Fighting floods with insufficient warning

Fighting floods with insufficient warning

Representative image (Photo by Biju BORO / AFP)

Consider a scenario wherein a flood forecast is issued as ‘Rising’ with a value above danger mark at a river point 24 hours ahead of time. The river point can be anywhere in Assam or Bihar or Karnataka or West Bengal.

In this compelling scenario, the forecast does say where the rising water would inundate. But the district administration and municipalities that receive the flood forecast have only 24 hours to react.

Consider another scenario wherein the same flood forecast is issued 7-10 days ahead of time but with probabilities of different intensities of flood events at that river point. The probabilities 10 days ahead could be something like this: The chance of water level exceeding the danger level is 70 per cent and the chance of inundation of a nearby village is 30 per cent.


Certainly, the flood forecast in the second scenario reflects higher level of confidence of the impending flood risk and enables local administrations to take better decisions and to be better prepared than in the first scenario.

The flood forecast of the type shown in first scenario is known as “Deterministic forecast” whereas that of second scenario is known as “Ensemble or probabilistic forecast”. The length of time from issuance of forecast and occurrence of flood event is termed “lead time”.

Advanced countries like USA, EU and Japan have already shifted towards “Ensemble forecast” coupled with “Inundation modelling” whereas India has recently shifted towards “Deterministic forecast”. However, the shortcomings with Indian flood forecast are glaring.

An accurate flood forecast with longer lead time is the outcome of the successful integration of meteorological and hydrological data; their technological integration of computer based modelling capabilities, advanced telemetry and deployment of advanced weather radars of adequate numbers, etc.

In India, the Indian Meteorological Department (IMD) issues meteorological or weather forecasts whereas Central Water Commission (CWC) issues flood forecasts at various river points.

The advancement of flood forecasting technology therefore depends on how quickly rainfall is estimated and forecast by IMD and how quickly CWC integrates the rainfall forecast (also known as Quantitative Precipitation Forecast or QPF) with the flood forecast and how quickly CWC disseminates that to user agencies.

Reports suggest that IMD has deployed advanced 30-35 Doppler weather radars for weather forecasting. Doppler weather radars can measure the likely rainfall directly from the cloud reflectivity over a large area due to which the lead time can be extended up to 3 days.

But the advantage of deploying advanced radars or any advanced technology by IMD becomes infructuous because the reality is that most of the flood forecasts at several river points across India are based on statistical methods that possess lead time of less than 24 hours. Yes, just 24 hours! This is contrary to the perception that India’s flood forecast is driven by Google’s most advanced AI techniques.

These statistical methods known as gauge to gauge correlations and multiple coaxial correlations fail to capture the hydrological response of river basins between a base station and a forecast station. They cannot be coupled with QPF.

Although radars extend lead time up to three days, statistical methods limit that to just 24 hours.

According to a scholarly study by NIT, Warangal, only recently, India has shifted towards using hydrological (or simply rainfall-runoff models) capable of coupling with QPF. So, lead time of third days is sporadic in India and at only select river points.

Further, the limitations of weather forecasting by IMD can also render any advanced infrastructure deployed by flood forecasting agency infructuous.

For example, USA being twice the size of India has deployed 160-180 next generation S-band Doppler weather radars called NEXRAD with a range of 250-300 km. So, India needs at least 80-100 S-band radars to cover its entire territory for accurate QPF. Else, the limitations of altitude, range and extensive maintenance enlarge the forecast error in QPF which would ultimately shift to CWC flood forecast.

Therefore, the outdated technologies coupled with lack of technological parity between multiple agencies decreases the lead time and increases the forecast errors thus throwing the onus of interpretation on hapless end user agencies.

The outcome is increase of flood risk and disaster.

Weather phenomenon is chaotic. Beyond three days lead time, deterministic forecast becomes less and less accurate. Hence, the developed world has leapt from deterministic forecast (which gives one value of water level for one model run with a warning like “Rising”) towards ensemble weather models that measure uncertainty by causing perturbations in initial conditions thus reflecting the different states of the chaotic atmosphere.

The probabilities are computed for different flood events with a lead time beyond 10 days.

India has a long way to go to master ensemble model based flood forecast.

Although, IMD has begun testing and using ensemble models for weather forecast through its 4-6 teraflop supercomputers called “PRATYUSH” and “MIHIR”, the flood forecasting agency has to catch up with this advanced technology and achieve technological parity with IMD in order to couple the ensemble forecasts to its hydrological models.

It has to modernize not only the telemetry infrastructure but also raise the technological compatibility with river basin specific hydrological, hydrodynamic and inundation modelling.

Only then can the country witness probabilistic based flood forecasts with a lead time of more than 710 days on par with the developed world. Only then will local administrations witness scenarios of probabilistic forecasts with ample time to decide, react, prepare and undertake rescue missions so as to reduce the flood hazard.

The writer is Director, Central Water Commission, Government of India. The views expressed are personal.