Tutorial
DecisionFX is offering software that not only provides defensible decision analysis capabilities but is also quite user-friendly. The following tutorial offers a look at the current version of RiskFX, to demonstrate the graphical user interface concepts employed in this tool.
RiskFX is a comprehensive risk assessment tool. It has the capability to perform either deterministic (i.e., single-run) estimates of risk or probabilistic (i.e., Monte Carlo) estimates of risk to quantify the uncertainty.
RiskFX has the capability to analyze a site for exposures to hazardous chemicals or radionuclides. The regulatory performance metrics for hazardous versus radioactive constituents are explicitly separate, and therefore the RiskFX user is asked to select from either a hazardous chemical or a radioactive constituent analysis. The following figure shows the popup window displayed for the user to make this selection.
Once the desired model type is selected the main display window appears, complete with a point-and-click pathways selection graphic and drop-down menus.
Once the user has selected the appropriate pathways for their land use and exposure assumptions, the next step is to provide the data inputs and assumptions. Under the Data drop-down menu are the following choices:
The user then steps through each of the various categories of data and parameters required to perform the desired risk analysis. Under Site Specification, for instance, the user is given the following input window:
In the Site Specification window the user is asked to supply risk limits (i.e., cancer risk limit for hazardous chemical carcinogens, hazard index limit for toxicants, and if this were a radionuclide analysis, a dose limit) for the purpose of performing inverse calculations that yield risk-based soil concentration thresholds for purposes of defining action levels or cleanup levels at a site. The user can also submit up to 9 desired output times to evaluate the transient behavior of the site under investigation.
The next menu option in the Data menu is Initial Concentrations. Here the user can choose practically any number of contaminants that exist at their site and supply an initial source term concentration. RiskFX accounts for cummulative (i.e., additive, not synergistic) risk effects from all chemicals under consideration. There are currently about 150 chemicals in the data base and about 50 radionuclides. The following window shows and example of the input screen.
The next window from the Data menu options is Basic Hydrology, shown in the following graphic:
The next window option for data input is Zone Properties, where hydrologic parameters for the source term, the unsaturated zone, and the saturated (aquifer) zone are specified:
The next input screen is for Distribution Coefficients, for specifying the propensity of the contaminants for sorption in the porous media under consideration (e.g., controls transport rates):
The next menu option under Data is the Dietary parameter specification:
The next screen is Non-Dietary input parameters associated with exposure:
The next set of screens deals with air dispersion and dust inhalation:
The next Data window deals with Time Fractions, relative to percent of time spent indoors and out:
The user has the ability to view and possible change the intake conversion factors, food transfer factors, Reference Dose factors, and Cancer Slope factors through the bottom menu item of the Data menu:
If a radionuclide analysis were selected, additional menu items would be available for setting parameters for additional pathways, including external gamma radiation and radon uptake within buildings.
A key feature not often found in other risk assessment software packages but present in RiskFX is the ability to quantify uncertainty in risk estimates through the use of probabilistic, or Monte Carlo, methods. The EPA is looking toward quantitative uncertainty analysis methods more these days as a result of criticism of the conventional Reasonable Maximum Exposure (RME) approach, which is considered by many as too conservative. The EPA had been concerned years ago with the computational burden associated with the implementation of probabilistic methods, and therefore was cautious of its routine use. With the advent of Pentium and PowerPC computers on desktop platforms, the issue of computational burden is no longer relavent. In RiskFX the Monte Carlo method is streamlined through the use of a sampling technique called Latin Hypercube Sampling (LHS). Instead of a traditional random sampling method for Monte Carlo requiring 10,000 separate simulations toward characterize uncertainty, the LHS method requires only 10s or 100s of simulations to characterize uncertainty, depending on the number of uncertain parameters. LHS also allows the user to have correlations amongst parameters, thereby reducing uncertainties even further.
The implementation of uncertainty in RiskFX is very straightforward. For any parameter which is considered uncertain the user double-clicks on the name of the parameter in a Data window (such as Thickness of Contaminated Zone in the Site Specification window). The following LHS dialog box appears for that parameter:
In this window, the user can specify a distribution type for the parameter, the distribution statistics (e.g., mean and standard deviation), correlation between other uncertain parameters, and type in a justification for the data or parameter distribution used. This last option is very important for documenting results. The regulatory community is quite hesitant to utilize uncertainty analyses for decision making if there is question about the defensibility of the probability density functions used to define uncertainty. Therefore, users are encouraged to document the pedigree of each data or distribution used in the simulation. An optional input summary file can then be produced which sequentially lists all data, distributions, and assumptions used in an analysis.
The number of distribution types available to the user is fairly complete, as shown in the next graphic:
Selection of the parameters for uncertainty analysis is generally done with the aid of a sensitivity analysis. RiskFX currently has a normalized sensitivity analysis method available to prioritize input parameters according to their relative sensitivity to the risk results. In the next upgrade a second sensitivity analysis method for global regression will be implemented to evaluate synergistic effects of the uncertain parameters in the Monte Carlo analysis.
Once the data and parameters have been established for an analysis the user selects the Monte Carlo analysis option from the Simulation menu, and specifies the number of simulation runs to perform:
Another option exists in this menu to save LHS reports, to confirm that the sampled parameters indeed represent the selected distributions and correlations.
Once the analysese are complete, RiskFX has a visualization program to immediately view the results of the analysis. Many options exist for display of the information generated from a given analysis. The following graphic illustrates some of these options:
With this plotting package a user can quickly deduce which pathways and contaminants are the major contributors to risk. Risk management decisions may be enhanced with this ability to segregate the various aspects of the risk analysis and exposure characteristics. A sample plot of the output graphics follows:
The next version of RiskFX , currently under production, will include a number of enhancements, which include: a tab style interface for parameter specifications for even greater ease-of-use; volatile contaminant migration and exposure; dermal contact; graphical display of probability density functions selected for describing uncertainty of the input parameters (for LHS input); and a more robust 3-dimensional groundwater flow and transport algorithm.
(this page still under development)
Return to Tutorials and Case Studies page