Premium
Analytic expressions for output reliability: Verifying automated calibration and confidence calculations of a groundwater flow simulator
Author(s) -
Zaadnoordijk Willem Jan
Publication year - 2003
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/2002wr001932
Subject(s) - calibration , monte carlo method , mathematics , hydraulic head , computer science , statistics , algorithm , engineering , geotechnical engineering
Auxiliary material for this article contains a data set of the analytic formulae for the partial derivatives of the piezometric head with respect to the logarithm (base 10) of the resistance c and the hydraulic conductivity k. Additional file information is provided in the README.txt. ftp://ftp.agu.org/apend/wr/2002WR001932‐README.txt ftp://ftp.agu.org/apend/wr/2002WR001932‐derivatives.ps ftp://ftp.agu.org/apend/wr/2002WR001932‐derivatives.tex Analytic relations between input parameters, output variables, and statistics for confidence are derived for semiconfined groundwater flow in a circular domain around a well. The statistics of the two calibrated parameters are derived analytically using a least squares target function. The statistics of the piezometric head are approximated using linear variance analysis (LVA, also known as first‐order second moment method (FOSM)) and quadratic variance analysis (QVA). LVA values for the capture zone radius have been calculated based on numerical derivatives. The results are used to verify TrCalCon, the calibration and confidence module of the groundwater flow simulation package TRIWACO. The LVA and Monte Carlo (MC) results of TrCalCon compare well with the analytic LVA and QVA values, respectively. The problem is only weakly nonlinear in the heads, so that the LVA is accurate enough for practical purposes, and QVA or MC analysis would not be needed. The test can be applied as well to other groundwater simulation packages with a module for automatic parameter optimization and reliability analysis. As presented, the problem is quite simple, and the parameter optimization is relatively easy. The analysis can be expanded easily to make the problem more intricate by adding parameters, changing parameter values, and including other measurements.