Global MODIS fraction of green vegetation cover for monitoring abrupt and gradual vegetation changes

Abstract

The presence and distribution of green vegetation cover in the biosphere are of paramount importance in investigating cause-effect phenomena at the land/atmosphere interface, estimating primary production rates as part of global carbon and water cycle assessments and evaluating soil protection and land use change over time. The fraction of green vegetation cover (FCover) as estimated from satellite observations has already been demonstrated to be an extraordinarily useful product for understanding vegetation cover changes, for supporting ecosystem service assessments over areas with variable extents and for processes spanning a variable period of time (abrupt events or long-term processes). This study describes a methodology implemented to estimate global FCover (from 2001 to 2015) by applying a linear spectral mixture analysis with global endmembers to an entire temporal series of MODIS satellite observations and gap-filling missing FCover observations in temporal series using the DINEOF algorithm. The resulting global MODV1 FCover product was validated with two global validation datasets and showed an overall good thematic absolute accuracy (RMSE = 0.146) consistent with the validation performance of other FCover global products. Basic statistics performed on the product show changes in average and trend values and allow for the quantification of gross vegetation loss and gain over different temporal scales. To demonstrate the capacity of this global product to monitor specific dynamics, a multitemporal analysis was performed on selected sites and vegetation responses (i.e., cover changes), and specific dynamics resulting from cause-effect phenomena are briefly discussed. The product is intended to be used for monitoring vegetation dynamics, but it also has the potential to be integrated in other modeling frameworks (e.g., the carbon cycle, primary production, and soil erosion) in conjunction with other spatial datasets such as those on climate and soil type.