The complementary relationship (CR) approach aids evapotranspiration estimation in the data scarce region of Tibetan Plateau: symmetric and asymmetric perspectives

Abstract:

While the land surface latent and sensible heat release in the Tibetan Plateau (TP) could greatly influence the Asian monsoon circulation, the actual evapotranspiration (ET_a) information in the TP has been largely hindered by its extremely sparse ground observation network. Thus the complementary relationship (CR) theory lends great potential in estimating the ET_a since it relies on solely routine meteorological observations. With the in-situ energy/water flux observation over the highest semiarid alpine steppe in the TP, the modifications of specific components within the CR were first implemented. We found that the symmetry of the CR could be achieved for dry regions of TP when (i) the Priestley-Taylor coefficient, (ii) the slope of the saturation vapor pressure curve and (iii) the wind function were locally calibrated by using the ET_a observations in wet days, an estimate of the wet surface temperature and the Monin-Obukhov Similarity (MOS) theory, respectively. In this way, the error of the simulated ET_a by the symmetric AA model could be decreased to a large extent. Besides, the asymmetric CR was confirmed in TP when the D20 above-ground and/or E601B sunken pan evaporation (E_pan) were used as a proxy of the ET_p. Thus daily ET_a could also be estimated by coupling D20 above-ground and/or E601B sunken pans through CR. Additionally, to overcome the modification of the specific components in the CR, we also evaluated the Nonlinear-CR model and the Morton’s CRAE model. The former does not need the pre-determination of the asymmetry of CR, while the latter does not require the wind speed data as input. We found that both models are also able to simulate the daily ET_a well provided their parameter values have been locally calibrated. The sensitivity analysis shows that, if the measured ET_a data are absence to calibrate the models’ parameter values, the Nonlinear-CR model may be a particularly good way for estimating ET_abecause of its mild sensitivity to the parameter values making possible to employ published parameter values derived under similar climatic and land cover conditions. The CRAE model should also be highlighted in the TP since the special topography make the wind speed data suffer large uncertainties when the advanced geo-statistical method was used to spatially interpolate the point-based meteorological records.