Time series electronic nose responses from tomato plants infested by Thrips palmi
Description
This dataset comprises time-series electronic nose responses of tomato plants (*Solanum lycopersicum*) subjected to various infestation levels of *Thrips palmi*. We examined the temporal dynamics of herbivore-induced volatile responses associated with insect feeding activity under controlled laboratory conditions. The tomato plants were divided into five treatments: control, low infestation, medium infestation, high infestation, and mechanical damage treatment. Sensor responses were taken at various time points, i.e., 1 h, 3 h, 6 h, 12 h, 24 h, 48 h, and 7 d after infestation. The electronic nose system was constructed with a set of metal-oxide semiconductor (MOS) sensors, such as MiCS-6814 (CO, NH₃, NO₂ channels), MQ135, MQ138, MQ3, TGS2600, TGS2602, and TGS822. The dataset includes both raw sensor responses and extracted features. Extracted features include peak response, area under the curve (AUC), rise slope, steady-state response, time to peak, and recovery characteristics. The dataset was created to assist research in non-invasive pest detection, plant defense responses mediated by volatiles, machine learning classification, precision agriculture, and insect-plant interactions.
Files
Steps to reproduce
1. Tomato plants (Solanum lycopersicum) were grown in a controlled environment until they reached a suitable stage of vegetative growth for infestation experiments. 2. Colonies of *Thrips palmi* were maintained under laboratory conditions and introduced to tomato plants according to predetermined infestation treatments, including low, medium, and high infestation levels. Additional treatments for control and mechanical damage were included for comparison. 3. Samples of volatile compounds released from tomato plants were characterized by a dedicated electronic nose system equipped with several metal-oxide semiconductor sensors, including MiCS-6814, MQ-series, and TGS-series sensors. 4. Sensor measurements were taken at multiple time points after infestation (1 h, 3 h, 6 h, 12 h, 24 h, 48 h, and 7 d) to capture temporal dynamics of volatiles associated with herbivore-induced plant responses. 5. Prior to feature extraction, raw sensor signals were baseline corrected and processed. For each sensor channel, features were extracted including peak response, area under the curve (AUC), rise slope, steady-state response, time-to-peak, and recovery features. 6. The processed data were compiled into tables, which included the treatment labels, infestation stages, sensor identities, extracted features, and time-resolved measurements. 7. The processed features were then analyzed using machine learning to assess the classification performance between infestation stages and treatment groups.