Reported landslide accident in SW India

AP is reporting that a landslide in S India has killed 39 people. The landslide occurred in the Ooty and Coonoor region of Tamil Nadu states, hitting 300 mud huts. The report quotes a "state flood control official" as saying that 14 bodies were recovered on Monday and a further 25 today.

Tamil Nadu is in the very south of India. The TRMM landslide warning system has this region as being likely or very likely to suffer landslides:

The relationship between global fatal landslides and rainfall in July

I thought it might be interesting to compare the occurrence of fatal landslides in July 2007 in relation to the distribution of rainfall in that month. Let's start off by looking at the July 2007 fatal landslide map (you will probably need to click on the map to get a decent view of it).

As a comparison, below are the statistics for July 2007 and July 2008, followed by the fatal landslides map for July 2008.



The number of fatal landslides is about the same, but the number of fatalities is very different. At first glance these two maps are also quite similar, but if you look carefully there are some key differences:
1. They are similar in that both have a clear cluster in Central China, although the location is subtly different;
2. In 2008 there were a number of landslides in Central America, whereas in 2007 there were none;
3. They both have a clear cluster along the Himalayas, but close examination shows that there are key differences between the two. In 2007 there were far more fatal landslides in Asia in 2007 than there were in July 2008. The southern edge of the Himalayas is particularly interesting - in 2007 there was a big cluster of landslides in northern Pakistan. This is not the case in 2008. In Nepal the 2007 distribution was concentrated in the west, whereas the 2008 focus was in the central part of the country. On the other hand, in 2007 there was a number of landslides in Bangladesh, but this was not the case in 2008, when only one occurred. Finally, in July 2007 there was a cluster of landslides in southern India, but only one landslide in this month is 2008.

So what drives the differences between these two years? The most obvious explanation is that of rainfall patterns between the two years. It is possible to look at this issue because TRMM provides monthly maps of rainfall anomaly - i.e. the difference between the recorded monthly rainfall and the monthly total averaged over several preceding years. Unfortunately, this map is not available for 2008 yet, but 2007 is online. Given that most of the landslides in July 2007 occurred in Asia, let's take a look there. First the landslide map:

Now the TRMM rainfall anomaly data. The units are mm / day difference from the long term July mean.

It should be fairly clear that there is a strong correlation between the distribution of fatal landslides and the rainfall anomaly in July 2007. In particular, the clusters along the southern edge of the Himalayan arc are focused almost exactly where the rainfall was unusually high. The apparent gap in central Nepal corresponds to an area of lower than average rainfall. Similarly the cluster in south-west India is also related to a high rainfall anomaly.

Finally, below is a NASA image of population density for Asia. Dark red colours represent high population density, orange is low. It is clear that in Asia the population is very unevenly distributed. Note the high population densities in places that were particularly badly affected by landslides in July 2007. Thus, it is clear that the distribution of rainfall probably determines the distribution of landslides, and the density of population then determines the fatality total. Of course this is a bit of an over-simplification, not least because landslides only appear on the database if they kill someone, which means that the population density plays a role here too, but I am sure that you get the general idea.

Finally, it should be noted that Dalia Bach, Yang Hong and the TRMM team have also been looking at this issue and have completed an analysis for landslides in general. See this web page for details:
http://trmm.gsfc.nasa.gov/publications_dir/potential_landslide.html

TRMM and landslide monitoring

One of the most admirable science projects at NASA at the current time is an attempt to create a global landslide and flood warning system using the Tropical Rainfall Monitoring Mission satellite data. TRMM using a precipitation radar system to provide near real time data on the intensity of rainfall in the tropical and subtropical regions. NASA then make these data available on the TRMM website.

The usefulness of these data can be illustrated using the last seven days of rainfall in Asia. During this time the monsoon has strengthened - we are just moving into the real wet season in south and SE. Asia. Furthermore, the first really destructive tropical cyclone of the year has just tracked across the Philippines and then made a more-or-less direct hit on Hong Kong, before dissipating in southern China. As well as causing the sinking of a passenger ferry, this typhoon also deposited large volumes of rain that have been responsible for extensive landslides and floods. Using TRMM it is possible to obtain an estimate of the volume of rainfall that has fallen in Asia over the last week (Fig. 1). This does come with a caveat, which is that the spatial distribution of rainfall is very variable at the local scale, which this instrument does not resolve. Nonetheless it does provide a really interesting indication of the rainfall distribution.

Figure 1: TRMM image of the rainfall distribution for the seven days to 27th June 2008 at 06:00 UTC. Note the high rainfall totals for the Philippines and S. China, generated by the passage of Typhoon Fengshan (click on the image for a better view).

The locations of high rainfall appear to correlate with areas of flooding - for example, in this time floods were reported in the Central Philippines, South China around the Pearl River delta and in parts of N. India. Unfortunately, the correlation with fatal landslides at least is far poorer. Figure 2 shows the locations of fatal landslides that I have recorded in the same time period. A caveat is that perhaps not all the landslides in China have yet been reported, so the map in Figure 2 might change a little. However, it is clear that a simple rainfall total algorithm is not enough to provide a really good indication of the distribution of fatal landslides at least.

Figure 2: Locations of reported fatal landslides in the 7 day period shown in Figure 1.

The team at NASA are working hard on their landslide algorithm and are doing a terrific job. This is one of the most interesting landslide projects underway at the moment. I am hoping that they will present some interesting progress in this work at the International Symposium on Landslides and Engineered Slopes in China next week.