Methods

Early efforts have resulted in two 14-day map products now being offered through the USGS map service, Integrated Vegetation Mapping http://ivm.cr.usgs.gov/viewer. The satellite-derived products are the Seasonal Greenness and the Percent of Average Seasonal Greenness.

The Seasonal Greenness (SG) (Figure 1) is a measure of growing season vegetation vigor or performance derived from the Normalized Difference Vegetation Index (NDVI). SG, now calculated on a 14-day basis, is based on the smoothed NDVI data distributed as part of Phenological Characterization [smoothed NDVI data]. SG is an integrated measure of NDVI from the start of the growing season to the time of observation. (Figure 2) The integration is performed on simulated daily NDVI. The Percent of Average Seasonal Greenness (PASG) is then calculated for each pixel using the following formula: (Current Observation SG / Mean SG) * 100 (Figure 3).

Areas that are above and below the mean SG are highlighted by this calculation. Areas below the mean may result from a variety of influences including standing water, drought, deforestation, or urbanization. Climate data and drought indicators are necessary to determine where drought is causing patterns of below average SG.

Previous studies have established significant relationships between satellite-derived vegetation indices and climate variables mainly over non-irrigated crops and grasslands. This drought project builds upon this research by investigating methods to integrate the information from climate-based drought indicators with satellite measures of vegetation greenness derived from multi-temporal NDVI data.

We are researching the use of regression tree analysis for determining the relationships between drought indicators (such as the Standardized Precipitation Index and the Palmer Drought Index) and the SD and PASG. Regression tree analysis is a data mining technique ideally suited for the analysis of complex data. The technique is flexible and robust, deals with non-linear relationships, and is fairly simple to understand.

Early results and data for our pilot study area should be available in 2003.