Maximizing multi-source data integration and minimizing the parameters for greenhouse tomato crop water requirement...

Area of Science:

• Agricultural Science
• Environmental Science
• Data Science

Background:

• Accurate crop water requirement prediction is vital for efficient irrigation management in protected agriculture.
• The FAO-endorsed Penman-Monteith model, while standard, faces challenges due to parameter complexity and empirical inaccuracies.
• Existing methods often struggle with precise water demand forecasting for greenhouse crops.

Purpose of the Study:

• To develop a novel, data-driven water demand prediction model for greenhouse tomato crops.
• To integrate multi-source data, including canopy coverage derived from image segmentation and environmental factors.
• To enhance the accuracy and reliability of crop water requirement estimations for scientific irrigation.

Main Methods:

• Utilized ExG algorithm and maximum inter-class variance for canopy coverage extraction via image segmentation.
• Employed Spearman correlation and random forest feature importance for optimal variable selection (Tmax, Ts, CC).
• Developed and compared average, weighted, and stacking fusion models using RandomForest, LightGBM, and CatBoost machine learning algorithms.

Main Results:

• The stacking fusion model demonstrated superior prediction performance with lower error metrics (MSE, MAE, RMSE) compared to other models.
• The selected feature combination (Tmax, Ts, CC) significantly reduced prediction errors and increased the R² value.
• The novel model achieved reductions in MSE, MAE, and RMSE by over 4%, 14%, and 3%, respectively, with a 1% R² increase.

Conclusions:

• The developed stacking fusion model offers a more accurate and reliable method for predicting greenhouse tomato water requirements.
• Integrating image-derived canopy coverage with environmental data provides innovative technical support for scientific irrigation practices.
• This approach effectively decouples and minimizes characteristic parameters for improved water management in protected agriculture.