Vegetation Health Product - Algorithm Description
One of the most important long-term (30-year) satellite-based data records characterizing land surface, air temperature near the ground and climate was created from the Advanced Very High Resolution Radiometer (AVHRR) flown on the National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites. Several global data sets have been developed from the AVHRR records since the early 1980s. They were the NOAA's Global Vegetation Index (GVI and GVI-2), National Aeronautics and Space Administration (NASA)'s Pathfinder and GIMMS (Tarpley et al 1984, James and Kalluri 1994, Kidwell 1997, Tucker et al 2004). These datasets focused only on the Normalized Difference Vegetation Index (NDVI), ignoring infrared measurements, which are very useful for monitoring land, climate and socioeconomics. Therefore, NOAA developed a new dataset entitled the Global Vegetation Health Product. The VHP has advantages over other long-term global data sets, being the longest (30-year), having the highest spatial resolution (4-km), containing, in addition to NDVI, data and products from infrared channels, originally observed reflectance and emission, many indices with suppressed noise, biophysical climatology and more importantly, products used for monitoring environmental and socioeconomic activities (Kogan 1995, 1997).
Both NDVI and BT data should be processed to minimize the impact of cloud contamination (by weekly Maximum NDVI value compositing), reduce the short-term noise (by time series smoothing) and correct for the difference of sensors on board the NOAA series satellites and long term noise (by applying EDF adjustment and linear adjustment algorithm). Detail data processing and noise reducing algorithm can be found in section 3 of VHP ATBD. (define what this is and state where to find it.)
After noise removal, weather-driven differences in NDVI and BT between the years become apparent: lower NDVI and higher BT in dry years and opposite in normal and wet years. This principle of comparing NDVI and BT for a particular year with their dry-wet range calculated from 30-year observations was laid down in the VHP algorithm development. The absolute maximum and minimum of NDVI and BT during 1981-2005 were calculated for each of the 52 weeks and for each pixel. They were then used as the criteria to estimate the upper (favorable weather) and the lower (unfavorable weather) limits of the ecosystem resources. Further, for estimation of weather impacts on vegetation condition, NDVI and BT values for a particular time were normalized relative to the absolute max/min interval. Following this procedure, NDVI and BT were rescaled based on equations (2.1-2.3). They were named the Vegetation Condition Index (VCI), Temperature Condition Index (TCI) and Vegetation Health Index (VHI) designed to characterize moisture (VCI), thermal (TCI) and total vegetation health (VHI) conditions in response to weather impacts
where NDVI, NDVImax, and NDVImin (BT, BTmax, and BTmin) are the smoothed weekly NDVI (BT), their multi-year absolute maximum, and minimum, respectively. The VCI, TCI and VHI approximate the weather component in NDVI, BT and their combination values. They fluctuate from 0 to 100, reflecting changes in vegetation conditions from extremely bad to optimal. The weighting factor (a) in equation 2.3 was determined by experience, currently, a=0.5.
Due to the limitation of climatology, based on equation 2.1, it is possible to have VCI <0. When this occurs, VCI is reset as 0 to indicate the worst situation on vegetation condition. Similarly, if VCI >100, VCI is reset as 100. These situations indicate real variation outside of the climatological range, and do not mean the VCI has poor quality. VCI data quality is mainly determined by NDVI quality. If NDVI is contaminated by cloud, over snow/ice or over desert, VCI will be in poor quality. The quality flags stored in Quality Assurance (QA) dataset in VH file records if the pixel is over desert, on coastal line, with too cold temperature (snow or ice). Users may use QA flag to further determine how to use the VCI product. The same procedure is applied to TCI.
Further information may be found in the Algorithm Theoretical Basis Document.