Reconstruction of a Daily Large-Pan Evaporation Dataset over China

Land surface evaporation is an important component of the earth's surface hydrological cycle, as well as in the atmospheric energy and water balances. In China, different instruments have been used over time to measure evaporation. A small pan with 20-cm diameter was used at most meteorological stations from 1951 to 2002. After 1998, large pans with areas of 3000 cm(2) each were adopted at most stations. Collocated small- and large-pan observations of evaporation have been made at major stations from 1998 to 2002. This makes developing a homogenous time series of evaporation measurements difficult. This paper describes how surface evaporation data collected in the past are merged with more recent data collected by automated weather stations. A statistical model was set up on the basis of collocated small- and large-pan observations of evaporation, using a second-generation statistical regression technique [partial least squares (PLS)]. The model succeeds in generating daily large-pan evaporation estimates from small-pan evaporation data and selected meteorological parameters. Interpolation is used to replace missing data. The reconstructed large-pan evaporation dataset covers 751 weather stations across China from 1951 to 2005. Cross-validation tests show that the relative error of the daily estimated evaporation derived from the PLS regression model is 19% with a monthly mean error rate approaching 0 mm. The spatial distribution and interannual-interdecadal trends of the reconstructed large-pan evaporation dataset and small-pan observations are similar.