A Monitoring, Evaluation, and Prediction System for Slight Water Stress in Citrus Seedlings Based on an Improved Multilayer Perceptron Model

Abstract

To address the lack of effective monitoring, evaluation, and prediction methods for water stress in citrus seedlings, we conducted 10 sets of water stress gradient experiments. Based on the experimental dataset, we constructed, trained, and improved an MLP classification model for citrus seedling water stress. In addition, we developed a monitoring, evaluation, and prediction system based on this model. The experiments demonstrated that 7 days of slight water stress can induce changes in overall root wilting and growth stagnation, and the chlorophyll content in the leaves can decrease by up to 11.78%. Furthermore, the optimal VWC for seedlings was [45%, 50%], the boundary of drought was [20%, 25%], and the boundary of waterlogging was [50%, 55%]. We validated the effectiveness of the system in assessing the growth status of seedlings over the past 7 days and predicting it after 7 days through testing sets and experiments on slight water stress. We found that the system achieved non-destructive remote monitoring, evaluation, and prediction of slight water stress in citrus seedlings, thus enhancing seedling quality. These research findings provide valuable insights into water stress management in citrus seedlings and other crops.

Keywords

• water stress
• citrus seedling
• machine learning
• precision agriculture
• IoT in agriculture