Normalised Difference Vegetation Index (NDVI)
The Normalised Difference Vegetation Index (NDVI) grids and maps are derived from satellite data. The data provides an overview of the status and dynamics of vegetation across Australia, providing a measure the amount of live green vegetation. The satellite data comes from the Advanced Very High Resolution Radiometer (AVHRR) instruments on board the National Oceanic and Atmospheric Administration (NOAA) series of satellites that are operated by the US (http://noaasis.noaa.gov/NOAASIS/ml/avhrr.html).
What is NDVI?
Live green vegetation absorbs visible light (solar radiation) as part of photosynthesis. At the same time plants scatter (reflect) solar energy in the near infrared. This difference in absorption is quite unique to live vegetation and provides a measure of the greenness of the vegetation.
NDVI is an index which measures this difference, providing a measure of vegetation density and condition. It is influenced by the fractional cover of the ground by vegetation, the vegetation density and the vegetation greenness. It indicates the photosynthetic capacity of the land surface cover.
NDVI is calculated from the red and near-infrared reflectances rRed and rNIR as
NDVI = (rNIR - rRed) / (rNIR + rRed)
Its value is always between -1 and +1. Vegetation NDVI in Australia typically ranges from 0.1 up to 0.7, with higher values associated with greater density and greenness of the plant canopy. NDVI decreases as leaves come under water stress, become diseased or die. Bare soil and snow values are close to zero, while water bodies have negative values.
Monthly NDVI maps
Monthly NDVI is a composite of the NDVI values from cloud-free observations in the month from the operational afternoon NOAA satellite. There are, in the absence of cloud, usually one and sometimes two observations per day. The satellite data are processed initially onto a 0.01x0.01 degree grid and then averaged to a 0.05x0.05 degree grid. The data is available within a few days after the end of a month.
The national map shown on the web is based on the 0.05x0.05 degree grid, sub-sampled at every fifth point to give an effective resolution of 0.25x0.25 degrees. The regional maps are based directly on the 0.05x0.05 degree grids, so there may be some slight differences in the fine detail between the national map and the regional maps.
All grids and maps may be updated from time to time as the processing algorithms are improved with, for instance, better satellite calibration or cloud masking. A date stamp at the bottom right-hand corner of each map indicates when the analysis was produced.
The 3-month and 6-month grids are averages of the one-month grids. They are generated monthly and are available within a few days after the last month of the averaging period.
NDVI standardised anomaly
The NDVI standardised anomaly is the departure of NDVI from the long-period average, normalised by the long-period variability. It indicates whether the vegetation greenness at a particular location is typical for a particular compositing or averaging period of the year, or whether the vegetation is more or less green.
Monthly anomalies are generated from the monthly NDVI dataset by subtracting the long-period mean and dividing by the long-period standard deviation for that month of the year, for each grid cell. The reference period is 1992 to 2008, excluding the poor quality months of April to September 1994 (due to low sun elevations, resulting in excessive shadowing on the ground) and September 2003 (due to satellite instrument scan anomalies).
For the 3-month and 6-month anomalies the NDVI values are averaged over the three or six months in question, and normalised by the mean and standard deviation calculated over all instances of those months over the reference period except for the poor quality months listed above.
The NDVI anomalies are produced from data from the AVHRR instrument carried by the series of polar orbiting satellites operated by the US NOAA. The NOAA-11, -14, -16 and -18 satellites are used, which are all in afternoon orbits. The AVHRR images each location in Australia in daytime at least once daily, in five or six spectral bands. Bands 1 and 2 make the measurements of red and near-infrared reflected sunlight, respectively, from which NDVI is calculated, while other bands measure the Earths thermal emission.
The AVHRR data from April 1992 to June 2008 were generated by CSIRO (http://www.eoc.csiro.au/cats/). Data from July 2008 onward were received and processed by the Bureau of Meteorology.
For key processing steps, CSIRO and the Bureau both use the Common AVHRR Processing System (CAPS) software developed by CSIRO. CAPS Modular Processing applies geolocation, calibration, cloud masking, sea masking and regridding. No atmospheric or angular corrections are applied.
While the AVHRR instruments are calibrated before launch, their calibration changes in orbit, typically with a rapid change immediately after launch and then drifting during their mission lifetime of several years. Furthermore, the AVHRR carries no on-board calibration system for the reflective channels. Therefore, consistency of calibration over the length of the NDVI time series was achieved by adopting the procedure developed for use in Australia by the Environmental Resources Information Network (ERIN). This procedure adopts the calibration published by Rao and Chen of NOAA in 1999 using a stable Libyan desert site for the NOAA-14 segment of the time series, and then assumes the reflectance stability of a set of Australian arid sites to detrend the calibration of the other satellites and match them to NOAA-14.
The cloud mask applied by CAPS is a modification of the CLAVR-1 scheme developed by NOAA, and uses the reflective and thermal bands. The data are regridded using nearest neighbour interpolation to a 0.01-degree geographic grid spanning the Australian continent. NDVI is calculated from the red and near-infrared (band 1 and band 2) reflectances as NDVI = (rNIR - rRed) / (rNIR + rRed).
A single satellite orbit over Australia is typically received at more than one ground station, as two to four swaths with substantial overlap. These are merged into a single file: by CSIRO as data in a low-level format ('stitching') before CAPS processing; and by the Bureau of Meteorology as regridded data after CAPS processing. The data from January 2006 to June 2008 comes from Alice Springs receiving station only.
Orbits are composited into three periods in each calendar month - the 1st-10th, 11th-20th and 21st-end of the month - by the NDVI Maximum Value Composite (MVC) method. Besides greatly reducing data volumes, compositing improves spatial completeness and temporal consistency at individual locations. After rejecting pixels with solar zenith angle exceeding 80°, these three sub-monthly composites are further composited by maximum NDVI value to produce the one-month composites on a 0.01x0.01 degree grid. These are averaged to a 0.05x0.05 degree grid, excluding sea pixels.
Most of these NDVI and NDVI standardised anomaly maps are produced as both colour and black/white GIF images, with low and high resolution versions available in each case. The low resolution colour GIF images are the ones usually displayed, with links to the other three types placed under the main image. Place names are found on the high resolution versions. Portable Document Format (PDF) versions of the images are also generated for high-quality printing. Please note however that the PDF version is not archived for reasons of space. PDF versions of older maps may be obtained from the National Climate Centre via feedback form, but charges may be imposed for their provision.
The map projections used are either Cylindrical Equidistant (CE) or Lambert Conformal (LC). The Lambert Conformal projection takes three parameters; the central longitude (in degrees east of the Greenwich Meridian) and two standard parallels of latitude (in degrees south of the equator).
|Map projection||LC 134° 10°, 40°||CE||CE||LC 140.8° 10°, 40°||LC 146.5° 10°, 44°||CE||CE||CE|
The Victoria and Tasmania maps are based on a finer resolution analysis than the remaining maps. Consequently there may be slight inconsistencies in the detail represented on the Vic./Tas. maps as compared against the Aus./SA/NSW maps.
Monthly NDVI grids may be downloaded from the Bureau's website. These grids are in an ASCII format suitable for ingesting into geographic information systems (GISs), compressed using the UNIX compress utility. The ASCII grids have appended to them their original AIFS ASCII grid header (a Bureau of Meteorology grid format), to provide additional grid metadata. Note that some GISs may require the user to change the grid file extension from '.grid' to '.txt', prior to ingestion into the GIS.
Areas of no-data are also indicated in the legend. These are either inland water bodies or areas where no data is available due, for example, to persistent cloud cover or very low sun elevation.
Map and grid availability
Maps from April 1992 are available, except for a gap from October 1994 to January 1995.
Grids are available from July 2008 onwards. Earlier grids are not presently available.
NDVI is not an absolute measure of primary production. Also, due to the short period used for calculating the statistics, the NDVI standardised anomaly analysis may be less representative of normal conditions than analyses using rainfall, which use more than 100 years of records, compared to the 17-year NDVI record.
The use of the arid-site detrending technique to establish the long-term calibration consistency implies that while the temporal stability of these NDVI data is expected to be comparable to the best available to date for AVHRR NDVI timeseries covering Australia, the data cannot be relied upon to make absolute statements about long-term trends.
While the US NOAA satellite program aims for continuous quality coverage of the afternoon orbit, contingencies in the operation of the spacecraft and in the processing requirements have introduced periods of no coverage or reduced data quality. Specifically:
- Data commences in April 1992, and is contiguous, except for a gap from October 1994 to January 1995 due to the lack of a sensor following the unexpected end of NOAA-11 operation.
- Coverage and quality are reduced during the winters of 1993, 1994 and 2000 due to very low sun elevations (and consequent shadowing on the surface) during the last year or so of operation of some satellites.
- Maps for September 2003, and possibly October, November and December 2003, display some artefacts due to a NOAA-16 sensor scan-motor problem.