In the first few days of August 2010, an epic storm was brewing over Leh in Jammu and Kashmir’s Ladakh region. A rare weather system, a product of meteorological activity in Europe and Russia, had descended over the Tibetan plateau north-east of Leh. On August 6,  the town was hit by a deluge that overwhelmed the desert region which normally gets less than 100 mm a year. More than 250 people were killed and more than 10,000 people affected.

It was an unusual system. Different elements mashed together to produce a seasonal monster. Normal storms are mostly alike; by contrast, each flood-producing storm is unique in its own way.

“Each flood (and storm) needs to be looked at individually to understand what’s going on,” Says Robert Houze, professor emeritus of atmospheric sciences at the University of Washington. Houze has been working on clouds and storms for more than 40 years, including how mountains influence cloud precipitating systems. He visited India in 1977 for the first time, and has been back many times since then.

His association with Pauline Austin (1916-2011), director of MIT’s Weather Radar Research Project for over 25 years, who pioneered obtaining “the first quantitative estimates of precipitation using radar signals,” helped Houze dig deep into the processes behind storms. “There is a wide range of ingredients that contribute to an individual flood,” he says. 

In his book—Darkest Hours, A Narrative Encyclopaedia of Worldwide Disasters from Ancient to the Present—American author Jay Robert Nash mentions  a storm surge of October 7, 1737, “when a massive cyclone, centring its fury at the mouth of the Hooghly river”, struck the Bay of Bengal. It “annihilated 300,000 persons”; another in 1864 drowned over 50,000 people.

According to Nash, one of the fiercest storms ever to hit India was the one of October 16, 1942. The winds topped 150 miles (250 km) per hour. They swept across the coast, turned inland, raged through the province of Bengal and reached Calcutta. More than 40,000 were killed in a single day.

There was also the Bihar flood of 1987, the Gujarat floods of 2005, the Maharashtra floods of 2005, the Brahmaputra floods of 2012, the J&K floods of 2014 and the Chennai floods of 2015.

The worst flash flood recorded in Indian history was when the Machchu-2 dam broke after heavy rain in August 1979. It washed away Morbi town and scores of villages in Gujarat killing thousands of people. According to research, only five per cent of tropical cyclones occur in the north Indian Ocean but they account for 95 per cent of casualties worldwide.

***  

A

thunderstorm typically arrives late in the afternoon. The ground is heated up, warm air columns start to rise. The higher you go, the colder it gets. As the air rises, it gets colder and colder and at some point, water vapour within the mass is saturated and begins to condense to form a cloud. The bottom of a cloud is the level at which the cloud becomes saturated. Above that level all the way up, water vapour condenses into precipitation particles.

As particles begin to form, they start to run into one another, and become larger and larger. The precipitation becomes denser and denser, eventually falling down as rain. In any thunderstorm, the main ingredients are water vapour and rising warm air. As warm air rises, it takes water vapour along with it, a process called convection.

Normal thunderstorms produce rain, accompanied by lightning and thunder. They crest the horizon, fall out of the sky, and are done after a short while.

When moisture from the sea reaches the foothills of mountains, upslope air motion produces constant lift for the low-level moist air and triggers thunderstorms. Flash floods occur when rain runs off the mountains.

Flood-producing thunderstorms come in different sizes and shapes. Usually called convective storms, they’re localised, intense bursts of water tumbling from the sky in short bursts, sometimes turning land into lakes. Another type is the widespread thunderstorm also called stratiform precipitaton. A larger, weaker weather system, it  produces rain over a longer period and causes flooding.  Storms also can be a combination of the former and latter, with intense bursts at the leading edge followed by weaker stratiform at the end. These are larger entities of intense precipitation. And they move.

These storms can occur in mountains, and in India they’re called “cloudbursts” whereas researchers around the world call them orographic thunderstorms—referring to mountains influencing storms. Convective storms over mountains are localised, with heavy rain falling for a short period of time.

When moisture from the sea reaches the foothills of mountains, upslope air motion produces a constant source of lift for the low-level moist air and triggers thunderstorms, many of which occur at the foothills or lower slopes. Flash floods occur when the rain runs off the mountains.

Houze, along with Professor Kristen Rasmussen of Colorado State University and Anil Kumar of the National Oceanic and Atmospheric Administration (NOAA), have been extensively studying the extreme events that lead to lethal flash floods.

Rasmussen became interested in weather at a pretty young age. In middle school, a mile from her school, a tornado landed. That was about the time Twister released. It kind of stuck with her. Among her research interests are extreme events and flash flood producing storms in India and Pakistan. “Extreme events are critical to understand and we should be able to forecast them because they cause a lot of damage and loss of life,” Rasmussen says.


Kristen_Rasmussen_headshot.jpg

Kristen Rasmussen

All over India, during the monsoon season, there are intense, high-precipitation systems, she continues. If these systems occur near mountains, you can look for similar atmospheric ingredients to other flooding scenarios around the world. This helps us understand these extreme events better and can lead to improved forecasts.

 “If you take a monsoon-related storm and put it on mountains, given the right situation it could turn into a high-impact event,” she says, with mountains helping focus some of the moisture and collecting the rain in a single day in a valley that has the potential to cause a flash flood.

“Thunderstorms happen all the time; they are not always high-impact events. When you combine multiple ingredients, even a garden-variety storm can become an extreme event. There are different types of anomalous events.”

Rasmussen, with colleagues, analysed the 2010 Pakistan floods, the 2010 Leh flash flood, and the 2013 Uttarakhand floods among other extreme events in different parts of the world.

For the last ten years, Anil Kumar has worked with Houze and Rasmussen in unravelling the mechanism behind anomalous storms and flooding in India.

Flash floods are  events where rising water occurs during or a matter of a few hours after rainfall. If the water level increases occur more than a few hours after the rain the event is considered a flood, not a flash flood.

He remembers being in Mumbai in July 2005 when intense rain flooded the city. He says that event was one of its kind across the world. He published a paper on it.  He says there was a localised vortex over Mumbai, and air masses from the tropics (warm) and mid-latitudes (cold ) collided. Moisture from the Arabian Sea supplied the fuel for the storm.“All these ingredients were present over Mumbai,” he says.

All researchers endorse the ingredient based approach in understanding and perhaps predicting flash flood-producing storms. Charles Doswell pioneered this approach through his paper, Flash Flood Forecasting: An Ingredients-Based Methodology, published with Harold Brooks and Robert Maddox in 1996. The methodology provides researchers and forecasters with the basic framework for understanding heavy precipitation.

In a presentation, Doswell noted, “Flash floods are defined to be flood events where the rising water occurs during or a matter of a few hours after the associated rainfall. If the damaging water level increases occur more than a few hours after the rainfall, the event is considered a flood, not a flash flood .

“Whereas many weather phenomena have specific geographical locations where they occur, rainfall is an event that occurs virtually everywhere. There will be occasions when it becomes intense and that intensity is maintained long enough to create the potential for flash floods. Hydrology plays a large role in the flash flood problem; a given amount of rainfall in a given time may or may not result in a flash flood, owing to such factors as antecedent precipitation, soil permeability, terrain gradients, and so on. Therefore, forecasting involves both a hydrological and a meteorological forecast.”

The key ingredients of such storms are ample and persistent supply of moisture; a focus for a thunderstorm to develop, such as mountains, which help uplift of air, which then condenses and forms precipitation; high rainfall rates and duration of rain; process or processes that cause thunderstorms to regenerate or repeatedly move over the same area; and soil surface properties (rocky, wet, saturated, dry and so on).

As Rasmussen explains, in the 2010 Pakistan floods, warm moist air from the Bay of Bengal and Arabian Sea rose up the mountains and sat there for many days. The uplift was widespread and led to mountain-enhanced stratiform precipitation, not very high rates but persisting over many days. The runoff flooded the villages.

*** 

O

n July 22, 2010, monsoon rain lashed Pakistan’s northwest, triggering flash floods in Baluchistan, Punjab, and Khyber Pakhtunkhwa provinces. The floods left an estimated 2,200 people dead, affected 20 million people and destroyed homes and crops across the country.

Barely two weeks later Leh was devastated by a flash flood on August 6. That was, Rasmussen continues, because there was a stationary high over the Tibetan Plateau. The high pressure developed in connection with the large scale warming patterns in Europe and Russia.

The system was due to a so-called blocking event. In his study of atmospheric teleconnections—connecting weather activity in one region with another region—between the 2010 Russian heat wave and Pakistan floods, William Lau of the University of Maryland and Kyu-Myong of NASA’s Goddard Space Flight Center in Greenbelt, Md., suggested that large scale weather patterns extending all over Europe set up a stationary blocking event high up in western Europe and Russia that led to fires there. That high also set up patterns in India and Pakistan with warm moist air being pushed up the slopes for days on end, stopped convection from happening. That high essentially blocked the low level moist air from breaking through to the atmosphere, and as a result moisture kept building and building like the fizz in a soda bottle.

The clock-wise wind pattern corralled the thunderstorms that occurred on the edge of the plateau into a larger propagating system. “That was very rare in the region,” she says.

The Leh flood was  unusual in two ways, Anil Kumar says: convection over the Tibetan plateau rarely grows into a Mesoscale Convective System (MCS).  They’re defined as systems extending over 100 km in one direction, larger entities consisting of a conga line of thunderstorms. MCS can sustain itself and last longer, some more than four hours. They happen in central India or near the Bay of Bengal but not in the northwest region over high terrain.

“The structure of the MCS was a combination of both convective and stratiform precipitation, with the convection at the leading edge and stratiform behind,” says Rasmussen.

The second unusual thing, according to Anil Kumar, is that a “travelling squall line system (a line of thunderstorms) of the type that affected Leh is rare in the Himalayan region.”

Houze, along with his colleagues, developed a model of storm climatology in 2007. Climatology, by giving knowledge of what’s usual and typical, exposes any outlier event.

When researchers like Rasmussen and Anil Kumar look for the ingredients that went into an anomalous event, they not only look at small scale local effects but also large scale weather patterns. Quasi-stationary weather patterns are a common ingredient they have seen around the world that helps create high impact floods. In its absence, that is, if the Russian heat wave had not happened due to the blocking event—it would have been different in Leh.

The 2013 Uttarakhand flood was different from the previous two floods. According to Anil Kumar, the associated rainstorm had the following characteristics: low-level moisture from the Bay of Bengal moved into the foothills; large-scale atmospheric circulation maintained the moisture supply over many days; moisture rose up the slopes and the lifting converted it into rain; there was prior rain that made soil susceptible to run off; it was essentially stratiform precipitation, not so much convective precipitation.

“There was an anomalous aspect to each of these floods.  The actual structures of the storms that produced the floods are different, emphasising why the ingredients-based approach is necessary,” says Rasmussen.

 ***

T

he anomalous aspects of storms came to the fore because Houze, along with his colleagues including Rasmussen, developed a model of storm climatology in 2007. They were interested in wherever the most intense or the deepest convective storms were happening around the world. They were interested in flash flood-producing storms and tornadoes. In 2015, they found that the most intense thunderstorms on the planet occur near the Rockies in the US, the Andes in South America and the Himalayas.

ROBERT HOUZE.jpg

Robert Houze

Using space-borne precipitation radar from the Tropical Rainfall Measuring Mission (TRMM) satellite, they looked for places where different types of convective and stratiform storms occurred. For the first time, they were looking at remote tropical and subtropical regions using a single radar to understand the differences. They had to use only TRMM data because they didn’t have access to the ground-based radar data and operational network data from India and Pakistan. Also, India doesn’t have weather stations in remote areas.

From their observations, they classified storms into two types: one by deep convection, basically very intense, vertically-oriented convection, with very strong updrafts. The other category is more horizontally organised type of storm with stratiform precipitation.

They then looked for extreme events in India and Pakistan and around the world, and compared those events to their climatology. In any given flooding event, they looked at the flooding pattern and characteristics of the storm in that event, and compared it to the climatology they developed. Climatology, by giving knowledge of what’s usual and typical, exposes any outlier event. 

Speaking of what set her on course to researching the region Rasmussen says that the 2007 study by Houze and others looked at several years of satellite-observed rainfall data to discern what type of storms were common in India. They concluded that deep convective storms occur in north and northwest India. Basically, there were no instances of weaker, long-drawn stratiform precipitation in the northwest.

We don’t know in advance whether it will be a strong or weak monsoon.  Knowledge of variability can be immensely useful for the country. As it turns out, these events can be forecast and there is skill in the forecasts.

However, when Rasmussen and her colleagues, including Houze, looked at TRMM data for the Pakistan floods, they found stratiform precipitation, which was not typical in this region from the storm climatology. Similarly, they found propagating organised convective systems over Leh, which was also very unusual.

“Because we had a storm climatology to compare to, we could put the storm into context and understand that these events were highly anomalous. This information could be used in the future to predict future high-impact flooding events,” she says.

“Forecasts of extreme events require an understanding of the storm climatology because we know what type of storms we can expect, and if we see an individual situation that looks different, we can potentially tell you whether it has the potential to be an extreme event or not,” she says.

*** 

P

eter J. Webster is a professor of earth and atmospheric sciences at the Georgia Institute of Technology, Atlanta, US. He has been working on the dynamics of atmopsheres and oceans with the aim of improving weather forecasts, especially in the monsoon regions, for more than 30 years. He was consultant in a project predicting heat stress in Gujarat and to the World Bank on flooding in Pakistan/Bangladesh.

His experience of working Indian and Bangladesh was “fulfilling and frustrating, ”­—“fulfilling” because he did a lot of work in equatorial dynamics and predictability of extreme events, and “frustrating  because rigid bureaucratic  mindsets make it difficult for institutions such as the Indian Meteorological Department (IMD)  to adopt new ideas.”

“If you look at the situation in South Asia—India, Pakistan, Bangladesh, Thailand, —there are a number of timescales we would like to forecast, ranging from the small scale rapid events (flash floods, tropical cyclones in the Bay of Bengal), rainfall events on the time scales of weeks to the variability of climate from year-to-year,”  Webster says.

We don’t know in advance, although we would like to, whether it will be a strong or weak monsoon. Even after the monsoon starts, there is going to be a lot of variability in precipitation. Knowledge of this variability can be immensely useful for the country as a whole, he adds.

As it turns out, these events are can be forecast, Webster says, and there is skill in these forecasts.

A good example was the skill shown for the floods in Pakistan in 2010, 2011 and 2012. Using information from the numerical models that come from the European Centre for Medium-Range Weather Forecast in the UK, it was shown that they were predicted with clarity 7-8 days in advance. Similarly, the 2013 disaster in Uttarakhand could have been were forecast a week in advance, he says.

Webster reckons that you could ask why  these forecasts aren’t used in South Asia. After all, they are available for scrutiny by the IMD, for example. And the IMD has its own numerical models.

IMG_1548.jpg

Peter J. Webster

“One has to have a mindset that forecasting is not going to be perfect---what a forecast gives is the probability of an extreme event. What we’re looking for is extreme events. What we look for is being able to say whether it’s a 50 per cent or 60 per cent chance of an event occurring in a small region of the planet some days in advance,” he says.

“What one has to do is develop methods that isolate the possibility of an extreme event (or a break in the monsoon) occurring. Such methods are applied through much of the world but not in South Asia.” According to Webster, IMD doesn’t accept the concept of probability.  Most places in the world do.

As Webster explains, when you run numerical models based on the laws of physics for the  atmosphere and ocean, they’re susceptible to errors of two types. One is that the models aren’t perfect. Two, the data that feeds the models and allows them to look into the future is faulty.

To offset this, you run the model many times and you try to “replicate the uncertainty”. What that gives you is a range of forecasts, many of which pick up an extreme event. Some of the model runs suggest an extreme event and others don’t. But what the forecaster finishes up with is a statistical estimate of the occurrence of an event.

The idea of probabilistic forecasts is difficult to understand but it’s so powerful, whether you‘re looking for flash floods, cyclones or heat waves. You can work out probabilities of where they’re going to occur.

Webster gives the analogy of crossing a road at rush hour. “You’re looking and thinking of your best chance of getting across the road. You could get up at 2 o’clock in the morning and cross when there is no traffic. In rush hour, on the other hand, you have make decisions, your brain comes up with probability, whether you’re going to be hit by a car, or if you’re going to be in a place where you shouldn’t be.”

For example, a tropical cyclone is forming in the Bay and moving northeast. Models give you the probability of where it will be in two days or three days. You have major cities: Bhubaneswar, Kolkata and Dhaka. One could evacuate all three or systematically work out the probability of a location where the cyclone will make landfall, say, 50 per cent, Bhubaneswar, 20 per cent Kolkata, 5 per cent Dhaka. One might start precautionary preparations in Bhubaneswar. The forecast of the following day would help in a better decision.

Webster says, “IMD doesn’t accept the concept of using probabilities. Only when a cyclone is identified as having formed do they run their model. ECMWF, on the other hand, makes 102 model runs over the Bay of Bengal out to 15 days. This is standard practice in most places.

The official said ‘Professor Peter, you must know that probabilities do not work in the subcontinent.’  I was flabbergasted. It showed how a narrow bureaucracy can stifle work that helps the poorest of the poor.

 “The idea of probabilistic forecasts is difficult to understand but it’s so powerful, whether you‘re looking for flash floods, tropical cyclones or heat waves. You can work out the probabilities of where they’re going to occur.”

Webster says if there is a 10 per cent chance of tropical cyclone, you probably are not going to evacuate; but, if you have 80-90 per cent, then you’re going to evacuate people.

 What they showed in Bangladesh is that if you put out a forecast saying there is an 80 per cent chance of flooding, people will respond. They will take their belongings; they will take their cattle, instead of going to the money lender. They also showed from economic studies that farmers can save $300-400 a year by following forecasts.

But the bureaucracy was wary of forecasting much in advance. They asked, what happens if we’re wrong?

Then Webster’s team countered, what happens if we’re right.

“It’s the mindset thing,” he says.

In the same context he talks about setting up a heat wave forecasting system in Ahmedabad. They had a teleconference involving people from all departments. He explained what they were going to do, beginning with a seven-day forecast.

“We explained that we would make our forecasts available to all those interested, including IMD. The regional director countered saying he would not be able to look at them unless he had authority form the highest level in Delhi. But, I said, these will be different to the IMD forecasts as they will state the probability of temperatures being warmer than a certain level. IMD forecasts give only the absolute temperature (not probability) and only out to three days in advance.

“But the official said ‘Professor Peter, you must know that probabilities do not work in the subcontinent.’ It was quite a statement and I was flabbergasted. Nonetheless, pressure was applied and IMD forecasts were used. The heat wave forecast system, so successful for three years, fell apart! This made me very sad as it showed how a narrow bureaucracy can stifle good work that help especially the poorest of the poor.”

A farmer may look at his farm in the event of an 80 per cent chance of flooding. He might harvest, or he might make an informed decision based on his circumstances. Although he cannot get everything, he might salvage whatever he can.

“He minimises risk. He hedges against catastrophic loss.”

If you divide India into 100 districts and have somebody constantly watching the areas and models, you will find the probability for extreme events in some regions. But you need somebody always looking.

Understanding probability forecasts is not just a problem in India, but it exists throughout the world. Webster talks of 2002. There was an enormous drought. Initial rains were very good and everybody planted. You cannot ask a farmer not to plant. For the next 30 days, the rains vanished.

“But the possibility was there that the monsoon would be delayed then. To be fair, even if IMD believed in probabilistic forecasts, educating people to use that information would be difficult. That applies to all parts of the globe.”

All the floods Webster has talked about, all the forecasts about flash floods are in retrospect, except for the Bangladesh flood forecasts that were done in real time for a number of years. There was surprise that models showed the probability of extreme events.

How do you turn that hindsight into future forecasts? For that to happen, he suggests paying attention, looking, really looking.

“If you divide India into 100 districts and have somebody constantly watching the areas and models, you will find the probability for extreme events in some regions. But you need somebody always looking. This could be automated to some degree. But that would take a change of mindset from the IMD,” Webster says.

*** 

I

ndia doesn’t have any cloudburst or flash flood forecasting. Flood forecasting is the job of the Central Water Commission (CWC), not IMD. Its rainfall prediction goes as an input into CWC’s flood warning.  IMD, at present, gives thunderstorm warning for the next three hours—whether it will occur or not—for 399 cities. But it is working on flash flood guidance—not forecast—with the World Meteorological Organization (WMO), says Mrutyunjay Mohapatra, head of the cyclone warning division at IMD.

WMO is opening two centres in South Asia, one in Delhi and the other in Pakistan. Delhi is responsible for flash guidance to India, Sri Lanka, Maldivies, Bangladesh, Bhutan, and Nepal. Preliminary work is going on; some experimental work is planned this year in India.

Guidance, according to Mohapatra, means it takes past rainfall, which will be superimposed over the particular area with digital elevation and other details. It will provide the impact of heavy rainfall over a particular region and intimate whether it has potential to lead to a flash flood. This can be utilised by the local forecasters to provide the forecast. Moving forward, there is also a provision to use radar and satellite data along with rain gauge records and forecast from weather numerical models.

“This helps create a composite picture for the region,” says Mohapatra.

General conditions for flooding can be identified a week or so in advance for regions of India.  Local forecasters should become aware of the likely locations of storms with respect to local topography and land surface properties.

IMD has taken up two mega projects; putting up radars in the Western Himalayan region and the north-eastern states over the next two years. As one seismologist commented, India is at a stage where “in our lifetime we have to collect more and more data so that the next generation picks it up and analyses it.”

India faces repeated floods and flash floods and needs to embrace the climatology that Houze, Rasmussen, and others  developed and adopt the ingredient-based approach in forecasting. Forecasts from the global models give an idea of an impending situation a week or so ahead and provide a general picture of likely types of storms that may develop.

Houze is an advocate for modernising India’s forecasting methodologies. He suggests looking at forecasts at every level—starting with the global forecast of a region a week in advance to see where moisture is concentrated, where high pressure is building and then look at local weather patterns, developing the details of how ingredients identified by Doswell are going to be strengthened, and finally look at land surface properties where clouds are likely to occur. From these factors they can decide if a situation is about to occur that might lead to flooding.

“Forecasters should become familiar with the research done by Peter Webster and colleagues. They have shown that general conditions for flooding can be identified a week or so in advance for regions of India. Forecast conditions can indicate departures from normal and it can be determined whether these anomalies are of a type that satellite climatology suggests are potential flood producers. In this way, local forecasters should become aware of the likely locations of storms with respect to local topography and land surface properties,” Houze says.

He also suggests increasing the number of personnel at IMD and other institutions, who have done their research and training at the world’s top universities.

Conditions don’t exist in India for collection and dissemination of the data, at least not on the scale required. Houze suggests an extensive network of sounding stations in India. Radars need to continue to be employed, especially along the Himalayas and coasts because that’s where heavy precipitation occurs. And a dense and reliable network of rain gauges is needed all over the country.

A major problem inhibiting monsoon research is that India doesn’t make digital data readily available to researchers worldwide. Many researchers cannot make sense of data denial. Webster says India is a little unfair with data exchange. It accepts all of the US and European satellite data (it is free) but won’t make Indian satellite data available.

By sharing data, India could benefit from international research on the monsoon. A global engagement of researchers might mitigate the repeated loss of life and crops and property that happens every monsoon.

The monsoon, after all, is not just about India. It’s part of a larger tropical circulation, starting from Africa and perhaps extending to the mid-Pacific. By denying the data of its part of the monsoon, India stands to lose in many ways.