Abstract
Spectroradiometer) images for ecosystem monitoring is currently widespread both in research and management. Vegetation indexes (VIs), such as NDVI (Normalized Difference Vegetation Index) and EVI (Enhanced Vegetation Index), are broadly extended for monitoring ecosystem functioning. These indexes are linear estimators of the fraction of photosynthetically active radiation intercepted by vegetation (fAPAR), the main control of net primary production. However, VIs are subject to errors. To handle such errors, the MOD13Q1 VI product includes a Quality Assessment (QA) layer with information about pixel quality. This QA layer represents a great advantage for final users, allowing filtering of pixels with VI values influenced by aerosols, clouds, snow, or shadows. However, the use of homogeneous filtering criteria throughout a heterogeneous region may cause the systematic loss of information in particular areas or times of the year. In this paper, we assessed the effect of different filtering criteria on spatiotemporal data of EVI for the period 2001-2010 in the Iberian Southeast. Our results showed no effect of filtering on EVI availability and magnitude values in low altitudes, but strong and significant differences in the mountains depending on the filter applied: aerosol, shadows or snow filters. Such effects of filtering on the EVI revealed that monitoring programs in these regions should include a filtering step before exploring for abrupt changes or longterm trends in the EVI time series.
