Dynamic vegetation models (DVM), which combine representations of biogeochemical processes with representations of processes contributing to the dynamics of vegetation structure and composition, were introduced to simulate vegetation change in global climate models. They are coupled to climate models for simulating past, present, and future vegetation cover distribution. Although this type of models normally work on very long time scales, some preliminary studies indicate DVMs also have impact on simulation of water and energy cycle at seasonal to interannual scales.

However, thus far, DVMs were mostly evaluated for their simulations of vegetation distribution and carbon flux, or for model intercomparison. Sometimes, model-simulated precipitation, surface temperature, or tree ring data was used to indirectly evaluate DVM performance. The systematic testing of this type of models for water and energy cycle studies at seasonal, interannual, and decadal scales using station data has not been reported so far. The vegetation simulated by DGVMs responds primarily to for factors: solar radiation, air temperature, and soil moisture, and atmospheric CO2 concentration. Therefore, an objective, meaningful evaluation should include every important component in the surface water, carbon, and energy cycle, not be based on one or two variables. It is necessary to validate the DVM-simulated variability using observational data to understand the uncertainty involved in DVM application and to understand how to apply DVMs for climate simulation.

In this study, we test the coupled biophysical model (SSiB-4, Simplified Simple Biosphere Model, version 4) with a dynamic vegetation model (TRIFFID, Top-down Representation of Interactive Foliage and Flora Including Dynamics). Selected sites with different landscape and climate over North American continent are tested using 50-year specified meteorological forcing from observation and reanalysis. How the four factors and initial condition control the vegetation dynamics, especially the simulation of the vegetation parameters with strong seasonal and interannual variability, such as leaf area index (LAI), is the main focus. The uncertainty in DVM simulation and how to apply observed data to reduce the uncertainty are also investigated. Meanwhile, the simulation with SSiB2 without TRIFFID is also carried out to compare with SSiB4/TRIFFID. Observational data from FLUXNET are applied as constrains for these sensitivity studies.