According to United Nations Office for
Disaster Risk Reduction, 79,732 people
were killed by 321 natural disasters like
floods and earthquakes in India between
2000 and 2019.Under its "Google Research"
imprint, one of the world's largest tech
companies is using machine learning to
address one of the most prominent disasters
in India and the world - riverine floods.
According to Sella Nevo, a software engineer
at Google who heads the Google Flood
Forecasting Initiative, artificial intelligence
can effectively help them send out flood
warnings way before traditional methods.
India and China collectively account for
the most loss of life and infrastructural
damage due to natural disasters. According
to United Nations Office for Disaster Risk
Reduction, 79,732 people were killed by 321
natural disasters like floods and earthquakes
in India between 2000 and 2019. In addition,
108 crore people were directly affected by
disasters in terms of livelihood, displacement,
Google is trying to change that. Under
its "Google Research" imprint, one of the
world's largest tech companies is using
machine learning to address one of the most
prominent disasters in India and the
world - riverine floods. In the absence
of mitigation measures and early warning
systems, a lot of lives that can be saved
are lost to natural disasters.
Using artificial intelligence, Google has
developed an early warning system. According
to Sella Nevo, a software engineer at Google
who heads the Google Flood Forecasting Initiative,
artificial intelligence can effectively help
them send out flood warnings way before
Pilot programme in India
The pilot programme for this AI-based flood
forecasting system was started in Bihar, India.
Nevo highlighted how Google assists government
with their goals. While the government assesses
potential floods, it cannot predict areas that
would be affected. By using satellite images,
Google essentially creates a model of areas
that are most likely to be hit by floods based
on the river's trajectory. A prediction of
the river's flood is created, based on which
warnings are issued.
The company can help send out flood warnings
to people two days in advance. With further
research, Google hopes to improve the capabilities
to five days in advance, giving a longer
window of time for people to seek shelter.
So far, Google has sent out 100 million Android
notifications to save lives of people in areas
that could face flooding, Sella Nevo claimed.
"Forecasting can prevent 30-50 per cent damage,"
What's the future of Google's flood forecasting?
By constantly improving AI by feeding it
expansive data, Google's "top priority is getting
these alerts to more people." So far, Nevo's
team has helped 360 million people in India and
The biggest roadblock is the lack of internet
access in many regions. To overcome this challenge,
Google is helping build a community-based
alert system with help from local NGOs, Nevo
said during the "Inventors @ Google" segment
on November 11.
And no, you don't need to sign up for anything
to be in tune with such warnings. The warning
system is integrated into Google Maps and Google
Search where just typing in the word "flood"
followed by your location will give the right
How does AI help?
The AI employs two primary models to predict
flooding. The "hydrologic model" helps scientists
predict the water level in a river at any
given time. "We have an error of 12 centimetres,"
The "invasion model" translates the river into
a map of sorts to identify areas that will be
flooded. The error rate in this model stands at
about 100 metres. While it may not tell you the
exact place where flooding will occur, it can
help save villages and neighbourhoods that are
most likely to get hit by flooding. For now,
the focus for Google and Nevo has shifted from
improving accuracy to "providing it to more people."
For now, the predictive model only takes into
account riverine flooding and not flash floods,
which Nevo hopes to pursue "may be in the future."
In response to a question by Indiatimes, Nevo
highlighted how Google and his team are making
the tool more accessible. First off, open sourcing
of the code may "help others scale up." To this
end, Nevo and team have "sent in hydraulics model
for open source" which is currently under review.
Nevo also believes that "machine learning
systems can adapt to changes in climate faster
than classic hydrologic systems." With help
from local disaster management agencies and
the Central Water Commission, under the aegis
of Indian government, Google hopes to save
more lives in the future.
Provided by the IKCEST Disaster Risk Reduction Knowledge Service System