It is a very different problem to predict the rainfall in a given area in a given day - you have to model the cloud dynamics - which is very poorly understood.
It is much easier to predict the average climate in a given place at any time. Because we have much more understanding of how the large scale climate works. And in certain places it doesn't vary much (eg continental interiors). The UK happens to be on the frontline between the cold polar air and warm, wet tropical air - and is inherently volatile in much more difficult to predict.
It is more straightforward to predict large scale patterns - eg warm air moving north, or cold air moving south, and make a prediction as to what that will mean to the climate.
What we can do it look at the fact that all over the world ice areas are melting. Storms are becoming stronger, which is a function of more heat in the system, species are moving northwards. Any number of indicators that on average in most areas it is warmer than it was. This is happening faster in the polar regions - probably because the large ocean mass in the middle of the earth helps to buffer things.
What I am trying to say - don't confuse an inability to predict where a thunderstorm will be (although often that is reasonably predicted) to a longer term larger scale change.
It is true that we can't predict the detail of what might happen in 20 years accurately - which is why they run ensemble runs with lots of different models and look for trends - repeated patterns. As computers get more powerful the ability to run models at finer and finer scales is improving understanding of the physics/chemistry involved. These are then used to model past situations to see if they can get the right results, and if they do then that knowledge of the system is then fed into predictive models. This is improving over time.
Regardless of what sceptics say - the majority of scientists working in the field would say that most indicators are pointing to the same general trend - rapid (on a global geological scale) warming. They then produce what if scenarios to try and gauge what that means. And most of those what ifs come out with a less than desirable result for the majority.
If we can buy time by slowing it down (and that is a big if) then we should at least try, and hope we come up with some geoengineering solutions.