BUSINESS TECHNOLOGY MEDIA COMPANY
Companies
Sectors

AI can save SA from devastating future flooding

Read time 4min 40sec
Junaid Kleinschmidt, head of AI and advanced analytics at Altron Systems Integration.
Junaid Kleinschmidt, head of AI and advanced analytics at Altron Systems Integration.

Investing in artificial intelligence (AI) technology can help South Africa better prepare for devastating weather events, such as the deadly floods that recently hit KwaZulu-Natal (KZN).

So says Junaid Kleinschmidt, head of AI and advanced analytics at Altron Systems Integration, who notes the latest flood damage is a clear indication the country needs to invest more in technology infrastructure for resilience and disaster recovery.

This month, days of heavy rain across KZN led to deadly floods, with government putting the death toll at 448.

Critical infrastructure, including major roads, transportation, communication and electrical systems, were also impacted by the flooding.

According to Sihle Zikalala, premier of KZN, rebuilding and repairs after the storms and flooding in the province will cost about R12 billion.

Today, climate change and the accelerated occurrence of natural disasters could arguably be one of the greatest threats facing humanity, says Kleinschmidt.

“We’ve seen the disastrous impact the floods have had in KZN, causing loss of life, severe damage to property and homelessness.”

Disaster costs vs tech

Kleinschmidt points out it is essential to accurately predict climate change well in advance to determine ecosystem shifts and sea-level rise, but more importantly to plan for its potentially devastating impacts on human safety, food and water security.

“Artificial intelligence has awarded us the technology that can, in 0.25 seconds, predict extreme weather.

“The technology is still in its early stages and is generally costly, but is something of a necessity that weather bureaus, meteorological research institutes, and both private and public sector organisations can unite to invest in, given that the cost of the disasters we are seeing far outweigh that of the technology.”

As an example, he says, AI deployed and run on the correct GPU infrastructure can predict the behaviour of extreme weather events across the globe, days in advance.

“At 100 000 times faster than traditional numerical weather models, this is a significant step towards building a digital twin Earth. Climate change models predict the province will be hit by more extreme weather, more frequently, in the future.”

ITWeb contacted the South African Weather Service to find out if it is considering investing in AI technology but the questions were not answered.

Computing company Nvidia says short-term and seasonal weather forecasting can play a large role in decreasing the socio-economic and human costs of extreme weather.

In 2019, the company says, meteorologists warned local and national leaders in the Philippines of a torrential rainstorm looming about three weeks out.

The forecast gave communities time to weatherise structures and evacuate before the category four typhoon hit, saving lives and reducing overall damage to the region, says the company.

Kleinschmidt notes the South African Weather Service does not use AI tools in its weather forecasts.

“The South African Weather Service has been using global best practice meteorological tools; however, these tools run on seven-year-old infrastructure, which has become inadequate.”

He says the weather service set out the process late last year of procuring and implementing a high-performance compute ICT infrastructure.

“However, as we understand, this performs the same traditional numerical weather prediction forecasting models. Some of these models are expensive, time-consuming and limited. So, the AI tools, as far as we understand, are not being used.”

He explains that AI tools can help in two ways – short-term prediction (seven days) and long-term prediction.

“Stanford University researchers developed a machine learning model that uses atmospheric patterns to predict extreme precipitation, which may lead to flooding. This tool can be used in the short-term for predicting when a disaster may strike; and in the long-term, for building infrastructure that is resilient to climate change – think of bridges, where houses are architected and designed, urban as well as remote geospatial planning.”

Deep dive

On the available AI tools, Kleinschmidt says there are convolutional neural networks (deep learning algorithm) models to analyse large-scale circulation patterns and high precipitation events associated with extreme weather.

He adds there are other AI models available, such as a physics machine learning model that emulates the dynamics of global weather patterns, and predicts extremes with unprecedented speed and accuracy.

With these models, he says, a seven-day forecast is done in a fraction of a second, five orders of magnitude faster than numerical weather predictions (100 000 times faster).

For SA to tap into these models, it will need high-performance supercomputers, specifically designed and engineered to perform AI, which can run models on terabytes of data, analyse these global datasets and predict patterns, and visualise the data using Earth system models and geospatial mapping.

“There is a lot of research in this space, but we see successful case studies in the US, predicting weather in California, as well as floods in the US Midwest.

“Over the course of a 40-year period, we have seen 5 000 billion tonnes of ice melt in the Artic, and there are studies happening right now when looking at Canada, Greenland and other Arctic areas. They use AI technologies to look at the greenhouse effects, which in layman’s terms, causes the planet to warm due to CO2 emissions,” he concludes.

See also