Eddy covariance records hold great promise for understanding the processes controlling the net ecosystem exchange of CO2 (NEE). However, NEE is the small difference between two large fluxes: photosynthesis and ecosystem respiration. Consequently, separating NEE into its component fluxes, and determining the process-level controls over these fluxes, is a difficult problem. In this study, we used a model-data synthesis approach with the Simplified PnET (SIPNET) flux model to extract process-level information from 5 years of eddy covariance data at an evergreen forest in the Colorado Rocky Mountains. SIPNET runs at a twice-daily time step, and has two vegetation carbon pools, a single aggregated soil carbon pool, and a soil moisture submodel that models both evaporation and transpiration. By optimizing the model parameters before evaluating model-data mismatches, we were able to probe the model structure independent of any arbitrary parameter set. In doing so, we were able to learn about the primary controls over NEE in this ecosystem, and in particular the respiration component of NEE. We also used this parameter optimization, coupled with a formal model selection criterion, to investigate the effects of making hypothesis-driven changes to the model structure. These experiments lent support to the hypotheses that (1) photosynthesis, and possibly foliar respiration, are down-regulated when the soil is frozen and (2) the metabolic processes of soil microbes vary in the summer and winter, possibly because of the existence of distinct microbial communities at these two times. Finally, we found that including water vapor fluxes, in addition to carbon fluxes, in the parameter optimization did not yield significantly more information about the partitioning of NEE into gross photosynthesis and ecosystem respiration.