The Vegetation Index; Other Vegetation Scenes - Lecture Note - Completely Remote Sensing tutorial, GPS, and GIS -
The Vegetation Index; Other Vegetation Scenes

The Landsat, SPOT and other systems participated in crop control and inventories of various kinds. As stated earlier, Landsat Thematic Mapper (TM) 4 and Multispectral Scanner (MSS) 6 and 7 bands (and SPOT Band 3) are the most sensitive for detecting IR reflectances from plant cells (modified by water content). TM Band 3 and MSS Band 5 (and SPOT Band 2), which measure reflectances in the visible red, provide data on the influence of light-absorbing chlorophyll. Ratio images using these bands help to quantify the amount of vegetation, as biomass, involved in signature responses. For the NOAA series, Bands 2 and 1 of their Advanced Very High Resolution Radiometer (AVHRR) sensor are roughly equivalent to TM Bands 4 and 3 (see Section 14 for a review of metsat systems). The ratio of TM Band 4 to Band 3, MSS Bands 6 or 7 to Band 5, or AVHRR Bands 2 to 1 is a simple approximation of the Vegetation Index (VI). Other VI variants depend on other combinations of these variables. Three of the most commonly used are shown here; read their captions for more information:

Simple Vegetation Index.
Normalized Difference Vegetation Index
Soil-Adjusted Vegetation Index; L is a measure of soil brightness

Most commonly used is the Normalized Difference Vegetation Index (NDVI), which is defined as (Near IR band - Red band) divided by (Near IR band + Red band). For TM this is (4 - 3)/(4 + 3); for AVHRR this is (2 - 1)/2 + 1).

NDVIs can be produced for the entire world. These can be for short periods (a month or so), entire years, or years apart. The top image below is a general global NDVI map for a summer. The bottom map shows a similar period but a different (time of) year and uses more contrasting colors to highlight the variations.

NDVI map for summer.
NDVI map for April.

When two NDVI maps of the world, one from northern winter and the second from northern summer, are compared, notable changes in the temperate and polar zones would be expected because of the seasonal contrasts; changes in the equatorial zone should be much less. This pair of images exemplifies this idea:

Northern winter and summer global maps of NDVI.

Changes in vegetation over shorter periods, say 2-4 weeks apart, can be displayed as NDVI maps. This series of maps applies to the 48 mid-continental United States, and progresses from mid-summer to late-fall:

Changes in NDVI for short periods of time within the U.S.

NDVI distributions for entire continents can be monitored in one view from geostationary satellites such as the meteorological satellite (this is the origin of the common descriptive term, "metsat"). Using AVHRR data (and supporting ground truth) grouped into 21-day periods for eight observing intervals (by NOAA-7) April 1982, through mid-February 1983, J.C. Tucker and his associates at NASA's Goddard Space Flight Center have produced, using Principal Components Analysis, a general classification of land-cover types for all of Africa. In the class map below medium blue = thickets and bushlands; dark blue = interspersed tropical forests and grasslands; purple = Sudan-type grasslands and woodlands; medium green = semi-arid wooded grasslands and bushlands; dark green = woodlands; yellow = deciduous bushlands and wooded grasslands; orange = semi-deserts to deserts; and red = tropical rainforests and montane forests.

A Vegetation Index (NDVI) map using two AVHRR bands on NOAA-7 showing the general distribution of various types and degrees of vegetation cover over the entire African continent, as integrated from April 1982 through February 1983.

This group and others have continued to apply metsat AVHRR to observe seasonal changes in biomass ("green wave") over all of Africa, as illustrated below for the following dates: A. April 12-May 2, 1982; B. July 5-25, 1982; C. Sept. 27-Oct. 17, 1982; D. Dec. 20, 1982-Jan. 9, 1983. This has proved invaluable in determining crop shortfalls and drought conditions in Ethiopia and in areas of the Sahel (northern desert regions) during periods in the last decade where starvation was a mass threat. This observational technique has now been applied worldwide.

Four maps of Africa showing seasonal changes during 1982-83 in vegetative cover as signified by NDVI determinations.
Key to NDVI values for the above image.

3-9: Knowing that the semi-desert south of the Sahara in northern Africa is called "The Sahel", in which of the four above panels has the "drought" moved farthest south; which panel marks the most significant northern encroachment of vegetation, as disclosed by a higher VI? ANSWER

Analysis of a mosaic made from SPOT images defines the broad categories of natural land cover. In the image below, greens indicate heavy vegetation (biomass), reds show mostly pampas grass cover, rose refers to very arid regions, and blue-white demarcates areas of barren, high, often ice-covered land:

SPOT land cover classification.

Satellite images covering large areas can give quick and direct impressions of the overall distribution of vegetation, without the need for special processing. Look at this Orbimage rendition in near natural color of Australia. It shows at a glance that most of the vegetated (forested) areas, in green, are in SE Australia, and along the coasts elsewhere. Most of the continent's interior is sparsely vegetated, as shown by the prevailing browns:

The continent of Australia.

AVHRR data allow monitoring of vegetation over scales ranging from continental, as above, down to regional or up to global. Here is a color mosaic made from Band 1 = blue; Band 2 = green; and NDVI = red for all of Canada. The browns that result indicate the wide extent to which that country is forested; in this case, greens denote areas of grasslands, such as the Canadian Great Plains.

Mosaic of Canada made from AVHRR data.

Much smaller areas can be studied with AVHRR, Landsat, SPOT, and other systems. The next three images, whose information content is given in their titles, cover a large drainage basin in the central Mexican Highlands. The Lerma River flows through the scene into Chapala Lake. This region is south of Guadulajara. This study was made by Joseph White of Baylor University.

Land Cover Classification of the Chapala-Lerma Basin
NDVI images for three different dates of AVHRR coverage of the Chapala-Lerma Basin.
LAI map using AVHRR data, Chapala-Lerma Basin.

At the other extreme, cumulative AVHRR data allow a plot of mean NDVI distribution on a global basis. This next illustration shows values averaged between 1982 and 1990, using the Los-Tucker model. Note that the highest values are in the Amazon region of South America and the largest low values are in the deserts of North Africa eastward through the Middle East into central Asia.

Global distribution of NDVIs averaged between 1982-1990.

Such data can be reworked to indicate a global parameter called HANPP (Human-Appropriated Net Primary Production). The map below, released in 2004, and developed by NASA scientists in cooperation with the Dept. of Agriculture, shows the extent to which plant food, wood, and fiber are harvested in different parts of the world. As expected, maximum amounts are gathered in the eastern U.S., southern Brazil, most of Europe, parts of Africa. and eastern China, Japan, India and Indonesia. Note the sparcity in Australia - this in part accounts for the low population in that island continent, confined largely to the coast.

Global map of regions of varying production and use of plant materials, including foodstuffs.

This type of assessment has been refined in 2007 by Dr. Marc Imhoff and colleagues at NASA's Goddard Space Flight Center. Their analysis is summarized in this pair of data plots:

Net primary production and Percent NPP needed by the local populations.

The top map shows the distribution of all vegetation, including agricultural crops, worldwide. The bottom map is very similar to the one above, showing in percentage, the amount of vegetation - mainly crops - need by the indigenous population. Note that many regions, such as India, require much more foodstuffs than can be grown locally, thus requiring imports.

Stressed vegetation can be revealed by NDVI calculations or more directly by other indices. In the summer of 2002, the United States was beset by one of the worse and widespread droughts in decades. Sparsity of rainfall in early August is discussed at the top of page 14-15. This image is a plot of AVHRR data processed to indicate vegetation stress:

Plot of degree of vegetation stress using AVHRR data for August 2002.

Vegetative stress will commonly turn green vegetation into brownish tints as trees and plants wither and perhaps die. This will cause reflectance (albedo) values for the vegetation to increase. That change is evident in this pair of Terra images of the Black Hills (heavily forested) and surrounding grass-covered high plains for the years 2000 and 2004. The latter year was one of worsening drought conditions over much of the western U.S. The higher reflectances extracted from image analysis for 2004 confirm this approach to detecting adverse conditions as rainfall diminishes.

Terra images of the Black Hills, South Dakota region and derived albedos for the years 2000 and (the drought-intensified) 2004.

Stress of another kind can have a significant impact on vegetation health and productivity. China, which needs to feed 1.3 billion people, is especially sensitive to droughts and other atmospheric factors that affect (stunt) growth. With its rapidly growing number of automobiles and industry, China has been experiencing thick blankets of smog and haze over large areas. This reduces beneficial solar radiation that promotes growth. The pair of images below, made with Terra's MODIS instrument, shows the widespread dirty haze spreading over much of eastern China as seen in natural color. The image beneath taken at the same time uses UV and Near IR bands to penetrate the haze, thus showing the extent of green (wheat mostly) in this late winter image. If this haze repeats often enough lowered productivity will result.

MODIS images of eastern China, in natural and false color, displaying a vast cover of haze but showing the greens associated with winter crops.

One thing that perturbs home owners in America is the effect of adverse weather on their lawns. Cristina Milesi, a fellow at NASA's Ames Research Center, chose as her Ph.D. thesis to study the distribution of lawns in the United States using both space and aerial imagery. She has published the map below which indicates in greens where the larger proportions of lawns are found in the U.S. She also produced a map showing areas that are impervious to rain - these are principally in cities, towns, and metropolitan districts. The patterns are closely coincident, demonstrating that most lawns in the U.S. - as one would expect - are found in urban and to a greater extent suburban regions of the country. Locate the green patch nearest you.

Map of the United States on which areas of significant lawn occurrence are shown in greens.
Map of impervious surfaces in the U.S.

The Volga Wheat Drought; Africa's Drought; the Salton Sea, California/Mexico

In the mid-70s, another example of crop failure on a grand scale, but not in Africa, made media headlines. In the Volga and other key wheat growing areas of the former Soviet Union, a severe drought in parts of Russia led to a threatening production shortfall that forced the leaders to seek help from outside wheat markets. They approached the U.S. and Australia governments, in particular, to furnish enough wheat and other grains to forestall possible starvation in several regions. Some critics claimed that the leaders were faking the shortage to take advantage of good prices elsewhere. But, the camera doesn't lie. Landsat images proved the veracity of the Russian plea for help, as is clearly depicted in this before and after image:

Partial Landsat images of the grain belt steppes of southern Russia around the Volga river taken at nearly the same time in 1974 (left) and 1975 (right), showing that the wheat fields in 1975 are much less red in this false color composite, indicating delayed growth and stress caused by a severe drought.

In the 1974 subscene that embraces a large bend in the Volga River, the fields are already in normal crop stages. A year later, and three weeks beyond the 1974 time, when mature crops should have increased the scene redness, instead much of the farmland is fallow (darker grays and tans), confirming the drought claims.

3-10: Does the drought appear localized or regional; what is the nature of the red colored area within the great curved bend of the Volga? ANSWER

Drought is one of the conditions affecting vegetation that can be sensitively measured by NDVI calculations. Repetitive space imagery, such as AVHRR, Landsat, and MODIS, has been the main data source for measuring NDVI on a regional and even continental scale. Such data can be coupled with TRMM and other meteorological satellites to correlate rainfall with vegetation distribution and health. This can also be a regional "first alert" when growth is abnormal and drought ensues. Here is an NDVI map of vegetation in all of Africa; the Sahel region, which is the zone of grasslands and farming between the Sahara and the Congo jungle, is indicated:

NDVI map of Africa.

If all goes well each year, the wet season is marked by a greening of the Sahel. Here are dry and wet NDVI maps of the Sahel showing a normal change:

NDVI and rainfall maps of the western Sahel.

But droughts can replace the pattern of normal growth. This next scene covers part of East Africa around the Horn extending from Ethiopia and Somali. The browns indicate abnormally severe stress conditions in the vegetation; yellow is a bit worse than normal; and green correlates with local areas not suffering from the effects of low rainfall.

NDVI map of part of east-central Africa, made from AVHRR data.

Drought intensified in 2004 and has threatened starvation for more than three million people in Kenya and neighboring countries. The rainy season normally starts in March and ends in June. In 2004 the rains fell mostly between mid-April and early May, amounting to less than 60% of normal, and exacerbating the previous years shortfalls. This MODIS-based NDVI map indicate the extent of the drought in June of 2004:

Drought in Kenya, 2004.

Winter wheat is a staple around Cape Town, South Africa. The next pair of MODIS images were taken exactly 1 year apart on July 21, 2002 and July 21, 2003 during the peak of southern hemisphere growing for this type of wheat crop. The greatly reduced green tones in the 2003 natural color image is a strong indication of poor crops owing to a pronounced drought in southern Africa; also affected are the natural grasslands that abound in this climate.

July 21, 2002 MODIS image of Cape Town in South Africa; normal wheat crop conditions Same area on July 21, 2003, showing effects of a severe drought.

As an aside from the theme of this page, look at this SRTM perspective image of Cape Town, which show how it and its harbor are nestled within a mountain range; the scene is rotated 90 counterclockwise from the above pair:

Perspective image made by combining a Landsat image with SRTM radar data to show the setting for Cape Town, South Africa.

Dry or desert-like regions do actually have considerable vegetation. Arid country is characterized by grasses and brushland. Here is a IRS scene in the Mojave Desert in which the main land cover types have been classified:

Classes of ground cover in the Mojave Desert, using data from the Indian Remote Sensing satellite.

The principal vegetation is the Creosote bush (Larrea Tridentata) which occurs in several settings.

Before leaving this agricultural theme, we want to look into one more example to show one of the most fertile and prolific growing areas in North America. which lies just a hundred miles or so south of the Mojave Desert. This astronaut photo shows much of southern California including the Mojave Desert, the coastal ranges east of San Diego, the Salton Sea and Imperial Valley further east, and a part of Mexico at the northern end of Baja California.

Astronaut photo looking north at the Gulf of California, the Salton Sea, and the Mojave Desert.

In this enlarged part of a Landsat scene, the Salton Sea and the Imperial Valley are hightlighted. The town of El Centro (bluish-black patch) lies about 16 km (10 miles) north of the Mexican border (very evident because of the sharp contrast with the agricultural activity). South of El Centro is Mexacali.

A nearly full Landsat TM scene of a part of southern California that contains the Imperial Valley, the Salton Sea, the Chocalate and Santa Rosa Mountains; the border with Mexico is marked by a sharp decrease in active farming at this time on the Mexican side; the Gulf of California, into which flows the Colorado River, is on the lower end of the image.

This agricultural region, one of the main producers of winter vegetables in the U.S., extends north and south of the Salton Sea, a saline body of water more than 49 km (30 mi) long, that fills a basin about 82 m. (269 ft) below sea level. This "Sea" was created by an overflow of water from the distant Colorado River (a small segment is visible in the upper right corner) shortly after the beginning of the 20th Century. Floodwaters poured through low dry washes, traveling westward more than 64 km (40 miles) to empty into the lowest part of the Coachella Valley. In this desert climate, that water is slowly evaporating and turning brackish (moderately salty) and is thus not suited for direct irrigation.

The lake-like water at the image bottom is Laguna Salada, which undergoes seasonal drops in level that at times reach a dry state, exposing playa lake beds. East of the Valley are the Chocolate Mountains, part of the Basin and Range system, against whose flanks are conspicuous alluvial fans. The bright strip in the right half of the subscene is the Algodones Dunes field, derived from beach sands left at the surface after an ancient predecessor to the Salton Sea had occupied the Salton Trough, a structural basin between the Coast Ranges (lower left) and the eastern mountains.

Today, canals from the Colorado River transport water to the sea. The biggest canal in this scene is the prominent All-American canal. Mild winters promote year-round farming (up to three harvests) in the Valley, with cotton, sugar beets, lettuce, and citrus being the main crops. Most fields are in full growth in this April scene, as indicated by the bright, uniform reds. Differences in land use practice and availability of water (no major canals) account for the pronounced decrease in agriculture on the Mexican side. This is visually quite striking when this ASTER close-up of the border land use is examined (El Centro is the town in the U.S.; Mexicali is in Mexico).

The U.S.-Mexico border south of the Imperial Valley; false color image from the ASTER sensor on Terra.

The individual fields, which show small differences in the two false color composites above, are hard to pick out in these traditional renditions. In the image below, there are three renditions of the Coachilla Valley farmlands north of the Salton Sea made by different band combinations from the ASTER spectral bands (see page 16-10). These are listed in the caption.

Varieties of false color composites of the Coachilla Valley farmlands in the area boxed in the lower right ASTER false color composite, made using these ASTER band combinations: Upper left: Band 1 = Blue; 2 = green; 3 = Red; Upper right: 1 = B; 2 = 3; 4 = B; Lower left:  4 = B; 1 = G; 2 = R.

3-11: How might a sequence of Landsat or SPOT scenes taken over the months or even years be beneficial to the economic and environmental management of this region? ANSWER

Clearly, the differences in expression using different combinations of just these bands suggests the ability to identify, differentiate, and classify the types of crops being grown. The additional bands on ASTER improve this capability, allowing greater accuracy in specifying crop types.

Before leaving the topic of conversation of arid land by irrigation into highly productive agricultural farms, which is so evidently successful in the Imperial Valley, here is a dramatic "before and after" pair of Landsat images that show the extent of change possible when water becomes abundant. The area shown in the images below is in southern Turkey just north of the Syrian border. The 1993 image shows that some croplands have been developed using subsurface water. Since then, the completion of the Ataturk Dam and Reservoir has permitted water to be spread over the area from irrigation ditches. The end result seen in the 2002 Landsat image - back to back cotton farms - is dramatic and obvious. Because such capability is still large for much of the world's arid lands to be developed using dam water, the prospect of famines over the global seems small for the foreseeable future.

Development of cotton farming over 9 years in southern Turkey, owing to abundance of water from damming.

Crop monitoring to control health and estimate harvest output has become a sophisticated system using various multisource data sets. Here is one scheme used by federal agencies and private commodity firms to make predictions or issue warnings about crop damage and other problems.

System for inventory and monitoring food and fiber at the global level

The LACIE, Agristars, and FIFE field experiments were early examples of proof-of-concept. Crop identification can be as high as 90% accurate. The programs require considerable onsite inputs but depend on satellites to provide high resolution multispectral data to extrapolate from the few training sites needed for classification of relatively small areas to maps of the vast regions that are needed to make reliable estimates at country or global scales. We shall consider this theme again in the first part of Section 13.