How much does the shallow cumulus regime contribute to the ECMWF model's shortwave radiation bias at ARM SGP?

Poster PDF

Description

Routine global evaluation using CERES observations suggests that the ECMWF model lacks shortwave cloud forcing over land. This appears to be consistent with an annual mean surface shortwave bias at the ARM SGP site over the period spanned by the CMBE record. Previous studies have suggested that lack of shallow cumulus cloud cover might contribute significantly to this bias, but results were based on relatively short time periods in the face of highly variable shallow cumulus cover from year to year. Also, the ECMWF data available from the ARM Archive is from the model cycle operational at the time of forecast, such that model cycles over a decade old are compared with more recent versions over the long record of observations at SGP. This study aims to systematically assess the impact of days dominated by shallow cumulus clouds on the surface shortwave radiation from the full available CMBE record. We use 146 days with non-precipitating shallow cumulus clouds between 1997 and 2009 to compare modeled and observed cloud amount and radiative impact. The current operational forecast model (CY36R4) is re-run for these days to ensure that the results apply to the model’s latest cloud and radiation schemes. The diurnal composite shortwave cloud forcing is reproduced quite well by the model. A small bias remains, but its contribution to the annual mean bias is small. A closer look at individual days of the time series reveals that the model fails to produce shallow clouds in about 13% of cases, but compensates for this deficiency by a stronger reduction of downwelling shortwave radiation on cloudy days.