Longwave radiation budget parameters computed from ISCCP and HIRS2/MSU products

Despite the large discrepancies between the International Satellite Cloud Climatology Project (ISCCP) retrieved cloud fields and the high-resolution infrared radiation sounder 2 (HIRS2) retrieved cloud fields, the outgoing longwave radiation (OLR) computed based on these two cloud fields are quite similar due to the cancelation effect of cloud top altitudes and cloud amount on OLR. The accuracy of temperatures and humidities on the computation of OLR is as important as the clouds. For example, the large regional discrepancies of the order of 30 W m−2 found over northern Africa are largely due to surface temperature differences. Most extensive discrepancies over the ocean are due to humidity differences. Harmonic analysis has been applied to OLR, clear sky OLR, and cloud radiative forcing at the top of the atmosphere for July 1983 using the ISCCP products. The results show that the diurnal variations of these parameters from 60°S to 60°N are larger over land than over the ocean. The clear sky OLR peaks around 1500 LST, usually over land areas, and is associated with the surface/air temperature maximum. The OLR over cloudy regions shows a similar diurnal phase as the cloud top pressure. The cloud radiative forcing at the top of the atmosphere has maximum value near 2100 LST over the U.S. Rocky Mountain areas and near 0000 and 0300 LST over equatorial Africa and southern portions of North Africa. The ISCCP-computed longwave cloud radiative forcing has smaller diurnal variations over the ocean than the HIRS2-computed longwave cloud radiative forcing. The difference between the twice-a-day monthly mean and the 8-times-a-day monthly mean is small when using the ISCCP cloud products. The global mean bias of OLR from 60°S to 60°N is around 0.74 W m−2 with a standard deviation of 1.57 W m−2. Locally, the bias can be as large as 5 to 10 W m−2. This appears to indicate that the twice-a-day measurements of the HIRS2/MSU (around 0315 and 1515 LST, which are the equatorial cross times at nadir track) can be used to derive monthly mean OLR without significant bias.