Phytoplankton community structure strongly shapes marine primary productivity and carbon export. While ecosystem models traditionally simplify this complexity using a fixed set of functional types, this approach may fail to capture community shifts under abrupt disturbances or long-term climate changes, leading to biased predictions. Trait-based models offer a dynamic alternative but are often computationally prohibitive for high-resolution or global-scale applications. To evaluate the practical implications of this trade-off, we compare a trait-based model against a fixed functional type model (NPZD framework) under scenarios of episodic bloom events and climate change. Both one-dimensional physical-biogeochemical models, constrained by BGC-Argo observations, are configured for the strongly stratified tropical North Atlantic. We perturb environmental forcings to simulate disturbances and climate change, analyzing divergent responses between the models. Specifically, we assess: (1) how predictions of productivity and export differ under stress; (2) whether surface and deep phytoplankton communities exhibit contrasting responses to climate forcing in stratified regions; and (3) if temporally varying parameters, derived from the trait-based simulation, can be used to improve the fidelity of the simpler NPZD-type model. Our findings aim to identify critical limitations of fixed-community models and develop strategies to enhance their predictive power for a changing ocean.