You might be thinking of numerical error, which rarely dominates the full uncertainty if one is doing the computation right. You can quantify the full uncertai

You might be thinking of numerical error, which rarely dominates the full uncertainty if one is doing the computation right. You can quantify the full uncertainty in both scale/analogue models and numerical models. In my experience the full uncertainty is much more often quantified in scale and analogue models than numerical models, by the way. Analogous to numerical error, if the systems under study differ slightly you can quantify the error introduced from the "distortion" as it is often called in the literature.