I am using FUND with GDP growth and population growth modified to conform to the SSPs (up to 2100, after which I assume stagnation for simplicity) and defining the climate sensitivity parameter with a log-normal distribution centered around 1.106 with standard deviation of 0.2646 (consistent with Nordhaus’ estimates based on the work of Olson et al). When drawing 100 000 sets of parameters (using @defism and generate_trials() without modifying any other probability function) and then calculating the SCC for each of these trials (using compute_scc()) by replacing the parameters of the trials in a loop, I find extreme values at the tails (negatives in the three figures under the 4th permille of the 100 000 estimated SCC values as well as a couple negatives and positives in the four figures or more). I checked the climate sensitivity values in the trials .csv the generate_trials() function generated associated with the abnormal SCCs and it does not seem they are responsible for this (no particularly low or high values). I used a similar method for DICE and PAGE and they give sensible results, making me think the SSPs are not at fault either. One of my thoughts is that the Monte Carlo sampling might very rarely be yielding combinations of parameters that result in unrealistic SCCs, but I am not sure. It could also be that I am somehow mishandling the model. Does anyone have any idea of what might be going on?