Global sensitivity analysis of parameters in land subsidence model

Yue LUO,Shu-jun YE,Ji-chun WU,Yan-hong ZHANG,Xun JIAO,Han-mei WANG
DOI: https://doi.org/10.3785/j.issn.1008-973X.2018.10.021
2018-01-01
Abstract:Sobol’ method was used to conduct a global sensitivity analysis of five parameters in the land subsidence model of Shanghai city in order to reduce the model uncertainty and enhance efficiency of the calibration (or inversion) on land subsidence model. The parameters were as follows: Kv (vertical hydraulic conductivity of the aquitards), Sskv (aquitard nonrecoverable skeletal specific storage), Sske (aquitard recoverable skeletal specific storage), PPmax (initial preconsolidation stress within the aquitards) and Ssk (skeletal specific of the aquifer). Total sensitivity, the first-order sensitivity and interaction sensitivity indexes were calculated on the response variable of mean square error (MSE). Then the impact of the parameter range and time series result simulated by land subsidence model on sensitivity analysis was discussed. 1) Ssk shows low sensitivity, which implies its less influence on simulated deformation. Ssk can be fixed at a given value when the model is calibrated or inversed. 2)Parameters sensitivity in different deformation phases shows different characteristics: Sske shows the most influence on deformation during elastic deformation phase; Kv, PPmax and Sskv show the most influence on deformation during plastic deformation phase. This implies the parameters in land subsidence model can be calibrated by each of two deformation phases. 3) The interaction sensitivity of parameters shows high value. This implies that the model may encounter equifinality when it is calibrated. The possibility of equifinality can be decreased when the range of PPmax is reduced for sensitivity analysis by analyzing the curve of groundwater level and deformation.
What problem does this paper attempt to address?