A couple of years ago, I had a chance to contribute to a book on automation in tree fruit production with focus on precision sensor technologies for pest and disease monitoring. In this article, I’m sharing parts of my original draft before it was modified to become part of the book. If interested in references used in preparing this article, please download the original draft (link at the end).
Ground and aerial platforms
Platforms used for detecting crop diseases and damages caused by pests include airborne remote sensing, space-based satellite imagery, and ground-based systems relying on proximal non-contact sensors. Wireless sensor networks (WSNs) are also increasingly used for the site-specific monitoring of plant health including insect and disease detection. Remote sensing can be a useful tool especially for monitoring disease development on the leaves of tall fruit trees or for use with disease symptoms not visible from the ground. Despite the fact that aerial and satellite imaging technologies have been in place for decades, the high cost and lack of enough resolution of acquired data have been limiting their application among fruit growers.
Recent advances in wireless networking technology have allowed for the development of low power, low-cost, multifunctional sensor nodes which can be used in ground-based pest and disease monitoring of large areas. As an indirect benefit, a reduction in disease problems associated with over-irrigation has been achieved as a result of using WSNs. Besides monitoring pests, disease, WSNs provide the opportunity to automatically detect water and heat stress, nutrient deficiency, measure plant cover, and to predict yield in real-time. WSNs can also provide a platform for monitoring and quantifying the possible effect of climate change on agricultural water consumption and increasing pests and disease. The application of WSNs specialty in crop production is currently limited by the availability of sensors, communication range and frequencies (Lee et al., 2010).
Multispectral and hyperspectral electronic imaging systems have been used for the study of crop disease along with aerial photography as the primary technique of remote sensing. Airborne multispectral and hyperspectral imagery and high resolution satellite imagery from commercial satellites (e.g. IKONOS, QuickBird, and SPOT 5) provide image data at fine spatial resolutions enough for mapping within-field plant variability. Airborne and satellite imagery have been successfully used for detecting crop diseases. For example, researchers at the University of Florida have developed a multi-rotor remote sensing (MRRS) platform equipped with multispectral cameras and a thermal camera capable of identifying stressed trees in citrus orchard.
Contact type sensors
Over the years, molecular techniques have evolved into the most robust tool for detecting the presence of plant diseases. Enzyme-linked immunosorbent assay (ELISA) and polymerase chain reaction (PCR) are the molecular techniques commonly used for disease detection. Among the shortcomings of the molecular techniques are that the sample preparation is very labor-intensive and time-consuming, and need an elaborate procedure. This means that these techniques cannot be used as preliminary screening tool unless significant compromise is made in terms of results accuracy. Moreover, the molecular techniques are moderately expensive, require consumable reagents, trained personnel and specific instrumentation, and their application is limited to well-known and harmful diseases.
Being the fastest growing technology, biosensors are expected to replace ELISA in a near future. Biosensors used in pathogen detection are chemical sensors which incorporate a biological recognition element. The biological sensing component is grouped into two main categories of biocatalysts and bioligands. The sensing element is selective making it highly desirable as a basis for analysis in complex mixtures in real-time. Biosensors are used to develop devices called “electronic tongue”, which, as the name implies, are to be used in liquid phase. Biosensors are still far from being commercialized and need to be improved on a number of different aspects including cost, durability of the biological element, sensitivity, and reproducibility.
Non-contact type sensors
Remote sensing is a rapid, non-destructive, cost-effective measurement method which allows for taking unlimited number of samples. It is a promising technology for the detection of disease and can help in taking measures to prevent physiological stresses and physical damages caused by pathogens. To date, many non-invasive techniques have been developed for plant disease detection including imaging and spectroscopic techniques, and volatile organic compounds (VOC) profiling-based technique. The spectroscopic and imaging techniques encompass a broad range of methods such as fluorescence spectroscopy, fluorescence imaging, hyperspectral imaging, and visible-infrared (visible-IR) spectroscopy. In fluorescence spectroscopy, the object (vegetation) is excited with a beam of light (shortwave ultraviolet spectra) and the emitted fluorescence from it is measured. In fluorescence imaging, fluorescence images are taken using a camera. In the hyperspectral imaging, the spectral reflectance is acquired for a range which may include the visible and infrared regions of the electromagnetic spectra. Visible–near infrared (Visible–NIR) spectroscopy, mid-infrared spectroscopy and hyper spectral imaging have been used in a number of studies for the detection of Huanglongbing (greening) in citrus orchards.
The VOC profile-based disease detection uses an electronic nose or GC–MS (gas chromatography–mass spectroscopy) analysis of volatile metabolite released by plants (healthy and diseased) to identify diseases. An electronic nose is simply made up of a series of gas sensors allowing for discriminating a range of organic compounds that may be present in the air. Electronic nose systems have been successfully used for identifying plant diseases.
Both categories of imaging and spectroscopic, and VOC profiling-based techniques have shown high degree of accuracy in non-contact detection of plant diseases. However, there are limitations in practical applications of these methods for real-time monitoring. Remote sensing techniques have been mainly used to evaluate the extent of disease damage as early detection of crop disease is difficult using these methods and even impossible in some cases. The VOC profile naturally varies within plant species as a result of an environmental or nutrient stress masking the changes due to the existence of diseases or stresses. To cope with this problem, distinct volatile biomarker specific for a particular disease has to be identified. The spectral reflectance from the vegetation is also affected by the environmental conditions. This can be possibly resolved by identifying vegetative indices or wavelength ranges that are not sensitive to environmental changes.
Sensor integration, data quality and algorithm issues
Despite the advances in the remote sensing technology, real-time automated monitoring of plant diseases under field conditions remains a challenge. A serious issue in remote sensing techniques is that multiple biotic and abiotic stressors may coincide and result in similar spectral reflectance. In this case, additional information about the diseases and other sources of stress are required to determine which one is responsible for which morphological and/or physiological changes in the crop. In practical applications, the stress detection algorithm must monitor soil, crop and diseases simultaneously, otherwise the technique may fail to discriminate plant conditions. The selection of statistical classification method, optimization of remote sensing techniques, and general algorithm for detecting a particular plant and disease depend on the data acquisition setup under field conditions. The algorithms need to account for a complex interaction between sensor and crop which is difficult to achieve in the field.
The spectroscopic and imaging technology has long been used for the identification of crop water stress and nutrient deficiencies. This technology can achieve superior control and management through integration with an autonomous vehicle which concurrently detects various sources of plant stress in real-time. Both VOC profiling and the spectroscopic and imaging techniques are well established for non-agricultural applications and have shown promising for non-invasive monitoring of plant diseases. Next step is to incorporate these techniques into autonomous robots. The spectroscopy-based imaging techniques have also been extensively used in evaluating the quality of postharvest vegetables and fruits. The methodology for data collection and statistical models developed for postharvest applications can be useful for real-time detection of plant disease because of their similarities.
Citations
Osroosh, Y., 2016. Precision technologies for Pest and Disease Management. Chapter Draft for Book “Automation in Tree Fruit Production, Principles and Practice”.