Bayesian inference of the groundwater depth threshold in a vegetation dynamic model: A case study, lower reach, Tarim River

Abstract The responses of eco-hydrological system to anthropogenic and natural disturbances have attracted much attention in recent years. The coupling and simulating feedback between hydrological and ecological components have been realized in several recently developed eco-hydrological models. However, little research has been carried out to study the estimation of threshold parameters in the eco-hydrological models. The aim of this paper is to infer the groundwater threshold parameter that connects the ecological and hydrological processes inside eco-hydrological models. A simplified vegetation dynamic model was set up, and used to simulate the vegetation dynamic along the lower reach of the Tarim River. Bayesian inference approach was applied with Markov Chain Monte Carlo sampling method used to infer the probability distribution of the groundwater threshold parameter. The result showed that the simplified model has the capacity to model the vegetation dynamic in the study area. The mean values of inferred groundwater threshold parameters for Yinsu, Kardayi, Alagan, and Yiganbjima are 4.99, 5.59, 5.74, and 5.85 m respectively, which are comparable with prior research, and showed significant spatial variability. The 90% confidence interval of the groundwater threshold parameter is larger than 1.27 m. This threshold parameter inference approach was suggested to be carried out before complex eco-hydrological modeling is performed.