Abstract ID: 3.12286 | Withdrawn | Talk | TBA | TBA

The exact nature of precipitation depends on processes spanning many scales, from cloud microphysics, through convection, to large-scale circulation. It is also influenced by the underlying terrain, where complex orography can interact with the precipitating airmass.

To model precipitation accurately, all these processes have to be included either explicitly, or using parameterizations. Thus, the representation of precipitation in various climate models differs greatly, depending on the model resolution and complexity. In traditional global and regional climate models, deep convection is parameterized, whereas the newest generation of km-scale climate models treats deep convection explicitly. The higher-resolution models also better resolve complex topography, which can add important, local scale detail. This improves the representation of various precipitation characteristics, especially on sub-daily timescales. However, even in km-scale models, some processes are either still not fully resolved, or can possibly be influenced by biases in other model variables, which leads to an imperfect representation of precipitation.

On the other hand, observational precipitation datasets come with their own shortcomings. Rain gauges are point measurements, representative of only small areas, and can suffer from severe undercatch. The resolution of many satellite-based products is insufficient for studying precipitation in complex terrain, and depending on the utilized wavelengths, these datasets struggle with light precipitation and/or precipitation over snow-covered regions. Radar measurements can suffer from clutter and beam blocking. Gridded observational datasets are based on sophisticated interpolation methods and expert knowledge. Reanalysis products are not true observations, and their quality also depends on the underlying models. The shortcomings of all these types of datasets are exacerbated in mountainous regions, where accurate precipitation observations are sparse and remote sensing is challenging.

As climate models represent precipitation ever more realistically, their evaluation poses a challenge: no single observational product can be taken as the ground truth. Here I share my experiences with evaluating high-resolution model data against various observations. I discuss what additional information about the observational datasets would be beneficial for model evaluation, and I would welcome feedback on how to evaluate models more meticulously.