16 January 2013

New research: extreme precipitation and landslides in 2010

Posted by dr-dave

ResearchBlogging.org As regular readers will know, since 2002 I have been maintaining a database of landslides that kill people worldwide (and this work was recently published in the Journal Geology).  In that dataset 2010 is the year with the highest level of losses from rainfall-induced landslides – it truly was a remarkable year.  In a recent paper (available online here and published behind a pay wall in the Journal of Hydrometeorology), Dalia Kirschbaum from NASA and colleagues (Kirschbaum et al. 2012) have used their own catalogue of mass movement events to examine the relationship between landslides and heavy rainfall.  The landslide catalogue that they have used is rather different to mine because it compiles information about all reported rapidly moving landslides, irrespective of their impact.  As such it is more comprehensive than my dataset, although it may be more subject to the vagaries of media and other types of reporting.  In this study, the catalogue has been compared with precipitation data from the TRMM satellite.  The TMPA dataset that they have used combines the TRMM data with rain gauge data to produce daily global rainfall dataset at a resolution on 0.25 x 0.25 degrees.  This dataset is known to represent large rainfall events quite well.

Perhaps unsurprisingly, the landslide dataset used here also show unusually high levels of landslide activity in 2010, with the increase above normal levels occurring primarily in Central America, the Himalayan Arc and Central-Eastern China.  In each case, the authors clearly show that the elevated levels of landslide activity were associated with rainfall levels that were above normal.  So, for example, this is the data for South Asia


The top graph presents the precipitation data for this region.  The red line is the precipitation pattern for 2010, whilst the green line is the average over 12 years.  It is clear that the summer monsoon in this season was much wetter than the medium term average.  The second (lower) graph presents the number of reported landslides in 2010 (in orange) and the average number of reported landslides for 2007-2009 in blue.  Clearly the landslide occurrence level was significantly higher than in previous years; the obvious potential cause is the higher levels of precipitation.  The other areas show similar patterns.

Now, in some ways it might be obvious that years with larger rainfall totals, and in particular more extreme events, are associated with more landslides.  Indeed, no-one is terribly surprised that the recent wet weather in the UK has been linked to a much higher occurrence of landslides across the country.  From that perspective, this paper is confirming what we already know, although the importance of exploring and quantifying an accepted truth can be underestimated.

However, this contribution is much more important for two reasons. First, the fact that we can combine different datasets in this way to produce a useful analysis provides a basis upon which to start to better understand spatial and in particular temporal patterns of landslides.  Kirschbaum et al. (2012) start the process of examining the relationship between the observed landslide patterns and global weather systems, such as the El Nino cycle.  They suggest for example that it is now established that in the year after an El Nino event, rainfall levels in the Himalayas are higher than normal, which explains the patterns in 2010.  This is really useful information that will help improve management and mitigation of landslides in the future.  Much more work of this type is needed, but this research opens the doors to it.

Second, the analysis suggests that combining these datasets provides a much better way to understand those areas that are at long-term risk from landslides.  This is likely to be a much more robust approach than the existing global landslide hazard and risk maps, such as those generated by the World Bank hotspots project.  Such approaches need more work, and this will take time, but in due course we should be able to identify those regions that are landslide-susceptble to a much better degree than at present.  Of course we also need to do something similar for earthquake-triggered landslides, but the low-level of occurrence of earthquakes will make this rather more difficult.


Kirschbaum, D., Adler, R., Adler, D., Peters-Lidard, C., & Huffman, G. (2012). Global Distribution of Extreme Precipitation and High-Impact Landslides in 2010 Relative to Previous Years Journal of Hydrometeorology, 13 (5), 1536-1551 DOI: 10.1175/JHM-D-12-02.1