Climate Change Vulnerability Mapping for Southeast Asia

A very interesting report was released last week by the Economy and Environment Program for Southeast Asia (EEPSEA) in which an attempt has been made to map the pattern of vulnerability across the region to climate change. The aim of the project, for which the report is available online, was to "identify which regions in Southeast Asia are the most vulnerable to climate change." This is a fairly bold thing to try to do, so lets take a look at what they have achieved and how they have done so.

The researchers have taken as a starting point maps of five hazards - cyclones, drought, floods, landslides and coastal inundation (sea level change). In each case they have used an external dataset to indicate the hazard associated with each of these events - so, for example, for landslides they have used the NGI landslide hazard maps produced in the World Bank Natural Disaster Hotspots project (Fig. 1), whereas for sea level change they have modelled the inundation associated with a 5 m increase in sea level.

Fig 1: NGI landslide hazard map used as an input into the Climate Change Vulnerability Map.

For each hazard a scale of 1-10 has been used, with 10 being highest hazard and 1 the lowest. For each cell on the map, the average score across the five hazards was then taken as an indication of the overall hazard. I will return to this below. The outcome of this analysis is termed the "Multiple Climate Hazard Index". The resultant map is shown in Fig. 2 below.

Fig 2: The Multiple Climate Hazard Map.

This map has then been combined with data for population density (incorporating areas that are ecologically protected) and "adaptive capability" (defined as "the degree to which adjustments in practices, processes, or structures can moderate or offset potential damage or take advantage of opportunities (from climate change)." The latter has used expert judgement to create an index based upon data on education, poverty, income inequality and suchlike. These three indicators have then been averaged to determine the level of vulnerability to climate change (Fig. 3).

Fig 3: The Climate Change Vulnerability Map.

Hmmmm! First, let me say that this is a brave thing to do - such exercises are really challenging given the complexity of the dataset. Such exercises are important and useful given the need to prioritise. I am hesitant to be too critical. I would however like to point out four things that are worth thinking about.

1. Perhaps most importantly, I don't see how this is a map of climate change vulnerability. An argument can be made that this is a map of vulnerability to meteorologically driven hazards, but most of the parameters do not appear to consider a changing climate. The data for floods, droughts and cyclones used historic data of occurrence. This does not consider change. The only parameter that considered climate change was the sea level inundation, but this used a terribly simplistic model (a binary switch at 5 m sea level rise).
2. The decision to average across the five hazards is strange. The problem can be illustrated with an extreme example. Imagine you live on a flat, tropical plain 20 cm above sea level. Most of the hazards are likely to be low - no landslides, no river floods, no droughts, no tropical cyclones. However, a comparatively small rise in sea level wipes you out. In the system used here, your hazard comes out low whereas actually it is very high. It might be more rationale to take the highest value of hazard, or a more subtle measure.
3. The decision to weight the parameters equally is also interesting and surprising. Given the vastly different impact of the hazards, it might be worth weighting the hazards appropriately.
4. The decision to average the hazard, the population and adaptive capability is also odd. I would have thought that these parameters should be combined so that they interact (r.g. through multiplication and/or division). Clearly, the case where the level of hazard is high, the population is high and the adaptive capacity is low is exactly where really serious disasters occur.

I suspect that this map needs another iteration or two, perhaps backed up with a sensitivity analysis, but as a first step the authors deserve praise.

October 2008 fatal landslide map

Rather later than is ideal (for which I apologise), here are the maps for fatal landslides in October 2008.

First, the statistics:
Number of recorded fatal landslides: 22
Number of recorded fatalities: 179
The average for the period 2002-2007 inclusive was 398 fatalities, excluding the impact of the Kashmir earthquake in 2005, so 2008 is considerably below the long term average.

This is the map of the fatal landslides that I recorded (click on the map for a better view):

As the SW monsoon ceases the concentration in S. Asia is now absent, so as usual the distribution has become far less concentrated.

This is the map of all of the fatal landslides that have I recorded in 2008 up to the end of October (again click on the map for a better view):


As usual, your comments and corrections are welcome.

September 2008 fatal landslide map

For the first time in recorded history the monthly fatal landslide map is actually on time!

Below is the map for September 2008. First the statistics:
Number of recorded fatal landslides: 59
Number of recorded fatalities: 1290

September has been a quite bizarre month, with fatal landslides occurring at a level that is much higher than expected. As such this continues the trend for August, which was also above average, but it bucks the trend for the rest of the year. At present I am not sure of the reason for this change - that will require some considerable analysis. It is very interesting indeed.

Comparison with previous years is difficult. In 2004 there were catastrophic landslides in Haiti that killed three thousand people. If this is included then the average across the period 2002-2007 is 808 - high, but still much less than in 2008. Excluding that single event in Haiti yields an average of 202 people. Note that 550 of the fatalities in 2008 were also in Haiti due to mudflows created by two tropical cyclones. Other large events were the Manshiyet Nasser landslide in Cairo, Egypt (107 fatalities) and the Shanxi flowslide (270 deaths).

So, here is the map for September 2008. As ever, you will need to click on the map to get a decent view of it:


A few things to note:

1. The very strong cluster along the Himalayan Arc is once again very evident. Nepal in particular is suffering an intense period of landslides.
2. There is also a clear cluster in the Caribbean as the Hurricane season continues to have a high impact. Note that in Haiti the cluster is a multiple event.
   

The map of the year to date is shown below. Again, click on the map for a decent view. The normal clusters are now well developed. Compare this map with that from 2007 (see here). The similarity between the two patterns is now very clear.

As usual I welcome any comments, corrections and clarifications.

August 2008 fatal landslide map

Better late than never (just a tad busy at the moment...), here is the fatal landslide map for August 2008. First the statistics:
Number of recorded fatal landslides: 61
Number of recorded fatalities: 537

As such August was substantially above the average for the last five years (212 fatalities). This is interesting as regular readers will know that until the end of July most months have been substantially below the five year average. July was particularly low. Therefore, this sudden increase in activity in August is slightly surprising and will be worthy of detailed analysis in due course.

So, first the map for August 2008. As ever, you will need to click on the map to get a decent view of it:

A few things to note:

1. The very strong cluster along the Himalayan Arc is now very evident. Nepal in particular is suffering an intense period of landslides.
2. There is also a clear cluster in the Caribbean as the Hurricane season starts to take its toll.
3. China is also experiencing a few events, but with no clear pattern.

The map of the year to date is shown below. Again, click on the map for a decent view. The normal clusters are now well developed. Compare this map with that from 2007 (see here). The similarity between the two patterns is now very clear.

I welcome any comments, corrections and clarifications.

July 2008 fatal landslide map

The map below shows the provisional distribution of fatal landslides for June 2008. The (again provisional) statistics are:
Number of fatal landslides: 55
Number of fatalities: 2127

Once again, the number of recorded fatalities this month has been substantially below the average for 2002-2007, which is 504 deaths. This is consistent with the monthly totals for the whole of the year - indeed at the moment 2008 has the lowest number of fatalities that I have recorded for any year, excluding the Wenchuan (Sichuan) earthquake of course. I will in due course write a post about the likely causes of this very interesting pattern.

Figure 1 shows the distribution map for July. As ever, click on the map for a bigger version. Please acknowledge any use of it or Fig. 2.

Fig. 1: July 2008 fatal landslide locations (click on the map for a larger version)

The effect of the monsoon in South Asia is now being clearly felt, with a very marked line of landslides along the southern edge of the Himalayas. The usual cluster of landslides in Central China is also clear. Once again there is a higher incidence of landslides in Central and S. America than is normal.

Fig. 2 shows all of the recorded fatal landslides in 2008 to the end of July, excluding those associated with the Wenchuan (Sichuan) earthquake. The typical annual patterns are now becoming clear, with clusters in Central China, S. Asia along the southern edge of the Himalayas, the Philippines and Central / S. America. There are certainly more recorded fatal landslides in the latter area so far this year than has been the pattern over the last few years, but there are also rather fewer in the Himalayas. The western part of the Himalayan chain (e.g. Kashmir) in particular seems to have far fewer fatalities than has been the case for the last few years.

Fig. 2: Recorded fatal landslides in 2008 up to the end of July. Each dot represents a single landslide. Click on the map for a larger version in a new window.

Please leave a comment about this map and dataset! I would really like to know what you think about the data, the maps and the general approach.
Of course, if you know of an event that I might have missed I would like to hear from you...

June 2008 fatal landslide map

The map below shows the distribution of fatal landslides for June 2008. The statistics are:
Number of fatal landslides: 43
Number of fatalities: 215

As has been the pattern all year, June was substantially below the average number of fatalities for may in 2003-2007, which is 334 deaths. Interestingly, during June there have been extensive reports of intense and prolonged rainfall events, but this does not seem to have translated into fatal landslides. The reasons for this are unclear to me.

June 2008 fatal landslide locations (click for a larger version)

The patterns that were starting to emerge in the annual map over the last few months are now clear. In particular the clusters in the Himalayas, western Central America, the Philippines and Indonesia are now very obvious. Whereas last month the cluster in the Himalayas was focused on the western part of the mountain chain, it has now extended to the east as the monsoon rainfall has intensified. The pattern of more landslides in Central America than is normal is continuing. China is continuing to get a fair number of fatal landslides. It is also interesting to see the occurrence of several landslides in SE. Europe.

Recorded fatal landslides in 2008 up to the end of June. Each dot represents a single landslide. Click on the map for a larger version in a new window.

May 2008 Landslide map

The map below shows the distribution of fatal landslides for May 2008. At the moment I am not including the landslides associated with the Sichuan earthquake as I cannot yet estimate the impacts of them. Clearly the numbers of landslides and the numbers of fatalities will dwarf the numbers below. I will try to ascertain the magnitudes in due course, though this is very challenging. I have included reported rainfall-triggered landslides in the aftermath of the earthquake, though again this is likely to be an underestimate.

The statistics are:
Number of fatal landslides (excluding those triggered by the Sichuan earthquake): 21
Number of fatalities: 280

May was once again very substantially below the average number of fatalities for may in 2003-2007, which is 336 deaths . In my post on the April figures I said that May "could be an interesting month!" I was right, but for the wrong reasons. The distribution is again fairly well-scattered, but as we come into the wet season it is increasingly focussed on Asia. The red triangle represents Sichuan.

May 2008 fatal landslide locations (click for a larger version)

The patterns that were starting to emerge in the annual map last month are now strengthening. In particular the inevitable clusters in the Himalayas, western Central America, the Philippines and Indonesia are now clear. The cluster in the Himalayas is still focused on the west side at the moment as this area gets heavy rain in the early part of the year. The monsoon rains have now started in the very southern part of the Indian subcontinent and will now start moving northwards. This will drive the growth of this cluster. There are certainly more landslides in Central America than is normal.

2008 fatal landslide locations to the end of May (click for a larger version)

So what of the month ahead? We, expect to see the number of fatal landslides increase markedly, especially in Asia. As the rains start in Central China landslides seem inevitable. As the La Nina conditions fade, we might expect to see fewer landslides in South America, but more in Asia.

Landslide sizes (numbers killed) around the world

One of the interesting aspects of our landslide fatality database is that it is possible to produce maps of the size of the impact of landslides (which in this case means the numbers of people killed) in each event around the world. I tried to do that for my EGU presentation, but my colleague Nick Rosser has produced a rather more attractive version (Fig. 1) showing the 2007 dataset.

Fig 1: Map of the size of the impact of landslides, as measured by the number of fatalities in each event, for 2007. Click on the map to open a larger version.

The variation in size is quite interesting. Note that in Europe, N. America and Japan the recorded landslides tend to be comparatively small, almost always resulting in less than five fatalities. South Asia and Central America on the other hand have a scatter of landslide sizes from small events through to the very large. China also has a high frequency of large landslides, but appears to have comparatively few small ones. Africa has surprisingly few recorded fatal landslides, but those that do occur tend to be quite large.

So what to make of this? Well, first and foremost it shows the different levels of mitigation of landslides around the world. In the more developed world (i.e. Europe and N. America), large landslides tend to be very well mitigated (i.e. engineering measures have been used to stabilise them). The few fatal landslides that do occur tend to be small and in fairly remote locations, resulting in low fatality counts. On the other hand, in South and South-East Asia, mitigation is much less likely and there is a high density of people in affected areas, meaning that fatality counts are much higher. There might also be an effect of the both the tectonic setting (i.e. the high mountains of S. Asia in particular might mean that the fatal landslides are larger) and the climate (i.e. torrential monsoon rainfall might tend to trigger very energetic landslides, which are more likely to kill, perhaps). Finally, the data also perhaps show that the current database still misses smaller landslide events in China and in particular in Africa. We are trying to improve this situation.

As ever your comments are really welcome.

April 2008 landslide map

The map below shows the distribution of fatal landslides for April 2008.

The statistics are:
Number of fatal landslides: 12
Number of fatalities: 33

As I predicted in March, April was once again very substantially below the average for 2003-2007, which is 96 fatalities per year. The distribution is very random, with absolutely no clear pattern. Two fatal landslides occurred in the Himalayan Arc, which is a sign of things to come over the next few months. La Nina conditions do still prevail, but the event continues to weaken. Typically the number of landslides starts to increase around early May. This is shown by the cumulative totals - for example, at the end of April 2004 I had recorded just 339 fatalities for the year to date (this years total is even lower - 301), but by the end of May 2004 this had jumped to 1136. It could be an interesting month!

April 2008 fatal landslide locations (click for a larger version)

Some patterns are starting to emerge in the annual map. Note in particular the emerging clusters in the Himalayas, western Central America, the Philippines and Indonesia. These are the normal locations for fatal landslide clusters. The cluster in the Himalayas is focused on the west side at the moment as this area gets heavy rain in the early part of the year. It will spread eastwards over the next three months.

2008 fatal landslide locations up to the end of April (click for a larger version)

March 2008 landslide map

The map below shows the distribution of fatal landslides for March 2008.

The statistics are:
Number of fatal landslides: 17
Number of fatalities: 77

Thus February was well below the average for 2003-2007, which is 129 fatalities per annum. The distribution is definitely atypical, with more than usual for this time of year in S. America but less in Asia. This would seem to reflect the current La Nina conditions, which have also led to a globally cool spell (as an aside this has got the global warming denier community amusingly over-excited). La Nina conditions do still prevail, but the event is now weakening. Thus, we might sell see below average numbers of landslides for April (but above April for S. America again perhaps).

March 2008 fatal landslide locations (click for a larger version)

2008 fatal landslide locations up to the end of March (click for a larger version)