The surface shortwave flux has a strong latitudinal dependence with typical annual mean values of 250 W/m2 in the tropics. The annual mean surface net longwave flux ranges from -30 to -70 W/m2. Estimates of these terms are available from in situ climatologies, from atmospheric reanalyses, and, since the 1980s, from satellite observations. These data sets have many potential sources of error that include: uncertainty in the satellite retrieval algorithms and in situ formulae, cloud representation in reanalyses, sampling issues and changing satellite sensors (Gulev et al., 2010). As for the turbulent fluxes, the uncertainty of the annually averaged global ocean mean shortwave or longwave flux is difficult to determine and in the range 10–20%.
High accuracy in situ radiometer measurements are available at land sites since the 1960s (see Wild, 2009 Figure 1), allowing analysis of decadal variations in the surface shortwave flux. However, this is not the case over the oceans, where there are very few in situ measurements (the exception being moored buoy observations in the tropical band 15°S to 15°N since the 1990s, Pinker et al., 2009). Consequently, for global ocean shortwave analyses it is necessary to rely on satellite observations, which are less accurate (compared to in situ determination of radiative fluxes), restrict the period that can be considered to the mid-1980s onwards, but do provide homogeneous sampling. Detailed discussion of variations in global (land and ocean) averaged surface solar radiation is given in Section 2.3.3; confidence in variability of radiation averaged over the global ocean is low owing to the lack of direct observations.