The future role of the weather forecaster

Authors


The recent article ‘The forecasting challenge of waving cold fronts: benefits of the ensemble approach’ (Young and Hewson, 2012) gave an illuminating snapshot of the current situation in operational weather forecasting, not only in the UK, but world-wide: a range of numerical weather prediction (NWP) models producing more or less diverging forecast solutions. The duty forecaster is expected to ‘add value’ to these model solutions by either deciding which one of them is ‘The Model of the Day’ or manually improving on the generally preferred model.

The forecasters’ task to find the ‘most likely outcome’ is made difficult due to the run-to-run variations which can be quite considerable and increasingly so with longer lead times. Young and Hewson discussed in some detail their search for this outcome in one particular event, using not only combinations of the latest two deterministic NWP runs but also products from the multi-member ensemble prediction system (EPS). They were successful, but the reader is left in the dark as to how often this is the case and whether the forecast is delayed in the process. And how is the ‘added value’ reflected in the automated NWP data-feeds?

This article also raises the important question: is the future of forecasters linked to their ability to improve on deterministic NWP? One of the rationales when the first EPS was launched at the ECMWF twenty years ago was indeed to help the forecasters cope with ‘forecast jumpiness’. However, there were quite divergent ideas on how that should be done, and those ideas are still with us today. One interpretation is that the EPS is subordinated to the operational state-of-the-art deterministic forecast (OPS): the EPS should indicate how much the OPS could be trusted. Young and Hewson's approach, although very elaborate, also sees the EPS as subordinated to the OPS.

However, an alternative view has been that the EPS is quite able to stand on its own feet. Practical experience and statistical verifications indicate that the average (or median) output from the ensemble is indeed the superior provider of ‘the most likely outcome’. The spread around this average tells how much trust can be put into this output, and, converted into probabilities, can warn about high-impact weather developments. It would therefore be quite possible to base the entire weather forecast guidance on the EPS. This is even more true today than it was 10–20 years ago thanks to the development of ensemble systems for short-range, high-resolution NWP.

An alternative vision of the future is one where the forecasters leave the ‘most likely’ scenario to the computer and concentrate their skill and experience rather on the uncertainties in the forecasts, in particular the risks of extreme or high-impact weather (see e.g. Persson and Riddaway, 2011). They need not base their forecasts on the probabilities emerging from any single EPS: there are other EPS systems, as well as other NWP models and different statistical interpretation schemes, whilst observations from weather stations, radar and satellites will highly influence the forecasters’ uncertainty estimations in the short range. Some of the methods tried by Young and Hewson might be useful in this context.

Experienced and well-educated forecasters are, however, not only needed to estimate the forecast uncertainties, but also to communicate them in one way or another (not necessarily as percentages). When I entered meteorology almost 50 years ago I was told that, thanks to computers, in 5–10 years there would be no need of weather forecasters. This might still be true if their future role is nothing more than manual modifications of ever-improving computer forecasts. But considering that estimating and dealing with risk information obviously is poorly understood by the public and even by paying customers, I would suggest that skilful weather forecasters will be needed in these capacities for at least another 50–100 years.

The fact that there have never been so many weather forecasters around as there are today might suggest that most of them indeed see their role rather as advisers to the recipients of the weather forecasts than competitors to the computer models.

Ancillary