Codes

Model Analysis and Decision Support (MADS v1.1; Vesselinov & Harp, 2011; LA-CC-10-055)

MADS (Model Analysis & Decision Support) is an object-oriented code that is capable of performing various types of model analyses and supporting model-based decision making. The code can be executed under different computational modes, which include (1) sensitivity analysis, (2) Monte Carlo analysis, (3) model calibration, (4) parameter estimation, (5) uncertainty quantification, and (6) model selection. The code can be externally coupled with any existing model simulator through integrated modules that read/write input and output files using a set of template and instruction files. MADS works directly with existing control files developed for the code PEST (Doherty 2009) and can also be internally coupled with a series of built-in analytical simulators. In addition, the code can be used as a library for internal coupling with simulators using object-oriented programming. The code provides (1) efficient parallelization (MPI and POSIX), (2) runtime control, restart, and preemptive termination, (3) advanced Latin-Hypercube sampling techniques (including Improved Distributed Sampling), (4) various gradient-based Levenberg-Marquardt optimization methods, (5) advanced single- and multi-objective global optimization methods (including particle swarm optimization, PSO, SQUADS), and (6) local and global sensitivity and uncertainty analyses (including ABAGUS). The code is written in C and tested on various platforms (Unix, Linux, Mac OS X, MS Windows). The code has been successfully applied to perform various model analyses related to environmental management of contamination sites at LANL. Examples include solutions of source identification problems, quantification of uncertainty, model calibration, and optimization of monitoring networks.

Executables, source code, example files, and brief manual can be obtained by sending e-mail request to software@lanl.gov. Please send us the following Software Request Questionnaire.