each photograph, which is difficult because some distortion is always present near the edges of individual aerial photographs.

(5) Analyses of Digital Satellite and Aircraft Images:

Since, the overall aim of the project was to detect and map changes in the areal extent of standing surficial water and wetlands over time, several standard, commonly used image-analysis techniques were considered. The Tasseled-Cap Transformation (TCT ) is a method to convert an input image acquired by a "multispectral sensor" (i.e.., one operating simultaneously in several spectral regions, such as visible green, visible red, and near-infrared) into an image which has three main output components that highlight "feature classes" such as bare soil, vegetation, and water (Kauth and Thomas 1976, Crist and Kauth 1986, and Crist 1985)

The calculation of TCT, which is a complex procedure involving the linear combination of all the bands of an image, results in six "synthetic output channels" for a Landsat-TM image, where band 1 corresponds to soil brightness, band 2 to vegetation greenness, band 3 to wetness, band 4 to haze in the atmosphere, and bands 5 and 6 to "system noise." Trial and error confirmed, for Brown County, that a combination of first two transformed bands provided a best result for separating vegetation, soil and water. The two transformed bands (synthetic channels) were then used to classify the image subset for the study area into three general classes (soil, vegetation, and water)( Figure 5). These classes were then labeled, evaluated, and verified by comparing them to both black-and-white and color-infrared aerial photographs, and also to NWI results. Areas of standing surficial water and wetlands in and around each lake in the study area were then calculated (Figures 6 and 7). The same procedure was applied for all three TM images (1986, 1991, and 1992).

Figure 5: TM Image Classification.

Figure 6: Open Surficial Water Calculated from TM Images.

Figure 7: Wetland Areas Calculated from TM Images.

The same general procedure was used to calculate open surficial water and wetland areas from the scanned CIR photograph (1989) as was used for the TM images, although the actual calculation of the TCT varied slightly. However, the procedure for the B&W photographs was quite different because they summarize reflectance/radiance of earth surface in only the visible portion of electromagnetic spectrum; i.e., they are not a multispectral product. Therefore, the differences in the representation of water, soil and

vegetation is often minuscule or even non-existent. As a result, it is very difficult to use any digital image-processing technique to process and analyze scanned B&W photographs. In order to accomplish the task of calculating water and wetland areas for 1954 and 1993, a manual image-interpretation method was used. This method involved visual interpretation of ground features by the image analyst and subsequent screen-digitizing of the standing water and emergent-wetland areas. Then, the acreage calculations were done (Figures 8 and 9).

RESULTS AND DISCUSSION:

The area of open surficial water calculated for each lake and for various dates is shown in Table 1(a). There are nineteen lakes in the study area, with six having no name provided on the corresponding USGS map. Therefore, these are simply listed as "noname" in the table. The noname lakes tend to be small (sizes range from less than five acres to around ten acres, except noname #1 which is several hundred acres in size).

Table 1(b) summarizes changes in the areal extent of the open surficial water from one year to the next, as does Figure 10. We found six lakes that have a range in open surficial water area between 250 acres and 700 acres, and another eight with variable areal extents between 50 acres and 200 acres. The remaining five lakes range between 5 and 50 acres in sizes, for the years analyzed.

Results of the calculation of spatial extent of open surficial water show that there is a negative change in the total acreage of thirteen lakes and there is apositive change in the total acreage of four lakes (Crystal, Philbrick, Enders, and Chain), whereas, Rat and Clear lakes remained almost unchanged between 1954 and 1986. The biggest decrease in the open water between 1954 and 1986 occurred in Noname #1 lake which

Figure 8: Open Surficial Water Calculated from Aerial Photographs

Figure 9: Wetland Areas Calculated from Aerial Photographs.

Figure 10: Variations in Total Lake Areas and Open Surficial Water Over the Study Period.

Table1: (a) Area of Open Surficial Water

(b) Change in Area of Open Surficial Water

	Lake_Names	1954 - 1986	1986-1989	1989 - 1991	1991 - 1992	1992 - 1993	1954 -1993
1	noname 1	-98.74	18.64	-20.64	-61.16	-129.71	-291.61
2	Moon	-45.31	17.08	3.83	129.21	-145.73	-40.92
3	Clapper	-13.07	81.50	-61.26	250.86	-254.14	3.89
4	Twin	-33.76	0.01	3.10	0.67	6.38	-23.61
5	Rush	-13.47	1.88	-0.55	1.33	-6.06	-16.86
6	Crystal	12.11	38.64	-36.42	42.26	-55.50	1.08
7	Philbrick	65.80	22.97	-1.84	72.50	-146.83	12.60
8	Skull	-27.50	-0.77	16.12	-1.11	-24.62	-37.88
9	Long	-52.24	27.52	-5.28	105.42	-125.94	-50.53
10	Rat	-0.41	13.84	-163.96	201.93	2.14	53.54
11	noname 2	12.01	-3.66	5.66	2.45	1.61	18.07
12	Enders	51.19	-14.56	26.57	32.47	-7.92	87.75
13	Clear	0.74	6.29	11.50	8.23	-27.56	-0.80
14	noname 3	12.68	0.53	-2.09	6.00	-2.25	14.88
15	Chain	5.17	-0.78	1.45	11.79	-16.68	0.94
16	Willow	-11.88	3.11	22.91	8.45	-31.98	-9.39
17	noname 4	2.22	18.31	24.84	6.89	-34.83	17.43
18	noname 5	6.23	3.17	0.16	4.45	-14.01	0.00
19	noname 6	5.78	12.31	-6.08	16.01	-28.02	0.00

is around 98.74 acres and the biggest increase of around 65.80 acres occurred in Philbrick lake. On the other hand, most of the lake areas have changed positively during 1986 and 1992. In 1993, most lakes show a reduction in the extent of open surficial water as compared to previous year. The areal extent of all lakes increased in 1989 as compared to 1986 except Enders, Chain, and Skull. Enders Lake showed a reduction of 14.56 acres whereas Chain and Skull lakes shrunk only about three-quarters of an acre. Between 1989 and 1991, eight lakes out of the total nineteen show a reduction in size, with the biggest reduction (for Clapper Lake) being about 61 acres. The second-largest reduction occurred at Crystal Lake, which dropped from 87.57 to 51.15 acres. All the lakes showed an increase in their size during 1991 and 1992 except Noname #1 and Skull Lakes. Noname #1 Lake was reduced 61.16 acres while Skull Lake showed a reduction of only one acre. A negative change occurred between 1992 and 1993 in almost all lakes; i.e., there was a reduction in the open surficial water extent. The reason for these changes may be a difference in the season in which the aerial and satellite images were compiled. For example, the 1992 TM image was acquired in April whereas 1993 image was collected in June. In April (spring season) there is very little green vegetation standing in wetlands; therefore, the satellites are likely to "see" open water. On the other hand, in June, the vegetation cover tends to be dense, so a lesser amount of water is detected by satellite. Another potential source of error was mentioned earlier; the fact that the area calculations for 1993 were the result of manual photo interpretation and screen digitizing. This may also, although doubtful, cause the reduction shown in the extent of open surficial water between 1992 and 1993. But, if open water measurements for 1993 are compared with those for 1986, half of the lakes increased in open water extent, including the big lakes ( e.g., Rat, Enders, Willow, which have areas of more than 250 acres). The rest are very close to the area in 1986. In summary, although, the areal extent of open surficial water fluctuated

Miller, S. 1990. Land Development and Use. In An Atlas of Sand Hills (2nd. edition), edited by A. Bleed and C. Flowerday, University of braska-Lincoln.

Mitsch, W. and J. Gosselink 1993. Wetlands (2nd. edition). New York: Van Nostrand Reinhold.

Mouat, D., G. Mahin, and J. Lancaster 1993. Remote Sensing Techniques in the Analysis of Change Detection, GeoCarto International. 2:39-50.

Reimold, R., J. Gallagher, and D. Thompson 1973. Remote Sensing of Tidal Marsh. Photogrammetric Engineering and Remote Sensing. 39:477-488.

Richards, J. 1986. Remote Sensing and Digital Image Analysis. Berlin:Springer-Verlag.

Robinove, C. 1982. Computation of Physical Values for Landsat Digital Data, Photogrammetric Engineering and Remote Sensing. 48:781-784.