There are plenty of comments made  on social media about claimed inaccuracies in forecasting cold weather.

However, many of these comments are made by those who are desperate for snow and cold. I feel their pain, but we have to remain scientific in our approach to forecasting cold events and be realistic about the possibility of the forecast changes from run to run.

As a bench forecaster I see no higher number of inaccurate forecasts for cold weather outbreaks as I do for what might be termed as 'normal' weather situations. Much of the comment is the result of disappointment that predicted chillier conditions won't arrive.

Let's consider what it is that makes a forecast?

We've seen a change from pre-computer days when forecasts out to five days ahead (and sometimes beyond) were based on empirical techniques. Rules and tips handed down between generations of forecasters (admittedly not that many generations  as forecasting is still a 'new' science) were used to make predictions which were the best that could be made given the information available.

Since the advent of computers and modelling it has become possible to forecast with a stunning degree of accuracy the weather of the coming hours, days and weeks. But it is managing of expectations which has not adapted to reflect the anticipated outcomes of the computer models.

Take, for example, postcode based forecasts. Absolute tosh! Those apps which we all have on our smartphones that tell us the 'weather where we are'? Nonsense! Models cannot predicted the weather 'where you are'. They can predicted the weather at a defined location at a point on a map, but there is absolutely no way that can be down to your street (unless you happen to be at the intersection of one of those points).  No, at best, and with a model costing many millions of pounds we can forecast operationally at points every 1.5km. And of course hills, coasts, lakes, factories, streets all have an impact on weather. Do you really think a model can actual 'see' these features? Nope,. it can't.

So, even with 'location based' forecasts all you are see is a best guess as to what the weather might do in the coming hours.

And what about forecasts of days and weeks ahead? Well, ask yourself, what is a forecast? Is it a prediction of what 'will' happen, or what 'might' happen? It's actually the latter. As one get's further head in time the variables impacting on that forecast increase. As a result inaccuracies are magnified and the forecasts ends up with a position where all he or she can forecast is the most likely scenario which may occur. Effectively, this is the basis of ensemble forecasting. take a set of start conditions, each slightly different, and then predict into the future. This gives a more objective statement as to what the most likely weather scenario may be.

I'm proud to say the development of ensemble modelling has been led here in the UK at the ECMWF and the quality of the output from their models is quite simply stunning. prof. Tim Palmer must take much of the credit for this. Read more about him on the Royal Meteorological Society's website And why not join the Society whilst you are there?

Now, onto the cold weather. Models are happiest when conditions fall within a certain range. That fits the algorithms under which they operate. The equations which go into the models are, mostly, not perfect. Once data is input which is beyond the normals the model would tend to see problems can be created. This can lead to perturbations in the models and some strange outputs. Thankfully these days these occasions are rare, but a quick hit of deep cold weather, such as we currently have ver eastern Europe, can cause problems.

Then consider the fine line between cold and milder conditions which can occur. At this time of year there is a sharp contrast between Atlantic mild and continental cold. A small error in modelling of a few miles can make the difference between -10C and +10C over the UK.

Now, these are not weathermen's excuses, oh no, I've got much better ones than those, but there are valid points which need to be made.

But back to those forecast expectations. Who is to blame for the public and other users of weather forecasts having an unreal grasp on what a forecast should tell them? Well, I happen to think meteorologists are largely at fault. The moment it was claimed that hourly based, postcode forecasts were possible was the moment we lost some professional credibility. It wasn't just in the UK of course, the USA led the way. It's totally reasonable for a forecast user the expect a forecast to be accurate if that is what is offered. Why should the user consider that the forecast may not be precisely for their location when that is what the name of the forecast is claiming?

And that dent in credibility then gets extended into the medium and long range. Our Weatherweb Premium subscribers are certainly the most experienced of forecast users. They understand the limits of forecasting days and months ahead, and as I frequently say, guidance is what we can give not a forecast.

If we can says that this scenario is more likely than that, then to me there is value in that forecast, and I guess to our loyal subscribers too. There are times when it is impossible to make a forecast as the atmosphere is too chaotic, and at those times one has to stand back, sometimes for hours, sometimes for days and let the weather do its thing. But eventually it comes back on-stream and a prediction is possible once again.

So, when you are looking at medium and long range forecasts use them wisely. Don't use them as you would a forecast for the next day or so, change your expectations. Listen to what the forecaster says is the most likely solution. This is generally good news for the weather forecasting profession because it shows we are still needed.

Above all, don't just go with a forecast because that gives you the weather you want to see!