Using Remote Sensing to Optimize IPM in Greenhouses

Using Remote Sensing to Optimize IPM in Greenhouses

Robert Starnes, University of California Davis

Team member Robert Starnes uses a remote control to move the robot along the rail. As the robot moves, the camera collects data from the plants below.

The key to effective integrated pest management (IPM) is the ability to accurately detect an emerging infestation, so necessary actions (releases of natural enemies, pesticide applications, changes in operational procedures, etc.) can be implemented when and where needed. IPM can be considered a two-pronged approach: 1) a strong emphasis on deployment of preventive actions (physical barriers, variety selection, cleaning procedures, temperature control, etc.), and 2) responsive actions, which are deployed after emerging pest populations have been detected.


While there has been widespread adoption of a range of preventive pest management actions, the adoption of IPM-based responsive actions (such as targeted releases of natural enemies) has been hampered by inadequate monitoring/scouting procedures to accurately detect early/emerging pest populations. Early detection of pests is key to minimizing their damage and therefore their economic importance.

Due to the size of most commercial greenhouse operations, the complexity of management operations, and constraints associated with time-consuming monitoring procedures, pest infestations are often detected too late, and fully established pest populations are difficult to suppress through the means of biological control or soft (slow-acting) insecticides. Development of practically feasible, cost-effective, reliable, and accurate technologies to detect emerging pest populations can be considered one of the most important challenges greenhouse operations are facing regarding IPM of insects and mites. If such technologies could be made readily available, then applications of softer insecticides and releases of natural enemies could provide much higher levels of crop protection, as they could be deployed more strategically (when and where needed).

Deployment of Advanced Remote Sensing Technologies

Researchers at the University of California Davis (UC Davis) are using advanced camera systems to detect stress when crops are infested with pests. I lead the team, and our research involves using drone systems to monitor almond orchards and strawberry fields. Recently, our team developed a rail system to be deployed inside greenhouses.

The fundamental principle behind the use of advanced remote sensing technologies to automate and optimize detection of pest infestations is that a healthy crop reflects light in a way that is different from crops being stressed by insects and mites. To make an analogy to humans, the crop plants “get a fever” when subjected to stress by pests. At that point, advanced camera technologies can be used in two important ways: 1) to detect that the plant is stressed (being infested), and 2) advanced analyses of the specific changes in reflectance features can be used as diagnostics to determine the type of stressor. This is possible because crop reflectance is being acquired in a high number of spectral bands, and reflectance values in each of these spectral bands are indirectly associated with the physiological processes in leaf tissues.

Pest-induced stress causes subtle changes in the plants’ physiological processes, and these physiological changes alter the biochemical composition of leaves, which in turn leads to detectable changes in leaf reflectance. It is important to mention that these subtle changes cannot be observed or detected by the human eye — our sensitivity to color change is not high enough. However, advanced camera systems, combined with robust procedures to calibrate and correct the leaf reflectance data, enable us to detect plant responses to very low levels of pest infestation. Thus, it is possible to accurately detect emerging pest infestations.

Rail-Based Remote Sensing

Most commercial greenhouses have either rail systems or pipes. Our team of researchers has developed a prototype to run along an existing pipe-heating system to avoid infrastructure changes to the greenhouse. The prototype is controlled by customized software to control both movement of the camera and leaf reflectance data acquisition. Thus, the prototype can move automatically above benches with vegetables or flowering plants, and the leaf reflectance data are transferred to a computer for processing and analysis.

A lighting system is also being added to the prototype, which will enable plant monitoring at night. This has multiple advantages:
• Due to shade from beams caused by subtle differences within greenhouse structures, light conditions are not uniform. These differences in the imaging conditions affect the quality of leaf reflectance data.
• Due to operational procedures, it is likely advantageous to have the remote-sensing system monitor the greenhouse crops at a time when workers are not handling crops or ornamental plants.
• The advanced-imaging technologies require large volumes of data, which, due to processing and analysis requirements, require some computer-processing time. So, the research team is working toward a system that can operate at night and provide mapping of hot spots (including diagnostics of stressors) the next morning, so greenhouse managers can decide when and where further inspections or actions are needed.

Detecting Spider Mite and Lygus Infestations in Gerbera

To illustrate the performance of the rail-based prototype, we conducted a study in which flowering gerberas were divided into three treatment groups: 1) control (non-infested), 2) infested with five adult two-spotted spider mites, and 3) infested with one adult Lygus bug. These treatment levels represent very low infestations, and the goal was to determine how early such low pest densities could be detected.

It is important to highlight that the remote sensing is not used to detect the pest individuals themselves, but rather the stress response by the gerbera plants to the imposed infestations. Two-spotted spider mites are among the most proliferous pests in greenhouse operations, and they have developed resistance to many of the available miticides. Effective biological control agents of two-spotted spider mites are commercially available, but their effectiveness is severely compromised if they are released too late and/or too early, so timing (both in space and time) of releases is critical.

Lygus is becoming one of the most concerning greenhouse pests in the cut-flower industry in California and elsewhere, and their destructive feeding causes flower disfiguration and abortion. Very few insecticides are effective against Lygus bugs, and worker safety is a big concern when potent insecticides are used to control this pest.

Leaf reflectance data were acquired before gerbera plants were exposed to two-spotted spider mites or Lygus, and at different time intervals thereafter. Even though average reflectance values appear similar among the three treatments in the visible portion of the radiometric spectrum (470 to 680 nm), there are important differences. These differences become more pronounced in wavelengths from 750 to 850 nm. Over time (after seven days of infestation divided with before infestation), values close to one would indicate little or no change.

It is seen that in spectral bands near 450 nm, Lygus infestation caused a decrease in reflectance, and two-spotted spider mite infestation caused an increase in reflectance. There was also a change in reflectance of non-infested plants, and that is due to growth of the plants, but that change was smaller than the change in plants infested with Lygus and two-spotted spider mites.

In some portions of the radiometric spectrum, there were clearly detectable differences between all three treatments, suggesting that leaf reflectance can be used not only to detect pest-induced stress, but also to determine the type of pest.

Our team’s next step is to study responses by gerbera to other insect pests and to broaden the scope to include important diseases and nematode pests. In addition, there are plans to study other greenhouse plants.

Broader Use of Remote Sensing

Using a rail-based remote sensing system in commercial greenhouses will allow for more efficient and effective detection and management of insects, mites, and diseases. Leaf reflectance patterns are not only indicative of arthropod infestations, they can also indicate abiotic stresses, such as nutrient deficiencies and salt stress, and they can be used to monitor crop plant health more broadly.

Our team at UC Davis is convinced that robotics and remote sensing will revolutionize greenhouse operations, and the prototype is just one of many being developed by private companies and university researchers worldwide.

While acquisition of remote-sensing data inside greenhouses is relatively easy, calibration and processing of data under commercial operations are associated with some important constraints. Also, crop varieties have unique reflectance profiles, and the reflectance data are also influenced by growth stage and management practices, so there are many variables to take into account when developing classification algorithms.

Ultimately, our team aims to use this information to develop a precise, practical, and economically feasible plant stress detection system for greenhouse growers that can serve as a decision support tool for improved and sustainable ornamental crop management.