Technical
Summary:
The term “Integrated Crop Management” is
used herein to indicate the application of remote sensing and geospatial
technologies to the detection, identification, and treatment of agricultural
pests such as weeds and insects as well as to the application of
harvest-aids with the objective of maximizing crop yields while reducing
costs. The proposed study will investigate and compare the use of
high-resolution satellite and aerial multispectral image data to monitor
physiologic growth; to detect, identify, and develop spatially variable weed
treatment maps; and to develop spatially variable harvest-aid treatment
maps.
The proposed effort will have both research
and practical components. The research components will compare
high-resolution satellite and aerial multispectral image data, data analysis
products, and resultant vegetation indices collected for the same fields for
assessing the physiologic growth of cotton over time, crop variability
within fields, and the presence of anomalies within fields. The project’s
practical components will provide automated data processing and improved
crop data management workflows that maximize the use of efficient methods
for data acquisition, data processing, field data collection, and data
product distribution. The research components will build on past
multispectral image data research that used high resolution aerial image
data and will help answer questions about whether new high-resolution,
multispectral satellite image data can be used with or substitute for
high-resolution, multispectral aerial image data to provide information
needed for integrated crop management. The practical components will provide
improved methods and technologies for data processing, distribution, and
integrated use to give farmers better tools and more timely information for
critical decision making.
The proposed project will involve close
cooperation with on-going research on pest management and cotton harvest-aid
application led by D. Reynolds, soils research led by M. Cox, and
computational modeling research led by R. King. Ground-truthing will be
conducted by MSU’s Data Acquisition Work Group (DAWG). Portable full-range
spectral radiometers will be used by the DAWG for spectral ground-truthing
of research areas. The proposed research will be coordinated with MAFES
disciplinary scientists, extension specialists in cotton, and extension
geospatial technology specialists. Industry collaboration is proposed with
DigitalGlobe and Mid South Ag Data to study the use of high-resolution
Quickbird II satellite image data and AgroWatch data products for precision
agriculture and crop management applications.
Objectives:
The overall research objective of the proposed project is to study and
compare multispectral image data collected for the same fields from
high-resolution aerial and satellite imagery for monitoring and assessing:
physiologic growth of cotton over time, crop variability within fields, and
anomalies within fields. The research will provide understanding about the
use of high-resolution aerial and satellite multispectral image data for
integrated crop management. Practical objectives involve developing and
applying automated processing techniques, improved data distribution
methods, and mobile computing methods. The effort will result in a seamless
digital workflow that minimizes the time required to go from raw
multispectral image to a treatment map for weed control or harvest aids. In
doing this, the project will accomplish research and practical objectives
including the following:
·
Research
Objective: Evaluate new high-resolution satellite and aerial
multispectral data sources for agricultural base map
development.
-
Research Objective: Investigate the
generation of vegetation indices from aerial and satellite multispectral
image sources and study the differences in data, analysis techniques, and
analysis products from the platforms contrasting both to the information
provided by portable full-range spectral radiometers.
-
Research Objective: Compare the use of
satellite and aerial multispectral data and data analysis products for
monitoring crop physiological growth and condition.
-
Research Objective: Compare the use of
satellite and aerial multispectral data and data analysis products for
detecting anomalies in research fields.
-
Research Objective: Investigate and
document the use of satellite and aerial multispectral data and data
analysis products for assessing crop variability within the research
fields over the growing season and compare crop variability evidenced by
image data to variability in monitored crop yield.
-
Practical Objective: Test and validate
the integrated use of mobile computing portable digital assistant (PDA)
devices equipped with GPS and GIS field scouting software for cotton
scouting and crop management.
-
Practical Objective: Test and validate
methods for integrating data and results between mobile computing GPS/GIS
platforms and decision support systems including expert systems and
spatially variable application mapping systems.
Procedures:
Data collection activities are planned for
two to three research fields in the Brooksville and Macon areas. The
selected fields are currently studied by MAFES scientists for pest
management and harvest aids. For these fields, extensive soil and crop
studies have been conducted and/or are ongoing. Ongoing MAFES research
activities include close monitoring of plant physiological growth and
condition. The proposed research will cooperate with other research
activities to collect important plant physiological growth and condition
data. The growth stage and condition data will be correlated to remote
sensing information to assess the use of both aerial and satellite
multispectral data to monitor crop growth.
Similar to past and ongoing research
efforts, at each of the research fields, the experimental design includes
the selection of measurement/observation sites, two meters in length in
quadrants of the field, which will be marked by GPS in the field for data
collection. Through the ongoing research efforts at these sites, data
collection efforts will be continued for the effort wherein:
“Data collected at
each site include spectral data using an ASD (Analytical Spectral Device),
nodes above white flower (NAWF), nodes above cracked boll (NACB), total
bolls, and percent open bolls. Data that will be collected on the whole
plot level includes percent defoliation, percent desiccated leaves, percent
green leaves, seed cotton yield, percent turnout, and fiber quality.” (D.
Reynolds written communication)
At these research fields, aerial
multispectral data have been collected in the past and will continue to be
acquired through routine RSTC data collection efforts. In conjunction with
remote sensing data collection, the DAWG will collect hand-held radiometric
field data and participate in the collection of data about the physiologic
growth and condition of the crop at observation sites as well as the
presence of weeds in the fields.
To provide uniform data and results, it is
important that field data collection methods and data processing procedures
are tested, validated, and documented as standard procedures. Collecting and
processing data using standard procedures enhances the ability to develop
scientifically sound, reproducible results. Closely coordinating with the
on-going efforts of MAFES, ARS, and RSTC scientists, the project will have
procedures that fall into the following categories:
Base Map Compilation
1.
Acquire DigitalGlobe Quickbird data and the
AgroWatch information products for developing a high-resolution,
georeferenced base map including bare-soils characterization of the research
sites. The selection of several research plots that are inside of a ¼ scene
would provide the best effective use of money in the acquisition of base
map, soils, and image data. Selection of sites at Brooksville and Macon
would provide ideal ability to effectively manage image and field data
acquisition.
2.
Compile for selected research fields existing
image and geospatial data including elevation data, sample grids, soil
information, historic crop yields, and other available data. It is currently
planned that the research will be conducted on 3 fields (2 in Brooksville
and 1 in Macon) currently under research by D. Reynolds.
Multispectral Image Data Collection
1.
Coordinate closely with RSTC about the collection
of GeoVantage aerial multispectral image data. Collect spectral data for
observation sites in the research plots using a hand-held ASD (Analytical
Spectral Device) every two weeks to coincide with collection of aerial
spectral images until harvest.
2.
Coordinate closely with MAFES, ARS, and RSTC as to
ideal data collection times for multispectral data for specific crops and
fields. Schedule satellite data acquisition with Mid South Ag Data and
DigitalGlobe to coincide with approximate time of aerial image data
collection for several specific time-frames related to important cotton
physiological growth stages during each project year.
Multispectral Data Processing
1.
Assure that aerial and satellite multispectral
data are properly georeferenced by plotting the data and visually assuring
horizontal accuracy of the data as compared to base map data. Conduct
routine processing by generating NDVI and other images such as a CIR and a
true color version of the image data. Work with the Remote Sensing
Technologies Center and DigitalGlobe to use the DigitalGlobe algorithm or a
similar algorithm, for generating the GVI vegetation index output for
satellite and aerial multispectral data for comparison with NDVI as the base
layer for the scouting map image.
The standard vegetation index to be used will be NDVI which has been
extensively used to monitor plant growth. NDVI uses the difference between
the near-infrared radiation and visible red radiation divided by
near-infrared radiation plus visible red radiation
“to quantify the density of plant growth on
the earth. The result of this formula is called the Normalized Difference
Vegetation Index (NDVI). Written mathematically, the formula is:
NDVI = (NIR — RED)/(NIR + RED)
Calculations of NDVI for a given pixel
always result in a number that ranges from minus one (-1) to plus one (+1);
however, no green leaves gives a value close to zero. A zero means no
vegetation and close to
+1 (0.8 - 0.9) indicates the highest
possible density of green leaves.”
(modified from http://earthobservatory.nasa.gov/Library/MeasuringVegetation/measuring_vegetation_2.html)
Some investigations
will also be made to explore the effect of past RDACS multispectral channel
center wavelengths and FWHM on the results and compared with the
GeoAdvantage channel parameters.
2.
Coordinate with the Remote Sensing Technologies
Center to develop efficient methods for uploading georegistered aerial and
satellite multispectral data to the data server and automating the
processing of vegetation indices.
Field Scouting
1.
Acquire, configure, test, and develop standard use
procedures for Compaq IPAQ PDAs, GPS, and Farmworks Site Mate and/or ArcPad
Software for field scouting. Accuracy characteristics of GPS-integrated PDA
will be compared to back-pack equipment and spatial characteristics of map
products will be compiled and compared with the accuracy requirements of the
crop management grid.
2.
Develop, test, and document standard methods and
procedures for using a PDA, Farmworks SiteMate, ArcPad, and GPS data to
prepare scout maps, capture locations, collect physiologic growth attribute
information, and other location specific information such as pests.
Data Management and Transfer
1.
Develop, test, and document standard methods and
procedures for coordinate system transformation, changing data file formats
as needed, and data transfer. This will build on previously developed
algorithms and scripts developed for RDACS processing.
2.
Develop, test, and document standard methods and
procedures for generating application maps and automate the process by
developing batch routines where possible.
Data Analysis And Results:
After collecting field and remote sensing data, a significant research
component will involve data analysis and generation of experimental results.
Analyses to be performed include, but are not limited to the following:
1.
ASD
field data and plant observations will be directly compared to the aerial
and satellite image data results for all test sites and image collections.
Graphical and statistical analysis of results will be conducted to ascertain
how well the aerial and satellite image data monitor plant physiological
growth. This will be coordinated with on-going studies in the Computational
Modeling activity of RSTC.
-
Vegetation index maps will be prepared
for each research field for each aerial and satellite data collection
activity. Hand-held spectral data for test sites locations in the field
where data are to be collected will be compared against the field-wide
image results and results for the pixels in closest proximity to the test
sites to ascertain spectral quality of the multispectral images and to map
crop variability within the fields. Field activities will be conducted by
the DAWG and all ASD collections will be accompanied by a photo of the
test site collected with a digital camera.
-
Aerial and satellite images will be
processed to identify anomalies in the field that may be weeds. For each
aerial and satellite image data collection effort, feature detection will
be attempted, anomalies will be documented, and the presence of weeds at
predicted locations indicated by anomalies will be verified. Additionally,
as areas develop in the fields with significant weed growth, problem areas
will be mapped and checked in the earlier multispectral image to identify
spectral and spatial patterns indicative of early-stage weed growth.
Identified patterns will be evaluated for use in subsequent seasons to
improve early-season detection and treatment of weeds.
-
Vegetation index maps and other combined
analysis products maps will be used to develop harvest-aid treatment maps.
These efforts will be closely coordinated with other research directed at
harvest-aid treatment. The use of image data and analysis products to
develop the treatment maps will be documented.
5.
Yield
maps will be generated and compared to the image data analysis maps created
of the research fields for the entire season. Areas with reduced yield
within the fields will be identified, commodity experts will provide input
as to the likely cause of reduced yield, and these areas will be checked on
the satellite and aerial image data and data analysis products to identify
any factors that might have contributed to the reduction in yield in those
locations.
Research Products And Deliverables:
This project will result in at least two journal publications, the first of
which will document the comparison of high resolution satellite and aerial
image data and data analysis products to monitor cotton crop physiological
growth stages and detect the presence of weeds. The second paper will
document the analysis and use of high resolution satellite and aerial
multispectral image data and data combination products for generating
spatially variable weed and harvest-aid treatment maps. It is anticipated
that the publications will be jointly developed and authored with others who
are investigating pest management and harvest aids in the Brooksville and
Macon research fields.
Justification:
Vegetation indices (computed from multispectral image data) such as the
Normalized Difference Vegetation Index (NDVI) can be used to assess lushness
and vigor of plant growth. Completed and on-going research at Mississippi
State correlate remotely sensed crop data to cotton physiological maturity
and percent open bolls. NDVI and other similar vegetation information
products provide significant benefits in assisting scouts in the early
detection of insect infestations in agricultural fields and may provide
significant benefit in tracking the growth cycle of a crop, determining
variability of a crop within a field, detecting anomalies in a field, and
developing spatially variable treatments for pests and harvest-aids for
cotton.
High-resolution multispectral image data and data analysis products can
assist scouts to locate and identify weeds for the development of effective
herbicide treatments. In most cases, the effective treatment of weeds should
occur very early in the season to prevent yield loss for the current year.
Remotely sensed data may provide assistance throughout the crop year to
manage a crop, soils and base-map information, and may provide information
that will help plan treatments to reduce pest problems for the following
year.
Whether for management of insects or weed control, or for developing images
that assist in developing harvest-aid treatment maps, multispectral image
data may be used to develop scout images that can be used to enhance field
scouting efforts. The image and field data integrated results can be
ingested by software for making decision about specific chemicals that
should be applied for pest management, to develop treatment maps such as
spatially variable application maps for pest control, and for designing
specific harvest-aid plans for a crop.
High-resolution aerial multispectral image data can be used to detect the
presence of weeds in a field. Until recently, the resolution of satellite
image data did not provide sufficient spatial data resolution for
effectively detecting and locating weeds in agricultural fields, but new
satellite image data may provide sufficient resolution to detect weeds at
various growth stages in agricultural fields. New satellite platforms such
as the DigitalGlobe Quickbird II can provide multispectral image data that
are similar in spatial and spectral resolution to the data that have been
used to study and develop precision agriculture applications. Other
improvements in technology allow aerial and satellite data to be provided
shortly after it is collected as a processed and geometrically correct
dataset.
For crop management, having rapid availability of georeferenced image data
means that images and derived image products can be integrated with other
base map data and used for scouting and decision making without delay. Rapid
turnaround of image data from aerial and satellite multispectral image data
collection can provide farmers with timely data that are needed in precision
agriculture applications. Comparing the use of Quickbird II data to data
collected by the GeoVantage aerial multispectral platforms will provide
insight needed to better understand the best uses of data from each
platform; increase scientific support for the appropriate use in
agriculture applications of data from either platform type; and foster an
improved knowledge-base for understanding the strengths, weakness,
limitations, and costs of the multispectral image data that may be collected
from selected aerial and satellite data-collection platforms which provide
timely image data of similar spatial and spectral resolution.
Extensive aerial remotely sensed multispectral image data have been
collected by the RSTC to study cotton crop development and for crop
management in Brooksville, the Paul Good Farm, and other locations. ITD/Spectral
Visions RDACS data have been used in numerous studies to correlate image
data to crop physiological maturity, to assess vegetative stress, to study
soils moisture conditions, to identify and manage treatment of pests, and
for many other purposes (http://www.rstc.msstate.edu). The aerial
multispectral data collection efforts will continue in 2002 and future
years, but will be collected using new sensor platforms and improved data
processing technologies that will provide georeferenced, high-quality image
data. The image data may be directly integrated with other spatial data and
used for field work, scouting, and other direct-use purposes or processed to
provide information useful for crop management and decision making.
Building on past work that used aerial multispectral image data, this effort
will use GeoVantage aerial multispectral image data collected for RSTC and
QuickBird II satellite image data from DigitalGlobe. Both data sets will
provide multispectral image data with high spatial resolution and similar
spectral data. The effective use of multispectral data for agricultural
applications requires that the intended use of the data be scientifically
proven and that the methods of use be adequately understood and described to
facilitate use of the data within the time-sensitive context of crop
management.
The proposed study will explore the scientific basis for best use of new
multispectral satellite and aerial data for ICM and will demonstrate
practical aspects of the technology – That it is possible to effectively
collect, process, and distribute the geospatial data for use within the time
constraints that are part of real world practices.
Literature Review:
Cathey, G.W. 1986. Physiology of Defoliation in Cotton Production. In
J.R. Mauney and J.M. Stewart, eds. Cotton Physiology. Memphis,
TN: The Cotton Foundation. pp. 143-154.
Dupont, J. K., Willers, J. L., Seal, M. R.,
and Hood, K. B. Precision pesticide applications using remote sensing.
Proc. 17-th Biennial Workshop on Color Photography and Videography in
Resource Assessment, Reno, NV. 1999.
Dupont, J. K., Willers, J. L., Seal, M. R.,
and Hood, K. B. Spatially variable insectide in cotton production: The role
for remote sensing. Proc. Second Internatl. Conf., Geospatial Information
in Agriculture and Forestry, Lake Buena Vista, FL. I-327-I-331. 2000.
Fromme, D.D. 1999. Utilizing COTMAN for Defoliation Timing and the
Management of Micronaire Values. Proc. Beltwide Cotton Conf. National
Cotton Council, Memphis, TN. 1: 529-530.
Kerby, T.A., J. Supak, J.C. Banks, and C. Snipes. 1992. Timing Defoliations
Using Nodes Above Cracked Boll. Proc. Beltwide Cotton Conf. National
Cotton Council, Memphis, TN. pp.155-156.
Maas, S.J. 1997. Structure and Reflectance of Irrigated Cotton Leaf
Canopies. Agron. J. 89:54-59.
Maas, S.J. 1998a. Estimating Cotton Canopy Ground Cover from Remotely
Sensed Scene Reflectance. Agron. J. 90:384-388.
Maas, S.J. 1998b. Remote Sensing Resources for Agriculture in the Next
Decade. Proc. Beltwide Cotton Conf. National Cotton Council, Memphis, TN.
1: 36-38.
McKinion, J. M., J. N. Jenkins, D. Akins,
S. B. Turner, J. L. Willers, E. Jallas, and F. D. Whisler. 2001. Analysis
of a precision agriculture approach to cotton production. Comp. Elec.
Agric. 32: 213-228.
Moran, M.S., T.R. Clarke, Y. Inoue, and A. Vidal. 1994. Estimating Crop
Water Deficit Using the Relation between Surface- Air Temperature and
Spectral Vegetation Index. Remote Sens, Environ. 49:246-263.
Qi, J., A. Chehbouni, A.R. Huete, Y.H. Kerr, and S. Soroochian. 1994. A
Modified Soil Adjusted Vegetation Index. Remote Sens. Environ.
48:119-126.
Roberts. D.A., M.O. Smith, and J.B. Adams. 1993. Green Vegetation,
Nonphotosynthetic Vegetation, and Soils in AVIRIS Data. Remote Sens.
Environ. 44:255-269.
Schepers, J.S. and D.D. Francis. 1998. Precision Agriculture- What’s in
Our Future. Commun. Soil Sci. Plant Anal. 29(11-14):1463-1469.
Schnug, E., K. Panten, and S. Haneklaus. 1998. Sampling and Nutrient
Recommendations - The Future. Commun. Soil Sci. Plant Anal. 29:
1455-1462.
Strickland, R.M., D.R. Ess, and S.D. Parsons. 1998. Precision Farming and
Precision Pest Management: The Power of New Crop Production Technologies.
J. Nematol. 30(4):431-435.
Turner III, C. 1997. G.P.S. Controlled Precision Spraying Minimizing Costs
and Environmental Impact. Proc. Beltwide Cotton Conf. National Cotton
Council, Memphis, TN. 1: 70.
Zhang, J.P., N.P.
Tugwell, M.J. Cochran, F.M. Bourland, D.M. Oosterhuis, and C.D. Klein.
1994. COTMAN: A Computer-Aided Cotton Management System for Late-Season
Practices. Proc. Beltwide Cotton Conf. National Cotton Council, Memphis,
TN. pp. 1286-1287.
Back to Top
Back to
Menu |
