The resilience of tropical forests and their ecosystem functioning under climate change hinges on the responses of forest mortality, which strongly regulates the length of time that carbon remains in forests. However, environmental drivers of tropical forest carbon turnover time remain unclear, largely due to lack of spatial data for woody carbon fluxes. Here, we demonstrated that temporal autocorrelation of high-frequency variability in long-term satellite observations is a robust predictor of forest biomass mortality across 102 Amazonian field sites (leave-one-out cross-validation R² = 0.63, RMSE = 2.34 Mg ha^-1 yr^-1). Based on the satellite-derived mortality maps, we estimated aboveground carbon turnover time varies from 14 to 79 years (0.01 and 0.99 percentiles) across intact Amazon forests with a median of 45 years. We found that spatial variations in carbon turnover time can be largely explained by environmental factors (R² = 0.69) and storm-related disturbance risk is the primary driver of carbon turnover time variations, followed by elevation, heat and water stress. Our results also highlighted the nonlinear response of carbon turnover time to these environmental drivers, which should be considered when predicting future carbon cycling in the tropics.