Sensitivity analysis using Sobol variance-based method on the Haverkamp constitutive functions implemented in Richards water flow equation


Citation

Goh Eng Giap, . and Kosuke Noborio, . Sensitivity analysis using Sobol variance-based method on the Haverkamp constitutive functions implemented in Richards water flow equation. pp. 19-33. ISSN 1394-7990

Abstract

Richards equation was approximated by finite-difference solution and implemented in FORTRAN to simulate water infiltration profile of yolo light clay. The simulation was successfully validated by published data of Philips semi-analytical solution. Global sensitivity analysis using Sobol variance-based method was also coded in FORTRAN and implemented to study the effect of parameter uncertainty on model output variability. Sobol sequences were used to generate quasi-random numbers to study the effect of every possible combination of different input parameters values based on each parameters uncertainty range on model outputs. First order sensitivity index (Si ) and total effect index (STi) were estimated based on quasi-Monte Carlo estimators. Various statistical parameters coded in FORTRAN such as kurtosis skewness 95 confident intervals etc. were used to provide a better understanding and description of the model outputs. Results found parameter constants ( B) and saturated volumetric water content (s) of Haverkamp constitutive functions to be dominant parameters with a combined 93 of model variability which could be explained by these parameters. The total effect index for every parameter was found to be greater than the first order effect index. In addition global sensitivity analysis tool was able to generate informative sensitivity indicators and a good statistical description compared to the local sensitivity tool.


Download File

Full text available from:

Abstract

Richards equation was approximated by finite-difference solution and implemented in FORTRAN to simulate water infiltration profile of yolo light clay. The simulation was successfully validated by published data of Philips semi-analytical solution. Global sensitivity analysis using Sobol variance-based method was also coded in FORTRAN and implemented to study the effect of parameter uncertainty on model output variability. Sobol sequences were used to generate quasi-random numbers to study the effect of every possible combination of different input parameters values based on each parameters uncertainty range on model outputs. First order sensitivity index (Si ) and total effect index (STi) were estimated based on quasi-Monte Carlo estimators. Various statistical parameters coded in FORTRAN such as kurtosis skewness 95 confident intervals etc. were used to provide a better understanding and description of the model outputs. Results found parameter constants ( B) and saturated volumetric water content (s) of Haverkamp constitutive functions to be dominant parameters with a combined 93 of model variability which could be explained by these parameters. The total effect index for every parameter was found to be greater than the first order effect index. In addition global sensitivity analysis tool was able to generate informative sensitivity indicators and a good statistical description compared to the local sensitivity tool.

Additional Metadata

[error in script]
Item Type: Article
AGROVOC Term: Simulation
AGROVOC Term: Water
AGROVOC Term: Infiltration
AGROVOC Term: Hydraulic conductivity
AGROVOC Term: Models
AGROVOC Term: Prediction
AGROVOC Term: Water content
AGROVOC Term: Soil water movement
Depositing User: Mr. AFANDI ABDUL MALEK
Last Modified: 24 Apr 2025 00:54
URI: http://webagris.upm.edu.my/id/eprint/8007

Actions (login required)

View Item View Item