Premium
Spatial Representativeness of Surface‐Measured Variations of Downward Solar Radiation
Author(s) -
Schwarz M.,
Folini D.,
Hakuba M. Z.,
Wild M.
Publication year - 2017
Publication title -
journal of geophysical research: atmospheres
Language(s) - English
Resource type - Journals
eISSN - 2169-8996
pISSN - 2169-897X
DOI - 10.1002/2017jd027261
Subject(s) - representativeness heuristic , series (stratigraphy) , grid , satellite , scale (ratio) , environmental science , sampling (signal processing) , meteorology , remote sensing , statistics , mathematics , geodesy , geography , geology , cartography , physics , optics , astronomy , detector , paleontology
When using time series of ground‐based surface solar radiation (SSR) measurements in combination with gridded data, the spatial and temporal representativeness of the point observations must be considered. We use SSR data from surface observations and high‐resolution (0.05°) satellite‐derived data to infer the spatiotemporal representativeness of observations for monthly and longer time scales in Europe. The correlation analysis shows that the squared correlation coefficients ( R 2 ) between SSR times series decrease linearly with increasing distance between the surface observations. For deseasonalized monthly mean time series, R 2 ranges from 0.85 for distances up to 25 km between the stations to 0.25 at distances of 500 km. A decorrelation length (i.e., the e ‐folding distance of R 2 ) on the order of 400 km (with spread of 100–600 km) was found. R 2 from correlations between point observations and colocated grid box area means determined from satellite data were found to be 0.80 for a 1° grid. To quantify the error which arises when using a point observation as a surrogate for the area mean SSR of larger surroundings, we calculated a spatial sampling error (SSE) for a 1° grid of 8 (3) W/m 2 for monthly (annual) time series. The SSE based on a 1° grid, therefore, is of the same magnitude as the measurement uncertainty. The analysis generally reveals that monthly mean (or longer temporally aggregated) point observations of SSR capture the larger‐scale variability well. This finding shows that comparing time series of SSR measurements with gridded data is feasible for those time scales.