This paper presents several methods which are based on multiple data sets to reduce the errors caused by the noise in the measured data. The comparisons show that the accuracy of inverse solution depends both on the noise level and on the number of consecutive observations. For low noise levels, both average and least squares methods perform well. When the noise level is high, integration methods based on the trapezoidal rule yield better accuracy. However, when noise dominates the record, all methods may yield an unacceptable error. To reduce noise levels, a Butterworth filter is used. Using filtered data, the accuracy of the estimated parameters is improved. A computed example shows that the errors in transmissivity and storage coefficient are different because they have different derivatives. Application of the inverse methods are demonstrated in a two-dimensional problem.
All Science Journal Classification (ASJC) codes
- Water Science and Technology