当前位置：
X-MOL 学术
›
Agric. Water Manag.
›
论文详情
Our official English website, www.x-mol.net, welcomes your
feedback! (Note: you will need to create a separate account there.)
Classification and Regression Tree (CART)-based estimation of soil water content based on meteorological inputs and explorations of hydrodynamics behind
Agricultural Water Management ( IF 5.9 ) Pub Date : 2024-05-25 , DOI: 10.1016/j.agwat.2024.108869 Tsung-Hsi Wu , Pei-Yuan Chen , Chien-Chih Chen , Meng-Ju Chung , Zheng-Kai Ye , Ming-Hsu Li
Agricultural Water Management ( IF 5.9 ) Pub Date : 2024-05-25 , DOI: 10.1016/j.agwat.2024.108869 Tsung-Hsi Wu , Pei-Yuan Chen , Chien-Chih Chen , Meng-Ju Chung , Zheng-Kai Ye , Ming-Hsu Li
In this study, we investigate the feasibility of using the Classification and Regression Tree (CART) algorithm to estimate soil water content (SWC) using commonly available meteorological parameters. We trained and validated CART models using data collected in a grassland terrain in northern Taiwan throughout the year of 2018, with the goal of providing precise information for agricultural irrigation and flood risk assessment. Results indicate the effectiveness of CART in SWC estimation, with error levels acceptable for agricultural purposes. The mean absolute error is less than 4% (v/v) for 53 out of the total 60 models in the 12-fold Time-Series Cross-Validation for SWC at depths of 10, 30, 50, 70, and 100 cm. Furthermore, the effectiveness of meteorological parameters in different sets of time shifting and parameter types are assessed. Our findings reveal that the responsiveness of SWC to parameters derived from precipitation varies with soil depth and season, with SWC dynamics in response to precipitation being more pronounced in shallower layers (≤50 cm) compared to deeper layers (≥70 cm). The influence of precipitation-derived and non-precipitation parameters on SWC dynamics is manifested in their distinct feature-importance characteristics in a CART model. This study highlights the importance of understanding characteristics of rainfall and underlying hydrological dynamics, such as evapotranspiration and soil texture, in order to make accurate SWC predictions using CART. Since CART serves as the basis for a variety of top-performing machines like random forest and gradient-boosted trees, the discoveries from this study can also help estimate SWC with these advanced algorithms. Overall, the results of this study provide practical guidance for refining machine-learning based SWC estimations, contributing to more effective agricultural water management and irrigation strategies.
更新日期：2024-05-25
京公网安备 11010802027423号