Spatial Optimization of Electrophysiological Signal Acquisition in Clivia Leaves Under a Controlled Leaf-Surface...

Plant electrophysiological signals can rapidly reflect the dynamic responses of plants to external stimuli, giving them strong potential for nondestructive monitoring and early state recognition. However, differences among plant organs, as well as spatial heterogeneity within the same organ, may substantially affect signal quality and stability because of variations in tissue structure and local physiological activity. To address this issue, this study used Clivia as an experimental model and established a controlled local leaf-surface salt-treatment paradigm to systematically evaluate the relative discriminative ability of electrophysiological signals recorded from different spatial positions on leaves. First, stepwise screening of longitudinal leaf regions and leaf hierarchy was performed using 0 mM and 100 mM NaCl agarose gel treatments to determine the optimal signal acquisition position. Then, based on the selected position, a five-level NaCl treatment recognition task was constructed, and LRPNet, a residual network integrating PoolFormer and an efficient channel attention mechanism, was proposed for multi-gradient classification of plant electrophysiological signals. The results showed that, within the current experimental framework, the basal region of the top leaf exhibited the highest relative separability and the best overall recognition performance. In the five-gradient recognition task, LRPNet achieved the highest mean Accuracy of 92.21% among the compared models. These findings indicated that plant electrophysiological signals exhibited pronounced spatial heterogeneity and that optimization of the recording location was not merely an experimental detail, but an important upstream factor that affected downstream recognition performance. This study provides a methodological basis for optimizing signal acquisition positions and improving electrophysiological signal recognition in plants. However, the present conclusions are mainly applicable to the controlled local salt-treatment paradigm established in this study and still require further validation through more rigorous physiological verification, cross-scenario testing, and more independent data-partitioning strategies.