VALID NOW. VALID LATER? HOW FUTURE STRESSES CAN INVALIDATE VALIDATED GROUNDWATER FLOW MODELS

Modelers generally develop numerical groundwater flow models in a linear, step-by-step fashion: (Step 1) define the purpose, (Step 2) define the conceptual model, (Step 3) identify and design the model, (Step 4) calibrate the model, (Step 5) verify the model, and then (Step 6) make predictions. Once the model is calibrated and verified, many modelers never look back when they proceed to make predictions. But is that model really verified for making predictions? Calibration and verification are focused on the conditions such as water levels, recharge, and discharge that the aquifer experienced in the past. A competently calibrated and verified model can be expected to make good predictions if the aquifer conditions used to make predictions fall within the range of those experienced during calibration and verification. However, a competently calibrated and verified model may not make good predictions if the aquifer conditions used to make predictions fall outside the range of those experienced during calibration and verification. We oftentimes expect our models to make these predictions with little thought as to how the new aquifer conditions may interact with boundaries. This has happened with several regional groundwater flow models in Texas where boundary conditions and parameter assignments that worked well when the models were calibrated produced unrealistic results when the aquifers were stressed beyond previous experience. A good calibration and verification (and perhaps a short-term validation) often make us feel comfortable about our models as we make predictions. However, modelers need to continue questioning the validity of their models with every phase of modeling.