A PSO-CatBoost Algorithm Based on Soil Sensor Data: A Novel In-Situ Intelligent Method for Estimating Field Capacity

This is the authors' abstract. We don't add key points for this paper.

Field capacity (FC) is a key parameter for smart irrigation, but it still requires manual measurement, which limits technological progress. Enhancing the in-situ intelligent analysis capability of soil moisture sensors is crucial for advancing this field. Among 11 common tree-based models, Categorical Boosting (CatBoost) achieved the best performance in FC simulation. Therefore, CatBoost combined with Particle Swarm Optimization (PSO) was employed for parameter optimization, resulting in a data-driven intelligent online FC measurement method. This approach dynamically characterizes soil moisture recession curves at various depths using sensor data, while improving temporal resolution and reducing the complexity of FC estimation. Using a dataset comprising 118 farmland soil moisture monitoring points from major grain-producing regions in China, the proposed method was validated through simulation. The simulation errors were compared with those from other emerging integrated algorithm models, and the results demonstrated that the proposed method achieved superior performance. When applied to FC simulation, the model achieves a remarkably high coefficient of determination (R 2 ) of 0.99, with mean absolute error (MAE) ranging from 0.41% to 1.35%, mean squared error (MSE) from 0.63% to 7.27%, and root mean squared error (RMSE) from 0.80% to 2.64%, respectively. These results confirm that integrating AI with soil sensors enables accurate, in-situ FC estimation. In practice, this method enables dynamic adjustment of irrigation thresholds, supporting automated decision-making and precision water management in modern agriculture.