Vegetation demography, size structured plant competition, mechanistic mortality, and plant trait filtering control the carbon dynamics and energy budgets of the Earth's surface. Therefore, an emphasis has been placed on incorporating these types of structures and processes in the land modelling component of Earth System Models. These processes have now been introduced to the Advanced Climate Model For Energy's Land Model (ACME-LM or ALM) as derived from its sister model the Community Land Model (CLM), with the coupling of the Ecosystem Demography (ED) model. The goal of this study is to evaluate how ALM(ED) captures carbon cycling dynamics, using a specialized testbed approach that is tailored to the needs of a demographic trait filtering model. The portion of the testbed covered here, emphasizes model benchmarking using a constellation of sites approach. In this framework, model output is compared to observations that capture ecosystem properties across size and trait dimensions (e.g. wood density), as well as isolating ecosystem carbon fluxes via specific carbon pools. These detailed carbon flux measurements are typically associated with sites of focused study, sometimes thought of as observatories, which benefit from both flux measurements and copious in-situ measurements such as plant census. Model evaluation has been conducted near Manaus Brazil (ZF2), Iquitos Peru, the Congo and a temperate forest site in northern California, among others. These new types of model-observation inter-comparison have necessitated new code developments, which incorporated new model diagnostics that includes the Net Primary Productivity of specific carbon pools, plant mortality rates partitioned by mechanism and plant growth increments, each of which are distributed by size and plant functional type class (PFT). For the central Amazonian forest near Manaus Brazil, ALM(ED) does an adequate job of capturing the total community level fluxes of carbon, including gross primary productivity (GPP), NPP, NPP to wood and NPP to leaf pools. However, the model has shown variability and bias against observations related to the distribution of biomass with size. In particular, we have placed emphasis on understanding why the model overestimates biomass in the very largest size classes and underestimates biomass in nearly all other size classes in the canopy. This phenomenon manifests at the three tropical modelled sites, but not in the temperate forest, which fails to simulate any biomass in the largest size classes. Part of the bias towards large trees may be explained by model sensitivity to meteorological forcing data sets. When ALM(ED) was driven with site-specific flux-tower derived climate data, as opposed to global gridded products such as CRU-NCEP and Qian et al., the bias in biomass and growth rates towards the largest trees was reduced, better matching observations. In summary, while ALM(ED) and CLM(ED) both improve upon the large tropical biomass observed in CLM4.5, more diagnostics are needed to evaluate the bias towards persistent large trees in ALM/CLM(ED).