Do we get to apply that standard to past IPCC predictions too ?
Reality is that the current temperature anomaly is -far- below IPCC 1990 predictions. So ... falsified ? It is, in fact, so that the temperature anomaly is below every IPCC prediction of more than 5 years back, best case scenarios, with the lone exception of the 2010 report's optimistic prediction for last year (2016 however is out of the 95% interval of that one too, unless something drastic happens).
There are reasons for this. Such as that there was a bit of a pause in global warming from late 90s to late 00s. But the fact that the models failed to see this coming, and only incorporated it afterwards of course is further indictment of those models. It kept warming, mind you, except perhaps for 1 year, just way, way slower than predicted.
There is a joke that the IPCC now must be correct for the next 22000 years. Why ? That's the only way to make their 95% certainty predictions actually be true in 95% of the cases, given one prediction every 5 years. And this is not the only consideration. Obviously things like the Kyoto accords were based on these predictions. They turned out wrong. So should we now abandon those accords and laws, at least until the evidence can be examined ? That would be very scientific, or at least, it would be very scientific the way I was taught statistics. To say that this would be politically problematic, a lot of work and very, very, very impractical is an understatement.
And of course, more recent science cannot be used until there is enough evidence to support the correctness of the model. Normally one would say, half to validate, half to be used. The problem is that you cannot trust researchers who've seen (and evaluated) against all data they've seen. So you need to use future data to validate models, but of course all IPCC models so far have crashed and burned in that comparison. So you cannot just say "let's just use the 2015 predictions", because if you just put down the temperatures from last year for the next 100 years you would have done better than the IPCC, so far (meaning even outirght climate deniers' models would easily have 95% accuracy one year out).
And that then violates intuition. The graphs show a clear and consistent rising of temperature for coming up on 160 years now. Obviously the truth is that we don't know why, or at least, not to the level of detail required to predict the effect of policies we implement. Now of course, both cause and effect seem obvious - BUT the models fail to predict effect given cause, so technically a mathematician should use the notions taught in philosophy of science, as applied to statistics, and reject the models. One would be obliged to point out that maybe the models are wrong because their assumptions are wrong, or at least inaccurate.
I would say that we should just train everyone in how to analyze the data, but I realize it's not very practical. Plus it would actually mean climate change would be tempered (one would hope that policies would be moderated given that the predictions they're based on turned out to be overly pessimistic).
All this is ignoring theoretical problems, the most obvious one is that these models are based on statistics. Statistics that studied cause and effect in a situation ... that no longer exists and very likely will never return. We are using models to predict the future based on values that lie outside of the range they were modeled on (predictions out of range is one of the statistical fallacies: you cannot model based on data between 0 and 10, then have confidence in predicting what happens at 15. Everybody does do that of course, but it's wrong, as obviously the distribution of your source data is different from the prediction data. But we don't have data with the same distribution, because the situation is new). Technically that means they're wrong, period. What do we do with that ?