LaDorna Jo Pfaff, May 30, 2006, Emporia State University, Kansas
Prepared for Earth Science 775, Advance Image Processing, http://academic.emporia.edu/aberjame/es775/syllabus.htm/ at Emporia State University, Earth Science Graduate Program, http://www.emporia.edu/earthsci/graduat.htm/ , Professor James S. Aber http://www.emporia.edu/earthsci/aberjame.htm/
Remote sensing images have been useful to study the changing extent and type of vegetation in riparian buffers. Combined with software to enhance the images, and ground support observations, timely assessments can be made of the of the riparian buffer. This study looks at three years of Landsat data, all taken in September, between 1973 and 1992. There was a period of drought in the mid 1970s, and a period of high precipitation in 1992. Other variables effecting the riparian buffer are noted, especially the introduction of the invasive tamarisk (salt cedar) tree, Tamarix ramosissima. Normalized Difference Vegetation Index (NDVI) is used to measure the vegetative biomass. The NDVI index is dependent on ground moisture, and therefore, precipitation. Changes in the size of the riparian buffer depends largely on the rainfall in the current and antecedent months. Changes in the shapes of the NDVI histograms may indicate a change in the type of vegetation cover.
Landsat introduction: In 1966 William Pecora, United States Geological Survey, announced plans for development of satellites carrying television cameras, return beam vidicom tubes (RBV), designed primarily to acquire Earth resource information. In 1968 Virginia Norwood, Hughes Aircraft, proposed a different kind of sensor: a scanner that created images in strips using an oscillating mirror. The Hughes scanner had multispectral capacity. The National Aeronautics and Space Administration (NASA) started to produce satellites in 1972. The early satellites had both MSS and RBV sensors. Later satellites had only the MSS sensors because the RBV systems failed. Since 1972 data has been streaming-in from the Landsat series of sensors: Landsat Multispectral Scanner (MSS) and the Landsat Thematic Mapper (TM).
Three satellite images of Canyon de Chelly, Arizona, are explored in this report. These images were obtained from Arizonia Regional Image Archive, ARIA, University of Arizona, Office of Arid Lands Studies:
September 20, 1986 TM images
September 4, 1992 TM images
On-the-ground photographs were obtained June 10, 1975.This was in the middle of a four-year drought.
Due to improvements in the later satellites, there are differences in sensors and band characteristics among these three images:The 1973 image is from MSS Landsat 1. It orbited at an altitude of 920 km Bands 1,2, and 3 were assigned to the RB video camera. Band 4, later renamed band 1 is 0.5-0.6 µm, green Band 5, later renamed band 2 is 0.6-0.7 µm, red Band 6, later renamed band 3 is 07-0.8 µm, shortest NIR (photographic infra-red) Note: this band is not in the subsequent sensors. Band 7, later renamed band 4 is 0.8-1.1 µm, NIR (near infra-red) The 1986 and 1992 images are from TM Landsat 7 which orbits at an altitude of 705 km. Therefore these images have greater ground coverage than the 1983 image. Band 1 0.45-0.52 µm, blue Band 2 0.52 -0.6 µm, green Band 3 0.63-0.69 µm, red - not as wide a red band as in the MSS image from 1973 Band 4 0.76-0.90 µm, NIR -- not as wide a NIR band as in the MSS image from 1973 Band 5 1.55-1.75 µm, MIR Band 7 2.15-1.35µm, MIR
The Canyon de Chelly and Canyon del Muerto were formed by meandering streams from the Chuska mountains. Like any mountain stream, they started to dig themselves into a canyon-in this case, narrow 304 meter-high vertical-walled canyons. (see photo). Uplifts that formed the broad-domed Defiance Plateau and the formation of the volcanic Chuska Mountains to the northeast accelerated the canyon’s formation.. Tsaile Creek forms Canyon de Chelly, and Spruce Creek forms Canyon del Muerto. Both creeks flow southwest from the Chuska Mountains and join the Chinle Creek at the western end of the Canyon de Chelly National Monument. Today the creeks are shallow unless there is a heavy rainstorm.
The meandering streams that carved the canyons, some which are ephemeral, are important riparian corridors. Flash floods are common after heavy summer rainfalls, and sandy areas become saturated to form quicksand. Summer rainfall begins early in July and lasts until mid-September. In general, the highest amount of rain occurs in August. But, rainfall is variable from year to year. Yearly averages are deceptive indicators of successful riparian buffers. The satellite images in this report were taken September 1973, 1986 and 1992. For specific years check the rainfall in the antecedent month-- the month before the satellite image was taken and compare this amount with the previous year’s statistics. There has been a 35-year record at Canyon de Chelly. 1973 was a time of drought, culminating in 1975. 1992 was a year of high rainfall, especially in the two months preceding the image acquisition.
1986: August = 0.62 inches (1.57cm), September = 1.57 inches (3.99cm), Annual total = 8.14 inches (20.68cm)
1992: August = 2.73 inches (6.93cm), September = 0.49 inches (1.24cm), Annual total = 10.19 inches (25.89cm)
To view entire period of record, go to the website: Western Regional Climate Center, Canyon de Chelly, Arizona, monthly precipitation (inches), obtained from the Web 2/28/06 http://www.wrcc.dri.edu/cgi-bin/cliMAIN.pl?az1248/
Air temperatures, while swinging between -30º Fahrenheit (-34ºC) in the winter to +104º F (40ºC) in the summer, have steady annual averages. From 1970 to 2006, the mean annual temperature was 53.60º F (12ºC) with a standard deviation of only 1.69º F.
Besides climate, there are other sources of change effecting the riparian buffer. If riparian buffers are maintained, they will support a wide range of native fish, provide drinking water for mammals and birds, and trap sediment for new plant growth. Beavers are native to the region. Before the 1900s their dams provide areas where sediment was trapped. As the dams fill in, the area can support the growth of new vegetation. In the early 1900s the beavers were extirpated for their fur. They have since been re-introduced. Overgrazing of livestock can cause desertification and increased erosion into the riparian buffer. A large contributor to change has been the introduction of non-native plant species. The Park Service initiated plantings as a conservation method to control erosion along the canyon’s waterways. In the 1950s on the Colorado Plateau they planted native Fremont cottonwood, and non-native plants: tamarisk (salt cedar), Russian olive, and peach-leaf willow. The non-native trees, especially tamarisk, has become invasive.
Tamarisk is a wispy needleleaf deciduous tree with deep roots, up to 30.4 meters. It leaves a salt residue in the soil and is unpalatable to wildlife and livestock. It quickly displaces native species, and is poor habitat for birds, animals, and livestock. The most critical problem is that tamarisk has much deeper roots and uses more water than the native vegetation it replaces. Thickets of tamarisk draw-down the water table. The tamarisk problem continues to worsen. In 2003 is was estimated that tamarisk species in the West consume 2.0-4.5 million acre-feet of water per year. This value is much higher than what native plants would use. It represents enough water to supply 20 million people, or the irrigation of 1,000,000 acres (4,049 square kilometers) of land. (San Souci, 2003)
Except for the pre-historic cliff dwellers (1250-1300), the canyon was largely unpopulated. In the1750s Navajo people settled the canyon to raise sheep and plant gardens and fruit trees on the alluvial benches in the riparian buffer. In 1863 most of the Navajos were marched off to Fort Sumner. The Native people were returned to a 3.5 million-acre Navajo reservation, but the area was too small; sheep overgrazing caused desertification. In 1931 the U.S. Park Service designated the area as a National Monument. Uncharacteristically, tribal rights and land ownership were maintained in the National Monument. In 1935 the reservation size was increased to 18 million acres. Today, besides raising sheep, the Native Americans use the riparian buffer for fishing, vegetable plots, and orchards.
The satellite images were imported from ARIA as band-sequential files and processed in Idrisi Kilamanjaro software. Realistic color in a "naturalistic-color" image in the 1973 data sets would utilize bands 1,4, and 2 (green, NIR, red). In the later images ( 1986 and 1992) bands 123 (blue, green, and red)are used. In the naturalistic scene water is black, vegetation is various shades of green, smooth surfaces are white and purple. Standard false color for the 1973 image would be bands 1,2, and 4 (green, red, NIR). In the later images standard false color composites use bands 2,3, and 4 (green, red, NIR). Standard false colors transform the vegetation into the reds: light pink to purple-red, shadows from the canyon walls appear dark gray, water appears black, and highly reflective surfaces, such as roads and clouds, appear light blue, green or white. An interesting way to view the shadows is by using the single NIR image. In this presentation the canyon shadows appear a deep bright blue.
The Normalized Difference Vegetation Index, NDVI, can be used to determine the density of green vegetation by measuring the visible and near-infrared sunlight reflected/absorbed by the plants. The chlorophyll pigment in the leaves of the plants strongly absorbs visible red light (0.6 -0.7 µm). On the other hand, the cell structure strongly reflects the near-infrared ( 0.7 to 1.1µm). A ratio is made using these the red band and the NIR band. The formula is NDVI = NIR-red/NIR + red. Values will range from -1 (no vegetation) to +1 (highest possible density of green vegetation). Much of the desert vegetation is ephemeral; it is quick to change when heavy summer rainstorms or drought occur. Even so, hardwoods and tamarisk can show a one to two-year lag in drought conditions because of their deep roots. Bottomwoods, such as riparian buffers, can sometimes show increased vegetation after the first year of drought. This is presumably from more understory growth in dry hollows and potholes.
In the 1973 image the MSS band widths are slightly different in width from the later sensors. There is also a difference between the spatial resolution than the spectral resolution. It will somewhat difficult to compare the three NDVIs because of these sensor difference. Seasonal climate and summer rainstorms are probably of greater importance than slight difference in spectral data. In ratio images, such as the NDVI index, it is important check for haze. The minimum value should be 0. Haze effects bands 1 and 2. In these data sets, all minimum values are 0, so no haze correction is needed. It is necessary to change the NDVI images from real to byte binary. To view the NDVI image, the NDVI palette is used. In the setup using the "NDVI palette" and a maximum display of 200, NDVI color coding is: active vegetation is red/brown, rock surfaces are light green, and water bodies are black. However, the shadows from the canyon walls are not clear-cut. To isolate the riparian buffer, an NDVI image of the three selected years was viewed with the “quant” palette. The vegetation with this palette is shown in green, sparse upland vegetation in the Chuska Mountains is shown as red. Surprisingly, the 1973 NDVI image shows more vegetation, compared with both the 1986 and 1992. This is at odds with both the overall precipitation record and what the analyst expected with the increasing invasion of tamarisk and Russian olive. Another way of showing vegetation in the 1973 image is to make a 3/1 ratio(photographic infra-red/green)image, and then use it in a composition with the red, NIR, photo infra-red/green formula. This image uses all bands in the early MSS systems.
Masks the riparian buffer's vegetation were made along Canyon de Chelly. The masks were made from the NDVI images using Idrisi’s Reclass and Supervised Classification:
Histograms of the vegetative zones in the NDVI images : The 1978 index shows a single sharp peak at 124. The 1986 image shows two peaks, a lower one at 116, and the steeper tall one at 124. The 1992 NDVI shows a smaller peak at 120 and the main sharp peak still at 124.
Note of caution: The satellite masked areas have not been correlated with ground-truth information from 1973, 1986, and 1992. Telephone information from 2006 with the Canyon de Chelly Park Service administrators confirm that there has been a large increase in vegetation on the canyon floor due to the invasive tamarisk and Russian olive exotics. It is imperative that satellite image work be tested with what is on the ground. For this exercise, the basic NDVI ratios do not seem to show the increasing extent of vegetation from the 1950s plantings of tamraisk and Russian olive. The vegetation during the three examined years appears to be hugely dependent on the immediate month or antecedent month's rainfall; Over the 35-year period of record, the average amount of rainfall in August is 1.27 inches, and 0.95 inches in September. September 1973's rainfall was slightly below average (0.77 inches). However, in August 1973 there was a high amount of rain (0.98 inches). The rainy August conditions may have increased the vegetated area in the riparian buffer of the canyon, as evidenced by the larger 1973 vegetative area. The 1986 image was taken later in September from the other two. The rainfall for September 1986 was 1.57 inches. This high amount of rainfall in September may have effectd to the imaged vegetated area. September 1992's rainfall was only 0.49 inches. This value is much lower than the average rainfall for September; thus, there is a low amount of vegetation on the 1992 NDVI image.
Tamarisk delineation and control
Can the extent or density of invasive vegetation, such as tamarisk, be mapped? Different vegetation has different spectral signatures. Is there a spectral difference between the native vegetation and the exotic species? In some systems, discussed below, there is confusion between tamarisk, alfalfa, pasture land, and some specific emergent wetland features. Tamarisk reflects in the 0.52 - 0.59 µm band, the green band. Can a remote-sensed tamarisk feature be separated to calculate the area of infestation? The first step in controlling the tamarisk invasion is to monitor and quantify it. Most techniques to separate tamarisk use higher-resolution imagery than the datasets in this report. Several avenues of monitoring tamarisk are highlighted below:
1. Kite aerial photography - (Aber, 2005). Using a kite with attached remotely-controlled digital camera, researchers can obtain birds-eye view of tamarisk thickets. Defoliation can be easily monitored by the extend of color change. Defoliation, occurring because of larval infestation, appears a distinctive reddish-brown color. In Pueblo, Colorado, this type of defoliation peaks in August.
2. Airborne color video imagery - (Corwin, 1996). The tamarisk invasives can easily be monitored on conventional color video imagery when the leaves turn yellow-orange to orange-brown in late November. These video images are combined with a global positioning system to give the coordinates. The aerial extend of the tamarisk can be mapped.
3. Multispectral - QuickBird satellite - (San Souci, 2003). Multi-resolution/multi-spectral images are fused (pan-sharpened) to make a higher-resolution image. The technique uses blue, green, red, NIR, and panchromatic images. In this approach, the lower-resolution multispectral bands (2.5m) are fused with the higher-resolution panchromatic image (60cm). This produces a pan-sharpened MS image used to extract the tamarisk feature. After applying enhancement algorithms, tamarisk can be separated from other vegetation with an approximately 80% accuracy. The mapped tamarisk areas are conservative, however, since the system can only separate areas greater than 10 square meters. This restriction eliminates measuring the narrow thickets along streams. New calibrations to remove variations due to differing soil brightnesses are being developed.
4. Airborne AVRIS sensors - low-altitude images with 3.7m ground resolution, using all 174 spectral bands can give an accuracy of 79% in mapping tamarisk vegetation.
5. Hyperspectral EO-1 satellite - Hyperion sensor (McGwire, 2002). Using 18 spectral bands of the 242 bands available, at 30m resolution, with a minimum noise fraction (MNF), tamarisk can be identified at 85% accuracy.
The second step in controlling the tamarisk invasion is trying to eradicate the plants. In Canyon de Chelly the National Parks Service initiated controlled burning in 2005 in the canyon's riparian buffer to eradicate the exotic invasive species. In other areas of the Colorado Plateau Field, tests in controlling tamarisk began in 2001. By introducing the Chinese leaf-eating beetle (Diorhabda elongata deserticola), the beetle selectively defoliates the target tamarisk trees. Defoliated tamarisks are not dead, and may green-up again the following year. However, multiple-year beetle infestation would eventually kill the tree.
Discussion and Conclusions
With the datasets provided by ARIA, the analyst hoped to monitor the extent of the riparian vegetation and compare it with the amount of rainfall in the month of the image acquisition. However, some difficulties were encountered. The shadows cast by the high, narrow canyon walls obscure some of the vegetation. Aerial extend of un-shadowed vegetation was mapped using reclass functions or supervised classification (cluster) functions. Masks of the vegetation were made using, alternately, the NDVI image and the false-color composite. After the mask of the un-shadowed vegetative area was defined, the aerial extend of vegetation can be computed. Histogram curves in those specific area show differences in the general reflectance and absorption of the vegetation. The varying peaks may indicate different species abundance, such as the increase of invasive plants. Information from the Park Service administrators in 2006 confirms that there has been a explosive increase in vegetation on the canyon floor due to the invasive tamarisk and Russian olive exotics. The selected masked areas should be correlated with on-the-ground observations from 1973, 1986, and 1992. The selected NDVI images in this report are, perhaps, better correlated to the preciptiation in the month of image acquisition rather than invasive vegetation.
Aber, J. S., D. Eberts, S.W. Aber, 2005, Applications of kite aerial photography: bio control of salt cedar (Tamarix) in the western United States, Transactions of the Kansas Academy of Science, Vol. 108, no. ½, p. 63-66
Arizona Regional Image Archive, ARIA, University of Arizona, Office of Arid Lands Studies, satellite image downloads, obtained from Web, 2/25/06, http://aria.arizona.edu/
Butler, B., Durango Bill’s Paleogeography Research, The Defiance Plateau, the Chuska Mountains, and Canyon de Chelly, obtained from Web 3/10/06, http://www.durangobill.com/PaleoAppendPart8.html/
Corwin, D.L. and K.M. Loaque, 1996, Applications of GIS to the modeling of non-point source pollutants in the vadose zone. Soil Science Society of America, xxiii, Madison, Wisconsin
Grahame, J. D. and T. D. Sisk, ed., 2002, Canyons, cultures and environmental change: An introduction to the land-use history of the Colorado Plateau. Obtained from the Web 03/17/06 http://www.cpluhna.nau.edu/
Leslie, E., National Park Service, Canyon de Chelly, telephone conversation 5/25/06
McGwire, K., 2002, Hyperspectral Monitoring of invasive, non-native plant species with EO-1 Hyperion Imagery, Desert Research Institute, Reno, Nevada, grant NASA-NCC5-486
San Souci, J.W., and J. T. Doyle, 2003,Tamarisk mapping and monitoring using high resolution satellite imagery, obtained from the Web 4/18/06, http://www.featureanalyst.com/topics/conferecne/SanSouci_Tamarisk.pdf/
Terraserver, topographic map, obtained from the Web 4/19/06, http://terraserver-usa.com/image.aspx?T=2&S=17&Z=12&X=25&Y=155&W=1/
Western Regional Climate Center, Canyon de Chelly, Arizona, monthly precipitation (inches), obtained from the Web 2/28/06 http://www.wrcc.dri.edu/cgi-bin/cliMAIN.pl?az1248/
Last updated May, 2006