Remote Sensing of Diseases

Annual Review of Phytopathology
Remote Sensing of Diseases

Erich-Christian Oerke
INRES, Plant Diseases and Crop Protection, Rheinische Friedrich-Wilhelms-Universität Bonn,
D-53115 Bonn, Germany; email: ec-oerke@uni-bonn.de
Annu. Rev. Phytopathol. 2020.58:225-252. Downloaded from www.annualreviews.org
Access provided by South Dakota State University on 02/12/22. For personal use only.
Annu. Rev. Phytopathol. 2020. 58:225–52 Keywords

The Annual Review of Phytopathology is online at
spectral imaging, detection, identification, disease control, e-nose
phyto.annualreviews.org
https://doi.org/10.1146/annurev-phyto-010820- Abstract
012832
Detection, identification, and quantification of plant diseases by sensor tech-
Copyright © 2020 by Annual Reviews.
niques are expected to enable a more precise disease control, as sensors
All rights reserved
are sensitive, objective, and highly available for disease assessment. Recent
progress in sensor technology and data processing is very promising; nev-
ertheless, technical constraints and issues inherent to variability in host–
pathogen interactions currently limit the use of sensors in various fields of
application. The information from spectral [e.g., RGB (red, green, blue)],
multispectral, and hyperspectral sensors that measure reflectance, fluores-
cence, and emission of radiation or from electronic noses that detect volatile
organic compounds released from plants or pathogens, as well as the poten-
tial of sensors to characterize the health status of crops, is evaluated based on
the recent literature. Phytopathological aspects of remote sensing of plant
diseases across different scales and for various purposes are discussed, in-
cluding spatial disease patterns, epidemic spread of pathogens, crop char-
acteristics, and links to disease control. Future challenges in sensor use are
identified.
225
INTRODUCTION
The production of crop plants worldwide is limited by abiotic and biotic factors. As pests, e.g.,
weeds, pathogens, viruses, and arthropods and other animal pests, may be controlled to some
extent by the intervention of the grower, assessment and control of crop losses due to pests are
crucial, even though losses due to abiotic factors may be considerably higher. The loss potential
of diseases (caused by fungi, oomycetes, bacteria, and viruses) and the actual losses have been
estimated to account for 16% and 11% of the attainable crop production worldwide, respectively
(87). It is generally agreed that plants have to be protected from damage caused by diseases and
other pests. Pest management has to be effective, efficient, and sustainable and can best be realized
by combining mechanical, biological, and chemical tools with other supportive technologies in
integrated pest management (IPM) systems (90).
Disease management in agricultural crops commonly assumes a homogeneous pattern of dis-
ease distribution, so crops are sprayed at uniform application rates. However, the heterogeneity of
crops caused by differences in soil conditions, topographical situation, neighboring fields, micro-
climatic conditions, and sources of pathogen inoculum often results in a heterogeneous distribu-
tion of diseases manifested as patches, gradients, or random patterns. As disease patterns may vary
from site to site (and year to year) and over time within epidemics in an individual field, farmers
have to extend the visual rating of disease (incidence and severity) to a representative number of
samples, e.g., 100 sugar beet leaves sampled along a diagonal transect within a field (151), to decide
whether or not to spray the field.
The heterogeneity of disease distribution in a crop may provide an option to minimize the
amount of fungicide sprayed during the growing season by applying the fungicide only where it
is needed. This reduction can minimize undesirable environmental effects from pesticides as well
as decrease selection for fungicide resistance in pathogens. A prerequisite for site-specific disease
control is the sensing of diseases, i.e., the monitoring of as many plants as possible, ideally all plants,
to determine whether any individual plant is diseased (and to what degree) as well as whether the
disease needs to be controlled. Fungicide use may be significantly reduced by applying fungicides
only when and where the first diseased plants occur or are likely to appear. Pesticide reduction
programs (now common in many EU countries) and technological development are drivers for
the use of sensors in crop protection. Current fungicide use in France could be reduced by 47%
without negative effects on either productivity or profitability in 59% of farms (57).
The diagnosis and quantification of plant diseases currently rely on visual rating of plants and,
in case of doubt, the use of specialized equipment for methods based on nucleic acid and serologi-
cal technologies. ELISA assays require washers and readers, whereas immune strips do not require
specialized equipment but are less sensitive (qualitative tests). Visual assessment may be inaccurate,
but the often lab-based direct methods are invasive and destructive and cover only a representative
sample of the crops of interest. Indirect remote-sensing approaches include thermography, fluo-
rescence imaging, and spectral techniques, all noninvasive and nondestructive methods that allow
repeated monitoring of, potentially, all plants of interest. Gas chromatography–mass spectrometry
and electronic nose (e-nose) techniques are intermediate in the time required for sample prepara-
tion and measuring, as they detect volatile organic compounds (VOCs) emitted by diseased plants
and pathogens infecting plant tissue.
Sensors may be used for the detection of diseased plants, to decide whether to control a
pathogen, and for the assessment of spatial patterns of plant diseases (147). Many authors, how-
ever, report on the damage of plant tissue caused by diseases rather than the disease itself. The
introduction of global navigation satellite systems, which can provide a link between field sites
and plant characteristics, geographic information systems (GIS), and imaging sensors to discern
226 Oerke
plant status, has enabled researchers and plant growers to use technical aids for remote sensing of
diseases by linking sensor data on the plant status with spatial information on the location of the
plant. Early reviews (e.g., 39, 49, 85) emphasized the potential of using the information contained
in electromagnetic radiation after pathogen interactions with plants for plant pathology. At that
time, sensors often were limited to airborne and spaceborne remote sensing. Ground-based digi-
tal sensors were introduced in the 1990s and later and considerably broadened the application of
remote-sensing techniques.
Remote sensing is the acquisition of information about an object or phenomenon by a record-
ing device not in physical or intimate contact with the object or phenomenon under study; hence,
it is noninvasive and nondestructive (118). It is suitable for monitoring across various temporal
and spatial scales as well as for time-series experiments. Proximal sensing, a subarea with sensors
operating close to the target of interest, is complementary to remote sensing sensu stricto, i.e.,
monitoring the surface of the earth from satellite and airborne systems. Since the review of West
et al. (142), technical advances, especially the introduction of imaging sensors and the develop-
ment of new methods for data processing (e.g., artificial intelligence), have considerably increased
the potential of remote sensing for plant diseases. However, other issues, e.g., the impact of la-
tent infections on the crop area to be sprayed for disease control, are still unresolved, and new
questions have arisen, e.g., how to make the best use of unmanned aerial vehicles (UAVs).
This review summarizes the current state-of-the-art for remote sensing of plant diseases caused
by fungi, oomycetes, bacteria, viruses, and nematodes. Remote sensing can be used as the first step
in site-specific disease control and also to phenotype the reactions of plant genotypes to pathogen
attack. Starting with basic information on the development of plant disease symptoms across var-
ious scales, the review focuses on the implementation of sensors in crop protection practice and
phytopathology research. Remote-sensing applications, specific requirements, current technical
constraints, and biological limitations intrinsic to plants and crops are addressed. Crop sensing
that can measure the incidence of diseases as the basis for decision-making on the necessity for
chemical disease control has the potential to substantially reduce the amount of fungicide applied
in the future. Sensor use in disease monitoring may have economic and environmental benefits,
provided that the spatial distribution of diseases is heterogeneous, similar to the patchiness of
weeds (47).
DEVELOPMENT OF DISEASES IN TIME AND SPACE: THE ROLE

OF BIOLOGY
Pathogens that colonize and parasitize plants cause modifications in the metabolism and biochem-
ical and physical status of plant tissue that result in visible disease symptoms. Successful resistance
reaction and susceptibility of the plant are both associated with changes in gene activity and phys-
iological and morphological alterations. Initial indicators of pathogen-induced plant stress de-
tectable by methods other than destructive gene expression profiling and metabolomics include
biochemical changes measurable as VOCs released from affected tissue. In many cases, the primary
metabolism, e.g., photosynthesis, is decreased, and the secondary metabolism, e.g., formation of
phenols, is initiated or increased. Changes in photosynthetic activity, transpiration, and increased
respiration may be detected prior to the occurrence of visible symptoms (76).
Visible symptoms, often characteristic of a specific disease, appear after a pathogen-specific,
environment-sensitive latency period varying from days to months. Symptoms on susceptible
crops can include changes in coloration and morphology; wilting; death of plant parts, the entire
plant, or tissue associated with specific/nonspecific necrotic spots; and fungal structures. Polycyclic
pathogens can spread to other tissues, nearby host plants, more distant areas within the same field,
www.annualreviews.org • Remote Sensing of Diseases 227

and neighboring fields via spores that are produced exclusively on symptomatic tissue (i.e., no dis-
persal occurs during the latent period). Symptoms of pathogen attack on more resistant crops may
differ considerably from disease symptoms on susceptible crops and may include invisible cellular
reactions, necrotic or chlorotic tissue, or fewer and smaller symptoms.
The incidence of diseases in the field is often rather heterogeneous over time and space, at
least in the early stages of an epidemic (68, 137, 140). The incidence of plant diseases is gener-
ally overdispersed, i.e., the statistical variability is greater than expected, indicating spatial het-
erogeneity in a data set (69). Differences in the physical environment (e.g., edaphic factors and
microclimate) can interact with crop development and the life cycles of pathogens to result in the
heterogeneity of disease incidence and severity in the field even under controlled conditions. In
the later stages of epidemics or at high infection levels, the pattern of diseased plants and disease
severity may become more homogeneous; however, early stages with low disease incidence are
relevant to decision-making in disease control.
The decision whether or not to apply a fungicide to control a sensitive pathogen depends on not
only the presence of a symptom but also whether the disease severity exceeds the action threshold
level calculated from economic considerations (114). In high-value crops, the action threshold may
be 0, making technical disease detection unnecessary. Monocyclic pathogens, i.e., those lacking
an additional generation during the crop growth period, do not need control at the time of the
first appearance of symptoms, as the damage is already done (e.g., smut fungi). In contrast, the
frequency of first disease symptoms caused by polycyclic pathogens is often low and the expected
increase in disease severity from the next generations of the pathogen may be controlled to prevent
disease severity from exceeding the economic threshold level, provided effective fungicides are
available. If fungicides are not available, diseased individuals may be eliminated from the crop to
avoid disease spread in the crop and inoculum production and survival for the next growth period.
The incidence of diseases caused by soilborne pathogens may be heterogeneous because the
causal organisms are affected by specific soil conditions, e.g., soil texture, organic carbon content,
availability of debris from specific host plants, and water status. Polycyclic spread of soilborne
pathogens is similar to that of airborne pathogens. Incidence and spatial patterns of disease from
these pathogens depend on differences in crop canopy based on plant status (influenced by soil
conditions, availability of water and nutrients), microclimate within the crop, weather conditions,
and other factors like the plant’s position within the field and the surrounding vegetation. In-
cidence and spatial patterns of vector-transmitted diseases depend on the interactions between
vector, plant, causal agent, and environmental conditions and are poorly understood and hardly
predictable.
Root diseases may be assessed indirectly by sensors as the plant’s shoot is affected by the activ-
ities of the pathogen via its effect on the functionality of the root system. These symptoms often
become visible in later stages of the disease, and frequently they are unspecific and the causal
pathogen cannot be controlled within the growing season. Seed-transmitted diseases caused by
monocyclic pathogens (e.g., smuts) are not of interest for disease sensing because control can be
achieved only by seed treatment before sowing.
The hierarchical scales of epidemics were recognized early in the history of plant epidemiol-
ogy, i.e., region > field > foci > plant > lesion (32, 139). In the context of using remote sensing
for decision-making in disease control, the ability to quantify symptoms on the leaf scale should
contribute to the decision of whether a focus of diseased plants exceeds the action threshold for
spraying a fungicide. The size of the disease focus relevant to site-specific control depends on the
capabilities of the sprayer system, e.g., variable rate application at the nozzle level, boom section,
and velocity of the sprayer. For example, clusters of take-all-infested wheat plants were 2–2.5 m
in diameter, and anisotropy of the spatial pattern could be linked to soil cultivation (33). Spatial
228 Oerke
heterogeneity has also been reported for airborne diseases, including powdery mildew and leaf
and stripe rust of cereals (17, 29). Initial symptoms of wheat stripe rust occur in foci and may
spread to the entire field if left untreated (109). For Rhynchosporium leaf scald of barley, an ag-
gregated pattern has been described based on visual inspections, suggesting that Rhynchosporium
commune is seed-transmitted and polycyclic (136). Geostatistical analyses of rice blast incidence
in China revealed spatial dependence at the regional and field levels (5 to 10 m) but not at the
county scale (36). In contrast, Sclerotinia sclerotiorum sclerotia and white mold in soybean fields
had random distribution patterns (153). In a coffee tree nursery, an epidemic of bacterial blight of
seedlings started with a random pattern and progressed to an aggregated pattern in later stages of
the epidemic (14). The spatiotemporal dynamics of citrus Huanglongbing depended on weather
conditions and the presence of the insect vector (113).
Although these recent examples illustrate progress in our understanding of epidemic progres-
sion, the spatial patterns of many diseases and their development over time are not yet known.
Systematic investigations aiming to map the diversity of diseases at the field and regional levels,
e.g., Lorestani et al. (66), remain rare. Because visual disease monitoring in the field is laborious
and incomplete, the number of samples is typically limited. Therefore, mapping of larger areas for
disease patterns with high spatial resolution depends on the availability of effective and reliable
disease sensors. The widespread availability of such sensors would enable a rapid increase in our
understanding of spatial development of plant epidemics.
SCOPES OF DISEASE SENSING

For a long time, the perception and classification of plant diseases depended on inspection by
human eyes and specialized knowledge of the operator (16, 86). The development of serologi-
cal and DNA-based technologies has significantly improved and facilitated the identification of
plant pathogens; however, these techniques are destructive and often time-consuming, and quan-
tification is relevant to the sample only. Technical sensing of disease symptoms and otherwise
detectable changes in the crop status caused by pathogens may be eligible for different purposes,
under different conditions, and by different users (Table 1).
Table 1 Sensing of diseases: potential application areas, conditions and users

Application Application area Environmental conditions User(s)
Control of plants for quality and Quarantine inspection Semi-controlled Exporters, importers,
safety regulatory authorities
Quality control of plant Controlled Food industry
products, especially for
postharvest diseases
Phenotyping genotype reactions of Crop breeding for disease Semi-controlled, field Plant breeders
plants resistance
Monitoring the health status of Specialty crops in the Semi-controlled, repeated Gardeners
crops greenhouse measurements
Perennial crops Orchards and vineyards Wine growers,
gardeners
Forestry Woods, forest plantations Management of private
and state forests
(Annual) field crops Field Farm industry,
contractors
Study of host–pathogen interactions Basic and applied research Laboratory, controlled, field Researchers

Sensors are expected to be objective, accurate, precise, rapid, and available 24 hours a day,
7 days a week (24/7). Sensors of plant diseases may be used in quality control (e.g., by the food
industry or quarantine authorities) once, or they may be integrated into autonomous systems for
the continuous monitoring of crops for plant diseases, i.e., checking and keeping a continuous
record of the crop health status. Systematic observation of a crop by technical sensors can allow
the operator to intervene when infections are detectable or exceed action threshold levels. Ideally,
sensors should be capable of (a) detecting a deviation in the crop’s health status brought about by
pathogens, (b) identifying the disease, and (c) quantifying the severity of the disease. Instruments
may be highly sensitive in detecting diseases but often lack diagnostic potential. The identification
of a disease requires the ability to differentiate among various/all potential diseases according to
disease-specific symptoms. The quantification of typical disease symptoms (disease severity) and
assessment of leaves infected by several pathogens are simple for imaging systems but a challenge
for nonimaging sensors and sensors with inadequate spatial resolution (Figure 1).
In some applications, disease detection is the perception of a deviation from a healthy

crop/fruit. In these cases, neither identification nor quantification of disease is required. In the
a Field monitoring with nadir sensing and b Monitoring of row crops under controlled
subsequent control activity conditions using sensors on tramlines
c Phenotyping with various sensors d Quality control of food, e.g., fruits
e-nose Cameras
Figure 1
Application areas of remote disease sensing. (a) Monitoring in the field (cereal canopy, orchard, forest) using sensor(s) to record
information on nadir plants. (b) Under controlled conditions using sensors on tramlines. (c) Phenotyping of single plants with various
sensors under controlled conditions. (d) Control of fruit quality with e-nose (left) and spectral cameras (right).
230 Oerke
context of quarantine regulations, the detection of diseases and identification of the responsible
pathogens are crucial to prevent the entry of invasive species and eradicate primary sources of
inoculum. Protocols of the International Plant Protection Convention for the detection of plant
pathogens integrate phenotypic, serological, and molecular techniques; these methods provide
complementary information (76). Physical sensors provide the opportunity to use autonomous
systems for quarantine inspections, which may be a first step applied to bulk material to iden-
tify suspicious material that can be further tested and validated using molecular techniques for
pathogen identification.
In the food industry, postharvest sensing of the quality and safety of (processed) plant products
includes checking ripeness, color, and suitability for storage; detecting defects, bruises, and dis-
eases of fruits and vegetables; and assessing mycotoxin contamination from fungi, e.g., Aspergillus
flavus in maize kernels (154). Highly controlled conditions favor sensor use and high-throughput
platforms. The sorting/grading steps require instantaneous decision-making based on the sensor
data.
The germplasm of crops may be phenotyped for disease susceptibility under controlled con-
ditions and in field stations. Susceptibility to major diseases is only one criterion among many
considered by plant breeders, who must also consider yield, crop quality, and tolerance to abiotic
stress. Phenotyping systems require sophisticated equipment but provide high-throughput and of-
ten monitor the plants several times during the growing period without requiring instantaneous
data evaluation. In some cases, single plants are used; in later stages of resistance breeding, small
plots have to be assessed and the requirements may differ. When inoculating the pathogen(s) of
interest, disease identification is not needed, but disease quantification is crucial. Assessment of a
plant genotypes’ reaction(s) to pathogen attack (e.g., infection types of cereal rusts) considerably
increases the system specification.
In the field, the differentiation and quantification of several diseases are necessary for IPM
decision-making. Disease detection in the absence of knowing the causal pathogen and disease
severity is insufficient because some pathogens cannot be controlled, control of other diseases/
pathogens may require different fungicides, and control is cost-effective only when disease inten-
sity exceeds the action threshold. When growing high-value crops under controlled conditions,
sensor use is aided by good infrastructure and sensors (e.g., spider cam) and may generate nadir
images or images from a lateral position (ground-based system in row crops) close to the crops.
Urban/vertical farming provides similar recording conditions and specifications. For decision-
making in IPM of field crops, the information on diseases may be used (a) during the same growth
period (for polycyclic pathogens); (b) in subsequent growth period(s) (for soilborne pathogens, e.g.,
nematodes, take-all); and (c) for estimating, and in later stages quantifying, the yield losses caused
by diseases. Airborne remote sensing may cover large areas/regions, but ground truth information
is required.
Host–pathogen relationships may be investigated using unprecedented spatial resolution for
metabolic changes brought about by disease development, early stages of disease, the range of
disease effects on noncolonized tissue depending on host genotype, etc. Automated sensor systems
may also increase the temporal resolution of information acquisition suitable for understanding
temporal changes in host–pathogen interactions.
RECENT REVIEWS ON DISEASE SENSING AND DATA PROCESSING

As technical options for sensor recording of diseases are rapidly evolving, several reviews on dis-
ease sensing are available, focusing on crop protection (16, 74, 76, 88), plant phenotyping (26,
60, 82, 117), and both (71, 107, 158). Spectral information and 3D data from remote sensing are

likely to promote the availability of structural and physiological attributes of tree crowns damaged
by insect pests and fungal pathogens (125). Commonly used sensor systems include digital imag-
ing, chlorophyll fluorescence, spectral imaging, thermographic imaging, and detection of volatile
compounds. Magnetic resonance, soft X-ray imaging, and ultrasound are less used approaches.
Noninvasive techniques have great potential for postharvest quality analysis and inspection of
specialty crops (37, 103). More specific reviews from 2017 to 2019 are given below.
Although not the focus of this review, innovative methods of data processing and analysis that
contribute to the recent improvements in disease sensing are briefly summarized. Algorithms must
be able to cope with the resolution, size, and complexity of the sensor data (13). These techniques
may be grouped into (a) correlation and regression analysis of disease detection and severity as-
sessment; (b) use of general spectral vegetation indices (SVIs) or development of disease-specific
indices; (c) data-mining algorithms applied to data processing and feature extraction for data di-
mensionality reduction; and (d) machine learning (ML) and classification techniques (76). ML
encompasses statistical methods trained to identify patterns in complex data sets. The status of
ML use in agriculture, including disease detection, has been reviewed by Liakos et al. (63). Appli-
cations in plant–pathogen interactions range from disease monitoring to the prediction of molec-
ular pathogen effectors (122). Robust ML performance requires knowledge of both methodology
and biology. The strengths of ML include the capacity to handle data of high dimensionality and
map classes with very complex characteristics. Relatively mature methods for application include
support vector machines, single decision trees, random forests, boosted decision trees, artificial
neural networks, and k-nearest neighbors (78).
ML approaches train algorithms using a training data set to analyze and predict results from
new data. Deep learning (DL) refers to artificial neural networks with a structure containing many
layers that allow the automatic determination of image features by the network, e.g., convolutional
neural networks (67). These methods of artificial intelligence require intensive training [i.e., large
data sets (big data), labeling of training data] and high processing time (depending on the amount
of information, e.g., wavebands) and are prone to noise from specific wavebands. Hyperspectral
acquisition techniques may be combined with DL architectures to solve specific tasks in different
application fields, such as disease detection (116). Traditional learning-based approaches rely on
the extraction of handcrafted features to be used in a subsequent classification. DL enables an
automatic and hierarchical learning process, using data themselves able to build a model with high
semantic layers until a representation suitable to the classification or regression task is reached.
The requirements for the use of methods from artificial intelligence and potential pitfalls, e.g.,
sparsity of training data and lack or redundancy of relevant information (known as the curse of
dimensionality or the Hughes phenomenon) have been discussed (116, 122).
SENSORS FOR PLANT DISEASES

Sensors may be classified according to (a) the range of the electromagnetic spectrum, e.g., visible
(VIS), near-infrared (NIR), short-wave infrared, thermal infrared, and radar; (b) the scale/platform
used, e.g., remote sensu stricto, airborne and spaceborne, UAV, ground-based/proximal, and mi-
croscopic; (c) the recording principle, e.g., passive sensors record radiation emitted by an object
(thermography) or the reflectance of solar radiation (RGB, spectral cameras)—active sensors
[SAR (specific absorption rate), LIDAR (light detecting and ranging), fluorescence] emit special
measuring radiation and record its modification due to interactions with the object of interest—
and (d) the type of data recording, i.e., imaging or nonimaging (88). The lower the distance
between sensor, object, and radiation source, the stronger is the influence of the geometry be-
tween them. Active sensors are less affected by varying environmental factors but strongly depend
232 Oerke
on the geometry between the emitter/detector and object, and the runtime of the electromagnetic
radiation is twice that of a passive sensor. In addition to optical systems, sensors for VOCs are
also noninvasive, nondestructive, and passive.
Spectral Information
Spectral information ranging from 350 to 2,500 nm is recorded by radiation-sensitive detector
systems that typically split the incoming radiation into its spectral components. Nonimaging spec-
trometers often have a high spectral resolution (measuring hundreds of narrow wavebands sepa-
rately) in the full range, but the spectral information results from the average of the sensor’s field
of view. Imaging systems varying in spatial resolution from some hundreds to millions of pixels
per image may record (a) one waveband or the sum over the 400–700-nm spectrum (panchro-
matic); (b) the three basic color components red, green, and blue (typical bandwidth 60–80 nm,
e.g., smartphone RGB cameras); (c) additional (NIR) bands (multispectral, discrete, and somewhat
narrow wavebands); or (d) narrow spectral bands over a continuous spectral range (hyperspectral,
spectral resolution <1 nm). Hyperspectral data contain more spectral information than RGB, as
each pixel is a vector with the dimensionality of the number of wavebands recorded (typically
several hundreds).
The use of RGB images for disease perception profits from the omnipresence of the sensor;
everybody using a smartphone has a sensor system ready for use. The information from the three
broad wavebands in the visible range is often not sufficient for the differentiation of disease symp-
toms, but the combination with spatial information and the availability of advanced image pro-
cessing methods make RGB images a powerful tool in disease perception. Barbedo et al. (10) es-
tablished a plant disease database for automatic disease detection and identification that includes
2,326 images of 171 diseases and other disorders affecting 21 plant species. They increased the size
of the database by subdividing each image into parts displaying at least one symptom, increasing
the number of images to 46,513 suitable for DL.
The potential of hyperspectral sensors, especially hyperspectral imaging (HSI) for proximal
assessment of (diseased) plants has been recently reviewed (53, 73, 80). Thomas et al. (132) dis-
cussed effects of scale and environment on the recording and analysis of hyperspectral data on
plant diseases that affect the translation of this technology from controlled conditions into the
field. Supplemental Tables 1 and 2 summarize relevant information from 2016 to 2019 on the
use of RGB and multispectral images and nonimaging and imaging hyperspectral information,
respectively, for plant disease detection.
Differentiation between disease symptoms that may occur on a crop independently from each
other or simultaneously is essential for operational systems under conditions in which many bi-
otic and abiotic stresses can affect the same crop or plant product and the cause of the symptoms
is not known (e.g., crop production, quality control). The uniqueness of spectral signatures of
plant diseases is not universally agreed upon in the literature (86). Stress-causing agents, and like-
wise various pathogens, often cause similar symptoms and modifications in plant physiology (86,
123). Chloroses and necroses are caused by not only plant pests but also abiotic factors. However,
spatial patterns of primary symptoms on leaves or within the canopy may be included as infor-
mation for disease identification; for example, the combination of subtle spectral differences with
information on the size and uniformity of abiotic necroses on the one hand and the small-scale
spectral variability of resistance reactions to Plasmopara viticola on the other hand made it possi-
ble to distinguish between the two types of brownish grapevine tissue (91). The differentiation
between various tissue characteristics, e.g., healthy tissue, asymptomatic but latently infected tis-
sue, and disease symptoms, is even more important for assessing the presymptomatic host plant
colonization by pathogens.

Despite the temporal dynamics, disease symptoms and the appearance of plant tissue, e.g.,
coloration and dead versus living tissue, are less affected by environmental factors than are
the plant’s physiological processes, e.g., photosynthesis, transpiration, and water status. These
strongly depend on temperature, irradiation, and relative humidity and, thereby, are sensitive to
short-term changes in these factors. Because of this stability, plant appearance seems to be more
suitable for disease sensing than modifications of physiological traits recorded by other sensor
systems.
Thermography
Infrared thermography is used to assess the surface temperature of leaves, plants, or crop canopies
depending on their water status, in particular, stomatal and cuticular transpiration (22, 51). Radi-
ation emitted in the thermal infrared (8 to 14 μm) is detected by thermal sensors and displayed
in false-color images. Thermal imaging may be applied on various scales ranging from proximal
ground-based equipment to airborne and spaceborne sensors (54). It enables imaging of the tem-
perature pattern of plant tissue and crops over a short period of time and is suitable for time-series
measurements and monitoring purposes.
Thermal detection of plant diseases is limited to diseases affecting the plant’s water status, e.g.,
root infections, reduced water transport within stems, modified stomatal aperture, and changes in
cuticular conductance (94). Powdery mildew, for example, can be hardly detected under normal
growing conditions. As this passive method is highly sensitive to early changes in transpiration,
thermography may be suitable for presymptomatic detection of pathogen activities within plant
tissue (e.g., 72, 89). As various pathogens (and abiotic factors) affect the tissue temperature in a
very similar manner, thermography lacks diagnostic potential.
Fluorescence of Plants
Chlorophyll fluorescence is a noninvasive assessment of photosystem II activity that is highly
sensitive to abiotic and biotic factors (8, 81). Active sensors may be used to monitor pathogenesis
of diseases in the field, where passive sunlight-induced fluorescence may be recorded as well (1,
99). Pathogen attack affects the plant’s photosynthetic apparatus, e.g., pigments, electron transport
chain, and enzymes of the Calvin cycle, directly by reducing the photosynthetic leaf area (necrosis)
and chlorophyll degradation (chlorosis) or indirectly through feedback regulation of the electron
transport chain. Specific information on the mechanism(s) affecting the photosynthetic electron
transport is available only under conditions allowing for predarkening of plants. The combination
of pulse–amplitude–modulation chlorophyll fluorescence systems using dark-adapted plants and
image analysis was suitable for quantifying the affected leaf area for high-throughput phenotyping
(101, 102).
Fluorescence spectra were useful to discriminate brown rust–infected tissue from healthy wheat
tissue as early as four days after inoculation (134). The fluorescence quotients were correlated
with polyphenol content and relative amount of fungal DNA detected in leaves. The intensity
and time course of fluorescence spectra induced by various light sources (LED, laser, sunlight)
may be recorded under controlled conditions as well as in the field (15). Similarly, the reactions
of barley genotypes differing in resistance to Blumeria graminis f. sp. hordei could be identified
by analyzing the quantum yield of photosystem II and nonphotochemical quenching (18). As the
photosynthetic electron transport is similarly affected by biotic and abiotic factors, the method
largely lacks diagnostic potential. The patterns of disease symptoms on the leaf and plant level are
often random and may be confused with effects due to arthropod damage.
234 Oerke
Electronic Nose
Plants emit VOCs when affected by diseases and arthropod pests (19). VOCs are also emitted by
healthy plants and play a role in growth, communication, defense, and survival (9). As abiotic stress
factors also modify VOC emission, a multitude of chemicals may be released under normal atmo-
spheric conditions and the composition of this cocktail is changed, sometimes with characteristic
successions. E-noses are able to detect (some of ) these VOCs; gas chromatographic headspace
analytics or customized commercial e-noses (e-sensing) are available for special applications. Al-
though often unspecific because plants may emit similar VOCs upon induction by different dis-
eases and different plant species may emit similar VOCs, plant-emitted volatiles may be used for
the detection and differentiation of plant diseases (and other pests and physical damage). The
blend of VOCs may be specific for plant–parasite interactions (24).
The development and application of e-noses in agriculture, horticulture, and forestry have
been reviewed (23, 61, 149, 150). Operational simplicity, nondestructivity, and bulk sampling are
advantages; drawbacks include a low sensitivity and specificity in comparison with microbiological
and molecular methods (20). Compared to the magnitude of sensitivity and specificity of optical
systems, these drawbacks are irrelevant. The following methods of data processing and analysis are
quite similar to those used for other sensors, e.g., DL methods. Phenylethyl alcohol, α-copaene,
and fluoro-aliphatic hydrocarbons have been identified as disease-specific VOCs of fire blight of
apple, gray mold of tomato, and powdery mildew of tomato, respectively (150). Headspace analyses
of VOCs differentiated basal stem rot–infected and healthy palm wood under lab conditions (44). A
smartphone-based plant VOC profiling platform has been created for accurate and early detection
of late blight in tomato leaves in the field (62).
E-noses are used in monitoring food quality and production processes and to detect posthar-
vest diseases of fruits and vegetables (Supplemental Table 3). E-sensing does not deliver spatial
information; however, it may be sensitive to latent infections and identify various stages of the
disease. Provided a high sensitivity under field conditions, future e-noses may guide the sensor
system (together with an actuator for pesticide application) to the original source of the VOCs in
the field by following the gradient of chemicals.
Sensing of Other Information

Some other technologies have been successfully used, especially but not exclusively, for research
purposes; some of them have also been tested for their potential in applications under field condi-
tions or for high-throughput purposes. These techniques often need expensive equipment, trained
personnel, extensive sample preparation, and controlled environmental conditions.
Nuclear magnetic resonance (NMR) and X-ray imaging techniques may be used for detecting
plant diseases and plant products (107). NMR imaging was used to investigate internal bruising
and Spraing disease symptoms in potato tubers (133) and differentiate belowground damage of
sugar beets affected by Heterodera schachtii and Rhizoctonia solani (43). Navakar et al. (84) used X-ray
imaging to identify fungal infections in wheat. Microfocus X-ray fluorescence has been tested for
basal stem rot detection in oil palms (79).
Sonic tomography distinguished five classes of basal stem rot severity of oil palms with
96% accuracy (48). Optical coherence tomography identified an additional subsurface layer of
melon seeds infected with Cucumber green mottle mosaic virus (58). Biophotonic inspection using a
1,310-nm swept-source, optical-coherence tomograph was suitable for identifying morphological
differences between healthy leaves and leaves with circular leaf spots on persimmon (148).
Biospeckle is a nondestructive method for monitoring the quality or vitality of agricultural crops
and can detect mechanical and disease defects of fruits (3, 157).

Sensor Fusion
The combination of information from various sensors may improve disease sensing; however, it
often suffers from differences in, e.g., spatial resolution and viewing angle. The combination of in-
frared thermography, chlorophyll fluorescence, and HSI in sensing Fusarium head blight of wheat
ears improved the differentiation of noninoculated and infected spikelets from <79% to 89% (72).
The addition of leaf spectroscopy of needles to intra-canopy distribution of LIDAR returns in-
creased the accuracy of detecting red-band needle blight from 80.9% to 96.7% (121). The benefit
from additional relevant information has to be balanced with higher operating expenditures and
extra time. The combination of several simple sensors is likely to be more successful than the com-
bination of a system producing a high amount of information with one or two sensors delivering
a single piece of information each.
SPATIAL RESOLUTION AND SENSOR PLATFORMS

Sensors of electromagnetic radiation may be used for disease measurement on various platforms,
e.g., satellite, airplane, helicopter, UAV, and machine-mounted, handheld, and autonomous sys-
tems. The distance between the sensor and object determines the spatial resolution of the infor-
mation and its usability for damage recording, disease assessment, and basic research. Proximal
remote sensing may be defined as recording and classifying data with an imaging sensor within a
distance <1 m (often much less) from the object (83). Airborne or spaceborne systems, i.e., clas-
sical remote sensing, have lower spatial resolution than proximal sensor systems but cover larger
areas of the landscape. Spatial resolution is defined by the size of a pixel in the object plane and,
hence, determines the smallest identifiable disease symptom. Using pixels that are larger than the
symptom results in mixed information, which is unsuitable for the detection of significant spectral
deviations.
For sensory measurement of diseases, sensitivity and specificity of the sensor technique/signal
are crucial. Noninfectious diseases caused by abiotic factors affect all shoot parts, and symptoms
may be assessed at the plant or canopy level from larger distances. Primary infections of plants
by pathogens, in contrast, are restricted to specific plant parts and begin at discrete sites. As small
symptoms are only a minor proportion of the plant tissue under inspection, the signal from a
nonimaging sensor or sensor with low spatial resolution is dominated by healthy tissue (and back-
ground), thus limiting the system’s sensitivity. Reliable spectral differences can be detected only at
higher overall disease severities. Using a spectrometer, diseased grapevine leaves could be differ-
entiated from healthy tissue only at downy mildew severities of >25% of the leaf area; in contrast,
in hyperspectral images, individual spots (2-mm diameter) of the sporulating pathogen could be
detected (91). Detection of low disease severity (an early stage in epidemics) requires a high spatial
resolution of the sensor system. Individual disease symptoms can be detected only when the pixel
size of the sensor is considerably lower than the size of the targeted symptom. If the symptom
is circular and camera pixels are square, spatial resolution should be better than one-third of the
symptom’s diameter (91).
Soilborne diseases of the root system may start in discrete parts; however, they only become vis-
ible and detectable on the shoot when a reduced supply of water and nutrients affects the complete
shoot (an exception is Verticillium wilt causing wilt only of plant parts connected to the infected
vessels). However, it is often hard to distinguish between wilting induced by pathogens and that
induced by water shortage. The spatial resolution for the detection of the primary infection sites of
airborne diseases such as powdery mildew, rust, downy mildew, and leaf spots must be significantly
better than for root-infecting pathogens or drought (88).
236 Oerke
Imaging sensors on satellites and airplanes have a spatial resolution of 1 to 100 m and 0.01 to
1 m, respectively, and are not suitable for the detection of early disease symptoms. UAVs flying at
an altitude of 10 m to 120 m have ground sampling distance (GSD) of 0.02 m to 0.25 m (21, 77).
Disease symptoms in the range of 0.001 m require a submillimeter spatial resolution (<0.0003
m) currently provided only by ground-based imagery. The level of disease (number of symptoms,
percent leaf area diseased) in plant tissue that can be differentiated from healthy plant tissue
using a sensor technique is crucial and determines the purpose(s) for which this technique may
be used. Unfortunately, information on disease sensitivity of sensor systems is very limited in the
literature (e.g., 28, 105, 126, 131).
Satellites and Aircrafts

Spatial requirements limit the suitability of airborne and spaceborne imagers and nonimaging
techniques to the detection and quantification of disease-induced crop damage in later stages of
epidemics. Satellite imagery was used to study the incidence and spread of Asian soybean rust and
Cercospora leaf blight of peanut, respectively (86). Using PlanetScope data sets (four bands with
a GSD of 3 m), the development of rice dwarf, blast mand glume blight and changes in leaf area,
pigment contents, and canopy morphology could be captured using broadband SVIs (112). Classi-
fication of rice diseases at a subfield scale had a kappa value of 0.47. The severity of Phytophthora
root rot of avocado trees, associated with reduced uptake of water and nutrients, canopy decline,
and defoliation was assessed by a WorldView-3 satellite image (GSD 1.2 m, canopy size per tree
>12.6 m2 ) with eight spectral bands (106). The classification accuracy of basal stem rot severities
of oil palms from WorldView-3 imagery and supervised learning algorithms was low (108).
Requirements for sensing (of diseases) in forestry are different from those in agriculture.
Because of the cultivation of long-living plants reaching considerable heights over large areas, the
use of airborne and satellite-borne sensors is common. Hyperspectral sensors on aircraft reach a
spatial resolution of 2–4 m, adequate for trees with crowns spanning several meters in diameter
(125). Ground-based sensors do not provide the level of performance needed for large-scale
inspections of forests. Remote sensing has the potential to aid in the detection of infected stands
and monitoring disease development and spread (120). Spectral unmixing of yearly Landsat
time-series and species distribution modeling revealed an annual increase of 7% in sudden oak
death in California from 2005 to 2016 (40). The control of forest tree diseases is often not
feasible.
Unmanned Aerial Vehicles

The potential of UAV-based spectral imaging, data processing, and applications for agriculture and
forestry has been reviewed (4). UAVs may carry lightweight sensors and are available for farmers
and researchers (in contrast to manned aircraft and satellites, which are available only after long
planning and are often expensive). Also, compact hyperspectral sensors may be integrated into
UAVs (4). Remote sensing with UAVs offers unprecedented spectral, spatial, and temporal resolu-
tion (70). Compared to drought-stress detection, the detection of diseases with UAVs is relatively
premature, but the fusion of thermal and hyperspectral data shows great potential. Thermal and
RGB imagery was successfully used for the detection of mistletoe-infected Eucalyptus species in
New South Wales, Australia. In a UAV flying 60 m above the ground with a tree height of 20 m,
i.e., a distance between sensor and top of the crown of approximately 40 m, the spatial resolution
was 1.5 cm and 10 cm for RGB and thermal cameras, respectively. In multispectral imaging for the
detection of citrus greening, the higher spatial resolution of a UAV-based sensor resulted in higher

classification accuracy compared to an airplane-mounted sensor (31). Laurel wilt of avocado and
flavescence dorée and esca of grapevine, which are all diseases that affect the entire plant, could
be detected with high accuracy by using UAV-borne sensors (2, 5, 25).
Ground-Based Systems
Technological advances in machine vision, global positioning systems, laser technologies, actu-
ators, and mechatronics have enabled the development and implementation of robotic systems
and intelligent technologies in agriculture, urban farming, and forestry (6). Sensing methods and
technologies for identifying plant disease symptoms have been successfully developed; still, the
majority of them require a controlled environment for data acquisition to avoid false positives.
An autonomous ground-vehicle robot developed for high-throughput in-field agricultural
row-crop phenotyping was used to measure plant height and canopy closure data by normal-
ized difference vegetation index (NDVI) assessment via HSI (138). The spatial resolution of the
ladybird system was 3 × 12 mm indicating that the system was able to assess the damaging effect
of plant diseases but not primary disease symptoms.
AREAS OF SENSOR APPLICATION

Quality Control
Several noninvasive and nondestructive techniques have been tested for their potential to monitor
or inspect plant tissue for fungal contamination of plant products like fruits, vegetables, bulbs, and
cereals (115). Soft X-ray imaging tested to detect fungal infections in corn and wheat is costly
and includes the dangers inherent to X-ray radiation. Thermal imaging lacks the potential to
differentiate between fungal species causing diseases and contamination of plant material.
HSI is successfully used in quality assessment of fruits, especially their physiological status.
Strawberry fruits diseased by Botrytis cinerea and Colletotrichum acutatum could be identified with
97% classification accuracy for each cultivar separately (115). Six of fifty-four features from HSI
were used to differentiate peach diseases caused by B. cinerea, Rhizopus stolonifera, and C. acutatum
in industrial application (128). A deep belief network model based on integrated information (494
features) gave the best classification results for the three diseases, with 82.5%, 92.5%, and 100%
accuracy for slightly, moderately, and severely decayed fruits, respectively. NIR spectroscopy was
useful to discriminate Fusarium-contaminated and healthy barley kernels (64, 65).
E-sensing allowed the discrimination of strawberry fruits infected by three postharvest fungi
at an early storage stage (95) and the detection of oranges colonized by Penicillium digitatum for
24 h (34). E-nose devices often have low operating costs and high ease of operation and provide
rapid results and response times, good precision, portability, flexibility, and high adaptability for
specialized applications (150). However, sensor drift, especially from temperature and humidity
effects, currently limits data accuracy and reliability (35).
Sensing may be conducted in a highly standardized, automated setup, with the number of (ma-
jor) diseases to be detected limited, and the material may be manipulated to optimize recording
conditions. Nevertheless, the demand for high sensitivity of the sensor (for the detection of low
disease levels) and very fast decision-making on graded objects makes disease sensing challenging
(Table 2).
Phenotyping
In plant breeding, plants are phenotyped in several thousand small plots to evaluate them for
various traits. High-throughput phenotyping platforms include environmental sensor networks,
238 Oerke
Table 2 Requirements for remote disease sensing under various application situationsa
Disease assessment
Crop Effects of Processing Post-sensing
Situation complexity environments time activities
Detection Identification Quantification
Quality Diverse fruits, − +++ (+) (+−) Short Grading,
control vegetables sorting
Quarantine Various plant (+) +++ +++ − Short to Separating
material (→PCR) medium
Phenotyping Individuals + +++ − +++ Long AUDPC,
indoor/ discussion
field
Plots ++ +++ (+) +++
Greenhouse Canopies, ++ +++ +++ +++ Short to Control online
lateral view medium or offline
Field 3D canopies +++ +++ +++ +++ Very short Decision
making,
spraying
Row crops +++ +++ +++ +++ Very short Decision
making,
spraying
Perennials ++ +++ +++ +++ Short Decision
making
Abbreviations: AUDPC, Area under disease progress curve; PCR, polymerase chain reaction.
a
+++, ++, +, and − denote significance as very high, high, low, and nonexistent/not necessary, respectively; parentheses indicate potential exceptions.
autonomous ground vehicles, phenomobiles, towers, field scanning platforms, UAVs, aircraft, and
satellites (110). At present, the phenotyping of crop genotypes focuses on recording plant height,
biomass production, and yield under abiotic conditions that affect all shoot parts (30, 41, 159).
Phenotyping for disease resistance is more challenging, as disease symptoms often occur as
tiny spots on some plant parts. Requirements for data on infection type, disease severity, and epi-
demic development (e.g., effects of partial resistance on sporulation) make the task even more
challenging. First approaches may be run under controlled conditions; sometimes the plants are
transported to the sensor system, which may include several sensor types and options to optimize
measuring geometry. The growth stage of the plant and the pathogen and the disease stage (i.e.,
time after inoculation) are known, and the objective is to identify differences between genotypes.
Plot experiments in breeding gardens may be possible because of support from sensors attached
to spiderweb cameras. In some programs, the assessment of disease effects on crop yield is most
important. The sensitivity of the sensors may not enable spotting initial disease symptoms and
decisions on disease control are not necessary, but monitoring and documenting disease develop-
ment over time are crucial.
A UAV equipped with a multispectral camera with a spatial resolution of 0.06 m at a 150-m
height of flight (155) was suitable to detect soil-N heterogeneity and maize performance. Thomas
et al. (131) used HSI for phenotyping the reaction of six barley genotypes to B. graminis f. sp.
hordei in miniplots under greenhouse conditions. Powdery mildew was detectable when symptoms
appeared. Grouping into genotypes with low (<5% disease severity), moderate (5 to 20%), and
severe (>20%) powdery mildew was possible; however, the baseline of the sensor system indicated
1% to 2% disease for inoculated plots at 0 dpi as well as for noninoculated plants throughout the
experiment. The authors discussed the problems, e.g., lighting, shading, 3D structure of cereal
canopies, specular reflectance of powdery mildew symptoms, and segments of the twisted leaves,
associated with measuring crop canopies instead of leaves.

Monitoring of Crops for Diseases
Sensing of plants for precision disease control is done in large fields or greenhouses where the
aim is to detect the occurrence of diseases at the early stages of epidemics, i.e., at low symp-
tom frequency. Lowe et al. (67) reviewed (hyperspectral) imaging of plant diseases, focusing on
early detection of diseases for crop monitoring. Techniques for the detection of biotic and abiotic
stresses differ in the level of accuracy. Remote sensing of diseases under production conditions is
challenging because of variable environmental factors and crop-intrinsic characteristics, e.g., 3D
architecture, various growth stages, variety of diseases that may occur simultaneously, and the high
sensitivity required to reliably perceive low disease levels suitable for decision-making in disease
control. The use of less sensitive systems may be restricted to the assessment of crop damage and
yield losses due to diseases.
Greenhouse crops. Plant species ranging from various kinds of vegetables to ornamentals and
fruit-producing species are grown, in rows or as canopies, under protected conditions. The con-
trolled environment, limited area, and arrangement of plants in rows, groups, etc., facilitate the use
of sensors for disease perception. Furthermore, the production of high-value crops makes moni-
toring for diseases by sensors more cost-effective than in extensive field crops. Proximal sensing
from spider cams, rail system–mounted sensors, and autonomous vehicles may be used for top
and lateral views. Sensors may be used for direct diagnostic purposes or as an early warning sys-
tem to spot anomalies in plant appearance followed by exact disease diagnosis using molecular
techniques. In an additional step, control options can include site-specific spraying or removal
of diseased plants. Similar conditions are expected for urban (i.e., vertical) farming, where crop
monitoring for diseases and pests is essential for intensive crop production. Technical (sensor)
equipment should be available.
Field crops. In field crops, diseases suitable for detection by sensors start in patches, are caused
by polycyclic pathogens, have action thresholds considerably higher than the lowest detectable
disease level of the sensor, are of economic significance, and may be controlled efficiently by
fungicides.
Perennial crops like grapevine, fruit trees (citrus, pome fruits), and plantation crops (e.g., coffee,
banana, palm trees) are often grown in rows and individual plants may have a high value, making
knowledge of the health status highly relevant. Depending on the plant species, the height and
3D architecture of the plants make disease sensing a challenge. Sensing of diseases with action
threshold level 0, e.g., apple scab and downy mildew of grapevine, is less useful than sensing of
diseases affecting roots and stems or diseases caused by invasive pathogens like Xylella fastidiosa
(156) or other pathogens of wood crops (2, 5, 25). Monitoring systems on different scales that
provide information for map-based control options, i.e., separating decision-making from sensor
information and subsequent control activities, may be available. The detection of infected plants
allows removal as early as possible to prevent further spread of the pathogen.
In annual crops, decision-making for disease control from disease sensing has to be very rapid
when working in combination with spray equipment. Sensor systems described in the literature,
however, are more suitable for damage/yield loss assessment than for early disease detection and
identification because of low sensitivity/lack of information on sensitivity. As the 3D architec-
ture of canopies exacerbates the sensing of disease symptoms, sugar beet and other broad-leaved
dicots are preferred crops for proximal disease identification compared with cereals for which the
orientation of leaves is much more complex. The 3D architecture is also important for differences
in disease patterns within plants; diseases from airborne inoculum may begin on the top leaves,
240 Oerke
whereas the first symptoms of soilborne pathogens are often restricted to stems and lower leaves,
which are hardly visible in cereal crops at stem elongation and later stages. Data acquisition is
additionally hampered by environmental factors, especially changes in irradiation (important for
passive sensors) and breezes causing movement of plants (131).
The use of airborne and satellite-borne imagery for the detection of damage caused by plant-
parasitic nematodes and other soilborne pathogens for management optimization is successful
because (a) impaired functionality of infested roots becomes visible in the foliage, (b) infestations
are often clustered in the field, (c) movement out of a cluster is slow because of low pathogen
mobility, (d) introduction of new infection loci into a field is rare, and (e) maps from one season can
be applicable for future crops (42, 88). Although several sensors of plant (and soil) characteristics
may be suitable for nematode detection, the identification of the causal agents requires ground-
truth information on the species involved.
For wilt diseases, symptoms often affect the complete plant and are very similar to physiolog-
ical wilting from drought. At least two measurements during the day are suitable to distinguish
between transient wilting from water shortage during high irradiation periods and permanent
wilting due to pathogen attack of roots. As chemical control of soilborne pathogens is largely
restricted because of regulations, the resulting information may be used for removal of affected
individuals during the growing period as well as for managing the cultivation in the next
season.
Detection of emerging diseases during the period of latency seems to be the current Holy Grail
of disease sensing. Detection of disease as early as possible is desirable because control is more ef-
fective and less active ingredient is used when there is a smaller amount of pathogen biomass.
Nevertheless, detection should be combined with disease identification to select an appropriate
control agent. For presymptomatic sensing, other reference data/spectra on nonvisible changes in
structure and physiology of plant tissue are required (in addition to those for typical symptoms).
The perception of latent infections is most interesting for diseases in which the pathogens’ spores
on the plant surface represent the typical symptoms, e.g., powdery and downy mildews, and rusts
caused by polycyclic biotrophic pathogens with short generation time and high spore production
needed for epidemic spread. During latency, pathogen spread is not possible and control should
be highly effective. For most necrotrophs, sporulation on dead tissue additionally requires suit-
able environmental conditions and pathogen spread depends on the presence of visible symptoms.
However, (hemibiotrophic) pathogens, e.g., Venturia inaequalis on apple leaves, may produce coni-
dia before visible symptoms occur (89).
Asymptomatic olive trees identified as infected by spectral plant-trait alterations developed
X. fastidiosa symptoms at almost double the rate of asymptomatic trees classified as not affected
by remote sensing with HSI and thermal cameras (156). Modeling the epidemic spread of wheat
stripe rust, Severns et al. (109) concluded that the timing of control is far more important than
the size of the treatment area. Even treatments of asymptomatic plants around an outbreak have
only a small effect on the spread of pathogens such as Puccinia striiformis.
In the literature, early detection at the presymptomatic stage of infection is often mixed with
the detection of symptoms at a low disease level, i.e., during the early stages of epidemics. For
presymptomatic detection, the recorded sensor signal should depend on the latent tissue colo-
nization, not on the presence of potentially large amounts of inoculum on the plant surface. From
a theoretical point of view, detection of infection stages on the plant surface that may not con-
tribute to the successful establishment of a functional host–pathogen interaction may result in
overrating disease intensity.
Latently infected plants surrounding visually diseased plants should affect which crop areas
must be sprayed with fungicides (142). A map-based fungicide application requires two passages

through the field and is generally regarded as too time-consuming. The development of online
systems combining disease sensing and spray application within a single passage has increased the
issue’s significance for practical fungicide application. For field crops like cereals, online systems
are only conceivable with the assessment of all information, including asymptomatic colonization,
on the status of all plants (as well as spray equipment with relevant application options). For green-
house crops, orchards, and plantations with high-value plants, the application of fungicide sprays
or the removal of plants infected by otherwise noncontrollable pathogens (e.g., wilt diseases, in-
vasive species) in a second operation seems to be acceptable. The assessment of soilborne diseases
may support management decisions in the next growth period.
Analysis of the potential profitability of targeted disease control in vineyards and apple or-
chards in central and southern Europe showed that a larger area provided a larger benefit and
lowered the effect of the price of the required equipment (135). The authors concluded that, al-
though canopy-optimized technologies may be profitable, profitability cannot be the only driver
for adopting robotic platforms for precision spraying in specialty crops. Evaluation of the to-
tal environmental costs of crop production processes, higher acceptance of low-pesticide crop
production systems by consumers/politicians, and evolving crop protection standards (e.g., the
amount of active ingredient applied per crop area) imposed by authorities and retail markets may
also affect the adoption of state-of-the-art technologies. Sensor-controlled variable-rate spraying
technology, which uses the level of green crop coverage as a parameter for adapting spray volume
to the site-specific biomass, has the potential to reduce the spray volume per area (129). Savings
ranged from 44% to 1% depending on the heterogeneity of the crop.
Sensing of Pathogens and the Environment

The action threshold for some diseases may be close to zero; for other diseases, the relationship
between environmental conditions and infection success of inoculum/disease occurrence is well
established. Especially in these cases, the sensing of the pathogen and environmental conditions
is a very promising tool to support decision-making regarding fungicide use. Pathogen inoculum
on the plant surface, in the air, and in the soil may be detected and quantified by a combination of
spore traps and polymerase chain reaction (PCR) (143, 145). This combination may give a rapid
warning of inoculum presence and is suitable to monitor changes in genetic traits of pathogen
populations, e.g., the race structure or frequency of fungicide resistance (144). The information
on environmental conditions from distant weather stations may be interpolated by GIS to generate
virtual weather stations on a small scale to determine the parameters, e.g., leaf wetness duration,
necessary for disease forecasting and decision-support systems (52, 100). Similarly, the extension
of information on leaf wetness of apple plants from areas with sensors to uncovered areas by using
ML algorithms on simple weather parameters permitted a risk evaluation of apple scab infection
for each orchard (124, 152).
Spectral information may also be used to identify pathogens and virus vectors. A nonlinear
support vector classifier used on hyperspectral data recorded from immobilized Bemisia tabaci, the
vector of important viral diseases of cassava, resulted in an accuracy of up to 98% when differenti-
ating virus-transmitting subspecies from other subspecies that are morphologically indistinguish-
able to human observers (27). Mixed cultures of soilborne phytopathogenic fungi (Colletotrichum,
Verticillium, Rhizoctonia, Fusarium) could be differentiated by using infrared microscopy (46).
Sensing of Host–Pathogen Interactions

Research represents the most diverse area of remote disease sensing. Sensors may be customized
to meet specific requirements; prototypes can be tested for potential applications. The same
242 Oerke
applies to methods of data processing often adopted from other disciplines. Spatial reference
points were used to investigate the temporal development of leaf rust and Septoria leaf blotch
on wheat leaves under laboratory conditions (11). This kind of geo-referencing was suitable to
trace typical disease symptoms back in time to early symptoms and sites of latent infections. The
progression of stripe rust on detached wheat leaves was assessed by using hyperspectral data sets
from a spectroradiometer and an imager (111).
Whetton et al. (146, 147) were among the first to use HSI for online mapping of the spa-
tial distribution of plant diseases at the field level. With disease coverages of <5%, stripe rust
and Fusarium head blight of wheat and barley could be predicted from spectral information with
moderate to good accuracy (R2 = 0.61–0.82); the derived high-resolution maps confirmed the
highest disease coverage at the field edges. A comparison of methods for fire blight detection in
apple showed that RGB and multispectral sensors provided low to moderate disease-detection
accuracy, whereas normalized SVIs from hyperspectral data resulted in moderate to high accura-
cies in disease severity evaluation, proving the potential of HSI as an early disease detection tool
(50).
NIR-spectroscopy transmittance measurements on intact onions and onions internally rotten
because of infection by Botrytis and Pseudomonas used multilaser light and revealed significant dif-
ferences in the light propagation properties (127). Attenuated total reflectance–Fourier transform
infrared spectroscopy (ATR–FTIR) with subsequent multivariate analysis nondestructively iden-
tified plant–pathogen interactions during disease progression through alterations in the spectral
fingerprint (119). ATR–FTIR fingerprint spectra of healthy, damaged, or sour rot–infected tomato
fruit indicated dynamic effects of the disease on the biochemical composition of the epidermis.
Detection of the causal agent Geotrichum candidum was based on spectral features distinct from
those of the tomato fruits.
High spatial resolution of imaging sensors enables time-series measurements of host–pathogen
interactions at the tissue scale. HSI validated a spatial continuum from healthy tissue, via slightly
affected zones, to the sites of primary infection for three sugar beet diseases (74) and downy mildew
of grapevine (91). These images may be used to visualize and quantify pathogen effects on host
metabolism with high spatial resolution, contributing to the understanding of tolerance mecha-
nisms in host genotypes. The progression of changes in plant metabolism and disease symptoms
can be visualized via spectral imaging, e.g., to estimate the age of Cercospora leaf spot symp-
toms (74) and the sporulating activity of this pathogen within symptomatic tissue (92). The high
information content of hyperspectral data (using hundreds of wavebands to specify a spectrum
compared to the few wavebands used for RGB) may contribute to deepening the understanding
of host–pathogen interactions (59, 141). The first attempts to link the activities of resistance genes
and enzymes to changes in spectral characteristics were published for the barley–B. graminis f. sp.
hordei interaction (55, 56).
The crossroads in the metro map approach of Wahabzada et al. (141), i.e., using different dis-
eases that exhibit highly similar spectral information, indicate that disease stages may not have
unique fingerprints, at least not at all stages. A complex of pathogens and plant tissues resulted
in similar spectral information, despite involving different pathogens. The full-range spectra of
brown rust and net blotch of barley, however, were different because of the divergent effects of
biotroph and necrotroph pathogens on plant tissue structure. Plants confronted with pathogenic
microorganisms have only a limited number of cellular responses available. This can hamper dis-
ease identification in practice. Nevertheless, subtle spectral differences between similar but dif-
ferent symptoms may be resolved by more information, e.g., metadata on size and texture. Some
reports seem to indicate that insufficient and randomly sampled spectral data on diseases may be
compensated for by elaborate data processing This may be true (and is very welcome) to some

extent; nevertheless, some reports do not demonstrate the general applicability of the developed
models.
FUTURE CHALLENGES
Remote-sensing techniques differ in their potential to identify plant diseases. Chlorophyll fluo-
rescence and thermography, although highly sensitive to changes in plant metabolism incited by
pathogens, lack the potential to identify diseases and differentiate them from abiotic symptoms
and effects from arthropod activities (93). Spectral information, in combination with spatial infor-
mation from images and information on VOCs emitted from affected plants, seems to be suitable
for disease identification or categorization. Nevertheless, thermography and fluorescence may be
used in crop monitoring for anomaly detection followed by an inspection of suspected plants or
areas. Additional sensor types can help increase the sensitivity and diagnostic capacity of platforms,
as physical limitations, e.g., spatial resolution from airplanes and UAVs, persist.
Information on the spatial patterns of major diseases in field-grown crops is urgently needed.
Sensor data from systems adequate for field monitoring should enable the identification of plant
diseases/disease stages with spatial heterogeneity appropriate for optimized applications of control
activities over time and space. Subsequently, lower-cost sensors highly sensitive to these diseases
have to be developed for site-specific control approaches.
Detection of a disease is not sufficient for crop protection. More studies are needed to demon-
strate the potential of sensors to differentiate among the diseases relevant to the crop of inter-
est or identify the diseases involved in mixed infections. As powdery mildew and downy mildew
produce similar symptoms on the same crop (although often on different sides of a leaf ), the
sensory differentiation between these diseases is likely to be challenging even in cases in which
there are metadata on weather conditions, identification of leaf sides, etc. Only a few studies
in the literature include information on the minimum detectable disease level. This informa-
tion, however, is crucial for objective evaluation of the results and for comparisons with other
studies.
The visualization of foot diseases in cereals is problematic, as obvious shoot symptoms become
visible only in later stages of disease development. In which ways can the geometry of the sen-
sor and the inspected plants be improved? Is it necessary and feasible to bring sensors closer to
the plant (part) under observation to deliver information relevant to disease management? Other
approaches may combine spectral information with LIDAR techniques useful for the assessment
of the 3D architecture of crops (12, 97). Nevertheless, the limited penetration depth of spectral
information may necessitate other supporting techniques, e.g., sensors operating within the crops
or transient opening of the 3D ground cover during sensor passage using mechanical devices to
split the cereal canopy.
Disease sensor systems have to be robust to several variables, including variability of (a) the
host plant tissue, e.g., greenness or waxiness of leaf tissue; (b) the pathogen genotypes causing
different host reactions, e.g., to rust fungi; (c) disease symptoms, e.g., downy mildew of grapevine
causing oil spots and whitish sporulation on the adaxial and abaxial leaf sides, respectively, and
different symptoms on different plant parts, e.g., rice blast on leaves, necks, and panicles; and
(d) environmental conditions. New diseases in the field may require the development of new sen-
sors, or existing systems must be capable of being modified to receive novel signals (e.g., new
spectra or VOC patterns).
The use of SVIs generated for vegetation classification and assessment of, e.g., plant biomass
and photosynthetic active pigments, seems to be less productive for disease sensing, as they
are nonspecific and ignore the majority of spectral information. Diseases may be detected by
244 Oerke
using these indices; however, it is the nonspecific effects of diseases on the physiological sta-
tus of plants that are being assessed. Disease-specific SVIs have been developed for diseases in
sugar beet (75), wheat (7, 45), and grapevine (91) and proved superior to SVIs developed for satel-
lite data. However, it takes time to produce specific indices for all diseases/disease stages of in-
terest, as, ideally, the specificity must also be tested against rare diseases occurring on the crop
species.
The potential of sensors to detect plant diseases prior to the appearance of visible symptoms
has been highlighted (e.g., 71, 98). The biochemical and physical changes within the plant tissue
that underlie the sensor detection are not known. The detected signal(s) should be derived from
infected plant tissue and not from the inoculum on the plant surface. Is it possible to differenti-
ate between sensory fingerprints of latent infections from different pathogens or to differentiate
latent pathogen infection from endophytic colonization of plant tissue by organisms used for bi-
ological control purposes? Systemic but asymptomatic colonization of the milk-thistle weed by
the biocontrol smut fungus Microbotryum silybum could be diagnosed in the field with an accuracy
greater than 90% by using a VIS–NIR spectrometer and supervised hierarchical self-organizing
models (96, 130).
Approaches for disease perception have to be tested on independent test data sets to generalize
the analysis procedures (e.g., data and algorithms). Big data or artificial intelligence may solve a
specific technical problem but does not guarantee a gain in knowledge with respect to the bio-
logical system. Underlining the impact of multidisciplinarity on the success of disease perception,
relevant biological phenomena have to be translated into algorithms for image processing (38).
Disease sensing integrates plant science, engineering, and computation. The use of best-fit mod-
els identifies the best technical solution, irrespective of whether there is a physical relationship
between biological phenomenon and sensor signal, and may indicate spurious correlations (e.g.,
water content and VIS–NIR range, peroxidase activity, and best wavelengths).
Diseased is considered to be the exception rather than the normal crop status. In the early
stages of epidemics, symptoms are rare and diseased plants are hard to detect. Anomaly detection
algorithms are used in environmental sciences for remote detection of targets with low probabil-
ities or significant changes occurring with respect to a previous situation and do not need prior
information about target objects (116). Procedures for change detection are likely to be useful in
regular disease monitoring, in contrast and in addition to the detection of disease in a single image.
In the first screen of an agricultural field, deviations from the normal status may draw attention
to plants or clusters of plants that need more detailed investigations. The definition of a healthy
index for sugar beet was the first trial to use this approach (104).
Sensing is a tool for disease assessment, not the prerequisite or the active control mechanism for
disease management. Sensors cannot replace the application of fungicides and mechanical devices
for disease control, but they may guide (and focus) the actuator(s) to the plants or areas requiring
a specific activity and thus help to reduce the amount of chemical applied. Sensing of polycyclic
diseases is reasonable only when effective curative control options, i.e., systemic fungicides, are
available for control. Sensors may be used for estimating the production losses due to disease, to
decide which plants have to be removed from the cultivated area, or to determine which areas of
the field must be left uncultivated in the next growing season because of the presence of soilborne
pathogens that cannot be controlled by means other than crop rotation.
DISCLOSURE STATEMENT
The author is not aware of any affiliations, memberships, funding, or financial holdings that might
be perceived as affecting the objectivity of this review.

LITERATURE CITED
1. Aasen H, Van Wittenberghe S, Sabater Medina N, Damm A, Goulas Y, et al. 2019. Sun-induced chloro-
phyll fluorescence II: Review of passive measurement setups, protocols, and their application at the leaf
to canopy level. Remote Sens. 11:927
2. Abdulridha J, Ehsani R, Abd-Elrahman A, Ampatzidis Y. 2019. A remote sensing technique for detecting
laurel wilt disease in avocado in presence of other biotic and abiotic stresses. Comput. Electron. Agric.
156:549–57
3. Adamiak A, Zdunek A, Kurenda A, Rutkowski K. 2012. Application of the biospeckle method for mon-
itoring bull’s eye rot development and quality changes of apples subjected to various storage methods:
preliminary studies. Sensors 12:3215–27
4. Adao T, Hruška J, Pádua L, Bessa J, Peres E, et al. 2017. Hyperspectral imaging: a review on UAV-based
sensors, data processing and applications for agriculture and forestry. Remote Sens. 9:1110
5. Albetis J, Duthoit S, Guttler F, Jacquin A, Goulard M, et al. 2017. Detection of flavescence dorée
grapevine disease using unmanned aerial vehicle (UAV) multispectral imagery. Remote Sens. 9:308
6. Ampatzidis Y, De Bellis L, Luvisi A. 2017. iPathology: robotic applications and management of plants
and plant diseases. Sustainability 9:1010
7. Ashourloo D, Mobasheri MR, Huete A. 2014. Developing two spectral disease indices for detection of
wheat leaf rust (Puccinia triticina). Remote Sens. 6:4723–40
8. Baker NR. 2008 Chlorophyll fluorescence: a probe of photosynthesis in vivo. Annu. Rev. Plant Biol.
59:89–113
9. Baldwin IT, Halitschke R, Paschold A, von Dahl CC, Preston CA. 2006. Volatile signalling in plant-plant
interactions: “talking trees” in the genomics era. Science 311:812–14
10. Barbedo JGA, Koenigkan LV, Halfeld-Vieira BA, Costa RV, Nechet KL, et al. 2018. Annotated plant
pathology databases for image-based detection and recognition of diseases. IEEE Latin Am. Trans.
16:1749–57
11. Behmann J, Bohnenkamp D, Paulus S, Mahlein AK. 2018. Spatial referencing of hyperspectral images
for tracing of plant disease symptoms. J. Imaging 4:143
12. Behmann J, Mahlein AK, Paulus S, Dupuis S, Kuhlmann H, et al. 2016. Generation and application of
hyperspectral 3D plant models: methods and challenges. Mach. Vis. Appl. 27:611–24
13. Behmann J, Mahlein AK, Rumpf T, Roemer C, Pluemer L. 2015. A review of advanced machine learning
methods for the detection of biotic stress in precision crop protection. Precis. Agric. 16:239–60
14. Belan LL, Ampélio Pozza E, Carvalho Alves M, Oliveira Freitas ML. 2018. Geostatistical analysis of
bacterial blight in coffee tree seedlings in the nursery. Summa Phytopathol. 44:317–25
15. Bellow S, Latouche G, Brown SC, Poutaraud A, Cerovic ZG. 2013. Optical detection of downy mildew
in grapevine leaves: daily kinetics of autofluorescence upon infection. J. Exp. Bot. 64:333–41
16. Bock CH, Poole GH, Parker PE, Gottwald TR. 2010. Plant disease severity estimated visually, by digital
photography and image analysis, and by hyperspectral imaging. Crit. Rev. Plant Sci. 29:59–107
17. Bravo C, Moshou D, West J, McCartney A, Ramon H. 2003. Early disease detection in wheat fields using
spectral reflectance. Biosyst. Eng. 84:137–45
18. Brugger A, Kuska MT, Mahlein AK. 2018. Impact of compatible and incompatible barley: Blumeria
graminis f.sp. hordei interactions on chlorophyll fluorescence parameters. J. Plant Dis. Prot. 125:177–
86
19. Cardoza J, Alborn HT, Tumlinson JH. 2002. In vivo volatile emissions from peanut plants induced by
simultaneous fungal infection and insect damage. J. Chem. Ecol. 28:161–74
20. Cellini A, Blasioli S, Biondi E, Bertaccini A, Braschi I, Spinelli F. 2017. Potential applications and limi-
tations of electronic nose devices for plant disease diagnosis. Sensors 17:2596
21. Chawade A, van Ham J, Blomquist H, Bagge O, Alexandersson E, Ortiz R. 2019. High-throughput
field-phenotyping tools for plant breeding and precision agriculture. Agronomy 9:258
22. Costa JM, Grant OM, Chaves MM. 2013. Thermography to explore plant–environment interactions.
J. Exp. Bot. 64:3937–49
23. Cui SQ, Ling P, Zhu HP, Keener HM. 2018. Plant pest detection using an artificial nose system: a
review. Sensors 18:378
246 Oerke
24. Dicke M. 2009. Behavioural and community ecology of plants that cry for help. Plant Cell Environ.
32:654–65
25. Di Gennaro SF, Battiston E, Di Marco S, Facini O, Matese A, et al. 2016. Unmanned aerial vehicle
(UAV)-based remote sensing to monitor grapevine leaf stripe disease within a vineyard affected by esca
complex. Phytopathol. Mediterr. 55:262–75
26. Fahlgren N, Gehan MA, Baxter I. 2015. Lights, camera, action: high-throughput plant phenotyping is
ready for a close-up. Curr. Opin. Plant Biol. 24:93–99
27. Fennell J, Veys C, Dingle J, Nwezeobi J, van Brunschot S, et al. 2018. A method for real-time classifi-
cation of insect vectors of mosaic and brown streak disease in cassava plants for future implementation
within a low-cost, handheld, in-field multispectral imaging sensor. Plant Methods 14:82
28. Franceschini MHD, Bartholomeus H, van Apeldoorn DF, Suomalainen J, Kooistra L. 2019. Feasibility
of unmanned aerial vehicle optical imagery for early detection and severity assessment of late blight in
potato. Remote Sens. 11:224
29. Franke J, Menz G. 2007. Multi-temporal wheat disease detection by multi-spectral remote sensing.
Precis. Agric. 8:161–72

30. Furbank RT, Jimenez-Berni JA, George-Jaeggli B, Potgieter AB, Deery DM. 2019. Field crop phe-
nomics: enabling breeding for radiation use efficiency and biomass in cereal crops. New Phytol. 223:1714–
27
31. Garcia-Ruiz F, Sankaran S, Maja JM, Lee WS, Rasmussen J, Ehsani R. 2013. Comparison of two aerial
imaging platforms for identification of Huanglongbing-infected citrus trees. Comput. Electron. Agric.
91:106–15
32. Gosme M, Lucas P. 2009. Cascade: an epidemiological model to simulate disease spread and aggregation
across multiple scales in a spatial hierarchy. Phytopathology 99:823–32
33. Gosme M, Willoquet L, Lucas P. 2007. Size, shape and intensity of aggregation of take-all disease during
natural epidemics in second wheat crops. Plant Pathol. 56:87–96
34. Gruber J, Nascimento HM, Yamauchi EY, Li RWC, Esteves CHA, et al. 2013. A conductive polymer
based electronic nose for early detection of Penicillium digitatum in post-harvest oranges. Mater. Sci. Eng.
C 33:2766–69
35. Gu WH, Sun Y, Tu K, Pan LQ. 2017. Evaluation of lipid oxidation of Chinese-style sausage during
processing and storage based on electronic nose. Meat Sci. 133:1–9
36. Guo FF, Chen XL, Lu MH, Yang L, Wang SW, Wu BM. 2018. Spatial analysis of rice blast in China at
three different scales. Phytopathology 108:1276–86
37. Hahn F. 2009. Actual pathogen detection: sensors and algorithms: a review. Algorithms 2:301–38
38. Hallau L, Neumann M, Klatt B, Kleinhenz B, Klein T, et al. 2018. Automated identification of sugar
beet diseases using smartphones. Plant Pathol. 67:399–410
39. Hatfield JL, Pinter PJ. 1993. Remote-sensing for crop protection. Crop. Prot. 12:403–13
40. He YA, Chen G, Potter C, Meentemeyer RK. 2019. Integrating multi-sensor remote sensing and species
distribution modeling to map the spread of emerging forest disease and tree mortality. Remote Sens.
Environ. 231:111238
41. Hickey LT, Hafeez AN, Robinson H, Jackson SA, Leal-Bertioli SCM, et al. 2019. Breeding crops to feed
10 billion. Nat. Biotechnol. 37:744–54
42. Hillnhuetter C, Schmitz A, Kuehnhold V, Sikora RA. 2010. Remote sensing for the detection of soil-
borne plant parasitic nematodes and fungal pathogens. In Precision Crop Protection: The Challenge and Use
of Heterogeneity, ed. EC Oerke, R Gerhards, G Menz, RA Sikora, pp. 151–65. Dordrecht, Neth.: Springer
43. Hillnhuetter C, Sikora RA, Oerke EC, van Dusschoten D. 2012. Nuclear magnetic resonance: a tool for
imaging belowground damage caused by Heterodera schachtii and Rhizoctonia solani on sugar beet. J. Exp.
Bot. 63:319–27
44. Hilmi NHZ, Idris A, Azmil MNM. 2019. Headspace solid-phase microextraction gas chromatography-
mass spectrometry for the detection of volatile organic compounds released from Ganoderma boninense
and oil palm wood. For. Pathol. 49:e12531
45. Huang WJ, Guan QS, Luo JH, Zhang JC, Zhao JL, et al. 2014. New optimized spectral indices for
identifying and monitoring winter wheat diseases. IEEE J. Sel. Top. Appl. 7:2516–24

46. Huleihel M, Shufan E, Tsror L, Sharaha U, Lapidot I, et al. 2018. Differentiation of mixed soil-borne
fungi in the genus level using infrared spectroscopy and multivariate analysis. J. Photochem. Photobiol. B
180:155–65
47. Hunt ER, Daughtry CST. 2018. What good are unmanned aircraft systems for agricultural remote sens-
ing and precision agriculture? Int. J. Remote Sens. 39:5345–76
48. Ishaq I, Alias MS, Kadir J, Kasawani I. 2014. Detection of basal stem rot disease at oil palm plantations
using sonic tomography. J. Sustain. Sci. Manag. 9:52–57
49. Jackson RD. 1986. Remote sensing of biotic and abiotic plant stress. Annu. Rev. Phytopathol. 24:265–
87
50. Jarolmasjed S, Sankaran S, Marzougui A, Kostick S, Si YS, et al. 2019. High-throughput phenotyping
of fire blight disease symptoms using sensing techniques in apple. Front. Plant Sci. 10:576
51. Jones HG, Schofield P. 2008. Thermal and other remote sensing of plant stress. Gen. Appl. Plant Physiol.
34:19–32
52. Kang WS, Hong SS, Han YK, Kim KR, Kim SG, Park EW. 2010. A web-based information sys-
tem for plant disease forecast based on weather data at high spatial resolution. Plant Pathol. J. 26:37–
48
53. Khaled AY, Abd Aziz S, Bejo SK, Nawi NM, Seman IA, Onwude DI. 2017. Early detection of diseases
in plant tissue using spectroscopy: applications and limitations. Appl. Spectr. Rev. 53(1):36–64
54. Khanal S, Fulton J, Shearer S. 2017. An overview of current and potential applications of thermal remote
sensing in precision agriculture. Comput. Electron. Agric. 139:22–32
55. Kuska MT, Behmann J, Grosskinsky DK, Roitsch T, Mahlein AK. 2018. Screening of barley resistance
against powdery mildew by simultaneous high-throughput enzyme activity signature profiling and mul-
tispectral imaging. Front. Plant Sci. 9:1074
56. Kuska MT, Behmann J, Namini M, Oerke EC, Steiner U, Mahlein AK. 2019 Discovering coherency of
specific gene expression and optical reflectance properties of barley genotypes differing for resistance
reactions against powdery mildew. PLOS ONE 14:e0213291
57. Lechenet M, Dessaint F, Py G, Makowski D, Munier-Jolain N. 2017. Reducing pesticide use while
preserving crop productivity and profitability on arable farms. Nat. Plants 3:17008
58. Lee C, Lee SY, Kim JY, Jung HY, Kim J. 2011. Optical sensing method for screening disease in melon
seeds by using optical coherence tomography. Sensors 11:9467–77
59. Leucker M, Wahabzada M, Kersting K, Peter M, Beyer W, et al. 2017. Hyperspectral imaging reveals
the effect of sugar beet QTLs on Cercospora leaf spot resistance. Funct. Plant Biol. 44:1–9
60. Li L, Zhang Q, Huang DF. 2014. A review of imaging techniques for plant phenotyping. Sensors
14:20078–111
61. Li S, Simonian A, Chin BA. 2010. Sensors for agriculture and the food industry. Electrochem. Soc. Interface
19:41–46
62. Li Z, Paul R, Tis TB, Saville AC, Hansel JC, et al. 2019. Non-invasive plant disease diagnostics enabled
by smartphone-based fingerprinting of leaf volatiles. Nat. Plants 5:856–66
63. Liakos KG, Busato P, Moshou D, Pearson S, Bochtis D. 2018. Machine learning in agriculture: a review.
Sensors 18:2674
64. Lim J, Kim G, Mo C, Oh K, Kim G, et al. 2018. Application of near infrared reflectance spectroscopy for
rapid and non-destructive discrimination of hulled barley, naked barley, and wheat contaminated with
Fusarium. Sensors 18:113
65. Lim J, Kim G, Mo C, Oh K, Yoo H, et al. 2017. Classification of Fusarium-infected Korean hulled
barley using near-infrared reflectance spectroscopy and partial least squares discriminant analysis. Sensors
17:2258
66. Lorestani EZ, Kamkar B, Razavi SE, da Silva JAT. 2013. Modeling and mapping diversity of pathogenic
fungi of wheat fields using geographic information systems (GIS). Crop Prot. 54:74–83
67. Lowe A, Harrison N, French AP. 2017. Hyperspectral image analysis techniques for the detection and
classification of the early onset of plant disease and stress. Plant Methods 13:80
68. Madden LV, Hughes G, Moraes WB, Xu XM, Turechek WW. 2018. Twenty-five years of the binary
power law for characterizing heterogeneity of disease incidence. Phytopathology 108:656–80
248 Oerke
69. Madden LV, Hughes G, van den Bosch F. 2007. Study of Plant Disease Epidemics. St. Paul, MN: APS Press
70. Maes WH, Steppe K. 2019. Perspectives for remote sensing with unmanned aerial vehicles in precision
agriculture. Trends Plant Sci. 24:152–64
71. Mahlein AK. 2016. Plant disease detection by imaging sensors: parallels and specific demands for pre-
cision agriculture and plant phenotyping. Plant Dis. 100:241–51
72. Mahlein AK, Alisaac E, Al Masri A, Behmann J, Dehne HW, Oerke EC. 2019. Comparison and com-
bination of thermal, fluorescence, and hyperspectral imaging for monitoring Fusarium head blight of
wheat on spikelet scale. Sensors 19:2281
73. Mahlein AK, Kuska MT, Behmann J, Polder G, Walter A. 2018. Hyperspectral sensors and imaging
technologies in phytopathology: state of the art. Annu. Rev. Phytopathol. 56:535–58
74. Mahlein AK, Oerke EC, Steiner U, Dehne HW. 2012. Recent advances in sensing plant diseases for
precision crop protection. Eur. J. Plant Pathol. 133:197–203
75. Mahlein AK, Rumpf T, Welke P, Dehne HW, Pluemer L, et al. 2013. Development of spectral indices
for detecting and identifying plant diseases. Remote Sens. Environ. 128:21–30
76. Martinelli F, Scalenghe R, Davino S, Panno S, Scuderi G, et al. 2015. Advanced methods of plant disease
detection. A review. Agron. Sustain. Dev. 35:1–25
77. Martins LM, Neiro ED, Dias AR, Roque CG, Baio FHR, Teodoro PE. 2018. Cotton vegetation indices
under different control methods of Ramularia leaf spot. Biosci. J. 34:1706–13
78. Maxwell AE, Warner TA, Fang F. 2018. Implementation of machine-learning classification in remote
sensing: an applied review. Int. J. Remote Sens. 39:2784–817
79. Meor Yusoff MS, Muhd Asshar K, Ideris AS. 2009. Identification of basal stem rot disease in local palm
oil by microfocus XRF. J. Nucl. Relat. Technol. 6:282–87
80. Mishra P, Asaari MSM, Herrero-Langreo A, Lohumi S, Diezma B, Scheunders P. 2017. Close range
hyperspectral imaging of plants: a review. Biosyst. Eng. 164:49–67
81. Murchie EH, Lawson T. 2013. Chlorophyll fluorescence analysis: a guide to good practice and under-
standing some new applications. J. Exp. Bot. 64:3983–98
82. Mutka AM, Bar RS. 2015. Image-based phenotyping of plant disease symptoms. Front. Plant Sci.
5:734
83. Nansen C. 2016. The potential and prospects of proximal remote sensing of arthropod pests. Pest Manag.
Sci. 72:653–59
84. Navakar DS, Singh CB, Jayas DS, White NDG. 2009. Assessment of soft X-ray imaging for detection
of fungal infection in wheat. Biosyst. Eng. 103:49–56
85. Nilsson H. 1995. Remote sensing and image analysis in plant pathology. Annu. Rev. Phytopathol. 33:489–
528
86. Nutter F, van Rij N, Eggenberger SK, Holah N. 2010. Spatial and temporal dynamics of plant pathogens.
In Precision Crop Protection: The Challenge and Use of Heterogeneity, ed. EC Oerke, R Gerhards, G Menz,
RA Sikora, pp. 27–50. Dordrecht, Neth.: Springer
87. Oerke EC. 2006. Crop losses to pests. J. Agric. Sci. 144:31–43
88. Oerke EC. 2018. Precision crop protection systems. In Precision Agriculture for Sustainability, ed.
J Stafford, pp. 347–97. Cambridge, UK: Burleigh Dodds Science
89. Oerke EC, Froehling P, Steiner U. 2011. Thermographic assessment of scab disease on apple leaves.
Precis. Agric. 12:699–715
90. Oerke EC, Gerhards R, Menz G, Sikora RS. 2010. Preface. In Precision Crop Protection: The Challenge
and Use of Heterogeneity, ed. EC Oerke, R Gerhards, G Menz, RA Sikora, pp. v–viii. Dordrecht, Neth.:
Springer
91. Oerke EC, Herzog K, Toepfer R. 2016. Hyperspectral phenotyping of the reaction of grapevine geno-
types to Plasmopara viticola. J. Exp. Bot. 67:5529–43
92. Oerke EC, Leucker M, Steiner U. 2019. Sensory assessment of Cercospora beticola sporulation for phe-
notyping the partial disease resistance of sugar beet genotypes. Plant Methods 15:133
93. Oerke EC, Mahlein AK, Steiner U. 2014. Proximal sensing of plant diseases. In Detection and Diagnostics
of Plant Pathogens. Plant Pathology in the 21st Century, Vol. 5, ed. ML Gullino, PJM Bonants, pp. 55–68.
Dordrecht, The Netherlands: Springer Netherlands

94. Oerke EC, Steiner U. 2010. Potential of digital thermography for disease control. In Precision Crop Pro-
tection: The Challenge and Use of Heterogeneity, ed. EC Oerke, R Gerhards, G Menz, RA Sikora, pp. 167–82.
Dordrecht, Neth.: Springer
95. Pan L, Zhang W, Zhu N, Mao SB, Tu K. 2016. Early detection and classification of pathogenic fungal
disease in post-harvest strawberry fruit by electronic nose and gas chromatography-mass spectrometry.
Food Res. Int. 62:162–68
96. Pantazi XE, Tamouridou AA, Alexandridis TK, Lagopodi AL, Kontouris G, Moshou D. 2017. Detec-
tion of Silybum marianum infection with Microbotryum silybum using VNIR field spectroscopy. Comput.
Electron. Agric. 137:130–37
97. Paulus S. 2019. Measuring crops in 3D: using geometry for plant phenotyping. Plant Methods 15:103
98. Peteinatos GG, Korsaeth A, Berge TW, Gerhards R. 2016. Using optical sensors to identify water de-
privation, nitrogen shortage, weed presence and fungal infection in wheat. Agriculture 6:24
99. Pinto F, Damm A, Schickling A, Panigada C, Cogliati S, et al. 2016. Sun-induced chlorophyll fluores-
cence from high-resolution imaging spectroscopy data to quantify spatio-temporal patterns of photo-
synthetic function in crop canopies. Plant Cell Environ. 39:1500–12

100. Racca P, Zeuner T, Jung J, Kleinhenz B. 2010. Model validation and use of geographic information
systems in crop protection warning service. In Precision Crop Protection: The Challenge and Use of Hetero-
geneity, ed. EC Oerke, R Gerhards, G Menz, RA Sikora, pp. 259–76. Dordrecht, Neth.: Springer
101. Rousseau C, Belin E, Bove E, Rousseau D, Fabre F, et al. 2013. High throughput quantitative pheno-
typing of plant resistance using chlorophyll fluorescence image analysis. Plant Methods 9:17
102. Rousseau C, Hunault G, Gaillard S, Bourbeillon J, Montiel G, et al. 2015. Phenoplant: a web resource
for the exploration of large chlorophyll fluorescence image datasets. Plant Methods 11:24
103. Ruiz-Altisent M, Ruiz-Garcia L, Moreda GP, Lu RF, Hernandez-Sanchez N, et al. 2010. Sensors for
product characterization and quality of specialty crops: a review. Comput. Electron. Agric. 74:176–94
104. Rumpf T, Mahlein AK, Steiner U, Oerke EC, Dehne HW, Pluemer L. 2010. Early detection and clas-
sification of plant diseases with support vector machines based on hyperspectral reflectance. Comput.
Electron. Agric. 74:91–99
105. Rutolo MF, Clarkson JP, Harper G, Covington JA. 2018. The use of gas phase detection and monitoring
of potato soft rot infection in store. Postharvest Biol. Technol. 145:15–19
106. Salgadoe ASA, Robson AJ, Lamb DW, Dann EK, Searle C. 2018. Quantifying the severity of Phytoph-
thora root rot disease in avocado trees using image analysis. Remote Sens. 10:226
107. Sankaran S, Khot LR, Zuniga Espinoza C, Jarolmasjed S, Sathuvalli VR, et al. 2015. Low-altitude, high-
resolution aerial imaging systems for row and field crop phenotyping: a review. Eur. J. Agron. 70:112–23
108. Santoso H, Tani H, Wang XF, Prasetyo AE, Sonobe R. 2019. Classifying the severity of basal stem rot
disease in oil palm plantations using WorldView-3 imagery and machine learning algorithms. Int. J.
Remote Sens. 40:7624–46
109. Serverns PM, Sackett KE, Farber DH, Mundt CC. 2019. Consequences of long-distance dispersal for
epidemic spread: patterns, scaling, and mitigation. Plant Dis. 103:177–91
110. Shakoor N, Lee S, Mockler TC. 2017. High throughput phenotyping to accelerate crop breeding and
monitoring of diseases in the field. Curr. Opin. Plant Biol. 38:184–92
111. Shi Y, Huang WJ, Gonzalez-Moreno P, Luke B, Dong YY, et al. 2018. Wavelet-based rust spectral
feature set (WRSFs): a novel spectral feature set based on continuous wavelet transformation for tracking
progressive host–pathogen interaction of yellow rust on wheat. Remote Sens. 10:525
112. Shi Y, Huang WJ, Ye HC, Ruan C, Xing NC, et al. 2018. Partial least square discriminant analysis based
on normalized two-stage vegetation indices for mapping damage from rice diseases using PlanetScope
datasets. Sensors 18:1901
113. Shimwela MM, Halbert SE, Keremane ML, Mears P, Singer BH, et al. 2019. In-grove spatiotemporal
spread of citrus Huanglongbing and its psyllid vector in relation to weather. Phytopathology 109:418–27
114. Shtienberg D. 2013. Will decision-support systems be widely used for the management of plant diseases?
Annu. Rev. Phytopathol. 51:1–16
115. Siedliska A, Baranowski P, Zubik M, Mazureka W, Sosnowska B. 2018. Detection of fungal infections
in strawberry fruit by VNIR/SWIR hyperspectral imaging. Postharvest Biol. Technol. 139:115–26
250 Oerke
116. Signoroni A, Savardi M, Baronio A, Benini S. 2019. Deep learning meets hyperspectral image analysis:
a multidisciplinary review. J. Imaging 5:52
117. Simko I, Jimenez-Berni JA, Sirault XRR. 2017. Phenomic approaches and tools for phytopathologists.
Phytopathology 107:6–17
118. Simonett DS. 1983. Manual of Remote Sensing. Falls Church, VA: Am. Soc. Photogramm.
119. Skolik P, McAinsh MR, Martin FL. 2019. ATR-FTIR spectroscopy non-destructively detects damage-
induced sour rot infection in whole tomato fruit. Planta 249:925–39
120. Smigaj M, Gaulton R, Suarez JC, Barr SL. 2019. Canopy temperature from an unmanned aerial vehicle
as an indicator of tree stress associated with red band needle blight severity. For. Ecol. Manag. 433:699–
708
121. Smigaj M, Gaulton R, Suarez JC, Barr SL. 2019. Combined use of spectral and structural characteristics
for improved red band needle blight detection in pine plantation stands. For. Ecol. Manag. 434:213–23
122. Sperschneider J. 2020. Machine learning in plant-pathogen interactions: empowering biological predic-
tions from field scale to genome scale. New Phytol. https://doi.org/10.1111/nph.15771
123. Stafford JV. 2000. Implementing precision agriculture in the 21st century. J. Agric. Eng. Res. 76:267–75
124. Stella A, Caliendo G, Melgani F, Goller R, Barazzuol M, La Porta N. 2017. Leaf wetness evaluation
using artificial neural network for improving apple scab fight. Environments 4:42
125. Stone C, Mohammed C. 2017. Application of remote sensing technologies for assessing planted forests
damaged by insect pests and fungal pathogens: a review. Curr. For. Rep. 3:75–92
126. Su JY, Liu CJ, Coombes M, Hu XP, Wang CH, et al. 2018. Wheat yellow rust monitoring by learning
from multispectral UAV aerial imagery. Comput. Electron. Agric. 155:157–66
127. Sun J, Kunnemeyer R, McGlone A, Tomer N. 2018. Optical properties of healthy and rotten onion flesh
from 700 to 1000 nm. Postharvest Biol. Technol. 140:1–10
128. Sun Y, Wei KL, Liu Q, Pan LQ, Tu K. 2018. Classification and discrimination of different fungal diseases
of three infection levels on peaches using hyperspectral reflectance imaging analysis. Sensors 18:1295
129. Tackenberg M, Volkmar C, Schirrmann M, Giebel A, Dammer KH. 2018. Impact of sensor-controlled
variable-rate fungicide application on yield, senescence and disease occurrence in winter wheat fields.
Pest Manag. Sci. 74:1251–58
130. Tamouridou AA, Pantazi XE, Alexandridis T, Lagopodi A, Kontouris G, Moshou D. 2018. Spectral
identification of disease in weeds using multilayer perceptron with automatic relevance determination.
Sensors 18:2770
131. Thomas S, Behmann J, Steier A, Kraska T, Müller O, et al. 2018. Quantitative assessment of disease
severity and rating of barley cultivars based on hyperspectral imaging in a non-invasive, automated phe-
notyping platform. Plant Methods 14:42
132. Thomas S, Kuska MT, Bohnenkamp D, Brugger A, Alisaac E, et al. 2018. Benefits of hyperspectral
imaging for plant disease detection and plant protection: a technical perspective. J. Plant Dis. Prot. 125:5–
20
133. Thybo AK, Jespersen SN, Laerke PE, Stodkilde-Jorgensen HJ. 2004. Nondestructive detection of in-
ternal bruise and spraing disease symptoms in potatoes using magnetic resonance imaging. Magn. Reson.
Imaging 22:1311–17
134. Tischler YK, Thiessen E, Hartung E. 2018. Early optical detection of infection with brown rust in winter
wheat by chlorophyll fluorescence excitation spectra. Comput. Electron. Agric. 146:77–85
135. Tona E, Calcante A, Oberti R. 2018. The profitability of precision spraying on specialty crops: a
technical-economic analysis of protection equipment at increasing technological levels. Precis. Agric.
19:606–29
136. Topp CFE, Hughes G, Nevison IM, Butler A, Oxley SJP, Havis ND. 2019. Rhynchosporium leaf scald
disease incidence: seed source and spatial pattern. Plant Pathol. 68:1179–87
137. Turechek WW, McRoberts N. 2013. Considerations of scale in the analysis of spatial pattern of plant
disease epidemics. Annu. Rev. Phytopathol. 51:453–72
138. Underwood J, Wendel A, Schofield B, McMurray L, Kimber R. 2017. Efficient in-field plant phenomics
for row-crops with an autonomous ground vehicle. J. Field Robot. 34:1061–83
139. Van der Plank JE. 1963. Plant Diseases: Epidemics and Control. New York: Academic

140. Waggoner PE, Aylor DE. 2000. Epidemiology, a science of patterns. Annu. Rev. Phytopathol. 38:71–94
141. Wahabzada M, Mahlein AK, Bauckhage C, Steiner U, Oerke EC, Kersting K. 2015. Metro maps of plant
disease dynamics: automated mining of differences using hyperspectral images. PLOS ONE 10:e0116902
142. West JS, Bravo C, Oberti R, Lemaire D, Moushou D, McCartney HA. 2003. The potential of optical
canopy measurement for targeted control of field crop diseases. Annu. Rev. Phytopathol. 41:593–614
143. West JS, Bravo C, Oberti R, Moshou D, Ramon H, McCartney HA. 2010. Detection of fungal diseases
optically and pathogen inoculum by air sampling. In Precision Crop Protection: The Challenge and Use of
Heterogeneity, ed. EC Oerke, R Gerhards, G Menz, RA Sikora, pp. 135–49. Dordrecht, Neth.: Springer
144. West JS, Canning GGM, Perryman SA, King K. 2017. Novel technologies for the detection of Fusarium
head blight disease and airborne inoculum. Trop. Plant Pathol. 42:203–9
145. West JS, Kimber RBE. 2015. Innovations in air sampling to detect plant pathogens. Ann. Appl. Biol.
166:4–17
146. Whetton RL, Hassall KL, Waine TW, Mouazen AM. 2018. Hyperspectral measurements of yellow rust
and Fusarium head blight in cereal crops: Part 1: laboratory study. Biosyst. Eng. 166:101–15
147. Whetton RL, Waine TW, Mouazen AM. 2018. Hyperspectral measurements of yellow rust and Fusar-
ium head blight in cereal crops: Part 2: on-line field measurement. Biosyst. Eng. 167:154–58
148. Wijesinghe RE, Lee SY, Kim P, Jung HY, Jeon M, Kim J. 2016. Optical inspection and morphological
analysis of Diospyros kaki plant leaves for the detection of circular leaf spot disease. Sensors 16:1282
149. Wilson AD. 2013. Diverse applications of electronic-nose technologies in agriculture and forestry.
Sensors 13:2295–348
150. Wilson AD. 2018. Applications of electronic-nose technologies for noninvasive early detection of plant,
animal and human diseases. Chemosensors 6:45
151. Wolf PJF, Verreet JA. 2002. An integrated pest management system in Germany for the control of
fungal leaf diseases in sugar beet: the IPM sugar beet model. Plant Dis. 86:336–44
152. Wrzesien M, Treder W, Klamkowski K, Rudnicki WR. 2019. Prediction of the apple scab using machine
learning and simple weather stations. Comput. Electron. Agric. 161:252–59
153. Wutzki CR, Berger-Neto A, Grabicoski EM, Henneberg L, Sartori FF, Jaccoud-Filho DS. 2019. Within-
field variability and spatial analysis of white mould and soybean crop attributes in southern Brazil. Trop.
Plant Pathol. 44:104–11
154. Yao H, Hruska Z, Kincaid R, Brown RL, Bhatnagar D, Cleveland TE. 2013. Detecting maize inocu-
lated with toxigenic and atoxigenic fungal strains with fluorescence hyperspectral imagery. Biosyst. Eng.
115:125–35
155. Zaman-Allah M, Vergara O, Araus JL, Tarekegne A, Magorokosho C, et al. 2015. Unmanned aerial
platform-based multi-spectral imaging for field phenotyping of maize. Plant Methods 11:35
156. Zarco-Tejada PJ, Camino C, Beck PSA, Calderon R, Hornero A, et al. 2018. Previsual symptoms of
Xylella fastidiosa infection revealed in spectral plant-trait alterations. Nat. Plants 4:432–39
157. Zdunek A, Adamiak A, Pieczywek PM, Kurenda A. 2014. The biospeckle method for the investigation
of agricultural crops: a review. Opt. Laser Eng. 52:276–85
158. Zhang JC, Huang YB, Pu RL, Gonzalez-Moreno P, Yuan L, et al. 2019. Monitoring plant diseases and
pests through remote sensing technology: a review. Comput. Electron. Agric. 165:104943
159. Zhao C, Zhang Y, Du J, Guo X, Wen W, et al. 2019. Crop phenomics: current status and perspectives.
Front. Plant Sci. 10:714
252 Oerke
PY58_FrontMatter ARI 6 August 2020 17:3
Annual Review of
Phytopathology
Volume 58, 2020
Contents
Gall-Inducing Parasites: Convergent and Conserved Strategies
of Plant Manipulation by Insects and Nematodes
Bruno Favery, Géraldine Dubreuil, Ming-Shun Chen, David Giron,

and Pierre Abad p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 1
Deep Roots and Splendid Boughs of the Global Plant Virome
Valerian V. Dolja, Mart Krupovic, and Eugene V. Koonin p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p23
Social Evolution and Cheating in Plant Pathogens
Maren L. Friesen p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p55
Tolerance of Plants to Pathogens: A Unifying View
Israel Pagán and Fernando Garcı́a-Arenal p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p77
Disease in Invasive Plant Populations
Erica M. Goss, Amy E. Kendig, Ashish Adhikari, Brett Lane, Nicholas Kortessis,
Robert D. Holt, Keith Clay, Philip F. Harmon, and S. Luke Flory p p p p p p p p p p p p p p p p p p p p p97
Epigenetic Mechanisms in Nematode–Plant Interactions
Tarek Hewezi p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 119
RPS5-Mediated Disease Resistance: Fundamental Insights
and Translational Applications
Sarah E. Pottinger and Roger W. Innes p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 139
Developing Public–Private Partnerships in Plant Pathology Extension:
Case Studies and Opportunities in the United States
Samuel G. Markell, Gregory L. Tylka, Edwin J. Anderson,
and H. Peter van Esse p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 161
The Geopolitics of Plant Pathology: Frederick Wellman, Coffee Leaf
Rust, and Cold War Networks of Science
Stuart McCook and Paul D. Peterson p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 181
Progress in Biological Control of Weeds with Plant Pathogens
Louise Morin p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 201
Remote Sensing of Diseases
Erich-Christian Oerke p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 225
v
PY58_FrontMatter ARI 6 August 2020 17:3
Origins and Immunity Networking Functions of EDS1

Family Proteins
Dmitry Lapin, Deepak D. Bhandari, and Jane E. Parker p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 253
Organic Amendments for Pathogen and Nematode Control
Erin Rosskopf, Francesco Di Gioia, Jason C. Hong, Cristina Pisani,
and Nancy Kokalis-Burelle p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 277
Modeling the Impact of Crop Diseases on Global Food Security
Serge Savary and Laetitia Willocquet p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 313
Functional Ecology of Forest Disease
Jonàs Oliva, Miguel Ángel Redondo, and Jan Stenlid p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 343
Ustilaginoidea virens: Insights into an Emerging Rice Pathogen

Wenxian Sun, Jing Fan, Anfei Fang, Yuejiao Li, Muhammad Tariqjaveed,
Dayong Li, Dongwei Hu, and Wen-Ming Wang p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 363
Managing Crop Diseases Under Water Scarcity
Cassandra L. Swett p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 387
Errata
An online log of corrections to Annual Review of Phytopathology articles may be found at

http://www.annualreviews.org/errata/phyto
vi Contents

Remote Sensing of Diseases

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Remote Sensing of Diseases

Uploaded by

Copyright:

Available Formats

Annual Review of Phytopathology

Remote Sensing of Diseases

Annu. Rev. Phytopathol. 2020. 58:225–52 Keywords

DEVELOPMENT OF DISEASES IN TIME AND SPACE: THE ROLE

www.annualreviews.org • Remote Sensing of Diseases 227

SCOPES OF DISEASE SENSING

Table 1 Sensing of diseases: potential application areas, conditions and users

www.annualreviews.org • Remote Sensing of Diseases 229

In some applications, disease detection is the perception of a deviation from a healthy

c Phenotyping with various sensors d Quality control of food, e.g., fruits

RECENT REVIEWS ON DISEASE SENSING AND DATA PROCESSING

www.annualreviews.org • Remote Sensing of Diseases 231

SENSORS FOR PLANT DISEASES

www.annualreviews.org • Remote Sensing of Diseases 233

Sensing of Other Information

www.annualreviews.org • Remote Sensing of Diseases 235

SPATIAL RESOLUTION AND SENSOR PLATFORMS

Satellites and Aircrafts

Unmanned Aerial Vehicles

www.annualreviews.org • Remote Sensing of Diseases 237

AREAS OF SENSOR APPLICATION

www.annualreviews.org • Remote Sensing of Diseases 239

www.annualreviews.org • Remote Sensing of Diseases 241

Sensing of Pathogens and the Environment

Sensing of Host–Pathogen Interactions

www.annualreviews.org • Remote Sensing of Diseases 243

www.annualreviews.org • Remote Sensing of Diseases 245

Precis. Agric. 8:161–72

www.annualreviews.org • Remote Sensing of Diseases 247

www.annualreviews.org • Remote Sensing of Diseases 249

synthetic function in crop canopies. Plant Cell Environ. 39:1500–12

www.annualreviews.org • Remote Sensing of Diseases 251

Volume 58, 2020

Bruno Favery, Géraldine Dubreuil, Ming-Shun Chen, David Giron,

Origins and Immunity Networking Functions of EDS1

Ustilaginoidea virens: Insights into an Emerging Rice Pathogen

An online log of corrections to Annual Review of Phytopathology articles may be found at

You might also like