Further description:-  Predictive modelling 

Glossary Entry
Principles and tools simplifying the 'real thing' to predict something, e.g. the behaviour of a contaminant, 
the effectivity of a remediation, etc.

Predictive Modelling

 

1. Summary

 

Before predictive modelling calculations are made several other steps need to be considered during the model application process (Figure 1). Every modelling project should be started with clearly defined goals and objectives. This point has to be stressed, since it influences every other consecutive step.

 

2. Model Application Process

 

The different steps are:

  • To collect and gather information e.g. concerning the geology, hydrology and geochemistry of the site. During this step the contaminated site is characterised with regard to the modelling goals. This information is incorporated into the conceptual site model. The established numerical model can be used in a later step to test and verify the conceptual site model
  • Selection of a computer code. The selection of an appropriate computer code depends e.g. on the decision whether two or three dimensional modelling is needed.
  • Preparation of input files and the incorporation of all governing equations. Assuming that all necessary information and equations have been included a simpler model should be applied preferentially over a more complex model.
  • The calibration process is undertaken until model simulations match the field observations to a reasonable degree. The subsequent sensitivity analysis should be used to test the overall responsiveness and sensitivity of the numerical model to certain input parameters.

 

3. Predictive Simulations

 

  • After calibration, the model can be used for predictive simulations. The model can be applied as a management tool for decisions in which the response of a system is predicted, e.g. concentrations in groundwater at some time in the future can be predicted.
  • However the model needs to be used with caution when applied, since uncertainties are always present and should be addressed. The uncertainties can be divided into two general categories: Those associated with model input parameters and those associated with numerical and conceptual difficulties. Methods to deal with uncertainty are sensitivity analysis and the Monte Carlo method.
  • The sensitivity analysis is used to rank important sources of variability and uncertainty. A sensitivity analysis can involve complex mathematical and statistical techniques such as correlation and regression analysis to determine which factors are most important for the model output..
  • The Monte Carlo Method considers each model input parameter to be investigated as a random variable defined by a probability density function (PDF). The PDF shows the probability of an uncertain quantity taking on a particular value (Zheng & Bennet 2002).

                                                                                                                    

 

 

Figure 1: Model Application Process (after Bear 1992)

 

4. Weblinks and Guidelines:

 

Bear, J., Beljin, M.S., Rose, R. (1992). Fundamentals of Ground-Water Modelling. EPA Ground Water Issue (online: http://www.epa.gov/tio/tsp/download/issue13.pdf)

 

ASTM (1999): RBCA Fate and Transport Models: Compendium and Selection Guidance (http://www.epa.gov/oust/rbdm)

 

Groundwater Services, Inc. (1996). Parameter Estimation Guidelines for Risk Based Corrective Action (RBCA) Modeling. NGWA Petroleum Hydrocarbons Conference, Texas

(http://www.gsi-net.com/Publications/param.pdf)

 

5. Useful (Groundwater)-Modelling Links:

 

Geotechnical and Geo-environmental Software Directory

http://www.ggsd.com

 

U.S. Geological Survey Software Links

http://water.usgs.gov/software/ground_water.html

 

USEPA Center for Subsurface Modeling Support (CSMoS)

http://www.epa.gov/ada/csmos.html

 

Evaluation of Selected Environmental Decision Support Software

http://www.epa.gov/swertio1/download/remed/doedss.pdf

 

6. Literature

 

Zheng, C., Bennet, G. D. (2002). Applied contaminant transport modeling (2 ed.). New York.

 

Hakanson, L. (2004). Break-through in predictive modelling opens new possibilities for aquatic ecology and management – a review. Hydrobiologia 518, 135-157

 

Schwarzenbach, R.P. (2003): Environmental organic chemistry (2 ed.) Wiley. New York

 

 

Authors
Stefan Gödeke
Universität Tübingen, Germany

Who does what?