Image: © Oană Liviu/Stock.adobe.com
AI is a ray of sunshine in the worsening climate crisis by providing accurate predictions of extreme weather events, allowing communities more time to prepare.
Human-caused climate change has resulted in widespread and rapid changes to Earth – particularly and disproportionally affecting those in the Global South.
A new report from the World Meteorological Organization (WMO) and other agencies warns that “we are far off track from achieving vital climate goals”.
It is noted in the report that 2023 was the hottest year on record by “a large margin”, with the oceans also heating to a record high, while the Arctic and Antarctic sea-ice extent reached a record low.
And this year looks to be following a similar pattern. Parts of the world are witnessing exceptionally high temperatures and facing many extreme weather events – from scorching heatwaves across Asia and droughts in Southern Africa to record‐breaking floods in southern Brazil and the “unprecedented” Category 5 Hurricane in the Caribbean.
There’s a higher than 85pc chance that at least one year in the next five will break 2023’s record and become the hottest year and an 80pc chance that the global mean surface temperature will exceed 1.5 degrees Celsius above pre-industrial levels in the next five years.
“The science is clear – greenhouse gas emissions are rising, global temperatures are shattering records and extreme weather is wreaking havoc with our lives and our economies.”
The report highlights the stark reality of the climate crisis, while also providing inspiration by looking at new technologies and innovations that could be “game-changers” for the planet.
AI revolutionises weather forecasting
The information and technology sector contributes between 1.5pc and 4pc of global carbon emissions, according to the World Bank. But the recent proliferation of energy-intensive artificial intelligence (AI) models will likely see that figure rising.
However, according to the new report, AI and machine learning (ML) can assist in speeding up weather forecasting, supporting day-to-day weather predictions and predicting extreme weather events faster, helping to mitigate some of the worst effects of the climate crisis.
AI and ML, in particular, are already being implemented in weather forecasting – Google-owned DeepMind claims its AI model Graphcast can predict weather conditions up to 10 days in advance and Huawei’s Pangu-Weather AI model can make much faster predictions than traditional numerical weather-forecasting methods.
Dr Alan Hally, the scientific lead of the AI transformation team at Met Éireann tells SiliconRepublic.com that Ireland’s national meteorological service uses both the traditional physics-based model, called numerical weather prediction (NWP), and the newer ML models together.
Ireland is part of the European Centre for Medium-Range Weather Forecasts, and through that, Met Éireann scientists get access to AI models that they can then compare to NWP models to get an understanding of how both models predict weather.
“AI based tools are just another feather in their bow really,” Hally says. The AI/ML models, he explains, have “learned from years and years and years of data” to make accurate predictions.
While running NWP models take significant computational resources and millions of lines of code, the newer ML models, once trained, can be run on a laptop, Hally says. Although training ML models takes a long time.
Once the model is trained, “then to run the simulation, you can do it in a matter of minutes as opposed to a traditional numerical weather prediction model which can take anywhere up to an hour,” Hally says. “So [the AI model] is faster and can be done with less compute.”
Once these models are adequately trained, these can help countries in the Global South to predict and mitigate extreme weather events faster, Hally adds.
The WMO report highlights the use of AI and ML models to predict large weather events such as hurricanes and cyclones. The report cites Cyclone Belal that hit Reunion and Mauritius this January, as an example. It said that the Artificial Intelligence/Integrated Forecasting System model was able to detect the cyclone two days before its impact. While at least four people died because of the cyclone, the BBC reported that more than 1,000 people were evacuated.
According to the WMO report, studies have shown that AI models can even predict the El Niño climate cycle up to three years ahead.
Traditional models still shining
However, the traditional models are still better for regional predictions and “probably will be for the next while”, says Hally. AI models are “still learning to deal with those types of events”.
Though the authors of the WMO report suggest that AI models may soon drive weather prediction. “Some evaluations have shown that AI/ML models are surpassing physics-based models in predicting some weather variables and extreme or hazardous events, such as tropical cyclones.”
Although, there are some challenges. Training AI models requires large, high-quality and consistent data sets and currently we only have uneven availability of data worldwide. Other challenges include unequal access and insufficient transparency, the WMO report notes.
While AI and ML models continue to push the boundaries in weather and climate forecasting, it is important to note that the use of technology in faster weather prediction is only an impact-mitigating measure and not a solution to the climate crisis.
The WMO strongly recommends a transdisciplinary approach to tackle the climate crisis – one that involves diverse actors including scientists, policymakers and local and Indigenous communities working together to develop solutions to the myriad effects of climate change.
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.
