Evolving connectionist systems (ECoSs): a new approach for modeling daily reference evapotranspiration (ET₀)

Area of Science:

• Environmental science
• Artificial intelligence
• Hydrology

Background:

• Daily reference evapotranspiration (ET0) is crucial for water resource management and agricultural planning.
• Artificial intelligence (AI) techniques are increasingly utilized for modeling complex environmental processes like ET0.
• Accurate ET0 estimation is vital for the Mediterranean region, particularly in Algeria, facing water scarcity challenges.

Purpose of the Study:

• To propose and evaluate novel evolving connectionist (ECoS) approaches for modeling daily reference evapotranspiration (ET0) in Algeria.
• To compare the performance of three ECoS models: DEFNIS_OF, DEFNIS_ON, and EFuNN.
• To assess the effectiveness of different combinations of climatic variables as inputs for ET0 modeling.

Main Methods:

• Application of three ECoS models (DEFNIS_OF, DEFNIS_ON, EFuNN) to model daily ET0.
• Utilizing climatic variables (Tmax, Tmin, WS, RH, SH) from Algiers and Skikda weather stations as inputs.
• Statistical comparison using RMSE, MAE, R, and NSE, benchmarking against the FAO-56 PM model.

Main Results:

• The DEFNIS_OF model, using all five climatic variables, outperformed DEFNIS_ON and EFuNN in ET0 modeling.
• High correlation coefficients (R) and Nash-Sutcliffe efficiency (NSE) values were achieved for DEFNIS_OF (e.g., R=0.954, NSE=0.910 for Algiers).
• DEFNIS_OF and DEFNIS_ON models showed good accuracy, while EFuNN exhibited lower performance.

Conclusions:

• ECoS approaches, particularly DEFNIS_OF, are effective for modeling daily ET0 in Algeria's Mediterranean climate.
• Input variable selection significantly impacts model performance, with a comprehensive set yielding superior results.
• The study provides a valuable tool for improving water management strategies through accurate ET0 prediction.