Challenges in simulating extreme precipitation: observational uncertainties and influence of increasing spatial resolution in global climate models

Abstract:

Determining the capabilities of global climate models is crucial, notably for precipitation extremes which are particularly challenging to model. However, there is a lack of analyses assessing how precipitation extremes are simulated at the global scale and in a multi-model framework. Furthermore, a number observational datasets are available for precipitation, yet there is little consensus on which dataset to use at the global scale. This work tackles the challenges of the evaluation of extreme precipitation in global simulations and addresses important issues for a better understanding of future changes.

First, we assess the consistency of extreme precipitation across a variety of observations. We inter-compare the representation precipitation extremes over quasi-global land in more than 20 products from a variety of sources: in situ-based, satellite observations with or without a correction to rain gauges, and reanalyses. We find that extreme precipitation intensity is relatively similar between in situ-based and satellites estimates. Given the level of uncertainties associated with the estimation of extreme precipitation in the observations, none of the datasets can be thought as the best estimate. We recommend avoiding reanalyses but considering in situ and satellite data to gather an ensemble of products for a better estimation of observational uncertainties.

Second, we provide an evaluation of six global climate models on their representation of extreme precipitation. We also assess the influence of increasing spatial resolution on models skill by evaluating pairs of high- and low- forced atmospheric simulations. Models are compared to different observations to enable consideration of observational uncertainties. We find that despite large observational uncertainty, simulated precipitation extremes tend to be significantly different than observations, and in particular in the tropics. Our findings highlight good agreement between models that precipitation extremes are more intense at higher resolution. Interestingly, models generally often show lower skill in the high-resolution compared to low-resolution simulations. This suggests that increasing spatial resolution alone might not be sufficient to obtain a systematic improvement in the simulation of precipitation extremes.