‘The flood came at 3 o’clock. The dam’s capacity was 9–11 lakh cusecs of water; the amount of water exceeded 15 lakh cusecs. Property, cattle and homes were inundated. We lacked a sophisticated flood forecast system to prevent this catastrophe.’

This was not a one-off event. Mahanadi river basin in Odisha is notorious for its floods – so much so that the locals call it the ‘sorrow of Odisha’. Heavy rainfall in the upper catchment area causes severe flooding in the downstream districts of the state. As recently as 2014, floods in the Mahanadi basin affected over 1,500 villages and almost 1.8 million people, resulting in damages of over 5 billion rupees.

Climate change is adding insult to injury. The State Water Plan prepared by the government in 2004 observes that ‘Variability of monthly rainfall is increasing which means that rainfall is concentrated in a particular period.’ These are worrying signs. Flood discharge of between 12 and 14 lakh cusecs has been recorded 10 times in Naraj district since the construction of the dam. When the dam was built, it was estimated that such flood conditions would arise only once every 1,000 years.

A systemic overhaul

In 2014, there was a growing sense of realisation within government circles that there was a need to develop an early warning system against floods. ‘We decided to take ACT’s help in developing this early warning model,’ says Muralidhar Panda, Deputy Director, Flood Forecasting and Flood Risk Management, Department of Water Resources.

The warning system in place in 2014 used observation networks of rain gauges and radars. It did not account for the dynamic conditions of soil moisture or previous day’s rainfall, upstream interventions, withdrawals, etc. The system provided an eight-hour forewarning at best. Meanwhile, passing the information to the relevant authorities was a tedious process. By the time the information was conveyed to villagers, it was already too late.

Government officials were eager to make improvements in the forecasting system but lacked the knowhow. During the same period, UK Aid was promoting climate-proofed growth in Odisha through the ACT initiative. Developing a robust flood forecasting model was a perfect fit.

ACT Team Leader Soumik Biswas recalls, ‘The government knew they had to do something. They couldn’t sit idle, but they didn't know where to go. UK Aid could provide them with the best available technical support without going through tedious and time-consuming processes involved in the government machinery. You have to think of the practicality. It immediately appealed to me because of the major impact it would have on the people. We cannot avoid the damage altogether, but if we increase the warning time by even eight extra hours it means people can move cattle, valuables and even property. They can rebuild their lives.’

The system needed an overhaul: a move from a reactive model based on the observed rainfall of the previous day to a proactive one based on the projection of rainfall for the next three days. 

The Soil and Water Assessment Tool model

Building a new flood forecasting model was risky. No one knew how it would perform. There were cost and time implications. Reputations were at stake. There was no shortage of scepticism within sections of the Department of Water Resources. Developing the model wasn’t enough: all the pieces had to come together to become greater than the sum of its parts.

Different models were carefully analysed for suitability before the Soil and Water Assessment Tool (SWAT) was selected. The SWAT model is a spatially distributed, physically based model. It requires site-specific information about weather, soil properties, topography, vegetation and the land management practices being followed in the basin.

The scope of developing the model involved:

1. Identifying the departments responsible for flood planning and real-time response to flooding and their minimum information needs;
2. Providing the indicative costs of producing a river basin modelling system;
3. Building the capacity (institutional, technical and financial) of relevant government planning and implementing agencies;
4. Testing and refining the model at the ground level.

‘Getting the data, the right data, and at the right time was the biggest challenge. We had to spend a lot of time to get the data that we wanted. Even when we got the data, extra efforts were necessary to sanitise the data properly and exclude outliers or simply impossible observations,’ says Soumik Biswas.

ACT knocked on the doors of different departments like the Central Water Commission and the Indian Meteorological Department and Water Resources for data. Subsequently, ACT cleaned the data to be able to begin the model design. After six months of work, in November 2015 ACT convinced the Department of Water Resources that developing a model to suit its requirements was possible through the SWAT model. The Department of Water Resources played a vital role in its development.

After being approved, the work on model development and validation began in full force. The work had to meet two separate and uncompromising criteria of providing a good projection with respect to flood timing and volume. Validations were performed in 23 locations. The model showed strong validations in all but three locations and was more accurate in predicting flows into Hirakud dam than both the Central Water Commission’s and the government’s own predictions. At that point (December 2016), the Principal Secretary gave the nod to launch the model.

On the ground

In flood-prone Tikhiri village, located on the banks of the Mahandai, people are unaware of these developments. However, they remember the devastation that floods cause almost every alternate year. Akshay Kumar’s daughter caught a fever during the 2011 floods. ‘We were cut off from everyone for 10 days. There was no medicine available. We rubbed neem leaves on her neck and used other ayurvedic measures,’ he says.

The school was shut down, there was no electricity and they slept on the roof of a former government revenue office – waiting for the water level to recede. It took them one month to rebuild their thatched huts. Rebuilding lives takes much longer – and then comes another flood.

When people are asked, ‘What would you do if you received a warning 36 hours in advance?’, the answers come thick and fast: ‘We would harvest crops first. It would be impossible to save everything, but we would call our relatives and try to cut and store as much as possible,’ says Sanjoy Meher. Another farmer adds, ‘We would send the children and women away to safer locations.’ ‘I would save drinking water,’ says a labourer. He adds, ‘During the previous floods, my daughter asked me for biscuits and I had nothing.’

Launching the model

ACT conducted an exposure training for operators at Hirakud dam in December 2016. Harmohan Pradhan, Chief Engineer, Upper Mahanadi Basin, who is responsible for the dam’s operation, agreed with the need for an overhaul. The chief minister formally launched the model in April 2017. They say this monsoon season is going to bring adequate rainfall. The model is ready for baptism by flood.

This article was previously published by ACT in On the Frontiers of Climate Change. Find out more about ACT's work.