Numerical Procedures for Subsidence Data Processing

NTRODUCTION Surface subsidence monitoring is the most effective means for investigating the nature of, and developing the techniques for pre¬dicting the surface movement and deformation process. However, because of imperfection in survey equipment and field practices, the surface movement data collected from field survey are not error-free. Also due to lack of proper numerical techniques for computing surface deformation from the measured surface move¬ment data, a relatively large portion of the subsidence data col¬lected could not provide information other than surface movement. In a paper by Luo and Peng (1992), the sources and magnitude of errors in the measured surface movements using an electronic total station in subsidence survey are discussed. If the errors in the measured surface movement components are left untreated before calculation for deformation indices (i.e., slope, strain and curva¬ture), the errors will be carried over to the deformation indices and enlarged if improper numerical procedures are used in the calcu¬lation. It should be noticed that in protecting structures affected by ground subsidence, accurate surface deformation indices are much more important than surface movement. In order to improve the accuracy of the collected surface movement data and the calcu¬lated deformation indices, a number of numerical procedures are also proposed in the same paper. The proposed techniques include: a data smoothing technique for data pre-processing; and a number of numerical methods and/or procedures for calculating surface deformation indices from the measured surface movement data. DATA SMOOTHING Because of various factors, the measured surface movement data always contain errors, small or large. In order to better utilize the collected data, a data smoothing method is proposed to correct the errors in the measured movement data. The method uses the same concept as the influence function method (e.g., Knothe, 1957) on which many existing subsidence prediction methods are based. The data smoothing method makes necessary corrections on the measured movement for the point of interest according to not only the measured movement at this point but also those at the surrounding points. The contribution of a surrounding point in making the correction for the point of interest is inversely propor¬tional, in exponential form, to the distance between them. The new magnitude of a movement component at the point of interest (Zi*) is calculated by