We compared four existing process-based stand-level models of varying complexity (physiological principles in predicting growth, photosynthesis and evapotranspiration, biogeochemical cycles, and stand to ecosystem carbon and evapotranspiration simulator) and a new nested model with 4 years of eddy-covariance-measured water vapor (LE) and CO2 (Fc) fluxes at a maturing loblolly pine forest. The nested model resolves the ‘fast’ CO2 and H2O exchange processes using canopy turbulence theories and radiative transfer principles whereas slowly evolving processes were resolved using standard carbon allocation methods modified to improve leaf phenology. This model captured most of the intraannual variations in leaf area index (LAI), net ecosystem exchange (NEE), and LE for this stand in which maximum LAI was not at a steady state. The model comparisons suggest strong linkages between carbon production and LAI variability, especially at seasonal time scales. This linkage necessitates the use of multilayer models to reproduce the seasonal dynamics of LAI, NEE, and LE. However, our findings suggest that increasing model complexity, often justified for resolving faster processes, does not necessarily translate into improved predictive skills at all time scales. Additionally, none of the models tested here adequately captured drought effects on water and CO2 fluxes. Furthermore, the good performance of some models in capturing flux variability on interannual time scales appears to stem from erroneous LAI dynamics and from sensitivity to droughts that injects unrealistic flux variability at longer time scales.