Page tree
Skip to end of metadata
Go to start of metadata

Conclusion

The IFS, ECWAM and NEMO forecast models and their ancillary analysis and parameterisation systems use state-of-the-art techniques.  Nevertheless, no model or system can capture with precision the state of the atmosphere or ocean, nor the handling of momentum, moisture, temperature and radiation fluxes, no matter how much the resolution is reduced or the detail of the physics increased.  There is a constant effort to improve models and techniques, but there will always be shortcomings in the results.  The forecaster continues to be employed to identify and correct, or at least minimise, any error by using their experience of the forecast area (be it local or large-scale) and importantly by understanding the models and their strengths and weaknesses.

The aim of Forecaster and NWP Model together

Together the forecaster and NWP model can produce predictions with a skill greater than either can alone.    At the time when ECMWF was founded, it was estimated that medium-range weather forecasts would lead to large economic gains for society.   Currently, forecasts are, on average, synoptically useful for over a week ahead, with extreme weather events generally forecast three to four days in advance.

An NWP model can provide a firm basis of the expected evolution.  It can, in some circumstances, use its high resolution to indicate a finer detail in time and space of an event.  However, the forecaster remains crucial to the forecast process.  It is in assisting customers with their decision-making processes that professional weather forecasters can really ““add value””.  This is because they have the benefit of their education and experience, a unique position in the centre of the information flow, and experience in the skill and characteristics of different forecast systems.  They can understand the shortcoming of the NWP models and can make allowances for them.  Furthermore forecasters are able to understand the importance of a forecast or threshold to the customer.   

Ensembles provide an explicit, detailed representation of uncertainties, and the potential for unusual events.  They enable forecasters to assess and communicate uncertainty.  It can be important for the user to know about the uncertainty in a forecast –  to be informed of other possible weather scenarios and the probability of the worst weather within these.

Weather forecasting has never been primarily about getting it ““right”” or ““wrong””, as if in some quiz show, but rather about providing information for decision-making to maximize advantages and minimize disadvantages.  The proportion of freely available, automatically generated weather forecasts has increased tremendously because of the expansion of the internet and mobile phone "apps".  Such forecasts might satisfy needs during normal weather conditions but not in situations of extreme or high-impact weather, particularly as some suffer from flip-flopping of the forecast.  A decision to evacuate an area will never be made purely on the basis of automated NWP output.  And there is not, and might never be, one single source of NWP information providing a single clear message, particularly in situations threatening extreme or high-impact weather.  Medium-range forecast information is not always used to its full potential; when decisions on the protective action to be taken against extreme weather are made, medium-range forecasts too often serve only as background information.  It is important that we continue to build trust in medium-range weather forecasts as it would make them an even more essential part of core meteorological activities within meteorological services, in particular for warnings of extreme events.

But forecasts only have value if customers use them - say to make a decision or take an action which would not otherwise have been made.  To make a good decision customers need to know the probability and, importantly, the impact or the consequences.  It is vital that the potential for high impact events is identified, no matter how low the probability.  It is better to warn of a possibility of a severe event rather than ignore it until too late.  This is not over-forecasting - this is stating the possibility of a severe event and customers can then make plans (or not) as they see fit.  

Improving the forecast system

One way to increase the trust in medium-range forecasts is, of course, to further improve the skill of the ECMWF forecast system.  By way of example the skill of the ECMWF deterministic forecast (HRES) has increased by a day per decade since 1979 and it is likely that this trend will continue, thanks to improvements planned for the next ten years by:

  • systematic increases in the resolution of the assimilation and forecasting systems,
  • enhancement of the representation of physical processes,
  • exploitation of better data for assimilation, in particular from satellites.

However, an overall improvement of the deterministic forecasts may not be enough to increase the use of the medium-range forecasts.  It is not enough that the forecasts are high quality, they must also be trusted. 

Building trust in individual forecasts...

...but sometimes a deterministic forecast cannot be trusted

The ultimate measure of a good forecast system is the quality of the decisions made that are based on it.   Although decision-makers have always had high confidence in the skill of the forecasts in general, there is less confidence in the skill of specific forecasts for a particular event (eg when extreme weather is likely).  It is therefore essential that every single forecast is so trusted that it can be used for decision-making.  A good forecast that is not trusted is a forecast without value, irrespective of how well it verifies retrospectively.  A fourth planned improvement is therefore to:

  • identify and implement better ways of quantifying the forecast uncertainty using ENS output.

The ENS is able to pinpoint those model solutions on which the highest reliance can be placed (which may or may not include the HRES forecast), and thereby increase end-users’’ willingness to make decisions (based on these, the most likely outcomes).  Whether or not the deterministic forecast is considered good guidance should not be relevant, but often this can be problematic for the customer who can be beguiled into accepting a deterministic forecast (HRES) because it appears so detailed. This is an area where the forecaster has a vital consultative role to play (see Advantages of Uncertainty).   

The optimum solution in all cases is to present the uncertainties in probabilistic or similar terms. 

The role of the forecaster in the medium-range

The demise  of weather forecasters ““within 5-10 years”” has been prophesied almost since the start of NWP.  However, there are probably just as many human forecasters in operational forecasting work today as ever before, and increasingly so in the commercial sector.   Clearly they continue to supply a perceived need.  What distinguishes professional meteorologists is their ability to make use of uncertainty.  In highly predictable weather situations anybody can confidently report ““what the weather is going to be”” just by reading off the computer output, but in difficult weather situations it is only the professionals who can express uncertainty in a helpful and informative way to best serve the customer.  This can be in broad terms:  ““This cold air might arrive over our area later in the week””...””The rain might exceed 50mm in places””...””We cannot exclude hurricane force wind gusts.””   But the probability of an event, or a weather parameter exceeding a threshold, expressed as percentage probability is also a powerful method of giving useful information to the end-users and allows them to take appropriate decisions based on their own knowledge of the risks they can or cannot accept.
It is fundamentally this uncertainty information that adds ““extra value”” to the forecast.




This User Guide should be cited as follows:           Owens, R G, Hewson, T D (2018). ECMWF Forecast User Guide. Reading: ECMWF. doi: 10.21957/m1cs7h




  • No labels