SSA-BP model for predicting water contents in stem integrating multiple environmental factors acquired via IoT

Stem is connected with the branches and leaves of the canopy upward in most trees. The water vapor of the stem can be exchanged with the atmosphere and then transferred to the canopy for the fixed photosynthetic products. Stem is also connected with the roots downward to transport water and inorganic salt in the soil into the important tissues of plants. Stem water content is one of the typical physiological water parameters for the maintenance of internal water balance, photosynthesis, carbon fixation, and normal cellular metabolism in standing living trees. An accurate prediction of stem water content is essential under the synergistic effects of different environmental parameters. In this study, the common urban tree species Acer pictum subsp. Mono was chosen as the research object. An Internet of Things (IoT) ecological information monitoring system was built to acquire the stem water content for the online display under various environmental parameters. A novel sensor was also designed to non-destructively collect the stem water content in real time, according to the principle of standing wave ratio. A systematic analysis was implemented to determine the variation in the stem water content at the growth stages, together with their correlations with the environmental parameters. An SSA-BP model was also proposed to predict the stem water content using the optimized back propagation neural network (BP) and the sparrow search algorithm (SSA). The root mean square error (RMSE) and mean absolute error (MAE) values of the model were calculated to evaluate the predictive performance, and to compare with the traditional BP model and the genetic algorithm (GA) optimized BP model. The results showed that the IoT ecological information monitoring system effectively obtained the stem water content and various environmental parameters. Among them, the stem water content gradually rose from the early hours of the morning to the peak at noon, and then began to decline during the germination and defoliation stages, indicating a significant pattern of "rising during the day and falling during the night". By contrast, the continuous increase of transpiration rate made the stem water content gradually drop to the valley value in the afternoon during the growth stage. Furthermore, the stem tissue began to rehydrate, as the transpiration weakened. Diurnal variation of stem water content presented an outstanding rhythm of "falling during the day and rising in the night". During the germination stage, the air temperature and soil moisture were the main influencing factors on the stem water content; During the growth period, the soil temperature and relative humidity were the determining factors; During the defoliation period, the highest correlation was found between stem water content and air temperature. The SSA-BP model shared the higher prediction accuracy for the stem water content during different growth stages, compared with the BP and GA-BP models. The SSA-BP model shared the higher prediction accuracy for the stem water content during different growth stages, compared with the BP and GA-BP models. Specifically, the R² of the SSA-BP model was 0.896-0.987, the RMSE was 0.314-12.971 mV, and the MAE was 0.232-5.030 mV. The predicted values of stem water content were highly consistent with the measured ones. A feasible system can be expected to accurately predict the stem water content. The finding can also provide a strong reference to reveal the internal water transport of plant stems and their environmental adaptation in agricultural and forestry plants.