USING PARAMETER SENSITIVITY TO INVESTIGATE TRANSMISSIVITY DISTRIBUTION INFORMATION IN TRANSIENT OBSERVATION WELL DATA

Estimates of the spatial variation of transmissivity from late-time drawdown response (Bohling et al.,2002) or temporal moments of the drawdown (Zhu and Yeh, 2005) are similar to estimates made from the entire drawdown curve. The analysis of the spatial sensitivity of drawdown to transmissivity variation, presented below, explores why late-time data is sufficient. The analysis also provides insight into where estimates of transmissivity are expected to be more reliable and provide a quantitative basis for evaluating the measurement support volume of observation well data. The analysis also suggests that very early drawdown data, if reliable, could improve estimates near the wellbore and along the direct path connecting the observation and pumping locations. The analysis technique could be used to the design of hydraulic tomography experiments.

The sensitivity of drawdown to a change transmissivity of each model cell is calculated by a version of MODFLOW-2005 that has been modified to use the adjoint state method. A brief review of the capabilities of this MODFLOW version will be presented. The sensitivity data is collected into a matrix; one column per cell, one row per drawdown measurement. Redundancy of measurement data is examined by comparing rows of the matrix and through Singular Value Decomposition. Spatial information content is examined by calculating a resolution matrix using the Singular Value Decomposition. The results are presented graphically.