Deadly landslides have been in the news on several occasions in recent weeks. A landslide in Yunnan, China, buried a school, killing 18 children. A series of landslides in Uttarakhand, India, claimed the lives of at least 40 people. In Dorset, U.K., three people died in two landslides triggered by heavy rain.
Given these deadly impacts, it is surprising that until recently we had almost no data on the global cost of landslides nor about where they cause the highest losses. To fill this gap, I have spent the past 10 years collecting data about deadly landslides from around the world.
The results are surprising. In a paper in the journal Geology, I analyzed the data from 2004 to 2010. In this seven-year period I recorded 2,620 rainfall-induced landslides worldwide that killed more than 32,000 people, a much higher toll than previously thought.
The total number of fatalities is even higher than that, as my analysis only considered landslides triggered by rainfall. If other landslides are taken into account, especially those triggered by earthquakes, the death toll rises to a remarkable 80,000.
This is in stark contrast to official figures in the United Nations International Disaster Database, which indicate only about 7,400 deaths from landslides and avalanches during the same period.
Why such a large discrepancy? The explanations lie both in what gets included and the nature of landslides themselves.
Let's address the first issue. A key factor is that the U.N. database includes many other hazard types, such as earthquakes, floods, and volcanic eruptions. Compiling such data is a huge task, so the researchers set a minimum impact threshold for inclusion. This is either 10 fatalities or a large economic loss, meaning that many small events are excluded. This is not a problem for earthquake-related data, as unfortunately almost all events that cause significant damage also kill more than 10 people, but it leads to substantial underreporting for landslides, most of which are small and localized.
A related issue is that the U.N. database records only a single cause for each death. All of the approximately 68,000 fatalities in the 2008 Sichuan earthquake in China, for example, are recorded as resulting from the quake. In a sense this is true, but more than 20,000 people were killed by landslides triggered by the event. This is a general problem: Landslides are typically triggered by another event, often intense rain or an earthquake, so their impacts are consistently underestimated.
A second key factor is that fatal landslides are concentrated in populous but remote mountain areas prone to heavy rainfall or earthquakes. The global epicenter is along the southern edge of the Himalayas; other hotspots are Indonesia, the Philippines, western China, some Caribbean islands, and Colombia.
In such places, obtaining information about landslides that kill small numbers of people has been hitherto impossible. The availability of digital and social media has made it easier.
The ultimate goal of such data gathering, of course, is to reduce landslide deaths. In the past three decades, several countries have set up successful programs to manage landslide risk. A good example is Hong Kong, which suffered a series of major accidents in the 1970s. Their program has reduced loss of life to a handful of fatalities per decade through a combination of engineering works—such as building retaining walls and installing drainage—public awareness, relocation of people most at risk, and an early warning system. While replicating this in full in less-developed countries is probably not feasible, some of the measures should be possible even with limited budgets.
An interesting question is whether deaths from landslides are increasing. There are good reasons to think they might be. My research shows that as population density increases, so does the number of fatal landslides. In part, this is probably because rising population forces people to live and work on unstable land and, of course, when there are more people in the landscape, it becomes more likely that any given landslide will hit someone.
Other factors will also be at play. Environmental degradation, especially deforestation, seems to be making landslides more likely. And the widely observed increases in rainfall intensity, which are probably associated with a warming atmosphere, may also be contributing.
My data set is still too small to determine whether there is a long-term upward trend. The impact of landslides varies considerably from year to year. Although 2010 has the highest number of recorded landslides, the number is considerably lower for 2011, and 2012 looks like it will be lower still. This probably reflects the state of global weather systems such as the El Niño/La Niña cycle and continental systems such as the Asian monsoon, which has been much weaker than normal in 2012.
The landslide data set has very similar characteristics to a weather one; in the latter case about 30 years of data are required before a trend can be determined. It is likely that I will need to continue to compile data for 20 more years before this question can be addressed.
Rainfall-induced landslides are for the most part a manageable hazard, and a coordinated effort to reduce landslide deaths in poorer nations could be highly effective. The first step needs to be a research program that seeks to better understand the occurrence of landslides and the mechanisms responsible for them. Unfortunately at present there is little indication that this is a priority.