Atmospheric models used for weather prediction and future climate projections rely on land models to calculate surface boundary conditions. Observations of near‐surface states and fluxes made at flux measurement sites provide valuable data with which to assess the quality of simulated lower boundary conditions. A previous assessment of the Community Land Model version 4.5 using data from the Niwot Ridge Subalpine Forest AmeriFlux tower showed that simulated latent heat fluxes could be improved by adjusting a parameter describing the maximum leaf wetted area, but biases in midday sensible heat flux and nighttime momentum flux were generally not reduced by model parameter perturbations. These biases are related to the model’s lack of heat storage in vegetation biomass. A biomass heat capacity is parameterized in Community Land Model version 5 with measurable quantities such as canopy height, diameter at breast height, and tree number density. After implementing a parameterization describing the heat transfer between the forest biomass and the canopy air space, the biases in the mean midday sensible heat and mean nighttime momentum fluxes at Niwot Ridge are reduced from 47 to 13 W/m2 and from 0.12 to −0.03 m/s, respectively. The bias in the mean nighttime canopy air temperature was reduced from −5.9 to 0.4 °C. Additional simulations at other flux tower sites demonstrate a consistent reduction in midday sensible heat flux, a lower ratio of the sum of sensible and latent heat flux to net radiation, and an increase in nighttime canopy temperatures.