PROFESSOR JAYA SHANKAR TELANGANA STATE

AGRICULTURAL UNIVERSITY
COLLEGE OF AGRICULTURE, RAJENDRANAGAR
HYDERABAD – 500 030
2020

TERM PAPER
ON
INTEGRATED PEST MANAGEMENT THROUGH REMOTE SENSING
AND SPATIAL ANALYSIS

ENT – 612 3(2+1): ADVANCED IPM

SUBMITTED TO: -

Dr. R. SUNITHA DEVI,

ASSOCIATE PROFESSOR,
DEPARTMENT OF ENTOMOLOGY.

SUBMITTED BY: -
J. CHANDRAKALA,
RAD /19 – 32,
Ph.D. PLANTPATHOLOGY– I YEAR.
INTEGRATED PEST MANAGEMENT THROUGH REMOTE SENSING
AND SPATIAL ANALYSIS

Introduction

Modern agriculture is influenced by increasing human population and the consequent

pressure for increased agriculture productivity. In addition, plant pests are increasing
worldwide due to globalization and a ready exchange of pests, weeds and disease material,
with increasing costs to nations Efforts to combat pests on crops while maintaining farm
profitability and productivity are not new, and Integrated Pest Management (IPM) approaches
are important additions in the fight. There are many definitions of IPM (the term was first
coined in 1967, the one adopted by the National Coalition on Integrated Pest Management is
used in this chapter: “a sustainable approach to managing pests (animal pests, pathogens, and
weeds) by combining biological, cultural, physical, and chemical tools in a way that
minimizes economic, health and environmental risks” (Jacobsen, 1997). Farmers and growers
practicing IPM use natural predators and parasites, pest-resistant crop varieties, cultural
practices, biological controls, various physical techniques, and try to minimize pesticide and
herbicide application In IPM, the emphasis is placed on acceptable pest levels and
encouragement of natural predators and parasites (biological controls) rather than pest
eradication, and on monitoring and regular observation of crop condition and pest
populations.
The emphasis on monitoring and thorough knowledge of field conditions prescribed
by IPM make new imagery sources and integrative geospatial technologies, particularly
global positioning systems, remote sensing, geographical information systems and spatial
analysis, powerful tools to assist in the management of pests (Barnes et al., 1996); and these
tools can be considered part of an IPM system. Global Positioning Systems (GPS) and
Geographic Information Systems (GIS) are by far the more commonly used of the geospatial
tools (Whipker & Akridge, 2006) and have been revolutionary technologies for agriculture.
The recent wide-scale adoption of GPS across all economic sectors exemplifies a
broad-scale technological transition from expert system to broad use. GPS is a satellite-based
navigation system made up of a network of 24 satellites placed into orbit by the U.S.
Department of Defense. Originally intended for military applications, the U.S. government
made the system available in the 1980s for civilian use (Johnson & Barton, 2004).
The technology has become increasingly used in precision agriculture applications; and
when differential correction is implemented, mapping- and survey-grade GPS have been used
to accurately map fields for variable fertilization application, targeted weed control and
currently form the backbone of many precision agriculture applications.
GPS integrates well with other technologies that rely on accurate and precise spatial
location information like remote sensing and GIS, computer and analysis systems designed to
store, analyze and output spatial data that are linked to nonspatial information. GIS are
increasingly used across a range of scientific fields, and are used now as databases to store
multiple “layers” of overlapping spatial data to integrating spatial data from numerous
sources, to performing complex analysis of spatial patterns across landscapes. There are
numerous examples demonstrating the benefits of adding precise and accurate spatial data
and analytical techniques to agricultural management.
To manage the pests, monitoring and forecasting has been an integrated part of the crop
production system in developed countries.

REMOTE SENSING IN PEST MANAGEMENT

What is Remote Sensing ?

Remote sensing is the acquisition of knowledge or information about objects from a distance
without coming into direct contact with them.

Remote sensing in entomology was first used in the field of forestry. Slow moving for
assessing pest aeroplanes were found useful for visually assessing the extent of damage by a
spruce budworm, in United States and Canada as early as 1921.

History of Remote Sensing

The first meterological satellite was launched by US on 1 st April, 1960 for weather forecast,
movement of hurricanes and other associated uses. It was named as earth resources
technology satellite (ERTS)- well suited for agricultural responses. These satellites are now
referred to as LAND SAT. First remote sensing satellite in India was launched in 1988,
named as IRS (India Remote Sensing).
Components of Remote Sensing

Remote sensing system has three essential components: 1. signal 2. sensor and 3. sensing.
The reflected radiation from the surface of an object is called signal. The signal is then
perceived by a sensor such as the human eye or camera or any other instrument.

Types of Remote Sensing

There are two types of remote sensing systems

 Passive remote sensing system
 Active remote sensing system.
Passive Remote Sensing System: A passive remote sensing system possesses only a sensor.
It does not produce radiations to irradiate the object to be sensed. For example human eye or
photographic camera without flash. These can only perceive objects when they are irradiated
by sunlight or electricity, but they don't produce radiations themselves.

Active Remote Sensing System: Active remote sensing system possesses the sensor and also
throws the light on the object to be sensed. In other words it also contains the source of signal
e.g. photographic camera with flash. In sunlight it can utilize natural radiation for sensing the
object but during dark it produces radiation to irradiate the object.

Spectral Reflectance of Vegetation

Plant pigments, leaf structure and total water content are three important factors affecting
spectral reflectance of vegetation. External factors, which influence spectral reflectances of
vegetation are: moisture stress, soil nutrients salinity, pests, seasonal variation and climatic
factors. The spectral reflectance of healthy vegetation crop is characterized by
High absorption i. e. low reflectance in blue and red regions of EM spectrum.
High reflectance in near IR due to internal cell structure.
Water absorption bands i. e. low reflectance in the mid IR.
Any deviation in reflectance from above pattern indicates that crop is not healthy and there is
some stress on the crop. For example, high value of reflectance in visible blue and red
regions, low reflectance in near IR and high reflectance in mid IR would show that there is
some stress on the crop.

Cause of Stress

Stress on vegetation or damage symptoms may be caused by any of the factors viz. moisture,
nutrients, pests and diseases. Simply by seeing the reflectance we cannot make out the cause
of the stress. The cause of the stress has to be ascertained or confirmed from the field. This is
called the "ground truth". The damage symptoms due to various factors should be
thoroughly differentiated and standardized for future comparisons.

REMOTE SENSING TECHNIQUES (DEVICES) USED IN

ENTOMOLOGY

Remote sensing techniques used in entomology include

 Photography and videography from aircraft and from ground

 Satellite- borne photography, multispectral scanning and thermal imaging
 Ground-based and air-based RADAR (Radio detection and ranging)

Photography

Photography has been used extensively as a tool to obtain crop loss information.
Panchromatic, colour infrared and black and white infrared films have also been utilized for
photographing damage due to pests. In aerial photography the activity of insects is detected
from the changes they produce in the appearance of plant foliage. Foliage may be changed if
insects leave deposits on leaves or if they induce changes of leaf colour, shape or density as a
direct or indirect result of feeding. Different types of damage can be differentiated in
photographs. Defoliators like spruce budworm cause thinning of foliage and discolouration
from green to yellow and yellow to red. This can be easily detected by aerial photographs.
Similarly, sucking insects cause deformation of treetops, twigs and branches that can be seen
through photographs.
Photographic methods generally rely on the detection of plant stress or damage, which could
be due to a variety of causes quite unrelated to insect activity. In such instances identification
depends on the spot visual inspection. However, some insects like aphids produce honeydew
on which sooty mould fungus grows. This fungus blackens the foliage, which is readily
detectable by aerial photography.

Satellite -Based Multi-Spectral Scanning

These collect data in visible and infrared portions of the em spectrum. These are also called
non-photographic techniques. These can form images over a much wider range of em
wavelengths (0.4-14 pm) than the photographic method. These do not generate an
instantaneous image of a whole scene. But these scanners use sensors with very narrow field
of view to scan an area systematically and an image is built up as the scan progresses. The
Indian Remote Sensing Satellite System i. e. IRS-1A and IRS-IB provides images in the
following bands i). 0.45-0.52 pm ii) 0.52-0.59 ym iii) 0.62-0.68 pm and iv) 0.78-0.86 ym.
Satellite based remote sensing has been used for two major purposes: (i) the detection of
changes in vegetation and (ii) the measurement of meteorological data. The spectral
reflectance data has been used to derive different vegetation indices. An index called the
ratio vegetation index, which is derived by dividing the near infrared reflectance by the red
reflectance. It has been found to be a good indicator of - defoliation due to insect-pests in
forest, areas. It is also helpful in exploring vegetation that could support locust breeding in
normally desiccated areas. Because reflectance in the red region is inversely proportional to
chlorophyll density and reflectance in the near infrared region is proportional to leaf density,
a large index is indicative of high green leaf biomass. Satellite remote sensing is an ideal tool
to investigate the environmental factors controlling the pest development such as rainfall and
air temperature. Flying moths of armyworm, Spodoptera exempta are brought together by
winds associated with rainstorms. This subsequently leads to dense outbreaks of caterpillars.
Meteosat data has been used to monitor these rainstorms and for quickly locating potential
outbreaks. Landsat data has been used to detect areas of short lived vegetation capable of
supporting desert locust, Schistocerca gregaria.
Measurements with multispectral scanner showed that groundnut with mite injury had a
different spectral signature than groundnut infected with late leaf spot. Similarly spectral
signatures of fields with low nematode population and high nematode population differed in
visible and near IR region.

RADAR

Radio detection and ranging (RADAR) perceives reflected radio waves from the surface of
the object. Its operating principle is that it transmits radio waves and receives the same
reflected by an object. The time interval between transmission and reception by the radar
determines the exact position of the object. Radar has been used for studying long distance
migration and flight behaviour of wind borne / air borne insects. These radars use a
wavelength of 3 to 10 cm. In case of small insects, millimeter radar is used. Typical
maximum detection range is about 1.5 to 2.8 km for individual insects and several tens of
kilometers for dense swarms. Flight behaviour of locusts, grasshoppers and moths of various
species of Heliothis, Spodoptera etc. has been studied through radar.

APPLICATION OF SATELLITE REMOTE SENSING

FOR PEST MANAGEMENT IN INDIA

The factors which favor the development of pests and their effect on plant parameters that
are detectable by remote sensors are shown in this below Figure 1.
 Figure 1: Development of crop pests and diseases leading to crop damage and its
detection through remote sensing.
The remote sensing can be found very useful in management of the following pests

 Desert locust
 Brown plant hopper
 Cotton whitefly

Desert Locust
The desert locust is endemic over Rajasthan, Gujarat and Haryana, which is a part of world
recession area. Successful breeding of locust in the recession area due to favourable soil
moisture, widespread rainfall, shade and green vegetation, results in devastating swarms.
Locust control is very difficult affair if early action is not taken. Vegetation index maps and
rainfall estimates through satellite remote sensing help to maintain a close watch on deseh
locust population development. Temporal (time related) and spatial (space related)
distribution of desert vegetation has been found to be closely related to desert locust
population and rainfall. The path of desert locust journey is controlled by large scale weather
systems like the position of the Intertropical Convergence Zone (ITCZ) and low level wind
circulation leading to vertical upward movement. The position of ITCZ can be known from
cloud signatures observed by INSAT. In this way, areas likely to be affected by the locust can
be detected in advance.

Brown Pant Hopper (BPH) on Rice

Long distance migration of brown plant hopper (BPH), (Nilaparvata lugens) is Remote
sensing helps associated with depressions. The continuous cultivation of rice season after
season is in quick assessment of responsible for heavy incidence of BPH. Staggered planting
and harvesting also pest damage in large favours higher multiplication of this pest. Hence a
uniform planting time would help areas in reducing menace of this pest. The satellite remote
sensing can be used to monitor and regulate the planting of rice in BPH prone areas.

Whitefly on Cotton Crop

The cotton crop is severely affected by cotton whitefly (Bemisia tabaci) in several parts of
India, especially in Andhra Pradesh. Adults and nymphs both suck sap from plants, which
leads to yellowing and shedding of leaves and reduction in boll formation. Sooty mold fungus
develops on honeydew produced by the insect. This diminishes photosynthesis and lint
quality. The cotton crop infested by whitefly in Andhra Pradesh could be detected by remote
sensing. Landsat false colour composites were used for assessment of the cotton crop. The
changes in the red colour on Landsat false colour composite in comparison to previous
normal year supported by ground truth helped in identifying whitefly infested areas. Areas of
moderate (50% crop loss) and severe (80%) crop loss due to whitefly were easily identified.
The remote sensing can be used for the management of cotton white fly in the following
manner

i) to implement the legislation to stop continuous cotton growing in a specified area

ii) to demarcate areas requiring crop rotation
iii) to monitor the time of harvesting so as to break the life cycle of the pest
iv) to estimate the area under alternate hosts of whitefly such as tomato, chilli, tobacco etc.
for assessing the carry over potential of the pest in the off season.

Remote sensing and digital image analysis are methods of acquisition and interpretation of
measurements from a remote target without physical contact between the measuring device
and the object. The object can be analyzed many times, non-invasively and without damage
( Jensen, 2000).

Remote sensing platforms can be field-based, or mounted on aircraft and satellites; and the
data they capture is often characterized by four resolutions:
(1) spatial (what the smallest resolvable unit is on the ground, also called the pixel), (2)
spectral (how sensitive the spectra is sampled), (3) temporal (how often the data can be
captured) and (4) radiometric (the ability to discriminate very slight differences in reflected
or emitted energy).

Common pixel sizes are wide-ranging: weather satellites have pixel resolutions
larger than 1 km; the AVHRR sensor, an early multispectral sensor still in use has a 1km
pixel size; the series of Landsat sensor have 30 m pixels, and there are a range of newer
commercial satellites (e.g. Quickbird and IKONOS) that have near and under 1 m spatial
resolution. Sub-meter resolution imagery is increasingly common, especially with the use of
aircraft-borne sensors.
The spectral information contained in imagery can include multispectral (<10
bands of spectra, covering the visible and NIR portion of the spectrum), hyperspectral (10s to
100s of bands, covering a wider range of the spectrum) and thermal spectra (covering longer
wave infrared emittance spectra).
 Management of crop health and detection of stress from pests, diseases and
weeds can make use of remote sensing technology. Weeds can sometimes be mapped directly
in imagery, and plant stress from disease or insect infestation can be expressed by a plant in
many ways. Stress can influence stomata closure and transpiration rates, and impede
photosynthesis. Other stress symptoms include morphological changes such as leaf curling,
change in leaf angle, wilting or stunting, and chlorosis, necrosis, or premature abscission of
plant parts (Nilsson, 1995).
Plants stressed by disease display changes in absorption and reflectance in the visible and
near infrared (NIR) light due to decreases in chlorophyll content, changes in other pigments,
and changes to the internal cellular structure of the leaves (Zhang et al., 2003).
Plant stress usually results in an increase in visible reflectance (due to a decrease
in chlorophyll and a resulting decrease in absorption of visible light), and a decrease in NIR
reflectance from changes in the internal leaf structure. Weed mapping relies on weed plant
spectra being different from the crop target (Ustin et al., 2002).
Platforms used for vegetation mapping typically are sensitive in the following
spectral regions: near infrared (NIR) (725 – 900 nm), green (550 nm), red (650 – 690 nm)
and thermal (8,000 – 12,000 nm) (Barnes et al., 1996).

The first Landsat-based sensor, the Multispectral Scanner (MSS) sampled the earth at 79 m
resolution with 7 visible, near-, mid- and thermal-infrared bands. This was quickly followed
by a series of other Landsat launches with the very successful Thematic Mapper sensors
which
had similar spectral resolution, but had 30 m spatial resolution.

Most recently, the Enhanced Thematic Mapper sensor onboard Landsat 7,

launched in 1999, provides comparable spatial and spectral resolution, and includes a 15 m
resolution panchromatic band. While more commonly used in forestry applications, there are
broad-scale remote sensing examples from the agricultural sector. Landsat 2 imagery was
used to discriminate between cabbage and potato fields for subsequent evaluation of clubroot
disease (pathogen: Plasmodiophora brassicae).
Nagarajan et al. (1984) used Landsat MSS data to detect leaf rust (pathogen:
Puccinia recondita f. sp. tritici) and yellow rust (pathogen: Puccinia striiformis) over large
areas of wheat in Pakistan in a preliminary study of the efficacy of remote sensing for disease
detection.
Nutter et al. (2002) used a combination of Landsat 7 and high spatial resolution
multispectral imagery to map damage caused by soybean cyst nematode (Heterodera
glycines) in crops near Ames, IA. They also found the imagery useful for indicating soybean
crop quality and yield. The Hyperion sensor, on board the new E0-1 satellite provides
continuation of broad spatial coverage with increased spectral sensitivity (over 200 bands
from 0.4 – 2.5 nm) that can help in plant disease or pest damage discrimination.
Sudbrink et al. (2003) used an aircraft-borne multispectral imaging camera to
monitor insect infestations on beet and cabbage plants. They report a significant negative
correlation between beet armyworm (Spodoptera exigua) hits and transformed normalized
difference vegetation index (NDVI) values on two dates across the growing season. This
association may be useful in the development of future sampling plans or site-specific
management techniques that direct insecticide applications for beet armyworm. They had less
success detection via remote sensing an infestation of cabbage looper (Trichoplusia ni).
The greater spectral discrimination provided by hyperspectral imagery combined
with high signal to noise ratio allows differentiation between vegetation characteristics over
small spatial areas (Zhang et al., 2003).
 Mirik et al. (2006) used field-based spectrometry and digital camera imagery to
investigate the relationship between reflectance and damage caused by greenbug (Schizaphis
graminum (Rondani) (Hemiptera: Aphididae)) on winter wheat.The report greenbug density
was positively correlated with visible reflectance, and negatively correlated with NIR
reflectance.
Fitzgerald et al. (2004) used AVIRIS imagery at 18m ground resolution to identify
the position, spatial extent and severity of strawberry spider mite (Tetranychus turkestani),
infestation on cotton crops in Shafter, CA USA. They used spectral mixture analysis (SMA)
technique, a tool based on the assumption that each pixel is a physical mixture of multiple
components (endmembers) and the spectrum of this mixed pixel is a linear combination of
the
endmember spectra. Remote sensing applications measure the effects of disease or pests, and
cannot identify the specific pathogen or stressor, but only map stress or mortality patterns,
giving clues to likely cause. Follow up ground verification must be performed for diagnostic
identification of pathogen or pest.

SPATIAL ANALYSIS

Natural variations in soil characteristics (texture, organic matter), biological characteristics

(soil microbial populations, available nutrients), weed populations, presence of pathogens and
insects, and the interactions among these factors combine to influence crop quality and yield.
While remote sensing can map some of these factors, not all of these can easily
be detected remotely, and often field measurements are critical for establishing spatial
heterogeneity of field conditions and characteristics.
Consequently, in cases where remotely sensed technology is not applicable or
remotely sensed data are unavailable, field sampling is required to study crop disease
dispersal or insect damage, and Spatial Analytical (SA) techniques can often lend added
understanding of the field and plant conditions, and processes controlling pest distribution.
Specifically, SA methods can be used to interpolate point samples to create a continuous
surface, or to describe spatial pattern, or to investigate the spatial co-occurrence and
relationships between complex factors. The tools can help farmers map the pattern of a pest
across a field from samples, help managers guide pest sampling efforts, and aid researchers
develop epidemiological hypotheses about pest establishment and spread, and the effects of
control (Van Maanen & Xu, 2003).
Measurements from the field are often gathered as point data (e.g. individual
plants), and can be interpolated into a 2-dimensional disease distribution map or 3-
dimensional surfaces using well-defined analytical methods called spatial interpolators.
Park & Tollefson (2005) used the kriging interpolator to predict the spatial
distribution of corn rootworm (Coleoptera: Chrysomelidae), and found that adult counts in
the ear zone at peak population densities during the one year could predict adult emergence in
the
following year. Similarly, Ellsbury et al. (1998) used a geostatistical approach to characterize
spatial variability in western (Diabrotica virgifera virgifera) and northern (Diabrotica
barberi) corn rootworm adult emergence patterns, and applied semivariograms based on
spherical spatial models to produce contour density maps of the adult populations in the
fields.

CONCLUSIONS
Precision Agriculture is built on the ability of GPS integrated with GIS to promote variable
management practices according to variable field characteristics and adoption of these spatial
tools in agriculture for Integrated Pest Management and variable rate application of
pesticides, herbicide and fertilizers will likely continue.
Remote sensing and spatial analyses are of additional value in planning crop
management practices, but these technologies are used less often for several reasons. First,
high spatial resolution imagery is not easily available for all rural areas and hyperspectral
imagery usually requires contracting; second, there is a lack of technical knowledge about
remote sensing and spatial analysis by farmers and consultants; and finally, much of the
benefit of remote sensing and spatial analysis depends on time-sensitive mapping and near
real-time image acquisition and product delivery.
Despite these challenges, remote sensing and spatial analysis can provide valuable
information in an IPM context, allowing for a complete understanding (via remote mapping
or spatial modeling) of the spatial complexity of the abiotic and biotic characteristics of a
field and its crops, and providing information about the disease and pest populations that are
present, or likely to occur.
The transition to the utilization of a full suite of geospatial tools for integrated
pest management is agricultural sector is mirrored in the realm of forestry, where increasing
and large-scale pest and disease attacks are increasingly reported, and where the spatial
pattern across landscape-scales of pest hosts, pest and pathogen population dynamics and
landscape structure interact to at times promote pest establishment.
As in agricultural settings, geospatial technologies are making forestry
management more precise and spatially comprehensive: and a better articulation of resources
across space yields new insights to yield, pest and control dynamics. New access to data and
technology will likely promote the transition of these tools from a research to an applied
domain across both sectors.

References:

Barnes, E. M., Moran, M. S., Pinter Jr., P. J., & Clarke, T. R. (1996). Multispectral remote
sensing and site-specific agriculture: examples of current technology and future
possibilities. Proceedings of the 3rd International Conference on Remote Sensing. 41:
533-536.
Fitzgerald, G., Maas, S., & Detar, W. (2004). Spider mite detection and canopy component
mapping in cotton using hyperspectral imagery and spectral mixture analysis.
Precision Agriculture. 5: 275-289.

Jacobsen, B. J. (1997). Role of plant pathology in Integrated Pest Management. Annual

Review of Phytopathology. 35: 373-391.

Jensen, J. R. (2000). Remote Sensing of the Environment: An Earth Resource Perspective.

Prentice Hall Series of Geographic Information Science. Prentice Hall, New Jersey,
544 pp.
Johnson, C. E., & Barton, C. C. (2004). Where in the world are my field plots? Using GPS
effectively in environmental field studies. Frontiers in Ecology and the Environment,
2, 475-482.

Mirik, M., Michels, G. J., Kassymzhanova-Mirik, S., Elliott, N. C., & Bowling, R. (2006a).
Hyperspectral spectrometry as a means to differentiate uninfested and infested winter
wheat by greenbug (Hemiptera : Aphididae). Journal of Economic Entomology, 99:
1682-1690.

Nagarajan, S., Siebold, G., Kranz, J., Saari, E. E., & Joshi, L. M. (1984). Monitoring wheat
rust epidemics with the Landat-2 satellite. Phytopathology, 74, 585-587.

Nilsson, H. E. (1995b). Remote sensing and image analysis in plant pathology. Canadian
Journal of Plant Pathology, 17, 154-166.

Nutter, F., Tylka, G., Guan, J., Moreira, A., Marett, C., Rosburg, T., et al. (2002). Use of
remote sensing to detect soybean cyst nematode-induced plant stress. Journal of
Nematology, 34, 222-231.

Park, Y. L., & Tollefson, J. J. (2005). Spatial prediction of corn rootworm adult emergence in
Iowa cornfields. Journal of Economic Entomology, 98, 121-128.

Sudbrink, D. L., Jr., Harris, F. A., Robbins, J. T., English, P. J., & Willers, J. L. (2003).
Evaluation of remote sensing to identify variability in cotton plant growth and
correlation with larval densities of beet armyworm and cabbage looper (Lepidoptera:
Noctuidae). Florida Entomologist, 86, 290-294.

Ustin, S. L., Di Pietro, D., Olmstead, K., Underwood, E., & Scheer, G. J. (2002).
Hyperspectral
remote sensing for invasive species detection and mapping. International Geoscience
and Remote Sensing Symposium: 24th Canadian Symposium on Remote Sensing,
Toronto, Canada.

Van Maanen, A., & Xu, X. M. (2003). Modelling plant disease epidemics. European Journal
of Plant Pathology, 109, 669-682.

Whipker, L., & Akridge, J. (2006). 2006 Precision Agriculture Services Dealership Survey
Results. 06-10, Purdue University, College of Agriculture, Department of
Agricultural Economics.

Zhang, M., Qin, Z., Liu, X., & Ustin, S. L. (2003). Detection of stress in tomatoes induced by
late blight disease in California, USA, using hyperspectral remote sensing.
International Journal of Applied Earth Observation and Geoinformation, 4, 295-
310.