Progress through computing power in the cloud

The temperature in an Alpine valley, the percentage of sunlight along the Norwegian coast, the wind at an airport in the Himalayas: these are all complex sites with highly localized weather effects. As a weather company, you want to perform well even in difficult locations like these. MeteoGroup’s newest innovation will make this happen and produces attractive and detailed images.

Our Weather Systems Team project manager, Wim van den Berg, enthusiastically discusses the innovations he and his team recently developed: ScaDo. “We are blazing a trail again with our newest forecasting method, which will enable users to make an effective estimate of basic parameters up to quite a high resolution.” ScaDo (Scalable Downscaling) is a forecasting method that can be scaled because it can handle virtually unlimited volumes of data.


Weather forecasts require comparisons
MeteoGroup continually strives to improve weather forecasting methods. Recently, for example, the MultiModel-MOS was entirely overhauled, and the meteorologists in our forecasting office are learning on a daily basis by comparing actual situations with forecasts. “The newest innovation, ScaDo, builds on the MultiModel-MOS,” according to Hugo Hartmann, senior meteorological researcher in the Weather Systems Team. “ScaDo interpolates, in a relatively inventive manner, that you can calculate a specific forecast for every place on Earth. That puts our clients in the driver’s seat.”

An important term used by our meteorological developers is ‘downscaling’. Briefly put, this means that you can prepare a forecast for every location based on the observations and forecasts for surrounding sites. Take as an example a mountaintop in the Alps that lacks a weather station. To arrive at a reliable weather forecast for that specific site, we look to all of the other weather stations in the region. Those observation posts are located at various elevations and at certain distances from the site in question, so we will have to assign all of these observations a certain weight. We refer to this as interpolation. For our Alpine mountaintop, for example, you would receive an equation (an algorithm) that forecasts the maximum temperature by assigning weights to the surrounding points: A (30%), B (20%), and C (50%). If our site is at an elevation of 3,000 meters, and the surrounding points are located at least 500 metres lower, then we will also have to apply general meteorological principles to arrive at a reliable forecast.

As Hugo reminds us: “Our MultiModel-MOS incorporates that feature. In other words, we’ve already taken care of it. The innovation primarily lies in the increased flexibility of the algorithms, because ScaDo takes better account of the constantly changing, specific weather conditions in the region.”

The sky is the limit in the cloud
This newest forecasting method dovetails perfectly with all of the options we now have for processing enormous volumes of data. Companies can no longer do this inhouse, as it requires unlimited resources. Those unlimited resources can be found in the cloud. When working in the cloud, the number of equations that can be made on an hourly basis is virtually limitless. “We enter an incredible amount of data into the system”, says Hugo. “The system then continuously examines how to downscale the forecast in an optimal way. This is a regression method known as kriging, the origins of which can be traced to the field of geostatistics. Using this method has enabled us to improve forecasts even further, and we’re seeing that reflected in the first quality assurance checks.”

Land use and model information
The ScaDo innovation is not limited to just working in the cloud; it also includes a large amount of additional topographical information. In addition to high-resolution elevation maps, this forecast method also includes detailed land-use maps. Elements such as wind and temperature are, of course, affected by whether an area encompasses grasslands, woods, or urban development. For example, wind speeds in a large city will generally be lower than they are in the surrounding countryside. On sweltering summer days, the heat island effect in cities is pronounced, while in outlying areas the temperature does not rise as high and drops more quickly during the night.


Wim van den Berg adds that “ScaDo also meets users’ wishes to generate reliable forecasts based on a small number of observations. The land-use and elevation maps and the self-learning principle that are included in the MultiModel-MOS anyway mean that the forecasts generated are as reliable as possible. ScaDo can make the difference even in locations without any observation stations nearby. Adding raw data from the model to the elevation and land-use maps yields results that are still valuable and usable.”

A long haul
Innovations do not just suddenly appear. A team of developers worked on this project for over a year. According to Hugo Hartmann, “It started with the newly invented algorithms and studying whether the interpolations worked properly. Afterwards, we added more intelligence, as well as the elevation maps and the landscape information. After taking these steps, of course, we had to demonstrate that the new method was superior to the old one. Fortunately, we succeeded in that and early this year we started developing the entire infrastructure. That was a project in itself, it took months on end. Everything was ready in the summer of 2018, and we went ahead with operational use.”

Naturally, putting the finishing touches on an excellent innovation does not mean that our work is done. As Wim says, “You always have to keep the future in your sights. Developments never come to an end. We are always on the alert for new possibilities and opportunities.”



Innovation improves weather forecasts, even in difficult locations.