Amazon forest accounts for around 40% of carbon stock in global tropical forests, playing a center role in uptaking anthropogenic carbon emission and mitigating global warming. Recent studies suggest that climate change has led Amazonia more susceptible to forest loss due to various drivers, and extreme storms is an important yet overlooked mode of tree mortality. Extreme storms produce downbursts and heavy rainfalls that may uproot or snap the trees. Field observations showed that extreme storms can explain ~50% of observed carbon residence time across the Amazon. In order to better understand how extreme storms influence forest dynamics, we used ED2 model to simulate forest growth in Amazonia with different settings of woody residence time (τ_w). We evaluated how spatial variability of τ_w derived from the relationship with extreme storms shaped modeled patterns of aboveground biomass (AGB) and productivity. We found that with spatially varying τ_w the model simulated lower AGB and stand-level wood density in western Amazon but higher AGB and stand-level wood density in eastern Amazon, which is closer to observed spatial gradients. Specifically, the model with spatially varying τ_w simulated larger proportion of early successional trees but smaller proportion of late-successional trees in northwestern Amazon where more frequent extreme storms happened, confirming the effect of this mortality on ecosystem processes. In addition, spatially varying τ_w enabled the model to simulate higher productivity and mortality in western but lower productivity and mortality in eastern Amazon, which are more comparable to observations, although current model parameterization tended to overestimate those rates. Our long-term simulations also predict larger carbon stock in Amazon forest at century-scale with spatially varying τ_w, which suggested the necessity of addressing the changes of extreme storms and feedbacks on ecosystem processes when predicting future climate influence on global carbon budget.