Translation:

Interview with Nick van de Giesen, professor of water management

Farmers in Africa need your computer for their weather forecasts
(Image) Rain like this in Lagos, Nigeria should be predicted more accurately.

A lot of computing power is needed to test models that should lead to better weather forecasts in Africa. Volunteers worldwide can contribute to this with their computer, says relevant professor Nick van de Giesen. "It is very important for farmers to know when it will rain."
(interview by Cor Speksnijder, November 24, 2019, 16:57)

From now on, every owner of a computer can contribute to improving weather forecasts for African farmers. He or she can provide unused computing power to test models that should ultimately ensure that precipitation in Africa can be predicted more accurately. Researchers at TU Delft have started the calculations, using this combined, donated computing power.

Precise forecasts of precipitation are crucial for African farmers. Current forecasts are mainly based on satellite data. They are not accurate enough to use at the local level. A satellite can register clouds, but cannot indicate whether it actually rains. For more precise forecasts, weather stations on the ground are needed.

The Delft professor of water management Nick van de Giesen, together with an American colleague, set up an organization (Tahmo) that in recent years, with financial support from the World Bank and IBM, has been placing some five hundred measuring stations in twenty countries south of the Sahara.

They want to test models with the data from ground stations and from satellites, which should ultimately lead to better and more accurate weather forecasts. This requires a lot of computing power. This can be delivered using the World Community Grid from IBM. By downloading a free software program, volunteers can relinquish computing power from their computers to test prediction models.

Rain forecasts for the next day in Africa are just as unreliable as those in the Netherlands for ten days, says Van de Giesen. "People hardly listen to the weather forecast. While it is very important for farmers to know when it will rain. They want to know when they can harvest best, when they can best let the harvest dry. Farmers are careful with fertilizer and do not want everything to wash away when they have just sprinkled. Or take farmers who have given their cows an anti-tick remedy - after a heavy rain shower they can start again."

Do precipitation patterns in Africa differ from Europe?
"The rainfall in Africa is comparable to summer rain showers in Europe. Showers that just start somewhere, that are hard to predict. Rain can be very local: one side of a city can be under water, while the other side is dry. With the current weather models for Africa, we have a resolution of tens of kilometers. Then you don't see what is happening. If you display a human face in one pixel, you will not see any details."

How accurately can the weather be viewed thanks to the ground stations?
"We ultimately aim for the same density as in the Netherlands. That would amount to 20 thousand stations under the Sahara in Africa. In some countries that will not work. In countries like Kenya and Ghana we are getting pretty close: there we have one station per 1000 square kilometers."

How do you ensure that the system continues to function in the long term?
"I have done many scientific projects in Africa, installed weather stations and trained people. Every time the project was over, it often meant the end of the observations, the measurements and the material. The ground stations must be maintained. That is why most stations are located at schools. We try to persuade teachers to check if everything is still okay. We will ask a fee from commercial companies or organizations that want to use our data. They are free for governments and science."

When can farmers see the first good weather forecast?
"I expect that within a year we can pass on the first forecasts to the meteorological institutes of Kenya and Ghana. We are now going to calculate data from the past two years to see if the models work properly."

Thanks, this is nice!

Thanks Adri for your translation.

Interesting! Until now we though data to process is from 2018-07-01 00:00 up to 2019-06-30 23:45.
The last sentence would mean that the now running year 2019-07-01 to 2020-07-01 have to be analysed too.

Ditto what Jan said!!!