There may be specifics in field that cannot be replicated in lab. For instance, consider my plot study of Hazel Pistol area, which had been clearcut and burned hot. The soils had apparently developed hydrophobic property. Related to a specific agricultural field, there might be remnants of installed terraces or even natural slope variances, network of rills, or other features as well as variances in soil type, cover, organic content, etc. that would be difficult to duplicate in lab. Of course, field studies may take years, and some erosion is episodic. We had good results in assessing small epheneral gullies, and small burn catchments using filter fences (methods described in Sediment From a Small Ephemeral Gully ...) in my researchgate. I remember an old study in the piedmont of SC, USA (I think author name was Hoover) which I have used as a citation on some of our gully control papers where the agricultural soils had macropores sealed by fine particles, therefore reducing hydraulic conductivity and infiltration. It is difficult to replicate field conditions such as this in a lab.

Another point, some areas are exposed periodically or episodically to wind driven rain. Rain gauges help collect the local amount of rain, but the forces of wind driven rain probably exceed the forces from gravity such as calm storm or rainfall generator in lab. If you obtain a map of the USLE R factor in the USA, you will see substantial variance with the SE USA with the highest R coefficients due to some extent from Gulf and Atlantic Cyclonic events and thunderstorms. The areas I worked in NW California and SW Oregon have a much lower R factor, yet more rainfall than most areas in SE USA.

I think your simulated results may not give the true results which lead to wrong information and become ineffective.