B51E-0068:
Stability and long term continuity of satellite data for rapid land surface monitoring of vegetation condition

Friday, 19 December 2014
Jesslyn F Brown and Daniel M Howard, USGS/EROS, Sioux Falls, SD, United States
Abstract:
Satellite-based normalized difference vegetation index, or NDVI, is a commonly used index in applications that require consistent and timely data, including monitoring drought, tracking vegetation phenological transitions, and assessing crop progress and condition. Although many other indices have been developed, the NDVI remains popular in the monitoring community. One reason is the value of NDVI for use in long-term studies necessitating multiple sensor data sources. It is calculated using a standard formula, red minus near-infrared divided by red plus near-infrared. But, this does not mean that all NDVI data are the same. Many factors, ranging from sensor design and raw satellite data ingest to initial data manipulation and post-processing, influence end-product quality and consistency. The purpose of this study was to perform a statistical scientific comparison between multiple NDVI data sources. The NDVI data analyzed in this study were derived from 8-day 250 meter (m) standard Moderate Resolution Imaging Spectroradiometer (MODIS) Terra and Aqua surface reflectance composites (Collection 5, MOD09Q1), 7-day 250 m expedited MODIS (eMODIS) Terra and Aqua NDVI products, and 14-day 1000 m Advanced Very High Resolution Radiometer (AVHRR) NDVI products of central U.S. over the 2003 – 2012 timeframe. Only composites falling within the growing season (between April and October) and temporally coincident were included. All composites were consistently post-processed using standard MODIS quality assurance data and evaluated to calculate spatial statistics (e.g. means and standard deviations) for 160 uniform 150 kilometer2 tiles. The results for seasonal, period-specific and overall correlations indicated the highest agreement was between standard MODIS Terra and Aqua composites, with a Pearson’s coefficient of determination of R2 = 0.98, and the lowest agreement was between eMODIS Terra and AVHRR, with a R2 value of 0.84. There was evidence of slight Terra sensor degradation within the time series.