Most of us take the accuracy of weather forecasts for granted, not giving too much thought to the science, technology and global cooperation behind the information we receive.

The use of satellite data in weather forecasting has demonstrably increased accuracy. What challenges exist to achieve still greater accuracy?

Impact of satellite data

European Centre for Medium-Range Weather Forecasts (ECMWF) Principal Scientist, Dr Tony McNally, from the EUMETSAT Numerical Weather Prediction SAF, says regular investigations into how well weather forecasts perform, both in “normal” conditions and “extreme” weather events, are carried out to evaluate the impact of satellite data on accuracy.

“Time and time again, these studies show satellite data are the most important source of observations in our system,” Tony said.

The German Meteorological Office (DWD) used Meteosat imagery as one of the inputs to diagnose the fast development of extreme phenomena on 29 May 2016 and issued red warnings, enabling local authorities to take steps to protect life and property and prevent a higher death toll from the ensuring floods. The image shows the circular-shaped thunderstorm which caused catastrophic flooding and resulted in severe damage in Baden-Württemberg and Bavaria.

“In routine conditions, northern hemisphere forecasts would be around half a day to a day less accurate without the input of satellite observations.

The SAF network
EUMETSAT has a network of eight specialised Satellite Application Facilities (SAFs) which are part of the organisation’s distributed ground segment. The SAFs are responsible for research, development and operational activities not carried out at the central facility at EUMETSAT headquarters, such as development of specific software, data or products. The SAFs are located within the National Meteorological Services of EUMETSAT’s Member States or other entities linked to a user community. You can find out more here. The Numerical Weather Prediction (NWP) SAF, with the UK’s Met Office as the lead entity, exists to increase the benefits from numerical weather prediction by developing techniques for more effective use of satellite data, and to improve the exploitation of data and products from EUMETSAT satellite programmes, and related programmes from other agencies.

“Southern hemisphere forecasts would be up to two or even three days less accurate.”

The impact on forecasts of severe weather events, such as storms, hurricanes and tropical cyclones, can be even more significant.

Models showing the expected paths of hurricanes which hit the Caribbean and the United States are generally highly accurate.

“But when we rerun some of those cases without satellite data, the results can be dramatic,” Tony said

An investigation of the forecast path of Hurricane Sandy, in 2012, showed that, without the input of satellite data, ECMWF model forecasts would not have provided critical guidance five or six days in advance that the storm would strike the New Jersey coast, which it ultimately did with devastating effect.

The AVHRR instrument on EUMETSAT’s Metop-A satellite captured this dramatic image of Hurricane Sandy over the northeast coast of the United States on 30 October, 2012

Similar studies during the 2017 Atlantic hurricane season have confirmed the key role played by satellite data in producing accurate forecasts.

“We obviously can’t stop the hurricanes but we can provide the best possible guidance that allows people time to make preparations and, if necessary, evacuate,” Tony said.

“We take a systematic and forensic approach to evaluating the quality of our forecasts, finding what went right and what went wrong in a particular forecast, to learn how to make the systems better.”

How much more accuracy is needed?

“The reliability of our medium-range weather forecasts, e.g. around day five, is currently very high and arguably sufficient for many people’s requirements,” Tony said.

“But the challenge, and what users need, is for us to extend that accuracy out to longer range forecasts – producing equally reliable predictions 10 days or more ahead of time,” he said.

“If you want a forecast for half-an-hour’s time, you could just look out of the window at the clouds,” Tony said.

“But if you need to make longer range forecasts, you need a global picture of what the atmosphere is doing right now – and that’s where satellites are particularly crucial. Only satellites can give you this global picture.

“For example, if you want to know what the weather will be like in Europe in eight days’ time, you need to look at what is happening over the Pacific Ocean. What’s happening over the Pacific Ocean now will affect Europe eight days from now.”

To achieve longer range forecast accuracy, our models will expand to take into account all aspects of the entire Earth system (atmosphere, land, ice, ocean and composition) on ever finer temporal and spatial scales.

These models will require new types of satellite observations providing highly detailed and accurate information, not just on meteorological parameters such as temperature and wind, but on variables such as aerosol, carbon dioxide, ocean state, ocean salinity and more.


In a few years’ time, EUMETSAT will begin launching its next generations of geostationary and low-Earth orbiting satellites, with state-of-the-art instruments.

These satellites, together with those of international partners and the European Copernicus Sentinel satellites, will dramatically increase the amount of weather and climate observation data.

But this is just one side of the equation when it comes to increasing weather forecasting accuracy.

Along with more and better satellites, scientists will have to be better trained to exploit these new measurements, Tony said.

The EUMETSAT NWP SAF provides training to scientists from National Meteorological Centres, satellite agencies (including EUMETSAT) and academic institutions who are either new to satellite data assimilation or wish to learn more about a specific area.

It’s a one-week course conducted as part of a larger training provided by the ECMWF.

“It’s a face-to-face course limited to around 30 people and is typically oversubscribed,” Tony, who is responsible for the course, said.

“We aim to explain the most important science issues related to satellite data assimilation, show how things are done in practice at the ECMWF and incorporate a ‘clinic’ element to the course, where participants can interact with our experts and ask questions about different problems or situations in their own area.”


Posted by Ruth Evans

One Comment

  1. Tillmann Mohr, DG of Eumetsat (ret.) 8 February 2018 at 13:46

    Very good article. However, I have a comment on the last para. before HOW. It’s stated that the Earth System models need not only “meteorological parameters such as temperature and wind, but on variables such as aerosol, carbon dioxide, ocean state, ocean salinity and others”.

    If one takes the definition of meteorology: Physic and chemistry of the atmosphere and the surfaces the atmosphere is interacting with, than all the parameters mentioned are meteorological ones and not only temperature and wind !!!!


Leave a reply

Your email address will not be published. Required fields are marked *