A modified inverse procedure to calibrate parameters of land subsidence model and its field application in shanghai

Land subsidence prediction depends on a theoretical model and values of involved parameters. An advanced inverse model (called for InvCOMPAC) has been developed for calibrating five parameters (K'v, S'skv, S'ske, S'sk and p'max0) in a compacting confined aquifer system (Liu and Helm, 2008). InvCOMPAC combines two steps: graphic analysis for initial set of parameters and Newton-Raphson method for adjustment procedure. Both steps were modified to improve the procedure involved, so the modified inverse model presented in this study was called MInvCOMPAC. Some improved procedures were introduced in both steps. For graphic analysis, smoothed data was used to get regular smooth hydrographs, compaction history curves and stress-strain curves, which could reduce the uncertainties for initial values of parameters. For Newton-Raphson method, three procedures were introduced to improve robustness and practicability of Newton-Raphson method. First procedure is to remove outliers from actual data by using smoothed data. This procedure can reduce risk of failure of numerical optimization. Second procedure is to use modified Newton method instead of Newton-Raphson, which can improve robustness of numerical optimization. Third procedure is to develop variable scale factor in modified Newton method instead of fixed scale factor. This procedure can simplify the modified Newton method. As a case study, MInvCOMPAC was applied to obtain values of five parameters mentioned above for Shanghai land subsidence model. Results of MInvCOMPAC were evaluated by the fit of observed and calculated data, and also compared to those of InvCOMPAC. Reference: Liu Y, Helm DC. Inverse procedure for calibrating parameters that control land subsidence caused by subsurface fluid withdrawal: 2. Field application. Water Resour Res. 2008 Jul 31;44(7). Acknowledgements Funding for this research from 973 Program No. 2010CB428803, and from NSFC No. 40872155, 40725010 and 41030746, is gratefully acknowledged.