In order to evaluate the effect of parameters’ measurements to the uncertainty of groundwater flow, a conditional model for groundwater flow based on the Karhunen-Loeve decomposition and perturbation expansion is proposed, in which the spatial variability of log conductivity, recharge and boundary conditions are considered together. The derivation of the numerical modeling for groundwater flow subject to conditioning points of multiple random fields are also studied. The numerical experiments show that for some special covariance functions and lower-order Karhunen-Loeve expansion, the proposed model has high efficiency and satisfactory accuracy comparing to Monte Carlo simulation. The conditional simulation reduces the overall head variance. For the cases with larger correlation length and more conditional points, the reduction to the overall uncertainty is more significant. The variance reduction also depends on the locations of conditioning points. The conditioning points should be located in the region that may bring the largest reduction to the uncertainty. In reality, the measurement scheme should be determined by the real flow and type of parameters.
