Research

4D-EARTH

Flooding and Pakistan

causes, impact and risk assessment

Menno Straatsma, Janneke Ettema and Bart Krol (ESA, ITC)
October 2010

In July and August 2010, Pakistan was hit by an extreme rainfall event, leading to a devastating flood. The impact of this flood is huge, about 1800 people were killed and about 21 million people have been affected by the disaster. In total 160,000 square kilometers of Pakistan's most fertile crop land is inundated by the floods,  200,000 herd of livestock are killed and massive amounts of grain are washed away. As we speak, flood waters are still spreading, inundating large areas of agriculture land and demolishing buildings on its way downstream. 

The main river in Pakistan, the Indus river system, drains the western part of the Himalaya, bordering the Ganges-Bramaputra basin in the east and several smaller catchments in the west (Fig. 1). Of the total surface area of the catchment, roughly 70 percent is semi arid. The foothills of the Himalaya are the wetter parts of the country. Table 1 gives a few key characteristics of the catchment {Khan, 2010 #2}. Due to the general dryness of the area, the water of the Indus and its tributaries is heavily used for irrigation leading to the longest irrigation system in the world.

 

Table 1 Indus catchment characteristics
Length 2880 km
Length catchment area 1152000 km2
Foothill area 448000 km2
Semi arid area  704000 km2
Annual rainfall 125 – 500 mm

 

Figure 1 Indus river and its tributaries
Figure 1 Indus river and its tributaries

 

Here, we intend to describe and summarize the meteorological causes that led to the rainfall event and the effects of the 2010 Pakistan flood. For this purpose, we use spatial information that is directly available through the internet.

Pakistan 2010 flood: meteorology

Rainfall anomaly
During the last days of July 2010, extreme rainfall occurred over the northwest of Pakistan, in the upper part of the Indus river system (Fig. 2). In certain areas, more than 4 times the normal monthly rainfall fell within 3 days. The instant rain intensity reached 300 mm over a 36-hour period according to the Pakistan Meteorological Department. Similar excessive precipitation rates were also observed over northern India. For the other parts of Pakistan and India, the rainfall was not heavier than in other years, or even below average such as in Balochistan.

Figure 2 Left: normal July precipition sum (average over 1971-2000) in mm; right: precipitation departure from normal for July 2010 in % for Pakistan (Source: CDPC / PMD)
Normal July precipition sum Precipitation departure from normal for July 2010

 

Indian monsoon
Every year, the Asian subcontinent is exposed to beneficial, but torrential rains in the months April to November. This pronounced seasonal variation of rain is linked to a persistent south-westerly airflow over the Arabian Ocean and Bay of Bengals, the so-called Indian monsoon. This flow of already warm and moist air picks up further large quantities of water vapour from the ocean surface. Since the Tibetan plateau enforces the air to ascend, and hence to condensate, large amounts of rain fall over Southeast Asia. Various factors combined, especially topography and upper-air wind, make that the annual monsoon is a complex regional weather pattern.

The 2010 monsoon season had a slow start, but made quicker progress northward than normal during June-July. In Pakistan, the onset of the monsoon was about 10 days earlier than normal (Fig. 3). However, a fast advance of the monsoon does not necessarily involve more extreme rainfall.

Figure 3 Onset of the monsoon in 2010 (Source: Indian Meteorological Department)
Onset of the monsoon in 2010 (Source: Indian Meteorological Department)

 

El Niño – Southern Oscillation
The intensity of the Indian monsoon rains is affected by the El Niño – Southern Oscillation (ENSO). This oscillation, with a period of 3 to 7 years is an intricate relationship between the ocean and the atmosphere, in which the Pacific ocean warms (El Niño), or cools (La Niña) in response to the strength of the trade winds. These changing ocean temperatures alter rainfall patterns all over the globe.

To date, 2010 is considered as a neutral or weak La Niña year. In a La Niña year, the Indian Ocean, and more specific the northern Arabian Sea is warmer than normal. The warmer ocean water could lead to more precipitation associated with the Indian Monsoon. Whether also the extreme events this year are related to the 1-2°C higher sea surface temperatures needs to be investigated in more detail. During the period the extreme rainfall occurred over northwest Pakistan, the sea was 0.5 to 1°C warmer than averaged over the same period in the years 1971-2000 (Fig. 4). 

Figure 4 Sea surface temperature departures [°C] for the period 11 July – 7 August 2010 compared to the base period 1971-2000 (Source: NOAA/CPC/NCEP)
Figure 4 Sea surface temperature departures [°C] for the period 11 July – 7 August 2010 compared to the base period 1971-2000 (Source: NOAA/CPC/NCEP)

 

Jet stream
At the time of the Pakistan flood, large wild fires occurred over Russia. Statistical analysis of the 60 years observational data shows that the extremely high surface temperatures associated with the wild fires have a return period of once every 400 years. These two extreme events could be linked by a large-scale atmospheric circulation, the so-called jet stream, which is a massive band of high wind speeds. The jet stream is too high to affect every day weather directly, but it does influence the large scale weather patterns as it acts as barrier between the cold Arctic air and the warmer tropical air. Sometimes the jet stream remains on the same location for a number of days. The result of such a circulation block is that the large scale pressure systems or weather stays put.

In the week of the extreme Pakistan flooding, a block was formed in the large-scale atmospheric circulation, which stopped the meandering of the jet stream. At that time, the stream was split in two with one section heading north over Russia and the other going south over the Himalayas into Pakistan (Fig. 5), which is a very unusual situation. It resulted in a pronounced high air pressure system over Russia. High pressure makes it hard for clouds to form, allowing the surface to heat and to give up its moisture. Additionally, this amplified high pressure has weakened the jet stream to the south, and a lot of moisture monsoonal air could get further north over Pakistan (the northward meander in Fig. 5). There unusual intense rainfall and consequent flooding was observed.   

Figure 5 Global plot of the wind vectors and mean wind speed [m/s] at 250 hPa, the height of the jet stream for 24-30th of July. Left: climatology over the period 1968-1996 is what scientists would normally expect over this period. Right: the exceptional year 2010 with clear northward meander over the Atlantic and southward meander over Europe and Pakistan (Source: NOAA/ESRL)
Figure 5 Global plot of the wind vectors and mean wind speed [m/s] at 250 hPa, the height of the jet stream for 24-30th of July. Left: climatology over the period 1968-1996 is what scientists would normally expect over this period. Right: the exceptional year 2010 with clear northward meander over the Atlantic and southward meander over Europe and Pakistan (Source: NOAA/ESRL) Figure 5 Global plot of the wind vectors and mean wind speed [m/s] at 250 hPa, the height of the jet stream for 24-30th of July. Left: climatology over the period 1968-1996 is what scientists would normally expect over this period. Right: the exceptional year 2010 with clear northward meander over the Atlantic and southward meander over Europe and Pakistan (Source: NOAA/ESRL)

 

Climate change
The occurrence of these extreme events, the Pakistan flooding and the Russian wild fires, at almost the same time raises the question about their possible linkage to climate change. The IPCC Fourth Assessment report published in 2007 states that ‘there is a tendency for monsoonal circulation to result in increased precipitation due to enhanced moisture convergence’ and that ‘the type, frequency and intensity of extreme events are expected to change as Earth’s climate changes’. Nevertheless, the report also states that ‘many aspects of the tropical climate responses remain uncertain’.

The above discussion of the different meteorological phenomena of the last week of July 2010, showed that the atmosphere behaved abnormally during this period. However, no single event can be linked decisively to climate change. The behavior of the atmosphere is very non-linear, which makes it hard to predict on seasonal, even on a daily basis. In a warmer world, the tropics will expand and in a response the path of the jet stream will change. But what this means for the interaction between jet streams, Rossby waves and the monsoonal circulation is still unclear.

Pakistan 2010 flood characteristics

Magnitude of the 2010 flood
The magnitude of a flood is expressed as the discharge, and is related to the return period in years, the inverse of the frequency. High magnitude floods have a longer return period and a lower frequency of occurrence than small magnitude floods. The relationships between magnitude and return period for peak floods have been established for two measurement stations in Pakistan by {Khan, 2010 #2} based on a 66-year measurement period (Fig. 6). The stations are Tarbela station, which is located in the north, and Taunsa that lies in the center (Fig. 1). The relationship between magnitude and frequency is also known by its inverse the magnitude-frequency (MF) relation.

The extremeness of the rainfall anomaly in the northwest is also visible in the extreme discharge when compared to the MF relation. The discharge of the 2010 monsoon flood for Tarbela was 24800 m3/s (832000 cfs). Based on the historic records, a flood with a 66-year return period would have a discharge of 18,500 m3/s. The 2010 flood discharge was 30 % higher than the historic values in Tarbela. From the MF graph (Fig. 6) it is clear that this flood far exceeds the expected discharge as the 2010 flood discharge in Tarbela is 6000 m3/s higher than the values ever measured. Based on the current MF relationship, the peak discharge of Tarbela would have a return period of 1000 years. For the Taunsa station, the 2010 flood is closer to the MF relationship, but still 1000 m3/s above the line. In Taunsa, the 2010 flood would have a return period of 86 years based on the historical data. From the MF relations, it can be concluded that the magnitude of the event had a higher return period in the upper part of the catchment, than in the middle part of the catchment.

Figure 6 Magnitude-frequency relationship Indus stations Tarbela and Taunsa based on a 66 year measurement period and extrapolated to 200 years.
Figure 6 Magnitude-frequency relationship Indus stations Tarbela and Taunsa based on a 66 year measurement period and extrapolated to 200 years.

 

Flood monitoring and propagation
A flood early warning system (FEWS) is operated by the Flood Forecasting Division of the Pakistan Meteorological Department (www.pakmet.com.pk) based on Delft-FEWS software {Werner, 2005 #1}. The FEWS system is operational for the Indus river below Tarbela and the tributary rivers, Chenab, Ravi, and Sutlej. Predicting a high discharge for the lower part of the river is fairly easy as it is mainly determined by inflow that is already measured upstream and the travel times between the stations is known (Table 2). For example a peak flood wave at Tarbela will take six days to travel to the Taunsa barrage and 13 days to Kotri. These travel times may differ based on the water depth. With a larger water depth the flood propagation will be faster and the time to reach the downstream stations will be shorter.          

While arrival times between cities along the river are relatively easy to predict, the propagation of the water in the floodplains is not. The modeling of such a large river over its total length is not an easy task. Especially at peak discharges, the floodplains widen to 10 or 20 km and there are strong interactions between the main channel and the floodplain. Additional uncertainty arises from the state of the embankments. When the embankments fail the water will inundate the embanked area. The prediction of flood propagation in these areas strongly depends on accurate values on the dimensions and accuracy of the dike breach parameters of the dike breaches and these are difficult to collect during a flood event.

Table 2 typical lead times in days for the Indus river {Werner, 2005 #1}.

Location Chainage [km] Lead times [days]
Tarbela dam 0 0
Kalabagh barrage 210 3
Chashma barrage 268 3.5
Taunsa barrage 505 6
Chenab confluence 665 8-10*
Guddu barrage 794 9-11
Sukkur barrage 940 10-12
Kotri barrage 1349 13-15

*) first value for floods from Indus, second value for floods from Chenab

The onset of the flood event was the intense rainfall on the last three days of July and the first days of August. In the north, this inflow led to a sharp peak in the flood hydrograph (Tarbela station in Fig. 7). Continuous heavy rains led to a double peak (at around August 13 in Chashma and Taunsa) in the flood wave for the upper catchments. This second peak, however, is not visible anymore in the hydrograph of the lower station (Guddu), showing that the second peak has caught up with the first one. In addition, it shows that the flood peaks widen as they move downstream. 

Figure 7 Flood affected districts on August 16 (UN-OCHA) and flood hydrographs at 4 different stations (Pakistan Meteorological Department)
Figure 7 Flood affected districts on August 16 (UN-OCHA) and flood hydrographs at 4 different stations (Pakistan Meteorological Department)

 

Seventy kilometer upstream up the Sukkur barrage a large breach in the river embankment occurred (Fig. 8). This led to the inundation of a large floodplain area (approximately 500 km2)  in the first three days. At the time there was a lot of uncertainty on the prediction of the flood propagation in this floodplain area. UNOSAT inundation maps that are issued on a regular basis show question marks around the edges of the flood extent. The dense network of small embankments, raised roads, and irrigation channels influences the water propagation, which complicates accurate parameterization of the flood by a hydrodynamic model. I

Figure 8 Dike breach above the Sukkur barrage leading to the inundation of a large floodplain area in just three days (UNOSAT map, August 13)
Figure 8 Dike breach above the Sukkur barrage leading to the inundation of a large floodplain area in just three days (UNOSAT map, August 13)

 

Flood risk
The total flood extent (date September 2) and the population density are given in Fig. 9. It shows that the area that is inundated by the Indus coincides with the districts that have the highest population density. The reason is the availability of water, which is, essential to sustain life in a semi arid environment. However, living in the proximity of the river also increases the flood risk as more people are exposed to the hazard. In total 1800 people were reported killed while 20 million were affected. The lives were lost mostly in the upper part of the catchment where the river gradient and flow velocities were highest, the affected people were concentrated in the middle and south of Pakistan where the inundated areas were largest. The number of houses damaged amassed to 1.8 million. 

Figure 9 Flood extent and population density (UN-OCHA September 7)
Figure 9 Flood extent and population density (UN-OCHA September 7)

 

It is difficult to compare the severity of different floods as the suffering of individuals is involved and this can not be related easily. Nonetheless, the 2010 Pakistan flood does not stand out in terms of people affected. Floods Ganges-Bramaputra delta affect more people in Bangladesh, including its 15 million people capital Dhaka. Also the number of casualties is low compared to the Banda-Aceh Tsunami in 2004, or many hurricane events. What is exceptional here is the extremely low probability of the event. With a return period of 1000 years the flood in the upper part of the catchment was without precedent and such an event would be devastating in all major  rivers in the world. In addition, the preparedness for flooding was low. The semi arid area mostly deals with water shortage and protection levels for the embanked area were less than the once every 100 year flood. A coordinated disaster risk management could contribute to limit future losses.