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.
