The easiest way to validate the results of any model is to upscale by testing the model assumptions and parameters under real life scenarios if it's safe and cost effective to do so.
You might need to consider the data calibration process to cater for systematic and systemic errors in estimating the model's error%