|
چکیده
|
This paper presents a detailed study of evapotranspiration estimation over the period of 1974–2017 in the Lake Urmia basin, Iran. Traditional empirical models like Penman-Monteith, Hargreaves-Samani, Priestley-Taylor, and Thornthwaite are compared in this study with some advanced soft computing techniques, including artificial neural networks, adaptive neuro-fuzzy inference systems, and support vector regression. The results show that the soft computing techniques always perform better, particularly ANN, in comparison to traditional models for both accuracy and adaptability over various climatic conditions. Among the traditional approaches, the Penman-Monteith model performs best. It also introduces a new ET model dimensionally consistent, developed through dimensional analysis, which works comparably to the Penman-Monteith model. This long-term trend analysis reveals a highly significant annual increase in ET of about 5.2 mm y 1, with a change point detected in the year 1995. The study further discusses the effect of land use changes on ET patterns, showing remarkable increases in agricultural and urban areas of about 23.7 % and 156.3 %, respectively, over the study period. Sensitivity analyses, in fact, show that accurate ET estimation is very important where temperature and solar radiation measurements are concerned. In this respect, different statistical techniques like wavelet analysis and principal component analysis will be used to create nuanced insight into ET dynamics within the Lake Urmia basin. Moreover, the paper investigates models’ performance for differing climatic conditions and their ability to capture extreme ET events. In this respect, the comprehensive approach to and the intercomparison of ET processes in semi-arid regions presented in this study are very useful.
|