Adjusting for Long-Term Anomalous Trends in NOAA's Global Vegetation Index Data Sets

Abstract

The weekly 0.144 resolution global vegetation index from the National Oceanic and Atmospheric Administration (NOAA) National Environmental Satellite, Data, and Information Service (NESDIS) has a long history, starting late 1981, and has included data derived from Advanced Very High Resolution Radiometer (AVHRR) sensors onboard NOAA-7, -9, -11, -14, -16, -17, and -18 satellites. Even after postlaunch calibration and mathematical smoothing and filtering of the normalized difference vegetation index (NDVI) derived from AVHRR visible and near-infrared channels, the time series of global smoothed NDVI (SMN) still has apparent discontinuities and biases due to sensor degradation, orbital drift [equator crossing time (ECT)], and differences from instrument to instrument in band response functions. To meet the needs of the operational weather and climate modeling and monitoring community for a stable long-term global NDVI data set, we investigated adjustments to substantially reduce the bias of the weekly global SMN series by simple and efficient algorithms that require a minimum number of assumptions about the statistical properties of the interannual global vegetation changes. Of the algorithms tested, we found the adjusted cumulative distribution function (ACDF) method to be a well-balanced approach that effectively eliminated most of the long-term global-scale interannual trend of AVHRR NDVI. Improvements to the global and regional NDVI data stability have been demonstrated by the results of ACDF-adjusted data set evaluated at a global scale, on major land classes, with relevance to satellite ECT, at major continental regions, and at regional drought detection applications.