Standard Vegetation Index
What is the Standard Vegetation Index?
The Standard Vegetation Index (SVI) is based on the fact that vegetation conditions are closely linked to weather conditions in the atmosphere closest to the ground. It shows us the effects of climate on vegetation over short-time periods.
What does the Standard Vegetation Index tell us?
Low SVI values indicate poor vegetation condition that could be the result of climate conditions. Events that can cause low values include moisture shortages and flooding or extreme temperatures. Other things, such as delays that farmers experience during planting season caused by wet soils can cause setbacks in vegetation condition early in the growing season. High SVI values might reflect ideal climate growing conditions so that vegetation greenness is higher than encountered in other years.
Key to SVI Interpretation
One key to interpreting SVI data is to remember that each pixel is being compared only to the data found in that single pixel over time. Comparisons can be made with surrounding pixels but keep in mind that each pixel is being generated only by the data found in that pixel through nine years of data.
When interpreting SVI, each pixel is a comparison of "vegetation greenness" over a 9-year period only at that location. For example, if you look at the 1989 images, October 1989 shows a wide coverage of green or (very good), while in July 1989 most of the U.S. is coded in orange or (very poor). Initially, this might seem opposite to what it should be in terms of "normal" growing season greenness for the two times of the year in the U.S. July should have the highest amount of vegetation vigor and October the least. However, these images are not showing you vegetation greenness comparisons between months of the same year. October of 1989 and July of 1989 cannot be compared to in terms of vegetation vigor. Vegetation at a pixel location can only be compared with the condition of vegetation at that same location (pixel) in the other eight years. For example, Nebraska has numerous pixels that are bright orange in July of 1989. You would interpret this by saying that based on eight other Julys, Nebraska's vegetation greenness in those specific pixels are very poor compared to the other eight Julys for each pixel location." Crops are still being grown in this area, but are not showing as much greenness or "vigor" as in other years. The causes of the poor vegetation vigor could be many. In this specific example it was drought that was causing the poor vegetation vigor in July of 1989.<
This data is very useful in terms of looking at the effects of drought, floods, and climatic changes on vegetation vigor over an extended period of time.
Technical Background & Methods
At present, the Advanced Very High Resolution Radiometer (AVHRR) on board the National Oceanic and Atmospheric Administration's (NOAA's) polar-orbiting satellites is the instrument of choice for collecting coarse-resolution imagery worldwide due to its twice-daily coverage of most areas, and wide angle of view. The Normalized Difference Vegetation Index (NDVI) has been widely used for vegetation monitoring. The NDVI, typically derived from AVHRR data, indicates vegetation photosynthetic activity. It has been extensively used for vegetation monitoring, crop yield assessment and drought detection. The NDVI is calculated as (NIR - RED)/(NIR + RED, where NIR is the reflectance radiated in the near-infrared waveband and RED is the reflectance radiated in the visible red waveband of the satellite radiometer. This ratio provides an indicator such that the higher the NDVI, the greater the level of photosynthetic activity in the vegetation. It has been demonstrated that a time series of NDVI derived from AVHRR data is a useful tool for monitoring vegetation condition on a regional and continental scale.
We extend the usefulness of the NDVI through a new measure that standardizes the NDVI (using t-scores) to time of year and pixel location in an AVHRR image so that comparisons among years can be made. T-scores are used to estimate the probability of vegetation condition relative to the possible range of greenness. The probabilities generated are the Standardized Vegetation Index (SVI) and describe the deviation of vegetation condition from the average based on monthly NDVI values, for each of nine years (1989-1997). Findings indicate that the SVI is useful for assessing changes due to climatic conditions at a spatial resolution of 1 km. We conclude that the Standardized Vegetation Index is capable of providing a near real-time indicator of the onset, extent and duration of vegetation stress. The objective of our work with the SVI is evaluate a new technique for comparing vegetation conditions over relatively long periods of time at the highest spatial resolution of the satellite
Nine years of maximum-value composite NDVI (bi-weekly) data, from 1989 through 1997, were used in our study. These data are available from the Earth Resources Observation System - Data Center (EROS-Data Center) in Sioux Falls, South Dakota. We used data from May through September to demonstrate our concept of the SVI. This nine-year AVHRR NDVI data set is not statistically significant by comparison to a 30- or 50-year climatic normal period used by climatologists. It does, however, encompass a very interesting time period relative to the El Niño Southern Oscillation (ENSO) in the equatorial Pacific. ENSO is a Pacific-basinwide phenomenon that forms a link with anomalous global climate patterns. The time period after 1979 has been biased and dominated by El Niño (Trenberth and Hoar, 1996). The 1988-89 La Niña was the strongest cold phase ENSO event in the last 50 years, while the 1990-1995 El Niño (warm phase) event was the longest on record (Trenberth and Hoar, 1996). The El Niño event of 1997 was the strongest on record since 1950 (Trenberth, 1997). The years 1990, 1995, and 1996 were average, and not dominated by either extreme ENSO phase.
The Standardized Vegetation Index (SVI) is based on the calculation of a t-score (Student's T) for each AVHRR pixel location in the conterminous United States. We chose the t-score because we do not know the true parameters of the NDVI population given that we only have nine years of data. This uncertainty can be accounted for by using the students-t distribution that has a wider "spread" than the normal distribution. We tested the assumption of normality at a random sample of pixel locations and found the data to be normal 72% of the time at alpha .05. The t-score is calculated from the NDVI values for each pixel location for each month for each year, during the years 1989-1997 as:
NDVIijk = highest NDVI value for pixel i during month j for year k,
ij = mean NDVI for pixel i during month j over n years
sij = standard deviation of pixel i during month j over n years, and
n = 9 years (the exception is for the month of September and October where only 8 years are used)
A t score was calculated for each land pixel location in the conterminous United States (2,889 by 4,587 pixel NDVI image). Monthly values were determined by taking the maximum-pixel value if images for two composite periods were available, or the actual NDVI value at each pixel location if only one composite image was available in a given month. Single-composite images that overlap two different months were not used in the study.
After calculation of the t-score for each pixel, the probability of that score was determined as:
SVI = Prob (ti,j,k<tx, ν )
ν = n - 1 (degrees of freedom)
Intuitively, the SVI is an estimate of the "probability of occurrence"
of pixel greenness. This per-pixel probability estimate constitutes the Standardized
Vegetation Index. Thus, the values of the SVI range between greater than zero and less
than one. "Zero" is the condition in which a pixel NDVI value is lower than all
possible NDVI values for a pixel for a month in all other years of this study (1989-1997).
The value "one" is the condition in which the pixel NDVI value for the
respective month is higher than all the NDVI values of the same month in the other years.
For mapping purposes, SVI values were grouped into 5 classes, each of which comprises a
different and consecutive range of values. SVI values were grouped into 5 classes, each of
which comprises a different and consecutive range of values. These classes are average,
good, very good, poor, and very poor. Pixel classification into probability class 1 (0 -
.025) indicates the pixel NDVI value is lower than the average NDVI value
for that same pixel for the same month in the nine years of the study period, indicating
very poor vegetation condition relative to that in the other years. Pixel classification
into probability class 5 (.975 - 1.0) indicates the pixel NDVI value for the respective
month is higher than the average NDVI value
s of the same month in the
nine year period indicating very good vegetation condition relative to that in the other
Comparison of the results of the SVI images to general climatic conditions in the United States during the growing season from May through September, for each year of the study, was implemented to demonstrate the reliability of the results.