Improving the representation of stomatal responses to CO2 within the Penman-Monteith model to better estimate evapotranspiration responses to climate change

摘要

An increase in atmospheric CO2 concentration (CO2) affects stomatal conductance (g(s)). Since g(s) is an important parameter in the Penman-Monteith model (P-M), an accurate estimate of g, can improve the accuracy of the estimation of evapotranspiration (ET) by P-M when CO2 increases. This paper identifies three models of the relationship between g(s) and CO2 (the linear model, L, the hyperbolic model, H, and the modified hyperbolic model, MH), and compares their estimations of ET when they are incorporated into P-M. A two-year experiment which included four levels (400, 550, 700, and 900 mu mol mol(-1)) and three levels (400, 550, and 700 mu mol mol(-1)) of CO2 in 2015 and 2016, respectively, was conducted in order to obtain validation data. Fifty papers that had studied the effect of increased CO2 on g(s), over a 45-year period, were analyzed to obtain calibration data. After being calibrated and validated, each of the three models was in turn included as a component of P-M to estimate ET under increased CO2, followed by the comparison between estimated and observed ET. The sensitivity of P-M to the parameters in three models was also analyzed. Our results show that g, decreased and the reduction rate gradually lessened as CO2 increased. Of the three models, MH gave the best estimate of g(s), having the largest values of the coefficient of determination (R-2), 0.97, the Nash-Sutcliffe efficiency coefficient (NSE), 0.96, and the modified affinity index (d), 0.93, and the smallest values of the root mean square error (RMSE), 0.03, and the Akaike information criterion (AIC), - 47.48. Each of the three models gave different ET results for different levels of CO2 when used as a component of P-M. The MH had better ET estimations than L and H when compared to observed ET across all CO2 concentration levels, with the highest R-2, NSE, and d, and the lowest RMSE and MC. The sensitivity of P-M to the empirical parameters varied among the three models, for instance, +/- 40 an