Evaluating the Performance of Data-Driven Models Combined with IoT to Predict the Onion Yield under Different Irrigation Regimes

Abstract

MAJOR international nutrition organizations are increasingly focusing on global agriculture production. In particular, food insecurity has emerged in Egypt due to climate change, population expansion, and rising food demand. Innovative techniques such as internet of things and machine learning are critical for farmers to make timely judgments that impact quality of agricultural harvests. A new open-source technology utilizing the Arduino Board was created to forecast the irrigation requirements of onion crops. The study implemented three different irrigation levels: 100%, 85%, and 70% of crop evapotranspiration (ETc) ML models, namely, Artificial Neural Network (ANN), Random Forest (RF), and Decision Tree (DT) models were constructed to predict onion yield based on meteorological and agronomic variables. These variables include minimum, and maximum temperatures, relative humidity, sun shine hour, solar radiation, growing degree days, vapor pressure deficit, plant height, leaf number per plant, readily available water. The results highlight that the highest onion yield values were recorded in 100% ETc (56.04 ton/ha), followed by 85% ETc (51.52 ton/ha). The lowest values were recorded at 70% ETc (43.36 ton/ha). The results also indicate that Arduino board optimize water usage by 13% and 28%, enhancing crop water productivity by 14.5 and 16 kg/m3 at 85% and 70% ETc, respectively. Additionally, the ANN model demonstrates a robust R2 value of 0.94 in predicting onion yield, while the RF and DT models perform at 0.90 and 0.86, respectively. Our results highlight the effectiveness of technology in enhancing agricultural decision-making and crop management.