Disaster Impact Before Ground Assessments

A couple of months ago I read a blog post by Andrej Verity on the subject of mixing pre and post disaster data in an attempt to model the impact of an event. When Typhoon Haiyan swept across the Philippines in November 2013 a data set was made that combined census data with measures of wind speed and storm surge to predict which areas would be most heavily affected.

Typhoon Haiyan, November 2014 — Credit Bruce Reyes-Chow

The concept appealed to me for several reasons. Responders in sudden onset disasters usually have to rely on qualitative information to guide their initial response before a systematic assessment is carried out. This can lead to over and under serving of places depending on how much exposure an area receives in both the humanitarian community and mass media. A case in point being Tacloban having a swath of aid agencies in the aftermath of Typhoon Haiyan at the expense of other affected places. An index could help to highlight those differences with an aim of distributing aid efforts more efficiently. As exemplified recently, with Cyclone Pam and Typhoon Maysak, assessments over the full extent of an affected area can take a while to come into fruition (over 3 weeks) for very practical reasons. The index certainly has a position to act as a stopgap before these are available. I have seen responders attempt to create this informally with maps of cyclones paths and populations trying to estimate damage in an ad hoc fashion. I hope a more rigorous approach will be a natural progression for people to adopt.

With Maysak and Pam I had time to trial compiling indexes for both. The results can be viewed on CartoDB here and here. The data and methodology have been made available on HDX.

Priority Index for Cyclone Pam, March 2015

The models were built using a combination of population, poverty levels and wind speed. From working as an information manager for shelter cluster during the Haiyan response I observed there was a big focus on not just affected population totals, but also on the areas with high poverty as they were the least likely to self recover.

I think the results generated could be useful, but currently the models are made on intuition. A useful starting place, but definitely something that can be improved upon.

The next steps should be to revisit these; to compare against formal ground assessments when available, fit the models mathematically to determine which parameters matter, to see if it is possible to develop consistency of parameters between countries and if priority indexes should be pursued at all.