Land subsidence model inversion with the structural noise of tectonic motion and unknown groundwater abstraction

The recent progresses in the monitoring techniques for land subsidence such as InSAR or GNSS provide the huge datasets available for the calibration of numerical models. However, the observed land surface displacement possibly includes the structural noise caused by the tectonic motion or land subsidence caused by unknown groundwater abstraction while the land subsidence model usually considers the effect only from the known groundwater abstraction. Then, the land subsidence model inversion is not straightforward because such structural noises may affect the inversion results if the observed land surface displacement is directly used for the calibration target.

The author proposes a new evaluation function based on the rotation energy of the estimated structural noise to avoid the interference from the structural noise in the inversion procedure. By replacing the typical evaluation function such as root mean square error with the proposed evaluation function, the inversion procedure can focus only on the reproducibility of the land subsidence caused by the known groundwater abstraction.

The performance of the proposed method was tested by the several synthetic land subsidence data composed of the displacement caused by the known and unknown groundwater abstraction calculated by the numerical model with the assumed physical parameter distribution, the typical tectonic motion, and the random noise. The proposed method successfully separated the land subsidence caused by the known groundwater abstraction and other components, and successfully reproduced the model parameters used in the numerical model to make synthetic land subsidence data.