资源环境科技发展态势分析平台

The representativeness of surface solar radiation (SSR) point observations is an important issue when using them in combination with gridded data. We conduct a comprehensive near-global (50 degrees Sto55 degrees N) analysis on the representativeness of SSR point observations on the monthly mean time scale. Thereto, we apply the existing concepts of decorrelation lengths (), spatial sampling biases (), and spatial sampling errors (epsilon) to three high-resolution gridded monthly mean SSR data sets (CLARA, SARAH-P, and SARAH-E) provided by the Satellite Application Facility on Climate Monitoring. While quantifies the area for which a point observation is representative, and epsilon are uncertainty estimates with respect to the 1-degree reference grid (G). For this grid we find a near-global average (G)=3.4 degrees, (G)=1.4W/m(2), and epsilon(G)=7.6W/m(2) with substantial regional differences. Disregarding tropical, mountainous, and some coastal regions, monthly SSR point observations can largely be considered representative of a 1-degree grid. Since epsilon is an uncorrectable error the total uncertainty when combining point with 1-degree gridded data is roughly 40% higher than the uncertainty of station-based SSR measurements alone if a rigorous bias correction is applied. Cloud cover and terrain data can at maximum explain 50% of the patterns of the representativeness metrics. We apply our methodology to the stations of the Baseline Surface Radiation Network. Overall, this study shows that representativeness is strongly dependent on local conditions and that all three metrics (, , and epsilon) must be considered for a comprehensive assessment of representativeness.

Plain Language Summary Station-based observations are the most robust way to measure the amount of solar radiation reaching the Earth's surface (surface solar radiation [SSR]). These measurements are point observations yet are often used in combination with satellite or model data (e.g., for validation or for derivation of other quantities). The latter have a typical spatial extent up to a few hundred kilometers. Thus, it is important that the point observations are representative of this large area. In this paper, we show that the representativeness of SSR point observations varies regionally and is strongly dependent on local conditions. We find that some regions are problematic in terms of representativeness. However, in most regions SSR point observations can be considered representative for a 1-degree grid. We quantified additional uncertainties, which have to be taken into account when combining point and gridded data.