On the effects of the temporal and spatial sampling of radiation fields on the ECMWF forecasts and analyses

The author's explore the implications of the temporal and spatial sampling of the radiation fields and tendencies upon the fields produced by the ECMWF system in operational-type forecasts, four-month seasonal integrations, and analyses. The model is shown to be much more sensitive to economics in the temporal than in the spatial description of the cloud-radiation interactions. In 10-day forecasts, the anomaly correlation of geopotential shows little sensitivity to a more complete representation of the cloud-radiation interactions, but temperature errors display a stronger dependence on the temporal representation. The difference increases with height, particularly in the tropical areas where interactions among convection, clouds, and radiation dominate. In pointwise comparisons over five days, the approximate temporal representation introduces only small differences in total cloudiness, surface temperature, surface radiation, and precipitation. In four-month seasonal simulations, the small errors seen in 10-day forecasts build up and a better temporal resolution of the radiation produces a colder stratosphere through cloud-radiation-convection interactions. The spatial sampling in the radiation computations appears beneficial to the operational model, inasmuch as, close to the surface, it smooths an otherwise wavy radiative forcing linked to the spectral representation of the surface pressure. The impact of the temporal/spatial sampling in the radiation calculations is usually much weaker in the analyses when and where observational data are available, but can be felt if the density of observations becomes smaller. On the contrary, the effect of the temporal/spatial interpolation is important on the sensitivity parameters derived from perpetual July simulations with perturbed SSTs.