资源环境科技发展态势分析平台

A newly reprocessed, bias-corrected version of hourly satellite observations that provides global coverage of precipitation at high space-time resolution is evaluated and compared with climate model simulations. Irregular subdaily fluctuations are the dominant component around the world, greater than variance of daily mean precipitation, and much greater than variance associated with the mean diurnal cycle of precipitation. Irregular subdaily fluctuations of precipitation are severely underestimated by models, even after taking into account the observational error bars implied by different space-time resolutions. Variance of daily mean precipitation is less severely underestimated. Although mean diurnal cycle amplitudes vary among the models, this component is but a small part of total precipitation variance. Therefore, the total precipitation variance is significantly underestimated by models in general. Further exploration of model-data discrepancies in precipitation at high-time frequency may lead to new and useful climate model diagnostics.

Plain Language Summary Satellite-estimated rainfall rates at high spatial resolution and high-time frequency can be compared with output from modern climate models. Here we examine the statistical variability of observed and model-simulated rainfall. Total variability divides naturally into three separate components, arising from (a) intermittent rainfall fluctuations at time scales less than a day, (b) variations of daily totals of rainfall, and (c) the average day-to-night cycle of rainfall. Observations show that in reality the first component dominates the total. Climate models, however, systematically underestimate the first component (though their errors in the other two are not as severe). Diagnosing the sources of these errors may provide a new way to improve climate model accuracy.