Seasonal patterns of tropical evergreen forest green-up in Amazonia, corresponding to drought and the dry season, have recently been detected by the Enhanced Vegetation Index (EVI) and Leaf Area Index (LAI) products of the Moderate Resolution Imaging Spectroradiometer (MODIS) satellite sensor. These observations provide additional evidence for solar radiation as the primary limiting factor regulating wet-tropical ecosystem processes. However, in situ structural mechanisms for forest canopy green-up are unclear and frequently inconsistent with observations. Here, we investigate the signal of seasonal green-up at several intensively measured sites, applying a rigorous series of filters to minimize error from atmospheric contamination that is common in tropical regions. We find that, while satellite-observed forest seasonality is sensitive to data-quality filtering, statistical noise reduction and spatial averaging, the signal is robust at sites where field observations are available, and in particular for the EVI. For the sites where field data are unavailable, it appears that additional filters to those commonly used to remove cloud effects and aerosols also reduce the seasonal magnitude of the LAI. These findings imply that seasonal tropical evergreen forest green-up remains sensitive to the methodology used in removing seasonal contamination from atmospheric conditions and that further field measurements and comparisons to remote sensing are required to reduce this uncertainty.