Temporal scaling dependency in precipitation simulated by the HCLIM regional climate model

Rainfall intensity-duration-frequency (IDF) curves play an important role in water management, for instance urban stormwater handling. They are often derived by fitting generalized extreme value distributions to observed annual maximum rainfall values and require long-term observations of rainfall at sub-daily scales to get robust estimates. However, mathematical models describing the IDF curves have been observed to have a fractal dimension, providing a possibility to avoid the challenging data requirement. In this case, we used a simple analytical formulation, where two key parameters specify the shape of the curves: the wet-spell mean precipitation and frequency. These two parameters also play a role for the fractal dimension of the temporal rainfall scales. 

This study aimed to take a step towards a better understanding of which conditions may play a role for the fractal shapes and how they are affected by geographical conditions. To this end, we have explored the dependency between wet-spell mean precipitation and frequency over different timescales, using data from convective-permitting (3km) climate simulations with the regional climate model HCLIM over northern Europe. The analysis was applied to HCLIM results with boundary conditions from the ERA Interim reanalyses, as well as the EC-Earth and GFDL-CM3 global climate models.

The simulated sub-daily fractal dimensions were influenced by geographical conditions for both the wet-spell mean precipitation and the wet-spell frequency. The results were consistent for the different boundary conditions representing current climate conditions (reanalysis and GCMs) and hence independent of the driving model. Similar simulations for the future, following the RCP8.5 scenario, hinted at changes in the fractal shapes at high latitudes. However, for large regions, there was little indication of change. This study is an example of using high-resolution regional climate model data for testing hypotheses where consistent, dense and temporally high-resolved long-term observational networks are lacking.