Reconstructing daily evapotranspiration data from multi-annual water budget using insolation and precipitation presence

This paper presents a simple method for reconstructing daily catchment-wide evapotranspiration (ET) where only two kinds of daily data are available, i.e., catchment-wide precipitation and streamflow at the outlet. Here, multi-annual ET of a catchment obtained from water budget is decomposed into daily values considering two daily factors of insolation at the top of atmosphere (r factor) and precipitation presence (h factor). Precipitation presence is adopted as a surrogate of humidity which is related to two effects of (1) atmospheric transmittance of insolation and (2) moisture diffusion near surface. It is hypothesized that daily humidity condition is described with precipitation presence of the day, the previous day, and the following day. On the basis of this idea, eight possible scenarios are proposed for humidity condition. This paper reports h values analyzed from long-term data of observed pan evaporation and precipitation for five stations in Korea. h values show significant difference between eight scenarios, implying that the proposed scheme is effective in distinguishing various atmospheric conditions in a simple manner. There exists a varying degree of seasonal and spatial variability in h. Such variability is geo-physiographically explained through analysis of other meteorological data such as relative humidity, cloud cover, sunshine duration, surface solar radiation, and wind speed. The proposed methodology is applied for reconstructing daily ET of a real catchment in Korea. Calculated daily ET values well agree with observed pan evaporation data. The proposed method is also well compared with the Priestley–Taylor method, where the latter requires daily air temperature. The proposed method successfully captures sudden daily ET fluctuations in accordance with precipitation presence which is hardly captured by the Priestley–Taylor method. Further, the reconstructed data nicely follow the trend of LandFlux-EVAL dataset.