Artificial Neural Network Modeling for Investigation on the Effect of Deficit Irrigation and Nitrogen Levels on Yield and Quality of Hay Remaining After Seed Harvest of Sorghum Sudangrass Hybrid

ABSTRACT

This study aims to develop an Artificial Neural Network (ANN) modeling to be trained to forecast the effects of different irrigation water levels and fertilizer doses on the hay yield and some quality traits of herbal parts of Sorghum × Sudan grass hybrid (Sorghum sudanense vs. Sorghum bicolor). The ANN model was developed on the limited field experiments implemented in Bilecik, Turkey, for 2 years in 2021 and 2022. Experiments were conducted in split-plot design with three replications. In the study, three irrigation levels (I100, I60, and I30) were placed in the main parcels, and four fertilizer levels (N0, N50, N100 and N150 kg ha⁻¹) were placed in the sub-parcels. Irrigations were made in three critical periods according to the amount of cumulative evaporation occurring in the Class A Pan. The results showed that irrigation and fertilization are important in terms of yield and quality characteristics. The yield increased depending on the irrigation and fertilization dose, and the highest value was obtained from the I100 × N150 interaction (28.10 t ha⁻¹). The highest protein yield was determined from the I60 × N150 (2.37 t ha⁻¹) interaction, and the Relative Feed Value (RFV) value was determined from the I30 × N150 (92.17) interaction. Irrigation Water Use Efficiency (IWUE) and Water Use Efficiency (WUE) values increased with decreasing irrigation amount, and the highest IWUE was determined from I30 and the highest WUE was determined from I60 irrigation subjects. According to the field experiments and ANN model, the I80 irrigation with 100 kg ha⁻¹ nitrogen doses would suit the feed yield and quality of Sorghum × Sudan grass hybrid.

KEYWORDS:

• Artificial neural network
• deficit irrigation
• fertilization
• hay yield
• protein quality
• roughage

Acknowledgements

The authors wish to thank the Scientific and Technological Research Council of Turkey (TUBİTAK) for funding this project under Grant No. TOVAG 122O683.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The work was supported by the Türkiye Bilimsel ve Teknolojik Araştırma Kurumu.