TY - JOUR
T1 - A Post audit and inverse modeling in reactive transport: 50 years of artificial recharge in the Amsterdam water supply dunes.
AU - Karlsen, R.H.
AU - Smits, F.J.C.
AU - Stuijfzand, P.J.
AU - Olsthoorn, A.A.
AU - van Breukelen, B.M.
PY - 2012
Y1 - 2012
N2 - This article describes the post audit and inverse modeling of a 1-D forward reactive transport model. The model simulates the changes in water quality following artificial recharge of pre-treated water from the river Rhine in the Amsterdam Water Supply Dunes using the PHREEQC-2 numerical code. One observation dataset is used for model calibration, and another dataset for validation of model predictions. The total simulation time of the model is 50years, from 1957 to 2007, with recharge composition varying on a monthly basis and the post audit is performed 26years after the former model simulation period. The post audit revealed that the original model could reasonably predict conservative transport and kinetic redox reactions (oxygen and nitrate reduction coupled to the oxidation of soil organic carbon), but showed discrepancies in the simulation of cation exchange. Conceptualizations of the former model were inadequate to accurately simulate water quality changes controlled by cation exchange, especially concerning the breakthrough of potassium and magnesium fronts. Changes in conceptualization and model design, including the addition of five flow paths, to a total of six, and the use of parameter estimation software (PEST), resulted in a better model to measurement fit and system representation. No unique parameter set could be found for the model, primarily due to high parameter correlations, and an assessment of the predictive error was made using a calibration constrained Monte-Carlo method, and evaluated against field observations. The predictive error was found to be low for Na
AB - This article describes the post audit and inverse modeling of a 1-D forward reactive transport model. The model simulates the changes in water quality following artificial recharge of pre-treated water from the river Rhine in the Amsterdam Water Supply Dunes using the PHREEQC-2 numerical code. One observation dataset is used for model calibration, and another dataset for validation of model predictions. The total simulation time of the model is 50years, from 1957 to 2007, with recharge composition varying on a monthly basis and the post audit is performed 26years after the former model simulation period. The post audit revealed that the original model could reasonably predict conservative transport and kinetic redox reactions (oxygen and nitrate reduction coupled to the oxidation of soil organic carbon), but showed discrepancies in the simulation of cation exchange. Conceptualizations of the former model were inadequate to accurately simulate water quality changes controlled by cation exchange, especially concerning the breakthrough of potassium and magnesium fronts. Changes in conceptualization and model design, including the addition of five flow paths, to a total of six, and the use of parameter estimation software (PEST), resulted in a better model to measurement fit and system representation. No unique parameter set could be found for the model, primarily due to high parameter correlations, and an assessment of the predictive error was made using a calibration constrained Monte-Carlo method, and evaluated against field observations. The predictive error was found to be low for Na
U2 - 10.1016/j.jhydrol.2012.05.019
DO - 10.1016/j.jhydrol.2012.05.019
M3 - Article
VL - 454-455
SP - 7
EP - 25
JO - Journal of Hydrology
JF - Journal of Hydrology
SN - 0022-1694
ER -
