Using Atmospheric Fronts to Evaluate the Representation of Precipitation in CMIP6 Models

Abstract:

Process-based evaluation of precipitation is key to understanding climate model biases. It is vital to ensure that precipitation is produced in the model due to the correct mechanisms (or weather system). Atmospheric fronts have been shown to be responsible for a large proportion of total and extreme precipitation in the mid-latitudes. Therefore, representation of precipitation associated with fronts in climate models needs to be tested.

We applied objective front identification to the historical simulations from the CMIP6 archive and linked them with their 6-hourly precipitation accumulations. We compared the model outputs to the results from observationally constrained datasets. The fronts were identified from ERA5 and linked to precipitation estimates from a number of sources including ERA5, CMORPH, and GPCP. The precipitation errors have been decomposed into components associated with the frequency and intensity of frontal and non-frontal precipitation. Models consistently underestimate the intensity of frontal precipitation and overestimate the frequency of frontal precipitation. A frontal amplification factor has been calculated, which is the ratio of frontal to non-frontal precipitation intensity, indicating that the non-frontal precipitation intensity is underestimated even more than the frontal precipitation.

The diagnostics from this analysis have been made into metrics which could be used to benchmark and evaluate model performance and aid in focussing future model development.