Accuracy analysis in CH4MODwetland in the simulation of CH4 emissions from Chinese wetlands

Process-based models have been widely used to simulate the CH4 budget in natural wetlands, but there are still large uncertainties in the simulation processes. Accuracy analysis of process-based models is important to evaluate the reliability of model estimates for different wetlands. In this study, we analyzed the three sources of the model bias from a process-based model (CH4MODwetland) when simulating CH4 emissions from different wetland sites and types in China. On a national scale, the root-mean-square error (RMSE) decreased from 70.6% at daily scale to 27.1% at annual/seasonal scale. On a national scale, more than 90% of the model errors were from random error. At most of the sites random error contributed more than 60% to the model errors, except the Guangzhou, Haikou, Wuliangsu lake and Yancheng Estuary sites, which showed lower RMSE values. The model simulated higher RMSE at the tidal marsh sites. The model had capability to simulate the daily and annual/seasonal CH4 emissions from all wetland types (model efficiency > 0), and performed better for coastal wetlands than peatlands and marshes, with model efficiency values of 0.75 and 0.99, respectively. When applied the model at annual/seasonal scale for peatlands, marshes, and coastal wetlands, the RMSE decreased a lot. The study indicated that CH4MODwetland is more accurate at annual/seasonal scale than daily scale for different wetland types across China. The non-stochastic portion of CH4 emissions from peatlands, marshes and coastal wetlands can be well predicted by the model.