A Concentration Prediction-Based Crop Digital Twin Using Nutrient Co-Existence and Composition in Regression Algorithms

Abstract

Crop digital twin is redefining traditional farming practices, offering unprecedented opportunities for real-time monitoring, predictive and simulation analysis, and optimization. This research embarks on an exploration of the synergy between precision agriculture, crop modeling, and regression algorithms to create a digital twin for farmers to augment the concentration and composition prediction-based crop nutrient recovery. This captures the holistic representation of crop characteristics, considering the intricate relationships between environmental factors, nutrient concentrations, and crop compositions. However, the complexity arising from diverse soil and environmental conditions makes nutrient content analysis expensive and time-consuming. This paper presents two approaches, namely, (i) single-nutrient concentration prediction and (ii) nutrient composition concentration prediction, which is the result of a predictive digital twin case study that employs six regression algorithms, namely, Elastic Net, Polynomial, Stepwise, Ridge, Lasso, and Linear Regression, to predict rice nutrient content efficiently, particularly considering the coexistence and composition of multiple nutrients. Our research findings highlight the superiority of the Polynomial Regression model in predicting nutrient content, with a specific focus on accurate nitrogen percentage prediction. This insight can be used for nutrient recovery intervention by knowing the precise amount of nutrient to be added into the crop medium. The adoption of the Polynomial Regression model offers a valuable tool for nutrient management practices in the crop digital twin, potentially resulting in higher-quality rice production and a reduced environmental impact. The proposed method can be replicable in other low-resourced crop digital twin system.

Keywords

• rice nutrient level
• fertilizer optimization
• nutrient analysis
• polynomial regression
• nutrient prediction
• environmental impact reduction