Publications

Fagua, JC; Ramsey, RD (2019). Comparing the accuracy of MODIS data products for vegetation detection between two environmentally dissimilar ecoregions: the Choco-Darien of South America and the Great Basin of North America. GISCIENCE & REMOTE SENSING, 56(7), 1046-1064.

Abstract
The daily images produced by the MODIS (Moderate Resolution Imaging Spectroradiometer) sensor aboard the Terra and Aqua satellites have been widely used to monitor global vegetation. Using these data, the Earth Observing System operated by US National Aeronautics and Space Administration (NASA) has developed a variety of MODIS products focused on the monitoring and evaluation of vegetation condition. These products have three possible sources of variation that can affect the sensitivity of vegetation detection: 1) orbital and mechanical differences between MODIS sensors aboard Aqua or Terra, 2) the preprocessing algorithms used to generate multitemporal cloud-free mosaics (MAIAC or original MODIS algorithm), and/or 3) post-processing algorithms applied by users to optimize vegetation index values derived from temporal sequences of imagery. We evaluated these sources of variation by comparing the results of a vegetation classification for two different ecoregions. The accuracies of vegetation classifications utilizing either the Aqua or Terra MODIS sensors, the MAIAC or original MODIS preprocessing algorithms, and two common post-processing techniques (Asymmetric Gaussian or Savitzky and Golay function) were compared to determine which set of techniques or sensors yielded the best results. The ecoregions we chose to use were the Great Basin of North America and Choco-Darien of South America. We compared four different MODIS data products (MOD13Q1, MYD13Q1, MOD09Q1, and MYD09Q1) as predictor variables using Random Forest as the classification algorithm to generate a land cover map. We found that the accuracy of the vegetation classifications (using Kappa as measure of accuracy) changed significantly depending on the MODIS platform (Terra or Aqua), the preprocessing algorithm (MAIAC or MODIS), and the two postprocessing algorithms for both ecoregions. Our result suggests that comparative analyses are needed to optimize the results when equivalent MODIS products are used in vegetation detection and classification.

DOI:
10.1080/15481603.2019.1611024

ISSN:
1548-1603