How not to present data on natural disasters

As part of the course on energy and climate change that I’m working through on Coursera we were referred to a site listing a summary of natural disasters. Knowing the number, type and frequency of various types of natural disasters is very useful information especially when thinking about what impact climate change may have on their frequency. So making the data readily available is a good thing.

We were referred to a site called PreventionWeb that summarizes data from The International Disaster Database. My issues are not so much with the underlying database but with rather with how PerventionWeb presents their summaries of the data. Those summaries can be found from their Disaster Data & Statistics webpage.

I looked at the disaster profile for Australia. But all the profiles seem to follow the same format. The profile pages provide a series of tables and graphics that force the user to make comparisons between the various types of disasters that simply should not be made.

The underlying problem is in how a particular disaster makes it to the the underlying database. The criteria for inclusion are rather simple. The disaster makes the list if ten or more people are killed, 100 or more people are affected, a state of emergency is declared, or a call for international assistance is made. There is no economic criterion for making the list.

Clearly the last two criterion will vary in their application by country so comparisons between countries are somewhat problematic. It also creates issues when creating summaries for the world as a whole or for regions of the world.

But the bigger problem comes in comparing types of disasters. Just about any drought will affect more than 100 people. A storm usually will not until it reaches a certain severity. I’m not sure what the cutoff is for the extreme temperature disaster but suspect that once that temperature criterion is met then it must make the list. These problems mean that the table comparing the number, cost, and impact of the various types of disasters are mostly meaningless.

As if that were not enough there are clear problems in data collection and comparability. In Australia there were two Earthquake events. They had an average economic impact of a nice round 500,000.00 – I believe this is in US dollars although the table is not labeled as such. But there were 20 wildfire with an economic impact of a very precise 124,642.20. Does anyone believe that the average is known to the nearest ten cents? This is clearly false precision. I only ask that they round the numbers to the accuracy they know they have in the data. Here we can surely say that he estimate the two earthquake events was not very precise. So how can I compare that number to what appears to be a much more precise number for wildfires?

I walk away with the feeling that this is data not very well presented.