Coupling uncertain patterns of climatic variables in estimating evaporation from open water bodies.

Since the variables involved in changing evaporation rate interact with each other, the result of first-order Sensitivity Analysis (SA1) is imperfect to reveal the dynamics of this phenomenon. The main goal of this study was to reveal coupling patterns of climatic agents in forming evaporation. The Partial Deviations (PaD) method, based on Back-Propagation Artificial Neural Network (BPNN), was used to reveal these patterns. While non-derivative methods survey only first-order sensitivity values, PaD can investigate higher-order sensitivity values. In this way, Summing Squares of partial Derivations (SSD) revealed magnitude of interactions. Also, gradient surfaces statistically reflected information on evaporation changes. To achieve better results, Wavelet-based denoising method was used to remove high-frequency component of inputs. The new method was tested at two neighboring sites (Ahvaz and Isfahan) in Iran. By feeding denoised input to BPNN, the uncertainty of first/ second order PaD values was reduced highly at Ahvaz station by 47.76 and 28.68%. At Ahvaz station coupling between one day-lagged evaporation with air temperature and humidity with magnitudes of 26.37 and 25.21%, respectively, had a major effect on the evaporation gradient. Similarly, the major effects on evaporation rate at Isfahan station belonged to coupling one day-lagged evaporation with air temperature and wind speed with magnitudes of 36.97 and 18.98%, respectively. The interaction patterns showed that rate of evaporation reversed for both stations in warm seasons because of an increase in atmospheric humidity. It seems cities near sea, despite having warm climate, show complex patterns of evaporation. Temperature, one-day-lagged evaporation, wind speed, and radiation aroused effect of other variables. The interaction of these variables caused an inverse rate of evaporation in some cases where the role of one day-lagged evaporation, as the ambient humidity memory, was more prominent than the rest. [ABSTRACT FROM AUTHOR]