UC San Diego

It is well known that the statistical methods ("parameterizations'') used to represent cloud processes for climate prediction distort the simulated diurnal rainfall cycle. In this study, the "superparameterization'' technique (embedding explicit models of convection) is investigated as a method of improving diurnal rainfall in order to study the mechanisms supporting it and to reduce uncertainties in climate prediction. Analysis of the effect of superparameterization on the simulated global diurnal rainfall cycle uncovers several unappreciated benefits of the technique: Diurnal rainfall becomes less well fit by a 24-hour sinusoid, more horizontally inhomogeneous within continents and oceans, and conforms better to transitions straddling coastlines. While these effects are robust to arbitrary configuration choices in the application of superparameterization, a favorable shift in the peak timing of continental rainfall is not. Regional deficiencies are documented associated with the Andean topography, and the Asian monsoon. Like conventional parameterizations that account for dilution, superparameterization favorably reduces the amplitude of the tropical land latent heating diurnal cycle. This removes an unrealistic deep barotropic subsidence wave that otherwise exerts undue remote control by land on the simulated diurnal cycle over the tropical ocean. The level of remote control of marine surface divergence is otherwise found to be undersimulated such that the tropical surface divergence signature of land-sea remote control does not validate well against scatterometer observations. Over the Gulf Stream, the pattern of warm season diurnal rainfall variability is captured in several simulations. Analysis suggests a giant sea breeze circulation fueled by convectively amplified cross-coastal radiative imbalances, may help explain the observed change in local rainfall timing with distance offshore. In the Central US warm season, the superparameterization approach is shown to capture packets of eastward traveling organized convection that underly a regional nocturnal convection maximum. The simulated rate and range of eastward travel validates against ground based radar data despite the fact that small scale storm propagation mechanisms (gust fronts and density currents) are artificially restricted to have only local effects under superparameterization. This is interpreted as evidence in support of an alternative "slow manifold'' view that these systems are primarily propagated by large-scale dynamics