The predicting potential of two different sensitivity coefficients in seven reference evapotranspiration models

The sensitivity of high data demanding reference evapotranspiration (ET 0 ) models to each climatic parameter can be affected temporally and spatially by the model type and the sensitivity coefficient chosen. Therefore, using different coefficients in order to comparatively study the sensitivity of various ET 0 methods can be helpful to choose the proper method especially in the lack of adequate climatic data accessibility. ET 0 sensitivity is usually studied in individual perturbation examinations, while collective perturbation case is more likely to happen in practice. In this study, the climatic parameters of 8 meteorological stations in Bushehr province, Iran, are used to test the predicting potential of two sensitivity coefficients (S and K) in individual and collective perturbation of input variables for seven ET 0 models. Values for the two sensitivity coefficients were obtained on daily, monthly, and annual scales. Ordinary Kriging interpolation method was used to investigate S and K spatial distributions. The predicting potential of each sensitivity coefficient in each of the ET 0 models was investigated by pairing the predicted values of diurnal ET 0 (ET 0 -Pred.) with those calculated by each model (ET 0 -Calc.). For the whole study area, S was the maximal for air temperature (S T , with an average value of 0.959), followed by sunshine hours (0.496), relative humidity, RH (− 0.236), and wind speed, U (0.184). Spatial variations of K T were not significantly different in most ET 0 models. The broadest range of variations was found in Jensen-Haise and Hargreaves-Samani models, in which K T of different stations were between 0.75 and 0.85 and 0.28 and 0.32, respectively. A similarity was found between air temperature (T) and S T and between sunshine hours (n) and S n temporal variations. However, such resemblance was not met between RH and S RH and between U and S U . Unlike the S sensitivity coefficient, daily variations of K for different variables were not similar to the daily variation pattern of each variable. FAO Penman-Monteith and Makkink were the two models showing best predictive behavior of both S and K sensitivity coefficients in all stations. However, predictive powers of S and K were not the same.