In any sort of predictive natural science, absolute certainty is hard to come by; this is why we use things like p-values and error bars. Normally, as you refine your data over time, the error bars shrink towards the middle and the initial estimate turns out to be roughly correct.
This has not been happening with climate science. Across a wide range of studies, the error bars have been contracting towards the top of the range. The overall theme of climate science over the last decade is that the worst-case scenarios are coming true.
The major reason for this appears to be that there was a systematic bias towards excessive optimism, driven largely by the fear of accusations of exaggeration. Nobody wanted their work dismissed as scaremongering, and journals were unwilling to publish any studies that weren’t conservative in their conclusions.
This is particularly the case with large, international collaborative efforts like the IPCC reports. Basically, nothing got into those unless pretty much the entire scientific community agreed with it. Again, this biases the results towards excessive conservatism
