In the clouds

Clouds always used to be the least understood element of the weather system, but that is rapidly changing . Computer clouds increasingly correspond with those in the sky, which promises weather forecasts at street level and more accurate climate scenarios.

Sniffing post

It´s known locally as the ‘sniffing post’. Indeed, the 213 metre mast in the village of Cabauw, in the Dutch province of Utrecht, was once used to monitor air pollution. Even after forty years, the tower stands out strangely in the flat landscape: 47 steel cylinders, each 5 metres high and 2 metres wide, are stacked to heights of 50, 100, 170 and 210 metres and anchored to the ground with wire stays. That mast put Cabauw on the world meteorological map as one of the top locations for atmospheric research: studying the complex system of clouds, dust particles, radiation, precipitation and turbulence in relation to the Earth´s surface.

In the mid-nineties TU Delft became involved in the atmospheric research, with organisations like the Royal Dutch Meteorological Institute (KNMI), the Netherlands Organisation for Applied Scientific Research (TNO), the Dutch National Institute for Public Health and the Environment (RIVM), the Institute for Marine and Atmospheric research Utrecht (IMAU) and Eindhoven University of Technology (TU/e). Professor Herman Russchenberg (Geoscience and Remote Sensing) recalls how all the equipment was initially brought to Delft to enable the radar to be used on the EEMCS building. “That campaign proved to be the start of Cesar (Cabauw Experimental Site for Atmospheric Research, ed.) because we had noticed how valuable it is to install all instruments together for long periods. It provides extra information.” Cabauw was a better research site than Delft due to its more inland location.

TU Delft operates two radar systems at Cabauw: a mobile trailer-mounted system for analysing rain clouds (Tara) and a 360 degree scanning radar with a higher working frequency (10 GHz), which also registers fog and drizzle (the IDRA drizzle radar).

At the symposium held on 26 October to mark forty years of Cabauw, Russchenberg said that understanding the water cycle (evaporation, clouds, precipitation) is the key issue for the coming time. “What’s happening now and how will an increase in CO2 change that? That is the main question.”

Cloud calculating

“We’ve actually thrown in the towel,” Professor Harmen Jonker (Geoscience and Remote Sensing) explains about the LES cloud programme. This Large Eddy Simulation programme is a detailed simulation of cloud formation. “In relation to LES researchers said: predicting cloud formation with clever models doesn’t work.  You just have to do the hard calculations using fluid mechanics equations.” One of the pioneers of the programme in 1992 was Professor Pier Siebesma, who currently also works at KNMI.

The programme calculates clouds in a box of 10 by 10 kilometres and several kilometres high. The spatial resolution is 10 to 20 metres and smallest unit of time is the second. It works well, but calculating the development of millions of cells requires a lot of computing power.

Over the last two years, the technology has become more accessible thanks to the use of graphics cards. One graphics card (developed for game computers and on sale at 200 dollars) proves to be able to calculate as fast as 64 supercomputer processors. Frits Post (Computer Graphics) and PhD candidate Eric Griffith have converted the Fortran code to Cuda, a language that controls GPUs (graphic processing units). Together with them, PhD candidate Jerôme Schalkwijk developed a modified version of the cloud programme: Gales (GPU-resident Atmospheric Large Eddy Simulation).

For a year now, this programme has been calculating the clouds around Cabauw based on temperature and humidity distribution, while special cameras continuously register the cloud cover. This results in split-screen films with the actual cloud on the left and the computer cloud on the right. Jonker: “The result is surprisingly good. Sometimes we do get it wrong and the cloud appears later than in reality. But apart from that, there’s real ‘weather’ in the programme. You’re starting to see the same turmoil on the monitor that you do outside.”

Differences between climate models

In the early days, research at Cabauw focused on the heat and moisture exchange between the air and the soil. The data that were collected in the seventies and eighties have now been processed in weather and climate models. The focus then shifted to clouds, the most poorly understood elements of weather and climate. Siebesma summarises these developments as follows: “Even with the first generation of climate models in the seventies, scientists recognised that clouds are the most uncertain factor in climate models. Based on the same rise in atmospheric CO2, different climate models consequently generated different global warming predictions, varying from plus 2 to plus 6 degrees. It was not then known which type of cloud caused this uncertainty. We now know it to be the low-hanging stratocumulus clouds. One model expects an increase of these clouds, another expects a reduction. Over the next five years we hope to discover which cloud processes are responsible for the differences between the various climate models.”

Siebesma says he is ‘in the middle of it all’. At KNMI he is a climate researcher and at TU Delft, together with Dr Stephan de Roode, he seeks to translate the results of the LES cloud simulation programme to a larger scale because, whereas LES works on a grid scale of 50 metres at most, the resolution of the weather and climate models is 50 to 100 kilometres.

Based on the cloud simulation programme, which uses the average temperature and humidity in its calculations, the researchers seek to enable the large-scale model to show the extent of cloud cover, how much radiation it will block and how much rain will fall from it. This should reduce the differences between the various climate models and the uncertainty in climate predictions. In the next IPCC report there will be just as much uncertainty as there was five years ago, Siebesma says. However, he predicts a significant improvement in another five years’ time.


The cloud simulations around Cabauw might be successful, but then the area concerned is of course only small. However, now that computing power in the GPUs is so cheap, Jonker and his colleagues have come up with the idea of using 256 graphic processors at once to run Gales in an area of 400 x 400 kilometres. The result is breathtaking. On a beautiful spring day you can watch a line of delicate cloud forming in the course of the morning. The IJsselmeer remains largely cloudless.

Compared with the actual weather, the patterns and movements are practically identical, although often not in exactly the same place. Jonker thinks this could be improved if the Gales calculations were based on information from a cloud radar. That would enable a unique forecasting method.

Russchenberg can see such a system in Rotterdam. A rain radar will be installed there next spring on top of the Nationale Nederlanden building. The objective is to be able to produce precipitation forecasts at street level to limit the inconvenience in the city. Russchenberg: “If you know what’s coming, you can take appropriate measures. Flooded streets and emergency response by the police and fire service cost several million euros a year. If you can cut that by half, the system will soon pay for itself.”

Enabling such fine scale forecasts throughout Europe requires hundreds of automated radar systems with a rain radar, a lidar (measuring cloud height) and a radiometer (measuring the radiation and hence the temperature of the clouds). This might seem science fiction, but a German firm is already working on a cheaper radar for mass production. “We’ll soon have a network of little Cabauws throughout Europe”, Russchenberg predicts.

Watch the clips on cloud formation at

Redacteur Redactie

Heb je een vraag of opmerking over dit artikel?

Comments are closed.