Detection of global energy budget trends using satellite and surface sites: Is the current surface site distribution sufficient?

Several recent studies (e.g., Liepert 2002, Wild et al. 2005) have focused on trends in the downwelling solar flux measured at surface stations at various locations. While it is clear that surface sites may detect local trends, a related question is whether the current distribution of surface sites is sufficient to allow detection of global radiative trends. In this study, we use satellite retrieved surface fluxes to address this question.

Two sets of surface sites are chosen for assessment. The first is the subset of sites represented in the Global Energy Balance Archive (GEBA) that have nearly continuous records from July 1983 until June 2003. The other set consists of the Baseline Surface Radiation Network (BSRN) sites. For comparison, we employ satellite retrieved shortwave fluxes from the GEWEX Surface Radiation Budget (GEWEX-SRB) version 2.5 data set averaged to 2.5° resolution. We first compare the surface measurements to the fluxes retrieved for the corresponding satellite product grid boxes to establish that the satellite values are a good proxy for the measurements. We then use the satellite data to evaluate the representativeness of the surface site locations.

A time series of global mean shortwave downward flux at the surface is created from the satellite data for the 20-year time period from July 1983 until June 2003. Additional time series for the same period are produced by averaging together the fluxes retrieved over the grid boxes in which the GEBA and BSRN sites are located. The time series are deseasonalized and linear trends are computed using least mean square estimation.

The global mean flux from the GEWEX-SRB v. 2.5 shows a slight positive trend (about 0.4 W m^-2 decade^-1) over the 20-year period. The SRB values averaged over the GEBA and BSRN site locations also show a positive trend, with values of about 0.5 and 0.7 W m^-2 decade^-1, respectively. However, because of the noise level in these time series, even the sign of these trends is indeterminate at the 95% confidence level for all but the SRB global time series. In addition, if satellite flux sampling is limited to those times and locations for which valid data is actually available in the GEBA archive over the 20-year period, the computed trend becomes -1.6 W m^-2 decade^-1. Further implications of measurement site locations, length of record, missing data points, and uncertainty estimates will be discussed based on additional trend examples.