Model Verification and Validation

The usefulness of geochemical or other complex models would seem to be determined in two stages. 

Model verification. Does the model do what it is intended to do That is, does the computer code solve the equation set correctly, and is the code free from serious bugs that will lead to unintended results  

Model validation. Does the model produce realistic results, given realistic input data That is, are the underlying conceptual models which the coded equations simulate adequate Do they actually represent natural processes  

It would seem to follow, then, that after verification and validation, a model could be used for predictive purposes. If only things were that simple. A useful discussion of these concepts is given by Nordstrom (1992). We present only a synopsis. 

Theoretically, verification of a computer code is an achievable task, given sufficient time, funding, and manpower. However, with increasingly complex programs, complete and unambiguous verification is becoming very difficult, or even infeasible. Nordstrom 

---

Experiment is the beginning and the end, the starting-point and final objective of any modeling. Despite the fact that the model is fundamentally not identical to the object of modeling, only experiment suggests the initial guess and provides primary data concerning the structure of the reaction scheme and the values of the kinetic parameters. In their turn, model verification and validation can be done only by comparison with experimental data. 

---