23 May 2010

Attabad – less than two metres to go before water flow

Posted by Dave Petley

Several people have asked for an update on the situation at Attabad, and a forecast of what will happen, so here goes.

First, the situation as I understand it is as follows.  In the last few days the availability of data on the situation on the ground has become difficult for reasons that are very understandable.  However, Focus have continued to take measurements of lake depth when possible, and NDMA have also posted information online.  Thus, a graph of the freeboard of the two measurements looks like this:

Whilst there is close agreement between the two datasets over the last ten days or so, and generally good agreement overall, the difficulties of forecasting the final point of overtopping are clear when just the last seven days of measurements are shown:

The next NDMA update will help – however my experience through the whole of this has been that the Focus data is the most reliable.  With a Fovcus measured increase of water level of 85 cm in the 24 hours ending yesterday, and a freeboard of 1.95 metres at lunchtime (local time) yesterday, the most likely time of overtopping is late tomorrow (24th) or early on the 25th.  This situation is not going to be any easier to manage should overtopping start at night.  However, the most recent NDMA data might suggest a slightly later date, and of course there is always the possibility of an earlier failure as discussed previously.

So, what is going to happen at the time of overtopping?  It is impossible to say with any certainty, but the most recent images from FOCUS, provided late on Friday, do help.  First, this overview image shows the dam, spillway etc, taken from the rock spur that is the location of the early warning team.  It is pleasing to see that there is no-one on the dam itself, which has been evacuated.  Note the seepage points.

Second, this image shows the spillway looking upstream towards the lake.  I was somewhat shocked when I saw this image:

The spillway is really very narrow indeed – only 7 to 8 metres deep at the base and 15 metres at the top, and about 14 m deep.  As such it seems too small to be able to handle even the current inflow rate (about 80 cubic metres per second), let alone the summer flood.  The channel is unlined, the walls are unprotected, it is not straight and it is too steep, all shown by this image inside the spillway:

So erosion and a release of a substantial volume of the lake water is I think very likely.  However, this may well not start immediately.  Indeed, at a rate of water level rise of 80 cm per day, which will reduce as outflow commences, it may take a couple of days for the spillway to carry the full inflow.  Thus, I would not be surprised if the spillway does not erode when water flow starts, but that it takes a few hours to a few days to initiate this process.  Once it begins, erosion will probably happen in three ways:
1. I would expect to see basal erosion, Once erosion starts along the whole of the base of the channel the process is likely to accelerate quite quickly.
2. I think we will certainly see erosion of the sidewalls, probably causing collapse of the banks, probably starting fairly soon after flow starts.  This can be very bad news indeed (temporary blockages can allow the water level to build up and then permit a rapid release) or it can be quite helpful in terms of widening the channel, increasing the capacity (see below).
3. We may see erosion on the downstream face as soon as water flow starts, which then could eat back into the dam, steepening the channel.  We could even see new slope failures on the downstream face, but the lower gradient here might help. 

That the materials on the downstream face are erodible is shown by the newly-formed channels below the seepage points.  Note also just how much seepage there is now – and remember that this is just one of several points:

It is also worth bearing in mind that this is about 2 cubic metres of flow per second; the current inflow is about 70 cubic metres per second, and the summer flow rate is probably  at least five times as great again.

The most likely scenario therefore is that water flow across the spillway is stable for a while, whilst eroding the sidewalls, then as the water volume increases the flow rate increases and basal erosion starts.  This will accelerate as more water is released, and the breach develops.  I still hope that this will happen reasonably slowly (i.e. over a few hours), but I believe that it could be rapid, especially in view of the inadequate spillway form.  Of course a very rapid failure could still be initiated by another landslide, or by sudden development of the seepage.

If you want information about what a collapse of a landslide barrier actually looks like, can I refer you to the Tangjiashan event, which I monitored two years ago.  This went from this (water flow through the spillway, which here was narrow but not steep, but was designed to initiate erosion):

To this (rapid failure of the landslide mass):

Peak flow occurred about 5 hours after erosion started.  In this case though the dam was broader and more resistance to erosion (the materials were more coarse-grained), which meant that the channel widened rather than deepening rapidly. Peak flow was about 9000 cubic metres per second.

I hope that we are equally as fortunate with Attabad.

Finally, I have stopped updating the monitoring site  for now as the monitoring data flow is too slow.  Updates will happen through this site.