py4sci

Page Navigation

Simulation Toolsets

Simulation Toolsets

Akuna provides six Simulation Toolsets (Figure 1), including a Single-Run (SR), Sensitivity Analysis (SA), Parameter Estimation (PE), Uncertainty Quantification (UQ), Risk Assessment (RA) and Decision Support (DS). Each toolset is launched separately, once Model Setup has been completed. Through Akuna, users specify parameters and their distributions. Input files are automatically generated using the auto-generated parameter suites, making it easy to launch simulation ensembles.

In addition, through the simulation toolsets, users can import data relevant to their site. In Akuna, a distinction is made between measured data (i.e., field data), and observation data (i.e., simulated data). Defining locations in the simulation domain to compare measured and simulated data is needed for model calibration (PE Toolset), but can also be used in other toolsets as well.

The RA and DS Toolsets are launched after the completion of other simulation toolsets. For example, groundwater risk and dose calculations can be executed in the RA Toolset once concentration predictions have been generated for a successful simulation run (e.g., SR, SA or UQ).

SingleRun ParameterEstimation DecisionSupport RiskAssessment UncertaintyQuantification SensitivityAnalysis

Figure 1. Akuna's Simulation Toolsets. Click on a Toolset name to be directed to its description.

Single Run (SR)

The Single Run (SR) Toolset provides a framework for executing a single forward simulation. This toolset is often used as a precursor to launching ensemble simluations (e.g., SA, PE, and UQ).

Sensitivity Analysis (SA)

The Sensitivity Analysis (SA) Toolset provides techniques to estimate the sensitivity of model predictions to conceptual model elements and model parameters. Currently, the global Morris-One-At-a-Time (MOAT) or “Elementary Effects Method” is fully implemented in Akuna. This method calculates sensitivities at multiple locations in parameter spaces. The particular locations are determined using a method developed by Morris (1991) that has been shown to be computationally efficient and robust. This method is qualitative, and is a useful screening tool to identify non-influential parameters and to rank parameters by their relative sensitivity.

Parameter Estimation (PE)

The Parameter Estimation (PE) Toolset provides techniques to estimate conceptual model elements and model parameters based on site observation data. It determines model parameters that produce simulated outputs that best match the observation data. The default algorithm is an implementation of Levenberg- Marquardt algorithm that permits better parallel performance and faster convergence through inexact Jacobian reconstruction. This toolset (as well as the other toolsets) provides for the import of measured data, which can be imported from a web-based data management tool, or from any data that resides on either the client or the server.

Uncertainty Quantification (UQ)

The Uncertainty Quantification (UQ) Toolset provides techniques to estimate the uncertainty of model predictions caused by uncertainties in conceptual model elements and model parameters. Most of the UQ methods such as Monte Carlo are computationally intensive and require large numbers of forward model simulations. Currently the Monte Carlo method is implemented in Akuna with options for Latin Hypercube and Random sampling methods.

Risk Assessment (RA)

The Risk Assessment (RA) Toolset uses predicted concentrations from the simulation toolsets to determine cancer risk from the groundwater pathway. The RA Toolset currently serves as a prototype for future development, so it does not consider additional exposure pathways. Only residential and industrial exposure scenarios for drinking water are considered.

Decision Support (DS)

The Decision Support (DS) Toolset provides a computational framework for scientifically defensible decision making. A major goal of the DS Toolset is to estimate the impact of conceptual model elements, model parameters and model predictions (and their respective uncertainties) on the decision making process. The DS Toolset also addresses the iterative nature of the environmental management, including theadjustment of decision objectives and environmental management goals. The current implementation of DS is a prototype, and focuses on monitoring network design.