Influence of using date-specific values when extracting phenological metrics from 8-day composite NDVI data

Abstract

Information of vegetation dynamics derived from remotely sensed data is essential for regional natural resource management. A common approach involves using an n-day composite time-series of normalized difference vegetation index (NDVI) satellite data for estimating vegetation phenology. The composite method was introduced to deal with problems associated with cloudy and noisy data but could possibly obscure the fine-scale timing of vegetation processes like leaf out and leaf drop, which can occur suddenly, often within a few days. Moreover, different methods for smoothing the NDVI curves also influence the accuracy of the vegetation parameters extracted from them. The auhtors investigate the difference between 8-day equalinterval (standard composite data) and 8-day date-specific NDVI data in their ability to extract phenological metrics of interest for ecologists and land managers. They also compare two different filtering algorithms - the Savitsky-Golay and the Gaussian filtering algorithms. Results from a typical savanna system in South Africa show that the Savitsky-Golay technique with date-specific NDVI values is the best method