Climix—a flexible suite for the calculation of climate indices

Climate indices have long been an important tool for the evaluation of climate impacts and are commonly used both in research and practical risk assessment. The term generally refers to advanced statistics derived from daily data, such as the longest spell of consecutive dry days in a year, or the maximal precipitation in a single day in a given month. Initially, these indices were calculated for station data, but with the advent and increased utility of global and regional climate models, they are now commonly calculated for gridded data. The increased spatial resolution, together with the increased length of simulations to hundreds of years, the increased use of ever-larger ensembles of climate simulations, and the interest in a wider selection of possible future climate scenarios renders some established, serial algorithms ineffective. This is compounded by the fact that modern computing architectures derive their growing power no longer from the speedup of single computing units, but rather from the integration of larger numbers of parallel computing units. To fully utilize this potential, it is not enough to implement a straightforward parallelization of existing algorithms. Rather, we need to rethink the computing task from the start in a parallel framework.

Here, we present parallel algorithms implemented in the Python framework Climix, that have proven useful in the calculation of climate indices for a large ensemble of climate simulations that provide the basis for the user-oriented climate service of the Swedish Meteorological and Hydrological Institute.

Climix is available as open-source software and allows the calculation of a large number of climate indices both from a command-line interface with good support for common HPC schedulers such as SLURM and via a flexible Python API.