Post-formation Behavior of Hattian Landslide Dam and Post-breaching Situation

Abstract

The Hattain Bala landslide dam, formed by the 8th October 2005 earthquake, provided a rare opportunity to observe and quantify the changes in the body of the debris deposit after its formation. The authors described post-formation behaviors of the debris mass, breaching-inflicted changes of the debris mass, and upstream and downstream reaches in their referred publications. This paper summarizes the findings from the observations of the two referred papers and the comments on the situation of the landmass after almost a decade of post-breaching of landslide mass.

1 Introduction

The October 8, 2005, Kashmir earthquake (epicenter 34°29′35″ N, 73°37′44″ E, focal depth 26 km, 7.6 Mw (USGS)) caused widespread destruction in northern areas of Pakistan and Pakistan-administered Kashmir, killing more than 86,000 people. The earthquake also triggered a massive landslide 3.5 km upstream of Hattian-Bala town near the southeastern end of the activated Balakot-Bagh fault.

The debris deposit resulting from the landslide blocked two tributaries of the Karli branch of the Jhelum River, forming Karli Lake (the large lake) and Tang Lake (the small lake). This landslide dam (called the Hattian Dam) became a significant concern for the people living along the lower reaches of the Karli and the Jhelum Rivers since the failure of such landslide dams usually results in catastrophic downstream flooding causing loss of lives, houses, and infrastructures.

The Hattian Dam survived the initial overtopping because a spillway was excavated across the Dam’s crest as a part of the flood disaster mitigation strategy; however, the danger of catastrophic flood outbursts in the case of the dam breaching was eminent. Sattar et al. (2011) observed deformations/erosions that had been occurring to the debris deposit since the debris dam was formed and estimated the potential outburst flood in case of breaching of the landslide dam. The simulated scenario of the flood water outburst indicated that about 40 buildings along the right bank of the Karli River, bounded by the Karli River on the west side and by the main road on the east side, were potentially in high-risk locations.

The Dam sustained over about 4 years and 4 months, and finally, the water of Karli Lake breached the northwestern part of the Dam on February 9th, 2010, following a continuous 5-day rainfall. As predicted earlier, the flood water engulfed many houses in Hattian Bala along the right bank of the Karli River, swallowing a boy. However, it could have been worse unless residents evacuated quickly.

Later, the water of Tang Lake also breached the northeastern lobe of the Dam during the monsoon rains from July to August 2010.

It was less likely that a flood of the same scale would occur again and cause any hazard for the downstream reach of Karli valley because a bedrock, which appeared on the scoured bottom of the breach channel, was not likely to be eroded any further. However, the remaining mass of the Hattian Bala landslide dam and its vicinity were not entirely out of danger after the breaching of the Dam.

This paper reviews the findings from the authors’ observations conducted before and after the dam failure reported by Sattar et al. (2011) and Konagai and Sattar (2011), and the post-breaching behavior of the landslide dam.

2 Features of Debris Deposit

The surface rupture that caused the 2005 earthquake was along the preexisting active faults or fault segments, collectively called the Balakot-Bagh fault. The Hattian-Bala landslide dam lies near the southern end of the Balakot-Bagh fault at approximately 33 km southeast of Muzaffarabad, the capital and largest city of Azad Kashmir, and 3.4 km from the fault (Fig. 1). Sattar et al. (2011) summarized the seismological and geological aspects of the region.

Fig. 1

(a) Map of Hattian area showing the source area of Hattian landslide mass, Hattian landslide dam, Karli Lake and Tang Lake with their catchment areas, approximate location of the active Balakot-Bagh fault and the main tributaries of the Karli River and Jhelum River. (b) Quick bird image (27/10/2005) of the Hattian dam Konagai and Sattar (2011)

About 65 million m³ of Hattian landslide mass originating from Dana hill ran down the valley slope and deposited its final volume of 85 million m³ along the Karli River. Immediately after the earthquake, the water levels in the lakes started to rise; therefore, a spillway was constructed across the Dam’s crest as a disaster mitigation measure against the potential breaching of the Dam. The water level in Karli Lake reached the maximum invert level (approx. 1354 m elevation above mean sea level, AMSL hereafter) of the artificial spillway in April 2007 and started spilling over. The 18 m deep excavation of the spillway limited the maximum storage capacity of Karli Lake to 62 million m³ from its initial 86 million m³ corresponding to the natural overtopping level, considerably reducing the possible flood magnitude in the case of breaching of the Dam. About 5 million m³ of water was impounded in Tang Lake.

The Water and Power Development Authority of Pakistan (WAPDA) monitored the landslide dam after its formation. Figure 2 shows the measured inflow of the large lake, the small lake, and seepage through the landslide mass from December 2005 till June 2006, and Fig. 3 shows the water levels in the Small and the large lakes measured during the same time (WAPDA 2006). From the measured water levels and seepage through the landslide mass, Sattar et al. 2011 concluded that the surface part of the dam body was permeable, while the deeper part of the debris deposit was impermeable. It was a distinctive nature of the landslide dam body that larger rocks gathered along shallower and toe parts of the dam body. The small lake reached its maximum capacity of 5 million m³ at the end of February 2006. The water level in the small lake decreased gradually after April 21st, even though the inflows to the lakes were constant, indicating that the interior of the debris deposit in front of the small lake had developed some piping erosion. The monitoring regime initiated by WAPDA ended after the water levels in the lakes were observed to be stable. The rock/earth wall lining the spillway extended only a short distance over the crest. Nonetheless, monitoring the dam site attracted less attention as the spillway performed well, and no immediate threat was imminent. Beyond the end of the spillway, the water flowed naturally over the body of the landslide mass.

Fig. 2

Discharge data from WAPDA (2006)

Fig. 3

Water levels of lakes measured from December 2005 till June 2006 (WAPDA 2006)

Gradual toe erosion was observed from November 2008 to June 2009 (Fig. 3). A significant retrogressive erosion developed about 300 m through the toe slope at some indeterminate time. Sattar et al. (2011) inferred from the observations that the retrogressive erosion developed because of an extreme rainfall event during the off-monsoonal dry season. Sattar et al. (2011) attributed the accelerated slaking of dried mud rocks of the debris mass to the observed backward erosion.

3 Observation of Changes to the Debris Deposit

Sattar et al. (2011) and Konagai and Sattar (2011) initiated a measurement regime to quantify the changes occurring in the landslide mass after its formation. Differential global positioning system (DGPS) measurements and laser scanning were included in the various techniques for quantifying deformations. The backward erosion mentioned above, which occurred somewhere between November 2008 to June 2009, resulted in the loss of approximately 65,000 m³ of soil mass from the downstream slope of the landslide dam (Fig. 4). This event raised concerns about the potential risk of breaching. The DGPS measurements conducted over a 6 months interval (June 2009–November 2009) showed that the crest zone of the Dam settled by order of 100 mm and the zone near the head of the eroded gulch was uplifted by a few millimeters (Sattar et al. 2011). Figure 5 shows the settlement observed during the period starting from June 2009 till November 2009. The settlements were observed to be greatest (up to 100 mm) along where the thickness of the debris deposit was greater, i.e. towards the eastern side of the deposit.

Fig. 4

Retrogressive erosion of debris mass (till Nov 2009) leading to development of erosion gully (June 2009)

Fig. 5

Settlements obtained by comparing DGPS data for June 2009 and November 2009 surveys. The perimeter of the erosion channel observed in June 2009 and November 2009 is marked near the toe of the debris mass (Sattar et al. 2011)

Meanwhile, gradual changes were being observed along the northeastern lobe of the debris deposit, where the water from Tang Lake was seeping into the deposit of large segregated boulders and was emerging at some distances downstream. With time, there was a growing suspicion that pipes were being developed in the interior of the debris mass, as indicated by the gradual surface deformations observed over the surface of the deposit (Fig. 6).

Fig. 6

View of the path of water flowing from the Tang Lake. The surface change can be observed between November 2006 (left) and June 2008 (middle). While the right most figure shows the breach channel formed in front of Tang Lake. Photo points are approximately N34.1420°, E73.7347° for the left and middle and N34.1427° E73.7326° for the right

4 Dam Break Analysis Prediction by Sattar et al. (2011)

In order to evaluate the inundation level of the flood wave from a breach of the Dam, Sattar et al. (2011) carried out a dam break analysis. The simulation was based on a down-cutting rate varying from a minimum of 10 m/h and a maximum of 100 m/h. The digital elevation model (DEM) used for the simulation was extracted from Shuttle Radar Topography Mission SRTM digital data, which covered the region from Hattian to Muzaffarabad city.

A two-dimensional flood routing model by O’Brien et al. (1993), Flo-2D, was used by Sattar et al. (2011) for inundation modeling of the dam breach scenario of the Hattian landslide dam. Flo-2d is a volume conservation model, and its constitutive fluid equations include the continuity equation and dynamic wave momentum equation (Saint Venant equations).

$$ \frac{\partial h}{\partial t}+\frac{\partial h{V}_x}{\partial x}=i $$

(1)

$$ \frac{1}{g}\frac{\partial {V}_x}{\partial t}+\frac{V_x}{g}\frac{\partial {V}_x}{\partial x}+\frac{\partial h}{\partial x}={S}_{0x}-{S}_{fx} $$

(2)

where, ‘h’ is the flow depth, ‘V_x’ is the depth-averaged velocity component and ‘i’ is the excess rainfall intensity, ‘S_fx’ is the friction slope based on Manning’s equation, ‘S_ox’ is the bed slope, and ‘g’ is the acceleration due to gravity.

The solution domain is discretized into uniform, square grid elements. The differential form of the continuity and momentum equations are solved with a central finite difference numerical scheme which solves the momentum equation for the flow velocity across the grid element boundary one element at a time. The incremental discharge for a grid element at a time step and the corresponding increase in flow depth are computed using the velocity components of the grid element.

The breach outflow hydrograph at the toe of Hattian dam for 100 m/h, 50 m/h, and 10 m/h down-cutting rate showed a peak outflow discharge of 13,830 m³/s, 5200 m³/s, and 700 m³/s, respectively.

The maximum flow depth at the confluence of the Karli branch and the Jhelum River (about 3 km downstream of the dam) was estimated to be 22 m (Fig. 7); however, the flow remained confined within the valley. Potentially hazardous low-elevation locations along the Jhelum River possessing a significant number of buildings were identified based on simulation.

Fig. 7

Google image^® of Hattian-Bala district overlain by maximum flow depth contours (in ‘meters’) of dam breach simulation by Sattar et al. Close-up of potentially hazardous locations along the Jhelum river with significant number of structures are shown

The simulation also demonstrates that the flood wave significantly attenuates along the Jhelum River when it reaches the main city of Muzaffarabad, 33 km downstream from the landslide. The simulation result was reported to the State Earthquake Reconstruction and Rehabilitation Agency (SERRA), Azad Government of the State of Jammu & Kashmir, on June 13, 2009, about 8 months before the breaching.

5 Failure of Karli Lake and Tang Lake

After 5 days of continuous rains in early February 2010, the water of Karli Lake breached the debris dam on February 9, 2010. This breaching drained about 36 million m³ of water from the lake, leaving 26 million m³ in the lake. The erosion stopped after the bedrock was exposed at the scoured bottom of the breach channel (Fig. 8). Using the laser scanning technique, we estimated that about 7.78 million m³ of soil was eroded during the breaching event. The flushed debris mass was deposited along the Karli River (Fig. 9). The thickness of the debris deposit was estimated to be up to 80 m close to the toe of the Dam and generally about 15–20 m along the 3.7 km reach of the river till it merges with the Jhelum River at Hattian village.

Fig. 8

(a) Front view of Hattian landslide dam, before and after failure, (b) Front view of the dam immediately after breaching (Konagai and Sattar 2011)

Fig. 9

Partially breached Hattian dam and debris deposit along the Karli River. (Photo taken on sixth June 2010 at N 34.156567°, E 73.744833°)

The TRMM-data-based estimate showed several-day continuous precipitation in early February, immediately before the breaching of Karli Lake. An assessment of the rainfall indicated a peak value of 32 mm/day on February 8^th, 1 day before breaching. However, several facts negated the possibility of breaching the Dam solely because of the rainfall. These facts included that the mountain slopes around Hattian Dam were covered with snow when the breaching occurred (Fig. 10), suggesting that only a part of this precipitation may have flowed into the lake. During the authors’ survey after the breaching, we saw a landslide mass deposited on the right bank of the emptied large lake. So, among the possible scenarios discussed was the rising lake water level due to the plunging of the unstable landmass into the lake. However, as will be discussed later, such a scenario for the dam breaching was not possible, considering the observations of the eyewitnesses of the event and the post-breaching photographs.

Fig. 10

Right bank slope (a) On 20/08/2008, before the breaching event (b) One day after the breaching event showing snow covered slope and mark of the water level of Karli Lake before breaching event

We examined the maximum flood levels during the breaching event by measuring the elevations of the flood marks visible along the downstream reach of the river. It turned out that the peak outflow discharge of 5500 m³/s at a down-cutting rate of 25 m/h best fits the measured flow depths and flow velocities along the Karli River (Fig. 11). As Sattar et al. (2011) predicted earlier, the flood water engulfed many houses in Hattian Bala along the right bank of the Karli River, swallowing a boy. However, it could have been worse unless residents evacuated quickly.

Fig. 11

(a) Flood inundation heights along the Karli River. (b) Houses affected by the debris originating from the Hattian dam. (c) Flow velocities estimated from super-elevations of the measured mud marks near the junction of Karli River and Jhelum River

The monsoon rains in late July 2010 caused a mega-flood in Pakistan, and the flood is considered the worst in the region’s record. Intensive rainfall was observed in Muzaffarabad and surrounding areas, including Hattian Bala. Continuous rain reportedly caused extensive landslides along the Neelum valley, and to a lesser extent, along the Jhelum valley. The Tang Lake (smaller lake) breached (Fig. 12) during the days of the most intensive monsoon rains of 2010 (July 27 to August 1); however, the exact day of the breaching remains unknown. The TRMM data indicated a peak rainfall rate of 105 mm/day for July 30 which could have been when Tang Lake breached. In the breaching process, the water from Tang Lake eroded the northeastern end of the landslide lobe, which had been covered with boulders segregated up at the time of deposit. The top 15 m depth of the lake water was drained during the breaching event leaving behind about 1.9 million m³ of water in the lake. As mentioned earlier, piping is considered to have been developed in the interior of the debris deposit in front of Tang Lake.

Fig. 12

Tang lake after breaching (Photo taken on 6th Dec-2010) GPS coordinates N34.13951° E73.73631°

It is inferred that the breaching process started as a backward erosion piping. The down-cutting rate of Tang Lake could have been slow because the boulders which were present along the water path were carried downstream without extensive crushing, differing from what we observed in the event of the Karli Lake failure. The slow down-cutting rate of Tang Lake can be attributed to the armoring effect provided by the surface boulder layer. No clear floodwater marks were found for this event, probably because the floodwater discharge was much smaller than Karli Lake’s discharge.

6 Discussions

The landslide mass damming Karli Lake failed after a rain/snowfall runoff in early February 2010. It remains to be clarified how the moderate precipitation caused the breaching, as the debris dam had survived much heavier rainfall before its failure. A possible mechanism that caused the breaching can be explained by considering the weak weathering resistance of the debris material. The debris originated from the Murree formation of Miocene-aged sedimentary rocks composed of reddish mudstones, shales, and greyish sandstones. The effect of weathering on the mudstones had been observed at various locations over the dam body, where mudstone boulders were slaked and reduced to fragments. Reddish watermarks found occasionally along waters seeping out of the dam body indicated that internal seepage flows had washed finer substances out.

The Hattian area experiences extreme weather conditions during summer and winter along with the monsoon season with heavy rainfalls, setting ideal conditions for wetting and drying cycles on the slaking process. Signs of a deteriorating dam body were continuously observed with the gradual surface deformations and later with the sudden detachment of a significant amount of soil from the Dam’s downstream face, as discussed above. In February 2010, relatively moderate precipitation extended for 5 days in the Hattian catchment, followed by dry and clear days in February. It thus may have accelerated the slaking process in a way that even moderate waters flowing over and seeping through the debris mass significantly accelerated backward erosion. Kiyota et al. (2011) discussed the slakable nature of the dam material through standard slaking tests and advanced unconventional direct shear tests on prepared soil specimens retrieved from the remaining dam body.

As mentioned earlier, we found a landslide mass deposited on the right bank of the emptied large lake during the post-event survey. According to eyewitness accounts, the breaching of the landslide mass triggered this large landslide, and the mass movement continued for 2 weeks after the breaching event. This coherent landslide mass was about 1000 m long in the runout direction and 800 m wide (as indicated by red arrows on the right bank of Karli Lake in Fig. 13). The landslide destroyed 174 houses and displaced about 1000 residents, making it more extensive damage than the one directly caused by flooding.

Fig. 13

Google earth^® image of the Hattian Bala dam after breaching of the dam and serval years post breaching. Red arrows and the yellow line across the dam body indicate the channel’s alignment over the dam body as of May 2011, and the channel alignment as of April 2018 is visible in the figure on the right. Red arrows along the right bank of Karli Lake indicate the coherent landslide mass in both figures

As contrasted with the Karli Lake failure, the Tang Lake failure did not produce any significant flood wave; indeed, it was not until the authors reported it that many residents at Hattian Bala knew that Tang Lake had failed. Perhaps, it is because the emptied water (1.6 million m³) was much smaller than that from Karli Lake, and the lobe of the debris mass that had been stopping the water of Tang Lake was covered thick with segregated boulders.

It was less likely that a flood of the same scale would occur again and cause any hazard for the downstream reach of Karli Valley because a bedrock, which appeared on the scoured bottom of the breach channel, was not likely to be eroded any further. However, the exposed bare walls of the remaining Hattian Bala landslide dam were not completely stable. Any collapses of the walls would clog up the breached channel and increase the lake volume at any time (in an earthquake, heavy rainfall, or otherwise). Konagai and Sattar (2011) discussed possible scenarios for these landslides and their impacts. They also reported post-breaching landform changes. These included:

• the erosion of materials from the debris deposit and raising of the river bed level at downstream river reach;

• the hazard posed by the unstable landslide mass at the right bank of Karli Lake;

Figure 13 compares the Google Earth^® image of the Hattian Bala landslide dam after the breach (May 2011) and a more recent image of the area (April 2018). The unstable landslide mass at the right bank of Karli Lake showed no clear sign of movement, as apparent from Fig. 13. The gorge formed over the landslide mass body by the overflowing water seems to have slightly changed its course towards the east, as indicated in Fig. 13. Some changes to the landmass near the toe of the landslide mass also appear insignificant. Beyond the toe of the landslide mass, the overflowing water changed its course over time.

The changes observed over seven post-breaching years are relatively insignificant, and the landslide mass appears stable and does not pose an imminent danger.

7 Summary

The October 8, 2005, Mw 7.5 Kashmir earthquake triggered a massive landslide 3.5 km upstream of Hattian-Bala town near the southeastern end of the activated Balakot-Bagh fault. The debris deposit resulting from the landslide blocked two tributaries of the Karli branch of the Jhelum River, forming Karli Lake (the large lake) and Tang Lake (the small lake). The Dam sustained over about 4 years and 4 months, and finally, the water of Karli Lake (large lake) of Hattian dam breached the northwestern part of the landslide dam on February 9th, 2010, after 5 days of rains/snowfalls, and the water of Tang Lake (small lake) breached the northeastern lobe of debris deposit during the monsoon rains of July–August, 2010. Until the failure of Karli Lake, the landslide dam had long been a severe threat to people living along the lower reach of the Jhelum River, and the authors had been conducting periodic observations and measurements of changes occurring in the landslide mass since the authors’ project started in 2008. The landslide mass started to show more evident signs of gradual deformations/erosions after the overflow started through the spillway, which was excavated to secure the safety of the debris deposit against possible failure of the Karli Lake. Evidence of weathering (slaking) of the mudstones of Murree formation was observed over the boulder-strewn landslide dam. The most significant change observed in the dam body before the failure of Karli Lake was a sudden erosion of an approximately 65,000 m³ soil mass from the downstream slope of the landslide dam during heavy rainfall in 2009. The rainfall estimate over the Hattian catchment area indicates that the precipitation of February 2010 that caused the breaching was less significant than the other heavy rainfalls that the area had experienced since the overflowing from Karli Lake started in April 2007. The breaching was most likely due to the slaking process developed during relatively moderate precipitation in the Hattian catchment, followed by dry and clear days in February, as suggested by Kiyota et al. (2011).

Sattar et al. (2011) estimated that the maximum flow depth at about 3 km downstream of the dam was 22 m and reported it to the State Earthquake Reconstruction & Rehabilitation Agency (SERRA), Azad Government of the State of Jammu & Kashmir, on June 13th, 2009, about 8 months before the breaching. As predicted, the flood water from the breached dam engulfed many houses in Hattian Bala along the right bank of the Karli River, swallowing a boy. However, it could have been worse unless residents evacuated quickly.

Landslide dams not only pose flood hazards or the associated landmass movements, but these debris deposits can also become an issue for the sustainability of any project on the major rivers if the landmass is relatively large. One example is the Attabad landslide dam blocking the Indus River at its upstream reach. It is to be noted that the Indus cascade, a big project in which the first author is involved, is currently under development in Pakistan. i.e., the cascade of dams, namely the Basha dam, Dasu dam, Patan dam, and Thakot dam; they are all located downstream of the Attabad landslide dam. Safe passage of the flood generated by breaching the landslide dam becomes a critical design consideration for these dams. A breaching scenario of the Attabad Dam thus forms the basis for the safety check floods for these dams. From the perspective of the sustainability of the under-construction hydropower projects, more minor landslides and debris flows associated with monsoon are also an important aspect that requires attention. According to estimates (Roca 2012), about 200 million tons of sediments are transported to the Terbela dam reservoir annually. Apart from the glacial melts, a portion of the source of sediments is the landslides and debris flow in Nallas during the monsoon, which eventually is transported as bed load and suspended sediments in the Indus River. The Indus cascade development must cope with the transported sediments to sustain the reservoirs.