Abstract
These papers (part I and part II) emphasize the need for sensitivity and uncertainty analyses. A number of techniques are applied, e.g. latin hypercube sampling, impact response surfaces and Sobol-analyses. Five examples are presented, four of them concerning the numerical model Swap. The data generation and analysis is performed with standard R packages. Although the computations can be made on any computer, the most time-consuming examples in this paper have been run on a High Performance Computer Cluster. With the relatively simple Impact Response Surface technique it is shown that variation of the saturated hydraulic conductivity has far less impact than changing the moisture content at saturation. Analyses according to the Sobol-Jansen method show that when the soil physical relationships are described according to Damiano, then the parameter b has a very large influence on the results. If the well-known Mualem - Van Genuchten equations are applied, most variation can be explained by the parameter n.
Original language | English |
---|---|
Article number | 105072 |
Journal | Computers and Electronics in Agriculture |
Volume | 170 |
Early online date | 13 Nov 2019 |
DOIs | |
Publication status | Published - Mar 2020 |
Keywords
- HPC
- IRS
- LHS
- Numerical model
- R
- Sensitivity
- Sobol
- Swap
- Uncertainty