If you are looking for models to estimate exposure values to compare to your toxicity endpoints, USEPA has a suite of models related to agriculture.

https://www.epa.gov/pesticide-science-and-assessing-pesticide-risks/about-water-exposure-models-used-pesticide

Hi,

I deal with human health risks, I find your question very interesting. If you have the distribution function (by agent), your median is what you refer as LC50 lethal concentration at which -- on the average -- 50% of those in the sample die). Why use a safety factor of 100 when you can use +3 sigma, +4 sigma ...? This is the easy part. The hard part is whether the responses are actually unimodal, not skewed, and so on. I would prefer to use the full distribution, show its empirical (data driven form) fit a curve to that distribution and judge what to do. The uncertainty in the right tail of the distribution is what matters. In other words, I would also be concerned with extreme outcomes (namely your 100 factor), which depends on the empirical distribution of the agent. Of course, if you are using a well-studied agent than LC50 and the 100 factor may well be justified in the literature. Eckel's point is very appropriate.

I hope I did not misunderstand the question.

Best, Paolo