We turn momentarily to a second SPOT subscene (10 km [6.2 mi] on a side), taken on May 11, 1986, over an area near Garden City in southwestern Kansas, a major wheat producing area in the Great Plains agricultural belt. The most striking feature is a series of circular patterns ranging from less than 0.4 to about 0.8 km (1/4 - 1/2 mile) in diameter (about the same sizes as the Kenyan farm plots). Some people are astounded when they see these features from an airliner flying across the Midwest.
Center Pivot Irrigation System, Northern California, Mount Shasta in the Background
The system uses a long water pipe that is mounted on motorized wheels and has one end connected to the water line at the center of the field. When operating, the irrigation system swings in a circle,sprinkling water as it rotates. (Several fields actually show a dark linear radius which is this interesting system.) The red circular fields are mostly plantings of winter wheat that are nearing harvest. Those circles in various shades of blue (note the darker patterns of moisture) are fallow at this time, but most of them have probably been seeded for spring wheat, or will be. Note the road patterns that outline squares. These roads follow survey lines that coincide with the one-mile squares, representing sections in the American Township and Range surveying system.
In some parts of the U.S. Great Plains, circular fields become the dominant type of agricultural pattern. This is strikingly illustrated in this Landsat subscene of a part of southern Nebraska:
This striking pattern of field growth is shown in this next scene - a natural color image made by Terra's ASTER instrument. The darker greens generally bespeak of standing stalks of corn. Lighter greens are various other crops. The yellows are associated with now mature spring wheat.
Circular irrigation is a worldwide practice, and can show up in unlikely places. A wadi (very intermittent stream) in the Great Desert of Saudi Arabia is the source of surface drainage and subsurface stored water from infiltration that are utilized in growing crops for distant villages:
Identification of crops using space imagery works especially well. Most fields are large (although those in parts of Asia tend to be small) - hence these will be defined by a fairly large number of uniform (homogeneous) pixels, thus favoring greater accuracy of classification. It proves very efficient to rely on Landsat-scale scenes to monitor the regional distribution of crops. This favors efficient crop yield estimates as well. Although a lower resolution sensor system, the AVHRR on some of the meteorological satellites has been effective in identifying and monitoring crops. Consider this AVHRR-based classification of crops and other classes in part of Oklahoma:
When an AVHRR classification is coupled with independent data on productivity, a reliable estimate of percent of a crop - here wheat - in various parts of a scene - here all of North Dakota - is achievable in a very short time frame:
The Great Plains and the High Plains, especially their western parts, are rich in natural grasslands. These have been widely used as rangelands on which to graze cattle. The areas involved are vast; hence, knowledge of their extent and the grasses vigor are important over the short run. Early in the ERTS program, a study indicated the effectiveness of monitoring rangelands as an aid to managing their productivity. Recent studies confirm this. A group at the USDA field station near Las Cruces, working with New Mexico State University students, tested high resolution Quickbird multispectral imagery as to its capability for distinguishing grasses types and extent of cover. This illustration shows two results: the map on the left was a "first cut" at separating grasses into 4 categories; the right map divided these into subclasses. Below it is a map that establishes four levels of grass cover:
Our final agricultural scene is actually two SPOT images of the same area, obtained two months apart during late winter and spring in northwest Africa. The purpose here is to illustrate changes in vegetation as the crop calendar progresses. The area is the Ghard plains, not far from the Atlantic coast in western Morocco, near Casablanca. This granery belt is a main producer for that country. The crops include beets, sugar cane, wheat, and rice.
Monitoring growing crops is an excellent example of using multidate imagery for change detection. Let's start by looking at the SPOT Band 1 (green) image (left image) acquired on March 14, 1986. A first reaction may be that the landscape is mostly farmland, with almost all plots being smaller than in the Kenyan and Kansan scenes. The next obvious characteristics are the small river (Oued Oum er Rbia) and the major roadway to El Jadida on the west and Settat on the east, met by several feeder roads. Note also some very dark areas in the image.
These dark areas are nearly black in Band 3 (IR) (the right image above). They are probably bodies of water, with some being irregular in shape that represent standing water from spring floods and others in field-like shapes are probably irrigated rice fields. Only a fraction of the fields (particularly along the river) are rendered in bright tones, indicating that most crops are not growing much yet.
But, when the same area is imaged in SPOT Band 3 on May 10, (the right image above) a larger percentage of fields are in medium gray tones, suggesting many crops are growing. The black areas of flooding are reduced, but rice fields remain about the same tone. The image shows towns in medium-gray tones. Tiny bright patches dot the one in the center, indicating numerous trees.
The pronounced differences in the extent of active crop and tree vegetation, are much more evident when comparing false color images first for the March scene (left) and then the May scene (right).
In the March image, an estimated 30% of the fields have actively growing crops, while in the May image, this increases to about 80%. Some of the areas thought to be water in the March scene show up as dark red (typical of rice fields) in the May view. In the March scene, a large part of the land not in crop growth appears in a dark brownish-green color which in false color is the expression of soils that are naturally reddish-brown.
The signature for rice will depend on the season in which the imagery is acquired. Rice fields during dormancy can take on a brownish color in this Landsat-7 ETM+ natural color image of an area in Guinea-Bissau, a West African country (the green areas are evergreen mangroves).
Image-processing techniques for change detection can help to emphasize major distributions of active versus dormant, or absent, vegetation. One simple approach is to make a ratio image (divide the DNs of one image by the other's DNs) using the same bands, usually in the IR, for two dates. This results in very bright tones wherever vegetation is prevalent in the dividend's date (high DN values) but not in the divisor's date (low DNs). (See page 1-15 for explanation of ratioing.) The opposite is true if the vegetation distribution is reversed because the dates are reversed. If vegetation is present to similar extents on both dates, the ratio image tones may be moderately gray. We attempted to do a ratio image on IDRISI for the March and May scenes but failed because the two scenes did not coincide (the images were taken on dates when the orbits were not exactly aligned). Co-registering images requires a computer algorithm that can execute a "rubber-stretch fit"). IDRISI, while it can generate ratio images, does not have a simple program for co-registering (aligning) two closely related images where one is slightly offset from the other (but it can do certain kinds of registration).
Ratioing of the vegetation-sensitive IR band to the corresponding red band will cause the vegetation to express in bright tones. To illustrate how well vegetation can be made to stand out by ratioing, we operate here on a same date TM Band 4/Band 3 image of an (unidentified) area in which a valley is heavily vegetated (appearing whitish in the ratio image) and the adjacent plateau is mostly rock and soil (dark).
More than any other class, it is vegetation that has strong seasonal variations (such as deciduous forests; croplands) which shows the largest differences in brightness in the infrared. Another change detection approach applicable to locating vegetation is to computer-register two scenes (IR is optimal; red band can be responsive) acquired on different dates (such as height of growing season versus winter) and then subtract the DNs of one from the other. Large changes give rise to big differences (can be assigned white in an image that spreads the differences out in the gray scale); small changes would lead to small differences rendered black; moderate changes in gray. Below is an image differencing rendition of mostly fields near Leicestershire, England, made by subtracting a February IR scene from a May IR scene. The prevalence of medium grays in this difference image means that the fields have "greened up" fairly uniformly. If the dates were summer - winter, the differences would be larger, and the tones whiter; the very white patch is probably a crop that matures early whereas the black patch is a barren field.
Another vegetation-rich environment which benefits considerably from satellite and aircraft monitoring by remote sensors is the general category of Wetlands. These are water-rich impoundments such as lakes, etc. or natural bodies of water, many at and in from the edges of ocean shores, lakes, or winters. Depending on whether the water is saline (salty seawater) or fresh, there will be a biotic assemblage that is characteristic of the particular conditions of growth. The monitoring is interested mainly in noting the shrinkage or expansion of the wetlands water and in the health and abundance of the supported vegetation. Often wetlands will follow a natural course that tends towards drying up as sediment fills the depressions holding the host water or the supply of water diminishes, or vegetation becomes so abundant as to supplant the water. Another, ever more common, cause is the deliberate draining of the wetlands by humans bent on reclaiming the area for farming, housing, recreations, etc. We shall show the ability of remote sensing to determine the characteristics of wetlands from satellites.
First up is this subscene from a Landsat TM image showing part of Cape Canaveral on the east-central coastline of Florida. Note the land area (with brownish tones) in the upper left quadrant. This is a mix of brackish (less salty than ocean water) water and land plants that are typical of this area.
The two dominant land plants are loblolly (slash) pines and palmetto plants, as seen is this ground photo.
This is the supervised classification of the small land bulge seen in the Landsat subscene
This next illustration has been degraded in the reproduction process, with the legend unreadable, but it indicates the amount of information leading to feature/class identification possible from an aircraft-mounted hyperspectral sensor. It is a classification of a neck of land tied to an "island" jutting into the Chesapeake Bay near the Bay Bridge on the eastern shore of Maryland.
A hyperspectral sensor system, AAHIS, built and operated by International Science and Technology, Inc., has 228 very narrow bands in the spectral range from 0.437 to 0.840 µm. Flown from an airplane, it also achieves high spatial resolution. Here is a false color image of the Nuupia Ponds on Kaneola Bay, Oahu, Hawaii. Beneath it is a specialized classification that distinguishes most classes present in the scene
Wetland conditions are sometimes characteristic of and suitable for growing crops that require considerable available water. Rice is such a crop and is a major staple for huge populations on Earth - mainly in Asia. Below are two classifications that include rice crops. The first is a Landsat supervised classification in China:
Multitemporal and multiband radar data can also yield quality classifications, as for this site near Dinghu, China: