Akuna: Platform & Integrated Toolsets
Akuna is the user environment supporting the creation of subsurface flow and transport models using Amanzi. Akuna is an open-source, platform-independent user environment that includes features for basic model setup, sensitivity analysis, inverse parameter estimation, uncertainty quantification, launching and monitoring simulations, and visualization of both model setup and simulation results. Features of the model setup tool include visualizing wells and lithologic contacts, generating surfaces or loading surfaces produced by other geologic modeling software (e.g., Petrel, earthVision), and specifying material properties, initial and boundary conditions, and model output. The model setup tool utilizes LaGrit for generation of both structured and unstructured model simulation grids. Integrating with WorldWind also allows a user to develop a model based on the initial visualization of their site in its actual geographic context, with displays of surface topography and geomorphic features.
After the model has been set up, Akuna facilitates launching a single forward run, performing formal sensitivity analyses, parameter estimation and uncertainty quantification, and visualization of results. Automated job launching and monitoring capabilities allow a user to submit and monitor simulation runs on high-performance, parallel computers. Visualization of large outputs can be performed without moving the data back to local resources. These capabilities makes high-performance computing accessible to the users who might not be familiar with batch queue systems and usage protocols on different supercomputers and clusters.
Akuna supports a common workflow for developing and applying a numerical model in support of environmental management. Many elements of this workflow are repeatedly and iteratively performed as part of the modeling process. Figure 1 provides a simplified chart of this workflow. In general, a conceptual understanding of the system to be analyzed is gained from site characterization efforts and monitoring data. This conceptual understanding is then translated into a mathematical model and further implemented in a numerical model, which requires tools to describe the model domain with its salient hydrogeochemical features, associated material properties, initial and boundary conditions, forcing terms, as well as information on how space and time are discretized for numerical solution.
These functions are supported by Akuna’s Model Setup toolset. Once an initial numerical model has been developed, Akuna’s Simulation Run (SR) toolset can be used to launch and monitor a single simulation, the results of which can be analyzed and visualized. If this initial run is considered reasonable, a formal local or global sensitivity analyses can be performed (using Akuna’s Sensitivity Analysis (SA) toolset) to identify the parameters that most strongly influence the system behavior, and to examine output variables that are sensitive to (and – if measured in the field - contain information about) the parameters of interest. These parameters may include material properties, but also initial and boundary conditions, and generally any aspect of the conceptual model that can be suitably parameterized. If measurements of sufficient sensitivity and accuracy are available, the model can be automatically calibrated using Akuna’s Parameter Estimation (PE) toolset. This step not only provides effective parameter values that can be considered consistent with the data collected at the site; it also provides estimates of the uncertainty with which these parameters were determined. The latter can be used (along with other information) in Akuna’s Uncertainty Quantification (UQ) toolset to evaluate the uncertainty of model predictions and provide the basis for a subsequent assessment of environmental and health risks in Akuna’s Risk Assessment (RA) toolset. Finally, the information from these model analyses enter Akuna’s Decision Support (DS) toolset, which evaluates and optimizes performance measures that help manage DOE’s legacy sites.
In practical applications, the workflow is not as linear as described above. The toolsets integrated in Akuna are transparent and can be flexibly invoked to accommodate any application’s particular workflow.
There are several major components that comprise the Akuna architecture as shown in Figure 2. Each of these is briefly described in this section.
Akuna Desktop User Interface (UI): The Akuna UI provides a front end to the simulation workflow. The cross-platform UI is written in Java and is built on the Velo knowledge management framework. The UI includes a data browser that provides access to all data, metadata, provenance, and tools associated with the workflow. Visit has been integrated to support remote visualization of large-scale outputs. A robust open-source content management system is used to manage workflow data and metadata . Shared as well as private workspaces are supported to enable collaborative modeling.
Agni: Agni is software located on the compute server that takes modeling requests from the Akuna client, executes them and reports information back to the UI. Agni includes a component for controlling local execution of the simulator as well as the analysis toolsets for sensitivity analysis, uncertainty quantification, and parameter estimation. In the future, tools for risk assessment and decision support will be added.
Simulator: Amanzi is the main simulator supported by the Akuna platform. Akuna and Agni are however designed to accommodate other simulators that can be plugged in using a set of defined interfaces.
ASCEM Observational Data Management System: The ASCEM Observational Data Management System (AODMS) provides data management capabilities to import, organize, retrieve, and search across various types of observational datasets needed for environmental site characterization and numerical modeling. The AODMS framework provides capabilities to organize, interactively browse on maps, search by filters, select desired data, plot graphs, and save selected data for subsequent use in the modeling process. Further description of this capability is beyond the scope of this paper, but readers are encouraged to view the AODMS site at http://babe.lbl.gov/ascem/maps/SRDataBrowser.php.