Constructing and calibrating numerical models can be a challenging task in terms of time, available data and professional abilities of the modeler. Since this work is usually done alone, there are many questions about the conceptualization of the physical environment, the construction of the model and the calibration with observed data. Based on the questions asked by some attendees of our webminars, we have written this article for all the confused hydrogeological modelers.

Calibration of numerical models

In relation to calibration, I want to ask:

1. Is it correct to consider a water balance based in measured data (such as recharge, infiltration, evapotranspiration, runoff) and hope to obtain a MODFLOW water balance with similar values by modifying the parameter less trustable (like hydraulic conductivity) and some boundary condition?

Answer: The water balance of measured data is oriented to the surface flow. There are parameters of groundwater cycle which are hard to quantify, like recharge. Also, not all runoff is relevant for a hydroeological model, just baseflow.

In conclusion, I think that these balances have to be separated. A superficial water balance has a faster transit and includes very dynamic physical processes like precipitation and runoff. When referring to groundwater water balances, processes are slower. MODFLOW calculates good water balances for uniform and transient regime and can it be compared with values of piezometric levels and baseflow.

2. When calibrating using hydraulic conductivity, sometimes it is necessary to use values that you would not associate with its granulometry/lithology; is this because we need to consider transmissivity to calibrate, in other words, not just permeability but also the thickness? Or is it a problem related to the raw information, specifically, that the permeability values obtained from the pumping tests correspond only to punctual information?

Answer: MODFLOW lets you insert values of hydraulic conductivity (K) and transmissivity (T). The transmissivity is associated with the thickness of the aquifer, but in MODFLOW it is associated to the thickness of the layer. My recommendation is to use K values that MODFLOW will transform to T values according to the layer thickness considered. It is important to remember that in Model Muse it is possible to add a big amount of layers easily, and in porous or fractured regional models more than 12 layers are used to ensure a correct representation of the hydraulic behavior with depth.

3. When calibrating with hydraulic conductivity, is there a more convenient flux direction to do it? (for example, from recharge zone to discharge zone)?

Answer: You are right, there is a direction in calibration. The lower areas or discharge areas are calibrated with drain conductance and the evapotranspiration rates; while, the higher areas are calibrated with the variation of hydraulic conductivity and recharge rate. Changes in vertical anisotropy apply to the higher and lower areas.

4. Could the model error be related to not considering the flux associated to a boundary indicated as impervious (for example, metamorphic rocks) when they actually provide groundwater flow to the model?

Answer: Nothing in nature is impervious, even metamorphic rocks have very low conductivity values (around 1e-10 m/s). Hydrogeological units should not be ignored because of their low hydraulic conductivity values. It is not hard to increase the size of a model and the boundary conditions. It is better to have a model bigger than necessary.

About the conceptualization and construction of numerical models

And more general questions:

5. Do you think it is correct to consider a river as a constant head boundary? (I am asking this because I have been told that it is incorrect, but it is the only way I can calibrate a model).

Answer: This is incorrect. A river has to be conceptualized as a River (RIV) or Drain (DRN). If the river is a river that "gains" and "losses" it needs to be conceptualized as River; if the river only "losses" it can be conceptualized as a drain. In terms of modeling, it is easier to implement a drain condition.

6. Is there any option to model basins with irregular geometries (with elevations ranging from hundreds of meters to almost zero elevation)?

Answer: It is important to remember that MODFLOW solves the inter cells flux. The more regular these cells are, the calculus is more precise. When working with andean basins or basins with extreme topography, it is better to use the MODFLOW NWT solver. It is recommended to leave a minimum thickness of the aquifer (300 m) in the discharge area of the basin.

7. In the introduction to the webminar it was mentioned that "Numerical models are the best tools available to understand the groundwater flow regime and the evaluation of the impact of a project on the groundwater resources". Considering this, what do you think of the following methodology as an option to model a hydrogeological basin?

Calculate the hydrogeological information (with data of geometry and piezometric surface).
Calculate a possible volume of groundwater extraction (with effective porosity and storage coefficient data).
Estimate basin recharge.
Estimate a recommended volume of groundwater extraction (considering recharge).

Answer: This is the same question I was asked when I presented my Master's Thesis. A global analysis of the basin as indicated above only lets you give a global diagnose of the state of groundwater. However, this diagnose would not work to evaluate the zones with more demand, the local state of reservoirs and aquifer management systems.