Sensitivity analysis and parameter estimation for an approximate analytical model of canal-aquifer interaction applied in the C-111 basin

I. Kisekka, K. W. Migliaccio, R. Muñoz-Carpena, Y. Khare, Treavor Boyer

Research output: Contribution to journalArticle

12 Citations (Scopus)

Abstract

The goal of this study was to better characterize parameters influencing the exchange of surface water in south Florida's C-111 canal and Biscayne aquifer using the analytical model STWT1. A three-step model evaluation framework was implemented as follows: (1) qualitative parameter ranking by comparing two Morris method sampling strategies, (2) quantitative variance-based sensitivity analysis using Sobol's method, and (3) estimation of parameter posterior probability distributions and statistics using the Generalized Likelihood Uncertainty Estimator (GLUE) methodology. Results indicated that the original Morris random sampling method underestimated total parameter effects compared to the improved global Morris sampling strategy. However, parameter rankings from the two sampling methods were similar. For the STWT1 model, only four out of the six parameters analyzed were important for predicting water table response to canal stage and recharge fluctuations. Morris ranking in order of decreasing importance resulted in specific yield (ASY), aquifer saturated thickness (AB), horizontal hydraulic conductivity (AKX), canal leakance (XAA), vertical hydraulic conductivity (AKZ), and half-width of canal (XZERO). Sobol's sensitivity indices for the four most critical parameters revealed that summation of first-order parameter effects was 1.0, indicating that STWT1 behaved as an additive model or negligible parameter interactions. We estimated parameter values of 0.07 to 0.14 for ASY, 11,000 to 14,300 m d-1 for AKX, 13.4 to 18.3 m for AB, and 99.8 to 279 m for XAA. The estimated values were within the range of values estimated using more complex methods at nearby sites. The Nash-Sutcliffe coefficient of efficiency and root mean square error for estimated parameters ranged from 0.66 to 0.95 and from 4 to 7 cm, respectively. This study demonstrates a simple and inexpensive way to characterize hydrogeological parameters controlling groundwater-surface interactions in any region with aquifers that are highly permeable without using standard pumping tests or canal drawdown experiments. Hydrogeological parameters estimated using this approach could be used as starting values in large-scale numerical simulations.

Original languageEnglish (US)
Pages (from-to)977-992
Number of pages16
JournalTransactions of the ASABE
Volume56
Issue number3
DOIs
StatePublished - 2013
Externally publishedYes

Fingerprint

Groundwater
Canals
Aquifers
Parameter estimation
aquifers
Sensitivity analysis
canal
sensitivity analysis
Analytical models
aquifer
basins
Sampling
basin
Hydraulic conductivity
hydraulic conductivity
methodology
surface interactions
ranking
sampling
Surface waters

Keywords

  • Canal-aquifer interaction
  • GLUE method
  • Morris method
  • Parameter estimation
  • Sensitivity analysis
  • Sobol's method

ASJC Scopus subject areas

  • Agronomy and Crop Science
  • Biomedical Engineering
  • Food Science
  • Forestry
  • Soil Science

Cite this

Sensitivity analysis and parameter estimation for an approximate analytical model of canal-aquifer interaction applied in the C-111 basin. / Kisekka, I.; Migliaccio, K. W.; Muñoz-Carpena, R.; Khare, Y.; Boyer, Treavor.

In: Transactions of the ASABE, Vol. 56, No. 3, 2013, p. 977-992.

Research output: Contribution to journalArticle

@article{61560da1792445a79c1807eac2e063c4,
title = "Sensitivity analysis and parameter estimation for an approximate analytical model of canal-aquifer interaction applied in the C-111 basin",
abstract = "The goal of this study was to better characterize parameters influencing the exchange of surface water in south Florida's C-111 canal and Biscayne aquifer using the analytical model STWT1. A three-step model evaluation framework was implemented as follows: (1) qualitative parameter ranking by comparing two Morris method sampling strategies, (2) quantitative variance-based sensitivity analysis using Sobol's method, and (3) estimation of parameter posterior probability distributions and statistics using the Generalized Likelihood Uncertainty Estimator (GLUE) methodology. Results indicated that the original Morris random sampling method underestimated total parameter effects compared to the improved global Morris sampling strategy. However, parameter rankings from the two sampling methods were similar. For the STWT1 model, only four out of the six parameters analyzed were important for predicting water table response to canal stage and recharge fluctuations. Morris ranking in order of decreasing importance resulted in specific yield (ASY), aquifer saturated thickness (AB), horizontal hydraulic conductivity (AKX), canal leakance (XAA), vertical hydraulic conductivity (AKZ), and half-width of canal (XZERO). Sobol's sensitivity indices for the four most critical parameters revealed that summation of first-order parameter effects was 1.0, indicating that STWT1 behaved as an additive model or negligible parameter interactions. We estimated parameter values of 0.07 to 0.14 for ASY, 11,000 to 14,300 m d-1 for AKX, 13.4 to 18.3 m for AB, and 99.8 to 279 m for XAA. The estimated values were within the range of values estimated using more complex methods at nearby sites. The Nash-Sutcliffe coefficient of efficiency and root mean square error for estimated parameters ranged from 0.66 to 0.95 and from 4 to 7 cm, respectively. This study demonstrates a simple and inexpensive way to characterize hydrogeological parameters controlling groundwater-surface interactions in any region with aquifers that are highly permeable without using standard pumping tests or canal drawdown experiments. Hydrogeological parameters estimated using this approach could be used as starting values in large-scale numerical simulations.",
keywords = "Canal-aquifer interaction, GLUE method, Morris method, Parameter estimation, Sensitivity analysis, Sobol's method",
author = "I. Kisekka and Migliaccio, {K. W.} and R. Mu{\~n}oz-Carpena and Y. Khare and Treavor Boyer",
year = "2013",
doi = "10.13031/trans.56.10037",
language = "English (US)",
volume = "56",
pages = "977--992",
journal = "Transactions of the ASABE",
issn = "2151-0032",
publisher = "American Society of Agricultural and Biological Engineers",
number = "3",

}

TY - JOUR

T1 - Sensitivity analysis and parameter estimation for an approximate analytical model of canal-aquifer interaction applied in the C-111 basin

AU - Kisekka, I.

AU - Migliaccio, K. W.

AU - Muñoz-Carpena, R.

AU - Khare, Y.

AU - Boyer, Treavor

PY - 2013

Y1 - 2013

N2 - The goal of this study was to better characterize parameters influencing the exchange of surface water in south Florida's C-111 canal and Biscayne aquifer using the analytical model STWT1. A three-step model evaluation framework was implemented as follows: (1) qualitative parameter ranking by comparing two Morris method sampling strategies, (2) quantitative variance-based sensitivity analysis using Sobol's method, and (3) estimation of parameter posterior probability distributions and statistics using the Generalized Likelihood Uncertainty Estimator (GLUE) methodology. Results indicated that the original Morris random sampling method underestimated total parameter effects compared to the improved global Morris sampling strategy. However, parameter rankings from the two sampling methods were similar. For the STWT1 model, only four out of the six parameters analyzed were important for predicting water table response to canal stage and recharge fluctuations. Morris ranking in order of decreasing importance resulted in specific yield (ASY), aquifer saturated thickness (AB), horizontal hydraulic conductivity (AKX), canal leakance (XAA), vertical hydraulic conductivity (AKZ), and half-width of canal (XZERO). Sobol's sensitivity indices for the four most critical parameters revealed that summation of first-order parameter effects was 1.0, indicating that STWT1 behaved as an additive model or negligible parameter interactions. We estimated parameter values of 0.07 to 0.14 for ASY, 11,000 to 14,300 m d-1 for AKX, 13.4 to 18.3 m for AB, and 99.8 to 279 m for XAA. The estimated values were within the range of values estimated using more complex methods at nearby sites. The Nash-Sutcliffe coefficient of efficiency and root mean square error for estimated parameters ranged from 0.66 to 0.95 and from 4 to 7 cm, respectively. This study demonstrates a simple and inexpensive way to characterize hydrogeological parameters controlling groundwater-surface interactions in any region with aquifers that are highly permeable without using standard pumping tests or canal drawdown experiments. Hydrogeological parameters estimated using this approach could be used as starting values in large-scale numerical simulations.

AB - The goal of this study was to better characterize parameters influencing the exchange of surface water in south Florida's C-111 canal and Biscayne aquifer using the analytical model STWT1. A three-step model evaluation framework was implemented as follows: (1) qualitative parameter ranking by comparing two Morris method sampling strategies, (2) quantitative variance-based sensitivity analysis using Sobol's method, and (3) estimation of parameter posterior probability distributions and statistics using the Generalized Likelihood Uncertainty Estimator (GLUE) methodology. Results indicated that the original Morris random sampling method underestimated total parameter effects compared to the improved global Morris sampling strategy. However, parameter rankings from the two sampling methods were similar. For the STWT1 model, only four out of the six parameters analyzed were important for predicting water table response to canal stage and recharge fluctuations. Morris ranking in order of decreasing importance resulted in specific yield (ASY), aquifer saturated thickness (AB), horizontal hydraulic conductivity (AKX), canal leakance (XAA), vertical hydraulic conductivity (AKZ), and half-width of canal (XZERO). Sobol's sensitivity indices for the four most critical parameters revealed that summation of first-order parameter effects was 1.0, indicating that STWT1 behaved as an additive model or negligible parameter interactions. We estimated parameter values of 0.07 to 0.14 for ASY, 11,000 to 14,300 m d-1 for AKX, 13.4 to 18.3 m for AB, and 99.8 to 279 m for XAA. The estimated values were within the range of values estimated using more complex methods at nearby sites. The Nash-Sutcliffe coefficient of efficiency and root mean square error for estimated parameters ranged from 0.66 to 0.95 and from 4 to 7 cm, respectively. This study demonstrates a simple and inexpensive way to characterize hydrogeological parameters controlling groundwater-surface interactions in any region with aquifers that are highly permeable without using standard pumping tests or canal drawdown experiments. Hydrogeological parameters estimated using this approach could be used as starting values in large-scale numerical simulations.

KW - Canal-aquifer interaction

KW - GLUE method

KW - Morris method

KW - Parameter estimation

KW - Sensitivity analysis

KW - Sobol's method

UR - http://www.scopus.com/inward/record.url?scp=84884187835&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84884187835&partnerID=8YFLogxK

U2 - 10.13031/trans.56.10037

DO - 10.13031/trans.56.10037

M3 - Article

AN - SCOPUS:84884187835

VL - 56

SP - 977

EP - 992

JO - Transactions of the ASABE

JF - Transactions of the ASABE

SN - 2151-0032

IS - 3

ER -