Open AccessEvaluation of several soft computing methods in monthly evapotranspiration modelling
Author(s) -
Gavili Siavash,
Sanikhani Hadi,
Kisi Ozgur,
Mahmoudi Mohammad Hasan
Publication year - 2018
Publication title -
meteorological applications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.672
H-Index - 59
eISSN - 1469-8080
pISSN - 1350-4827
DOI - 10.1002/met.1676
Subject(s) - soft computing , evapotranspiration , adaptive neuro fuzzy inference system , arid , hydrology (agriculture) , mean squared error , empirical modelling , irrigation , irrigation management , computer science , environmental science , water resource management , statistics , mathematics , artificial neural network , fuzzy logic , simulation , machine learning , geology , ecology , artificial intelligence , biology , paleontology , geotechnical engineering , fuzzy control system
Evapotranspiration assessment is one of the most substantial issues in hydrology. The methods used in modelling reference evapotranspiration ( ET 0 ) consist of empirical equations or complex methods based on physical processes. In arid and semi‐arid climates, determining the amount of evapotranspiration has a major role in the design of irrigation systems, irrigation network management, planning and management of water resources and water management issues in the agricultural sector. This paper presents a case study of five meteorological stations located in Kurdistan province in the west of Iran. The ability of three different soft computing methods, an a rtificial n eural n etwork ( ANN ), an a daptive n euro‐ f uzzy i nference s ystem ( ANFIS ) and g ene e xpression p rogramming ( GEP ), were compared for modelling ET 0 in this study. The FAO 56 Penman−Monteith model was considered as a reference model and soft computing models were compared using the Priestley−Taylor, Hargreaves, Hargreaves−Samani, Makkink and Makkink−Hansen empirical methods, with respect to the determination co‐efficient, the root mean square error, the mean absolute error and the Nash–Sutcliffe model efficiency co‐efficient. Soft computing models were superior to the empirical methods in modelling ET 0 . Among the soft computing methods, the ANN was found to be better than the ANFIS and GEP .