Fig.1. The Taylor diagram showing (a) the performance of all model data from EC-towers were used.; (b-e) the performance of all model under different latitudes, where N and S represent the northern and southern hemispheres, respectively. Color dots represent the models in the corresponding legend. Taylor diagram is a polar graph in which the cosine of the angle between the X-axis is the correlation coefficient between the GPP of the model and EC-tower. The radial direction is the ratio of model to EC-tower GPP standard deviation. The grey arcs represent RMSE normalized by standard deviation for each model. The n is number of EC-towers.
The ability of a model to capture temporal changes is one of the keys to evaluating the performance of GPP models. The double mass curve (cumulative GPP predicted by model versus cumulative GPP observed at EC-towers) computed per site can reflect the ability of models to simulate the temporal changes of GPP (Fig.2). From the distribution of double mass curve, results show that, among the 10 models compared, the distribution cumulated GPP for the LUE-EF was the most concentrated around the 1:1 correspondence against the cumulated GPP from EC-towers, which indicates its greater ability to simulate the patterns in temporal variability in GPP. Fig. 2 also shows the distributions of relative bias in ratio (PB), which ranged ±0.4 for all models, indicating that the models had a large heterogeneity in simulating the temporal change of GPP. The biases were however narrower for the LUE-EF and LUE-NDWI models, respectively containing 120 and 110 of the EC-towers within ±0.2 in their PB, the largest number of EC-towers amongst all models. This indicates that these two models have a strong comprehensive ability to capture the temporal changes of GPP.
<Fig 2 roughly here>