Not all topics warrant certainty, but given the power of news outlets forming opinion and thereby influencing the population's behavior, I do think certain topics demand responsible reporting which doesn't report anything at all if uncertain. It's the same as a good police detective not disclosing speculations to the media because they pursue many leads but only conclude one, and you don't want a mob go lynch people. The same logic applies to news outlets forming people's opinions. That's why I think there's a need for, admittedly few, certain-report-only news agencies. That way, if you read posts on TheSun or E!Online, you get accustomed to not taking it seriously, forming doubt that this is most likely speculation. Once something gets confirmed unquestionably, it can migrate to one of the few vetted-only news outlets, if that's something they cover.
That said, our weather models are pretty good but not good enough to make certain predictions that far away into the future, but they can for the next few hours.
It's like a software company's model of code branches. The Apple/Google/whatever filesystem team works on something, it gets pushed into their level of production branch, then it percolates up to the shared production kernel branch, and after a couple more layers it hits the common branch, which is what public production binaries are made from and consists of kernel, userland, foobar modules all merged together. Not all software shops operate this way, but it's what size of a project can demand after it hits certain amount. The linux kernel works this way too, to name a successful non-commercial project. You can argue this doesn't prevent regressions, and that's true, but it's hard to deny there would be more regressions (aka false reporting) with unfiltered (aka unvetted) reporting.