Stochastic engine and active subspace global sensitivity analysis codes

Matlab codes for running a full stochastic sampling of a multi-parameter model with the aim of creating an ensemble of model evaluations that are showing a user defined behavior (i.e. are behavioral). The code uses 2-stage acceptance in which a pre-selection step is performed using a surrogate model. The stochastic engine is setup for working with active subspaces and Gaussian process emulation as surrogate models, although the GPE-version requires external codes. It is directly compatible with HydroGeoSphere and local matlab models, but can be easily modified for any other model. The package also includes codes to perform a global sensitivity analysis using active subspaces. Via the template data and start file, a dummy problem is directly operational. This package has been used by the following publications: Erdal & Cirpka 2019 (doi.org/10.5194/hess-23-3787-2019), Erdal & Cirpka 2020 (doi.org/10.5194/hess-2019-629, in review) and Erdal et al. 2020 (Sampling Behavioral Model Parameters for Ensemble-based Sensitivity Analysis using Gaussian Process Emulation and Active Subspaces, submitted to Stochastic and Environmental Research and Risk Assessment).