The main purpose of those dams which were built right after 1949 in Huai river basin was flood control, not power generation. The dam you link, was built in 1958, but the power generators were installed almost ten years later at 1968, the electricity was just not the the priority for those dams. And it was not just one dam failure, but multiple dam failures, 62 dams in total failed in this extreme weather event, Banqiao was just the largest one, and only one with power generators at the time. After dams lost, there were constant floods in downstream areas, and almost all dams were rebuilt in 80s and 90s.

The 150k-230k death toll most likely is estimate for whole flood event, not just for this dam. We can say that if the constructions of the dams or coordinates between dams were better death toll can be much lower. But if dams were not existed at all, for that level of heavy rainfall, some people at some specific locations may be alive, but for the whole basin, I don't think there will be a huge difference.

Chernobyl is so different from a modern reactor, that it's not really useful to judge to current risks with existing reactors, let alone planned new Gen III+ reactors which are orders of magnitudes safer.

In fairness, the same is probably true for dams, as well - we've gotten a lot better at engineering them.

Recent California flooding makes me question this.

Yea, people talk about not being able to fully predict nuclear disasters, then turn around and claim sun, rain, and wind are totally predictive enough to rely on. I get that the risks vary, but man, a lot of people really don't understand statistics and probability.

And in particular the recent near-failure of the dam in Oroville. That could've significantly impacted hundreds of thousands of people.

The UN estimates for Chernobyl are based on the known bad LNT model for radiation exposure. Anything approaching 4,000 is unlikely if you apply more realistic models.

To clarify for others - LNT refers to "Linear No Threshold", the most conservative [0] model for assessing deaths from radiation exposure. Under LNT, if some amount of radiation leads to a 1% chance of death, 1/10th that amount leads to a 0.1% chance of death. This is most likely overly conservative, as cancer rates in the states do not seem to vary with radiation exposure as would be predicted by LNT.

So 4,000 is a good upper bound, if anything.

0: https://en.wikipedia.org/wiki/Linear_no-threshold_model

Thank you for the further explanation. However, I think that it could also be referred to as a "wildly exaggerated upper bound." Although I understand what you're saying, it's a terrible upper bound to use for risk assessments, or anything that you need to make sound engineering decisions based on. The true number of deaths from Chernobyl, outside of the actual plant fatalities, is likely closer to 0 than to 4,000.

Are those UN estimates time bound to any period of time and if not, what assumptions do they make of the future?

Also, per the Wikipedia link they included following famine in the numbers killed. Is that accurate?

The Chernobyl death toll estimates have been widely criticized as wrong because of the influence of the Soviet government trying to hide the problem. Some estimate the death toll to be at least an order of magnitude higher, and the area is still uninhabitable today many decades later.

The UN's estimate from 2005 is unlikely to have had much Soviet influence in its preparation.

How much data/information that the UN used was able to be influenced by the Soviet Union and its records? genuinely curious, I have no idea where the UN estimates come from nor a particular thought/agenda on the matter