PREP/SPOP: Uncertainty and Sensitivity Analysis Monte Carlo Program and Input Preparation
LISA: A Long-Term Isolation Safety Assessment Code.
Division of Radiochemistry, Joint Research Centre of the European Communities I-21020, Ispra, Italy, through the OECD NEA Data Bank, Issy-les-Moulineaux, France.
The PREP program (data PRE Processing Unit) prepares samples for a Monte Carlo simulation using the distribution functions of the input variables. Any degree of correlation among the variables can be requested in the input. The SPOP program (Statistics Post Processing Unit) is used for performing uncertainty analysis and sensitivity analysis on the output of the simulation runs. SPOP provides for each output variable:
- a histogram and cumulative distribution,
- confidence bounds on the cumulative distribution,
- mean and confidence bounds on the mean, and
- estimate of the sensitivity of each output variable to card input variable.
The sampling in PREP can be carried out purely randomly or by using the Latin Hypercube Technique (LHS). The technique of Iman and Conover is used to eliminate spurious correlations or to impose pre-specified correlations. The prepared samples are stored on an external file for further use. SPOP uses Kolmogorov statistics for estimating confidence bounds on the cumulative distribution and the Tchebycheff's theorem for the mean end confidence bounds on the mean. The estimation of sensitivities is carried out using:
- Smirnov’s test,
- Spearman’s test,
- Cramer test,
- two sample T test,
- Kendall’s test,
- Mann-Whitney’s test,
- Klotz’s test, and
- sum of squared ranks’ test.
The PREP test problem set up for 100 runs and 15 variables and Latin Hypercube Sampling requires less than a second on an _X64 processor; the SPOP test problem set up for 1000 runs, 22 variables, 250 time points and with statistics computed at 6 time points requires a few seconds on an _X64 processor.
PREP and SPOP should run on any modern computer architecture. Also compatible with IBM-MVS, VAX-VMS, CDC-NOS, CRAY-COS systems.
A Fortran 77 compiler is required.
T. Homma and A. Saltelli, LISA Package User Guide, Part I: PREP (Statistical PREProcessor) Preparation of Input Sample for Monte Carlo Simulations Program Description and User Guide, preprint of EUR 13922 EN (1991).
P. Prado, A Saltelli and T. Homma, LISA Package User Guide, Part II, LISA (Long Term Isolation Safety Assessment) Program Description and User Guide, preprint of EUR 13923 EN (1991).
A. Saltelli and T. Homma, “LISA Package Use Guide, Part III: SPOP (Statistical POst Processor) Uncertainty and Sensitivity Analysis for Model Output Program Description and User Guide,” preprint.
A. Saltelli, Seminar on “Sensitivity Analysis of Model Output,” seminar presented November 20, 1992, at IPSN/DES/SESID Bureau Fontenay-aux-Roses, Paris.
A. Saltelli, Informal note dated 10 December 1992.
Included in the package is a CD with files in tgz Linux format (uncompressible with GNU utilities or WinZip) which includes an installation procedure, source, scripts, data files and documents listed above in Section 10a.
KEYWORDS: MONTE CARLO, RADIATION ENVIRONMENT, WASTE MANAGEMENT