Countries in the Horn of Africa are bearing the brunt of climate change. Devastating floods and droughts are taking their toll on people’s lives. Shruti Nath, a climate scientist at the University of Physics at Oxford in England, believes that artificial intelligence could help better predict extreme weather conditions and save lives in the region.
Using satellite data of cloud banks and cloud top temperatures, the AI makes predictions and attempts to identify upcoming extreme weather conditions.
“So for AI, there’s a part where it predicts. But to predict, it first has to train. And we train on historical observational data. So the input to our AI device is weather forecasts. It takes those weather forecasts and trains itself to match the observed reality based on the state-of-the-art satellite observations and the station data that we have,” she explains.
Like all AI models, it is constantly improving. The code rewards accurate predictions and penalizes inaccurate ones.
“So as it trains, the model learns more and more, and it is rewarded when it provides accurate predictions that match the observed reality. If it doesn’t, it is penalized “.
In countries like the United Kingdom, supercomputers are used to predict the weather.
For example, the Met Office supercomputer performs 16 trillion calculations per second.
The cost of these supercomputers, the data collection stations and the radar banks that power them is high, and they are not available in developing countries.
In contrast, the AI prediction code produced by the University of Oxford Physics can be used from a laptop.
“That’s the beauty of it, because the computation is so affordable that we literally have people running it on their laptops. So you can have this model and once it’s trained, you can just run it on your laptop and it’ll produce 50 predictions about all the possible ways the weather could evolve, which, as you said, would otherwise take banks and banks or supercomputers,” Nath says.
The project is still in its pilot phase, but the results are impressive, according to people on the ground in the Horn of Africa.
Isaac Obai, Food Systems Project Manager at the World Food Program, is in the UK for the project’s biannual meeting.
He said: “If nothing is done to disseminate information or early warning messages about these extreme weather conditions, we will see more and more people becoming vulnerable, more and more people being exposed to these weather conditions extremes and we will have a lot of lives lost, a lot of lives affected, as well as livelihoods Ultimately, the severity of the weather conditions is increasing and the frequency is increasing. climate, they are also becoming more aggressive. If nothing is done, many people will be affected, lives will be lost, livelihoods will be affected and poverty levels will increase significantly.
Predicting the weather in this region of Africa is notoriously difficult due to its changeable nature and the lack of weather stations that observe and record data.
With the 48-hour warning provided by AI weather forecasts, people in danger can be notified through text messages, emails, and even radio and TV broadcasts.
Maislin Gudoshava is a meteorologist at IQPAC (African Regional Forecasting Organization).
“So we are convinced that AI models can really improve the forecasting system. Not only on a short-term basis, but also on a short-term basis. By short-term, I mean a few days, but also the whole season. If we could really do that, I think it could really improve our early warning systems,” she says.
The pilot programme is currently being implemented in Kenya and Ethiopia, but there are plans to expand it across the region.
If it proves successful, it could be used in other parts of the world where extreme weather conditions, due to climate change, are devastating.
The program is a collaboration between the World Food Program, Oxford University of Physics, the Kenya Meteorological Department and IQPAC (the region’s meteorological service).
The project received support and backing from Google.
