This preview shows page 1. Sign up to view the full content.
Unformatted text preview: × 300 1000 × 1000 Δxi (discretization cell size) (m) 10/3 1 Li (domain length) (m) 1000 1000 λ 100/3 10 Δxi/ λ (grid resolution) 1/10 1/10 Li/ λ (sampling number) 30 100 ΔH/Lx (mean gradient) 1/1000 1/1000 σ (standard deviation of lnK) 1, (lnk- correlation scale) (m) 2 , 1.6, 2, ln(10) 1, 2 , 1.6, 2, ln(10) A statistically isotropic log-normal random field K(x,y) was generated on a one million node grid
(1000 × 1000 cells), with either smooth (gaussian) or noisy (exponential) covariance structure. A
good fit was obtained when comparing theoretical vs computed spatial autocorrelation function of
lnK(x,y) for the gaussian covariance with σlnK = 1 on a 1000 × 1000 grid. Smaller 300 300 fields
were then extracted from the center of the domain, and single replicate simulations of seawater
intrusion were conducted, with variability ranging from σlnK = 1 to ln10.
In order to increase numerical accuracy, we used an iterative continuation method (or homotopy
method) with respect to the σ parameter, where σ is the standard deviati...
View Full Document
This document was uploaded on 01/19/2014.
- Winter '14
- The Land, In French