Cedric Stephens | Tracking risks through artificial intelligence

ADVISORY COLUMN: RISKS & INSURANCE

Last week’s column – headlined “Hospital hazard management” – was written to try to change the direction of the conversation about the non-compliance with fire-safety rules by some public institutions that The Sunday Gleaner started.

Today’s will attempt to alter the direction of the discussion that Prime Minister Andrew Holness amplified last Tuesday in Parliament about the financial impact of the heavy rainfall that has been affecting the island during the latter part of October. Estimates suggest that the costs will rise sharply above the initial $2.7 billion.

It is raining heavily at the time of writing. Water is accumulating in my backyard. This has not happened before. It is coming from the rear of my neighbour’s premises. I live near to a large floodwater channel that forms part of the Sandy Gully drainage system. Google Maps says that the system is located on the Liguanea Plains.

A St Elizabeth melon farmer contacted me. The absence of crop insurance to protect his on-the-ground exposure, he argued, is a major deterrent to investment in agriculture. His farm suffered significant losses because of flooding caused by the October rains. And this year’s record-breaking hurricane season still has many more days to run before it ends.

Government’s response to the many challenges caused by the flooding seems stuck in 20th-century reactive mode. Establish ‘no-build zones’ in flood-prone areas and on gully banks. Allocate more funds to each member of Parliament, and disburse rehabilitation grants to the neediest of citizens.

These efforts, though laudable, are inconsistent with policies to promote online learning, the distribution of tablet computers to schoolchildren and talk about 5G. Why don’t we try something vastly different to manage the flood hazard, which is expected to become worse in future due to climate change?

The Ministry of Health and Wellness, according to my information, is using a statistical model to help its officials to manage COVID-19. Wikipedia says that “a statistical model can be thought of as a statistical assumption (or set of statistical assumptions) with a certain property: that the assumption allows us to calculate the probability of any event”.

The making of assumptions about the probability of a random event – like flooding – can be beneficial in planning in the absence of better information. Would the country and its citizens be in a better position if they were to rely on a statistical model that was developed for the island about the risk of flooding? I do not know the answer to this question.

The October 14, 2020, issue of TechWire Asia reported that Google AI, artificial intelligence, is predicting flooding in India and Bangladesh. “The tech giant’s AI-powered Google Flood Forecasting Initiative can protect approximately 200 million people in India and around 40 million people in Bangladesh. The recent monsoon season floods that were unleashing havoc across Southern Nepal, Northern India, and Bangladesh recently have had devastating consequences, causing technology giant Google AI to double down on the use of its artificial intelligence-powered flood-forecasting tools in the region,” the report stated.

“More than 1,200 people have died, and more than 41 million people have been adversely affected, in the worst flooding to hit the region in decades, according to the Australian Red Cross. Over a third of the country of Bangladesh ended up submerged under water, so Google (which also applied its AI capacities to study the coronavirus) widened its Google Flood Forecasting Initiative to include all of Bangladesh and also all of India.

“Despite the significant landmasses of the affected nations, covering both countries entirely will provide greater details on timing patterns and water depths and will be able to dispense alerts to locals in a shorter period in up to nine new localised languages.

“The new predictive AI model that Google launched will allow for the lead time to be doubled over the previous models, enabling more notice to be given to the governments of both India and Bangladesh – which could give tens of millions of people in those densely-populated areas more time to prepare, with time savings of up to a day or more.

“Experts have long been warning that a densely populated delta country like Bangladesh is one of the most at-risk victims of climate change, and the severity of the recent, frequent floods can attest to that. Google started its Flood Forecasting Initiative in 2018 to predict when and where flooding will occur, tapping artificial intelligence to make sense of large volumes of data.”

For context, according to Wikipedia, about 26,000 square kilometres (10,000 sq miles) of Bangladesh is flooded – that is, around 18 per cent of the country – killing over 5,000 people, and destroying more than seven million homes each year. “During severe floods, the affected area may exceed 75 per cent of the country, as was seen in 1998. The floods have caused devastation in Bangladesh throughout its history, especially in 1966, 1987, 1988 and 1998.”

If the US tech giant can help to predict flooding in India and Bangladesh and help to protect their citizens, there are absolutely no good reasons why our officials should not investigate if its Flood Forecasting initiative, using AI, could be applied locally to protect Jamaican citizens and help policymakers better manage the country’s flood hazard. The ball is in now your court, Prime Minister Holness.

EndNote: Over 12 months ago, I shared information in writing from an Indian start-up with an official in what was then the Ministry of Industry, Commerce, Agriculture & Fisheries an idea involving the use of satellite technology to monitor the growth of crops and help facilitate the development of insurance schemes for agriculture. To this day, he has failed to even acknowledge the receipt of my email.

Cedric E. Stephens provides independent information and advice about the management of risks and insurance. For free information or counsel, write to: aegis@flowja.com