During the fag end of 2015, Chennai experienced severe floods resulting in the death of about 500 people and economic losses of about INR 50,000 crores. The flooding stranded the city and was termed a 'man-made disaster' resulting from irresponsible water management and rapid urbanisation. The northeast monsoon of the year left most parts of South India marooned, exposing how vulnerable our cities are to such catastrophes.
"That's when the Office of the Principal Scientific Adviser took a major initiative to develop a real-time, integrated, urban flood forecasting system that was non-existent in our country. It called for a meeting of experts among multiple government agencies and academia," recalls Prof Subimal Ghosh. He is currently a Professor at the Department of Civil Engineering, Indian Institute of Technology Bombay (IIT Bombay).
Soon after, with a team of scientists from various institutes across the country, he swung into action to develop the first-ever expert system in India to forecast floods.
An expert system is a computer-based program that makes decisions and predictions based on a set of data. The researchers designed the flood forecasting system in a record year and a half.
"It was highly interdisciplinary, and a team of 30 scientists from 8 institutions took the responsibility. I led the project, and this is one of the best learning experiences of my life," shares Prof Ghosh.
In a recent study, published in the journal Current Science, the researchers shed light on the development of the automated flood forecasting expert system. The study was funded by the Office of the Principal Scientific Adviser to the Government of India and had a committee of experts chaired by Dr Shailesh Nayak, Ex-Secretary, Ministry of Earth Sciences (MoES). The research team comprised of scientists from the Indian Institute of Science, Bengaluru, IIT Madras, Anna University, India Meteorological Department (IMD), National Centre for Medium-Range Weather Forecasting (NCMRWF), Noida, National Centre for Coastal Research (NCCR), Chennai, Indian National Centre for Ocean Information Services (INCOIS), Hyderabad, and Indian Space Research Organisation (ISRO), Hyderabad.
The expert system designed by the researchers has six components—some running in parallel and a few dependent on each other. It has computational models that can forecast regional weather and surges in the tides and storm. It also has data from sensors that measures water levels in the rivers. Hydrological models, which consider reservoirs and river flows, are also included in the system, along with flood models that calculate which areas would be inundated. The system produces visual maps of forecasted inundation. All these components are automated and need no manual intervention at any stage.
The data for the expert system include rainfall and weather parameters; tidal and ocean depth data obtained from the Adyar, Cooum and Kosasthalaiyar river mouths in Chennai; water levels in these rivers; reservoir levels of Chembarambakkam and Poondi reservoirs; historical rainfall data, and the current land use, topography and drainage data. The output is an inundation forecast of 6-72 hours with ward-wise three-dimensional maps showing the depth of the predicted flooding and its area.
The complex calculations based on these parametres happen in a jiffy in the expert system.
"Normally NCMRWF releases the forecasts at 3 PM, and within the next two hours, our expert system will release the first forecast for the next three days," shares Prof Ghosh. "If there is a forecast of heavy floods, the real-time computing operations start, and the forecast will be updated every 6 hours," he explains. The level of details thus obtained can help rescue and alert operations.
The expert system also has a databank to speed up the process of forecasting. It has 796 scenarios resulting from rainfall extremes with different severity of water flow and tides, and past rainfall.
"Flood simulations for large cities take a very long time. Hence, we have generated the extreme possible cases in a data bank," reasons Prof Ghosh. As soon as the input forecasts arrive, a search algorithm finds the closest scenario from the data bank and releases the first forecast.
The researchers have validated their system with data from the December 2015 floods. They found that their inundation map with flood depths was 80% accurate within one metre as compared to the real flooded regions.
"We also experimentally verified its performance during the winter monsoon in Chennai, and found that it was working well," says Prof Ghosh.
Currently, the entire expert system is transferred to NCCR, Chennai, which is maintaining it. It is now working on an experimental basis and will be fully operational after a year. The researchers believe that their system can also be used for flood forecasting in other cities too.
"This is quite a robust framework, and MoES is using it for developing a similar forecast system in Mumbai. We are delighted that MoES finds our approach very useful and is implementing for other cities," concludes Prof Ghosh.
This article has been run past the researchers, whose work is covered, to ensure accuracy.