Determining the drivers of tree mortality in Amazonia is a complex task, yet essential to reliable prediction of carbon storage in a warmer climate. Past studies have shown an east-west gradient of forest disturbance and rainfall amount across Amazonia. This study uses remote sensing and dynamic vegetation modeling to take a deeper look at drivers of tree mortality and community composition shifts associated with varying mortality rates. Our analysis, using 20 years of Landsat 5 images, showed that ever-wet Amazonia (located in north-west Amazonia, i.e. NWA) was more susceptible to windthrows than the Central Amazon (i.e. CA), which has a a well-defined dry season. The higher frequency of windthrows in NWA forest correlates with higher frequency and intensity of deep convection events in this region, observed using Tropical Rainfall Measuring Mission (TRMM) data. While a combination of factors including: soil characteristics (and by proxy rooting depth) and community composition exacerbate the regional gradient of disturbance, wind was the mechanistic agent of disturbance. Using an individual-based gap model for tropical forests populated with the most representative NWA tree species and increased mortality rates, we found a decrease of biomass in this region and a slight increase in NPP compared to a control simulation, a pattern that is similar to observations. The model predicted which species had the largest response in basal area change due to elevated disturbance, but there was a non-significant shift in community composition in the NWA forests. However, analysis found strong differences in community composition between the modeled NWA and CA regions, consistent with observed results. When CA forests were subject to higher mortality rate similar to the current NWA region, dissimilarity in community composition continued to persist. In addition the model identified species-specific maximum tree height and maximum diameter as the most influential predictors of mortality, thus leading to species-specific susceptibility to windthrows. Including dynamic mortality mechanisms in Earth system models will allow us to examine changes in forest biomass turnover associated with wind-related tree disturbance. This need has become more relevant due to the projected increase in rainfall intensity in NWA.