Generalized gene expression programming models for estimating reference evapotranspiration through cross-station...

Area of Science:

• Hydrology and Agricultural Meteorology
• Soft Computing Applications in Environmental Science

Background:

• Local meteorological data scarcity hinders accurate crop water requirement assessment.
• Exogenous data and soft computing offer potential solutions for data-deficient regions.
• Reference evapotranspiration (ET0) is a critical parameter for irrigation management.

Purpose of the Study:

• To evaluate the generalizability of Gene Expression Programming (GEP) for estimating ET0.
• To assess the efficacy of using cross-station and exogenous data for ET0 modeling.
• To compare GEP model performance against traditional empirical equations.

Main Methods:

• GEP models were developed using meteorological data from 10 stations in Turkey.
• Models were validated using data from 18 stations in Iran, with distinct latitude and time periods.
• Performance was compared between GEP models and established empirical equations.

Main Results:

• Generalized GEP models demonstrated promising accuracy in simulating daily ET0 across distant stations and time periods.
• Empirical equations showed reduced accuracy when calibrated with exogenous data compared to their original forms.
• Despite reduced generalization in different climatic contexts, GEP models outperformed classic empirical equations.

Conclusions:

• GEP is a robust technique for estimating ET0, especially when utilizing exogenous data from ancillary stations.
• The study highlights the potential of soft computing for overcoming local data limitations in agricultural water management.
• GEP offers a more accurate alternative to traditional empirical methods for ET0 estimation in data-scarce environments.