| The acquisition of remote sensing data having an investigated quality level constitutes an important step to advance our understanding of the vegetation response to environmental factors. Spaceborne sensors introduce additional challenges that should be addressed to assure that derived findings are based on real phenomena, and not biased or misguided by instrument features or processing artifacts. As a consequence, updates to incorporate new advances and user requirements are regularly found on most cutting edge systems such as the Moderate Resolution Imaging Spectroradiometer (MODIS) system. In this dissertation, the objective was to design, characterize and assess any possible departure from current values, a MODIS vegetation index (VI) algorithm for restoring the continuity 16-day 1-km product, based on the new 8-day 500-m MODIS surface reflectance (SR) product scheduled for the forthcoming MODIS Collection 6 (C6). Additionally, the impact of increasing the time resolution (by reducing the compositing period) from 16 to 8 days for the future basic MODIS C6 VI product was also assessed. The performance of the proposed algorithm was evaluated using high quality reference data and known biophysical relationships at several spatial and temporal scales. Firstly, it was evaluated using data from the AERONET-based Surface Reflectance Validation Network (ASRVN), FLUXNET-derived ecosystem gross primary productivity (GPP) and an analysis of the seasonality parameters derived from current Collection 5 (C5) and proxy C6 VI collections. The performance of the 8-day VI version was evaluated and contrasted with current 16-day using the reported correlation of the Enhanced Vegetation Index (EVI) with the GPP derived from CO2 flux measurements. Secondly, we performed an analysis at spatial level using entire images (or "tiles") to assess the Bidirectional Reflectance Distribution Function (BRDF) effects on the VI product, as these can cause biases on the SR and VIs from scanning radiometers. Lastly, we evaluated the performance of the proposed algorithm for detecting inter-annual VI anomalies from long-term time series, as compared with current MODIS VI C5. For this, we analyzed the EVI anomalies from a densely vegetated evergreen region, for the period July--September (2000--2010). Results showed a high general similarity between results from both algorithms, but also systematic differences, suggesting that proposed algorithm towards C6 may represent an advance in the reduction of uncertainties for the MODIS VI product. |
