We don't know — the spell text leaves it open, which in 5E generally means "DM's call".
The word "predict" just means to foretell, to say in advance. Nothing in plain English suggests that predictions are accurate — in fact, people make predictions that turn out to be wrong pretty much constantly.
Now, common sense suggests that since it's magic, it should be pretty good. In a comment to another answer, @SeraphsWrath notes that the cantrip is Transmutation, not Divination — which suggests that the spell does not really look into the future at all. But, this is a multi-purpose utility spell and some of the other things don't really fit — if these were individual spells, the non-weather-forecasting "create harmless sensory effect" use might be illusion, and lighting campfires might be evocation. So, I think there is a lot of room to make more or less of the spell school if you like.
Personally, as a DM, I tend to play this rather strongly. Let's face it: Druidcraft is probably the weakest "spellcaster class theme" cantrip, and druids don't get a ton of cantrips — and there are lot of solid mechanically-adventageous other options. Your player took this to enhance the flavor of playing a druid.
I haven't come across a situation where I had to consider Control Weather, but I did have a case where a magical future event was likely to change the weather, and on that instance, I described the resulting sun orb as weirdly colored and tinged with magic. I allowed an Arcana roll to investigate, and used it as an opportunity to give the party some hints as to the situation they were walking into.
I think this same thing might be appropriate if you know Control Weather is likely to be in play.