Abstract
The authors describe a tropical cyclone risk model for the Philippines,
using methods that are open-source and can be straightforwardly
generalized to other countries. Wind fields derived from historical
observations, as well as those from an environmentally forced tropical
cyclone hazard model (using environmental forcing from the recent
historical period) are combined with data representing exposed value and
vulnerability to determine asset losses. Exposed value is represented by
the LitPop dataset, which assumes total asset value is distributed
across a country following population density and nightlights data.
Vulnerability is assumed to follow a functional form previously proposed
by Emanuel, with free parameters chosen by a sensitivity analysis in
which simulated and historical reported damages are compared for
different parameter values. Use of different vulnerability parameters
for the region around Manila yields much better agreement between
simulated and actually reported losses than does a single set of
parameters for the entire country. Even then, however, the model
predicts no losses for a substantial number of historical storms which
did in fact produce them, a difference the authors hypothesize is at
least in part due to the use of wind speed as the sole metric of TC
hazard, omitting explicit representation of flooding due to storm surge
and/or rainfall.