Deadly Floods in Pakistan Were “Predictable,” Study Says

By Andrew Freedman

Program Needed to Convey Warnings to Vulnerable Populations

A view of heavy floods caused by monsoon rains in Punjab Province, near the city of Multan, Pakistan, on August 15, 2010. Credit: UN Photo/Evan Schneider.

In Pakistan last summer, severe flooding killed nearly 2,000 people and affected more than 20 million, destroying infrastructure such as roads, bridges and power stations, killing 20,000 cattle, and damaging fertile farmlands that are major sources of income and food for the Pakistani people. The floods struck with little to no warning and shook the confidence of the Pakistani people in their government, making many inside and outside the ravaged country wonder if anything could have been done to take advantage of decades of worldwide advances in observational networks and weather forecasting computers.

A new study to be published in the journalGeophysical Research Letters asks exactly this provocative question — were the floods predictable? And if so, what can be done in the future to ensure that warnings are disseminated to those who are most at risk? The study, by Peter Webster of the Georgia Institute of Technology, along with two colleagues, also tackles the issue of whether the floods — which some have cited for their potential connection to climate change— were truly abnormal relative to events in recent years.

By analyzing rainfall records from different datasets, including satellite-based measurements covering the 1981-2010 period, Webster and his colleagues conclude that the average May to August rainfall for 2010 was “somewhat greater than in magnitude to previous years.” Where the 2010 event really stands out, Webster says, is when you look at the rainfall rates.

Pakistan lies at the western edge of the South Asian monsoon, yet in 2010, heavy monsoonal rains repeatedly pushed into northern portions of Pakistan, in an unusual weather pattern that may have been linked to the extraordinary heat wave in Russia at the time. According to the study, some Pakistani observing stations recorded rainfall amounts of more than 300 mm — or about 12 inches — during a four-day period. Six major “pulses” of heavy rainfall occurred, starting in early July, culminating in the most intense rains between July 27 and 30, the study states.

Webster and colleagues utilized 15-day computer model projections from the European Center for Medium-Range Weather Forecasts, known as the ECMWF model, in order to determine the predictability of rainfall in Pakistan. The forecasts from ECMWF, which is an intergovernmental organization based in Reading, England, are not currently made available to Pakistan, although Webster says efforts are underway to change that.

Their analysis shows that rainfall in Pakistan “is highly predictable out to 6 to 8 days, including rainfall in the summer of 2010.” Had such rainfall forecasts been available to Pakistani authorities, and paired with a hydrological model to determine the potential for flooding in particular parts of the country, many of the deaths might have been prevented, Webster says.

“It seems to me that you could have saved an enormous amount of property and probably a lot of lives” with advance warning, he says.

As indicated in the study, the forecasts would not have been able to prevent the floods, but they could have enabled crucial preparations to be made. “If these forecasts had been available to the regions of northern Pakistan, government institutions and water resource managers could have anticipated rapid filling of dams, releasing water ahead of the deluges. A high probability of flooding could have been anticipated,” the study states.

Webster says the floods resulted from a “conspiracy of factors”, including the high rainfall rates falling in a mountainous region of northern Pakistan, where deforestation is high, and recent drought conditions have further reduced the ability of hillsides to absorb heavy rainfall.

This image from the forthcoming study shows (A) total observed precipitation (mm/day) during 28-29 July 2010, and (B) the ECMWF ensemble mean of the forecast initialized four days previously (July 24, 2010) for the same time period. Note how the ECMWF forecast accurately captured the potential for very heavy rainfall, particularly in northern Pakistan.

Webster has extensive experience studying extreme rainfall events in South Asia, having helped establish an early-warning network in Bangladesh in cooperation with ECMWF, the Asian Disaster Preparedness Center, and the Bangladeshi Government. In that program, known as the Climate Forecast Applications for Monsoons project, long-range weather and climate forecasts are disseminated to end-users, down to individuals at the village level, so they can anticipate the potential for flooding. Webster is seeking to establish a similar program in Pakistan and other developing countries in Asia.

According to a press release from the American Geophysical Union, when flooding struck Bangladesh several years ago, the flood forecasting program helped reduce damages and deaths. Accurate forecasts were estimated to have saved residents up to $450 per farm, or “about the equivalent of an average annual salary in that country.” 

“The proof of concept was Bangladesh,” Webster says. “What we try to do is get forecasts to the people who need them.” To that end, he is working with ECMWF to expand data sharing to other countries, which requires the approval of the European Commission. Webster says a forecasting system in Pakistan would cost a few million dollars to establish, but just $100,000 or so per year to run.