Can Big Data help predict/prevent natural calamities?
On 18th June, tragedy struck Uttarakhand. Were we caught unaware by nature's fury or was something more sinister at play?
I believe that nature has its own checks and balances and gives enough indications that something is wrong to whosoever might care to listen. In this case, we couldn't make sense of signals that nature sent us.
According to experts, major devastation was caused by the crumbling of the 'Kedar-Dome', a glacier like body of ice and rock. Even a novice will agree that such a phenomenon cannot take place overnight. Something sinister would surely be at work for many years if not for decades and centuries. Did we ignore the warning signs or did we not have a mechanism to collect and analyze these signs? Both these questions demand an entirely different solution approach.
This blog post discusses the second question which can be unraveled by the technology available to us today.
Today we have elaborate seismic devices to measure even the minutest of disturbances, instruments to measure glacial movements and river courses.
Yet the events of Uttarakhand question our ability to make sense of our numerous signals that are available to us.
What if we started a serious initiative to make sense of all the inputs? Would we succeed in sending warning signals well before disaster strikes?
As a country, we are routinely ravaged by floods, earthquakes, storms and other natural calamities. I trust we have excellent documentation about all of these incidents. Can we not mine this data to reveal tell-tale signs of disaster? This historical data along with the observations and signals from our everyday life help prevent the next Uttarakhand disaster?
Can Big Data then not be the answer to preventing natural disasters?
Lot of pilgrims who visited the Ganges at Haridwar few days before the tragedy remarked that the river seemed not muddy but rather blackish indicating that it contained the silt and rock from the Kedar Dome itself. Surely this input alone could have averted this tragedy?
Even simple measurements like direction of the wind, pressure of water in flowing rivers and dams, the silt quotient in water, temperature and humidity measurements can prove to be very important. Even photographs shared randomly on the internet can reveal tell-tale signs.
In the past, we have been restricted by the amount of information available and our ability to hold and analyze the same. However these factors no longer pose any restrictions on us.
There are possibly three main reasons why Big Data has not been thought of a solution for predicting natural calamities -
Firstly, Big Data has largely been successful in fields where the ROI is easy to measure and there is a ready business case available. Thus areas like Capital Markets, Healthcare and Retail have readily embraced big data analytics. Big Data for predicting natural calamities has more of a humanitarian angle to it and is likely to be funded by governments. Justification of a Big Data driven Disaster Prevention mechanism is likely to be a hard-sell vis-à-vis the tried and tested disaster management mechanisms.
Secondly predicting the next natural calamity is more complex than predicting the right time to buy the cheapest airline tickets. But this is more due to" unknown unknowns" --i.e. our inability to ask the right questions rather than any technological limitation.
Thirdly as described elaborately described by Viktor Mayer and Kenneth Cukier in their compelling read 'Big Data', the world needs to move from being concerned with 'Causality' (i.e. the question of how) to being concerned with 'Correlation' (i.e. the question of what). This paradigm shift is unlikely to be a smooth transition especially when dealing with societal concerns.
However the truth remains that if adequate infrastructure is put in place to capture the vital signs of naturally sensitive zones and then make sense of all these various inputs, we will finally be able understand the signals that nature sends us and prevent or mitigate a tragedy before it strikes.