Reference Evapotranspiration Modeling Using Artificial Intelligence-Based Approaches: A Bibliometric Analysis

Abstract

A bibliometric analysis was performed over the period 1992–2023, to pinpoint important trends, emphasis, and geographic distribution of international reference evapotranspiration (ETo) modeling research using intelligence-based approaches. Data was mined from the databases of the online version of Web of Science. The data was analyzed using the Excel program, and the bibliometric mapping was performed using the VOSviewer software. A total of 1148 articles met the required criteria. The findings indicated that the number of articles had increased rapidly over the past five years and that English was the prevalent language (99.5%). Researchers in 99 countries have published in this field of research. China ranked first with 356 articles (31.0%), followed by the United States of America with 193 articles (16.8%). Egypt and Saudi Arabia are two of the top 15 countries in the world. These articles were published in 289 journals; the Journal of Hydrology was the most productive journal (95 articles, 8.3%), followed by Computers and Electronics in Agriculture (67, 5.8%). The most productive author is Kisi Ozgur from Turkey (68 articles, 4.2%). Taking into account all the institutions working on ETo modeling (1555 institutions), the Chinese Academy of Science was ranked first (80 articles, 7.0%), followed by the Egyptian Knowledge Bank (55 articles, 4.8%). Artificial neural networks and machine learning were the most commonly used intelligence-based approaches for ETo modeling. Reinforcement learning was the least popular technique for ETo modeling, which could be due to its complexity and data requirements.