For the sake of clarity, and because it's your primary point, I do think it's worth to point out even in the article you cite nuclear is the cleanest, but safet

For the sake of clarity, and because it's your primary point, I do think it's worth to point out even in the article you cite nuclear is the cleanest, but safety depends on how you perform the estimate. I found the comparisons in analysis they performed here: https://ourworldindata.org/safest-sources-of-energy#why-do-biomass-and-nuclear-estimates-vary-how-are-nuclear-deaths-calculated https://ourworldindata.org/safest-sources-of-energy#why-do-b... interesting. One estimate results in nuclear being more deadly per TwH produced than wind, hydro, and solar (about 2-3x), and the other results in it being less deadly by the same amount. It is important to note that a significant portion of deaths from nuclear seem to come from a few large events; with this in mind, one study includes Fukushima, and the other includes Chernobyl. [1] A larger factor in the discrepancy is one study's very conservative estimate of "deaths from occupational effects", which I interpret as the net added risk versus background from radiation exposure for all workers involved in the maintenance and fueling of the plant. The true relationship between radiation exposure and death risk seems to be under dispute: the more conservative estimate now appears to be an overestimate because it assumes a linear relationship, while the article states consensus is shifting towards a threshold model where one assumes that very low exposure to radiation (the type of exposure most nuclear plant workers experience, I presume?) does not increase cancer risk. [1] perhaps this is an instance of a more general difficulty in statistics: that of estimating the mean of heavy-tailed distributions (see e.g. https://arxiv.org/pdf/1906.04280.pdf https://arxiv.org/pdf/1906.04280.pdf). I'd guess that deaths from nuclear events follow a heavy-tailed distribution, which implies the mean could not be a good estimate of the true impact. But as I have not actually taken the time to investigate the analysis of nuclear deaths performed in the studies, I can't say how well existing analyses do this.