We present an open-source Python-based framework that can be used for objective evaluation of Atmospheric Rivers (ARs) in climate models. With a comprehensive pool of pre-defined metrics, this package is useful for systematic AR evaluation across a range of model simulations, such as the MIPs in CMIP5/6 models and large ensemble experiments. The required inputs for this software package are merely the integrated vapor transport, the climate variable of interest (e.g., precipitation, wind), and the user choice of AR detector (ARDT), which makes it easy and straightforward to implement for the AR researchers.

The initial version of the metrics include mean bias and ratio of standard deviation for AR characters (size, length, width, centroid latitude and longitude, landfalling frequency), and temporal correlation and RMSE for AR associated precipitation variability and extremes. These metrics are calculated for a list of user defined ocean basins and landfalling regions over the globe. With TempestExtremes ARDT, we’ve tested these metrics in historical simulations from 5 CMIP5 models and 3 CMIP 6 models, compared with the ERA5 reanalysis as the reference data. The results are presented in the form of profile plots, as a standard diagnostic tool in this package, which is intuitive to observe and compare.

We are in the process of including global metrics and expanding regional metrics for more AR characters, such as landfalling duration, seasonal peaks, consistency of seasonal cycle. With this tool we are also able to address specific scientific questions that require analyzing large amounts of data , e.g., how sensitive are the metrics to the choice of ARDTs? How IVT weighted metrics differ from non-IVT weighted metrics? Is the AR variability dominated by dynamics or thermodynamics? How sensitive are the ARs to various modes of climate variability? As such, this metric package will be a valuable addition to the AR research community.