Evaluating and Improving the Performance of Common Land Model Using FLUXNET Data

Abstract:

Common Land Model(CoLM), combined the best features of LSM, BATS, and IAP94, has been widely applied and shown its good quality. However, land surface processes are crucial for weather and climate model initialization, hence it’s necessary to constrain land surface model performances using observational data. In our preliminary work, eddy covariance measurements from 20 FLUXNET sites with over 100 site-years were used to evaluate CoLM while simulating energy balance fluxes in different climate conditions and vegetation categories. And the results show CoLM simulates well for all four energy fluxes, with sensible heat flux(H) better than latent heat flux(LE), net radiation (Rnet) the best. In terms of different vegetation categories, CoLM performs the best on evergreen needle-leaf forest among the 8 selected land cover types, and shows significant priority at evergreen broadleaf forest. Although a good agreement of simulation and observation is found on seasonal cycles at the 20 sample sites, it produces extreme bias mostly at summer noon, but not shows consistent bias among different seasons. This underestimate was associated with the weakness in simulating of soil water in dry seasons and incomplete description of photosynthesis as well, that’s why we will first focus on implementing mesophyll diffusion in CoLM to improve the physical process of photosynthesis.