How often does the precipitation over 10 minutes exceed 50 mm somewhere on a given network of roads?

For long-term planning of the highway infrastructure, engineers in the Ministry of Infrastructure and Water Management of the Netherlands are considering the trade-offs between the risk posed by extreme precipitation in a changing climate and the cost of measures to reduce this risk for the entire network of highways and its critical elements, such as tunnels. This leads them to questions such as "How often does the precipitation over 10 minutes exceed 50 mm somewhere on a given network of roads?"

Naturally, this frequency is higher than the frequency of exceedance of the same depth at a site; it depends on the size and shape of the domain and on the spatial dependence of extreme precipitation. 

In the present study, statistics describing the spatial dependence of extreme precipitation are estimated from 11 years of gauge-adjusted radar precipitation data collected over the Netherlands.  At each radar pixel, annual maxima of precipitation depth are computed for durations ranging from 15 min to 12h. From these maxima, the values of the extremal coefficient function (ECF) for selected spatial domains are estimated.

From these values, a simple model is derived for converting return values of precipitation depth at a single site to return values of the highest precipitation depth within an arbitrary spatial domain, for durations from 10 min to 12 h. The model describes the duration-dependent statistics of the parameterized footprints of heavy precipitation events.

Confidence intervals are predicted using bootstrapping. The model is checked for fitness for its application to the design and maintenance of the drainage of highways, and the scope for further improvement is discussed.