Can a Data Scientist Drown a City in 3 Feet of Water?

Yes, Let’s dive into the details.

The Red River of the North has flooded repeatedly throughout the centuries. It is highly prone to spring flooding because of its northward flow, the nearly-flat former lake bed of the valley, and ice formation on the river. As spring approaches, the snow melts from south to north in the same direction as the river flow. Due to this, the United States and state governments made additional improvements to the flood protection system in North Dakota and Minnesota and built the Flood walls which can hold the water level up to 51 feet.

In January 1997, the National Weather Forecasting Agency released a report that the water level in the Red River will rise to 49 feet this spring close to a record. People in the city are not worried because the floodwalls can hold the water up to 51 feet. Very few brought flood insurance.

But Shockingly in April 1997, the water level in Red River raised to 54 feet and it drowned the entire city. 75% of people lost their homes, around 50,000 people were evacuated and estimated damages for the Red River region were US$3.5 billion.

What went wrong?

The National Weather Forecasting Agency later revealed that “ There is (+/-) 9 feet uncertainty in our forecasts based on average flood forecasts we made in the past. We did not reveal this with the public because we were afraid that the public might lose confidence in our forecasts “.This statement reveals that there is a 35% chance of the city being drowned.

If this uncertainty was discussed with the public then the government may have enforced the city floodwalls using sandbags to hold the 54 feet water level, the flood may be diverted into depopulated areas and the people have taken flood insurance.

The main problem is that the forecasters explicitly mentioned that the water level may rise to 49 feet without any reservations. Many people thought that the 49 feet may be the maximum it may raise. But a small mistake lead to catastrophic failure which we have seen above.

Realizing the catastrophic failure the National Weather forecasting Agency started explaining the uncertainty in their forecasts clearly with the public along with the forecasts made.

So a Forecaster / Data Scientist should always put away their pride and dilemma aside and should discuss the forecasts and uncertainty in terms of Confidence Intervals, and never as a single number.

My name is Tharun and I hope you enjoyed reading. Thanks.

References:

1. The Signal and the Noise: Why So Many Predictions Fail-but Some Don’t
2. https://en.wikipedia.org/wiki/1997_Red_River_flood
3. <a href="http://www.freepik.com">Designed by welcomia / Freepik</a>