Improving vegetation phenological parameterization of a land surface model

Abstract. The growing degree day (GDD) model and the growing season index (GSI) model are two common approaches used in various land surface models (LSMs) for simulating phenophases. The capacity of these two models for simulating phenolphases was evaluated by coupling them to a LSM (DLM: Dynamic Land Model) and validated by observation data from the 22 selected eddy covariance flux towers representing six typical plant functional types. The main findings are threefold: (i) the simulated phenophases using DLM-GSI were much closer to the observations derived from the green chromatic coordinate data than using DLM-GDD. The start of the growing season (SGS) was estimated to be earlier by DLM-GSI and later by DLM-GDD. Meanwhile, the end of growing season (EGS) was estimated to be later by DLM-GSI and earlier by DLM-GDD; (ii) compared to the GDD model, the GSI model significantly decreased the absolute bias of the phenophases simulated by DLM for all sites. The DLM-GSI model simulated biases for SGS and EGS decreased by 48.2 % and by 39 % on average, respectively; and (iii) the accuracy of modeled GPP using the DLM-GSI model is much higher than using the DLM-GDD model for all sites. The DLM-GSI model reduced the root mean square error of simulated GPP by 8.0 % and increased the corresponding index of agreement by 7.5 %.