PROJECT TEAM
James D. Hipple, Geography-ICREST (University
of Missouri-Columbia)
Tim Matisziw, Geography-ICREST (University of Missouri-Columbia)
Tim Butchart, Geography-ICREST (University of Missouri-Columbia)
Bill Wiebold, Agronomy (University of Missouri-Columbia)
Newell Kitchen (USDA)
Kenneth Sudduth (USDA)
Joe Sorrels (J-Mar Agri Group)
Kevin Maynard (J-Mar Agri Group)
Dale Ludwig, CEO (Missouri Soybean Association/ Missouri Soybean
Merchandising Council)
PROJECT STATUS
There are a number of concurrent activities
under way in ICREST’s Multi-scale modeling of soybean productivity
in the Midwest project. These tasks include 1) data processing,
2) model development, 3) business plan development, and 4) field
trials.
1.Data
Processing
All of the remotely sensed data (Landsat 5 & 7) from 1999 has
been processed and is presently being analyzed. Tasseled Cap coefficients,
NDVI, mSAVI, and other vegetation indices have been computed and
integrated into the database containing the field data provided
by J-Mar Agri Group for close to 20,000 acres of crop land in the
Bootheel Region of Missouri. Data were obtained from Dr. Bill Wiebold
and Kenneth Sudduth for several of the Belmont fields of J-Mar Agri
Group. The data set includes soil electroconductivity, protein content
and oil content obtained through NIR methods, and fatty acid profiles
obtained through wet chemistry for 1999’s soybean crop. The
soil electroconductivity measures were collected in April 1999 and
the protein, oil and fatty acid profiles were obtained from insitu
measurements in late September 1999. The electroconductivity dataset
consisted of around 7517 data points while the soybean protein data
consisted of approximately 50 data points. All of the data points
from both datasets have been complemented with tasseled cap (greenness,
brightness, third components), normalized difference of vegetation,
and modified soil adjusted vegetation index values throughout the
1999 growing season associated with them to aid in geostatistical
analysis of the points. Other components, such as crop yield and
soil type have also been integrated into this database. Our data
analysis is primarily for soybean crops, but data sets include yield/productivity
measures for other crops, including millet, wheat and corn.
We are presently preparing for field work in the fields concurrent
with planned Landsat 7 overflights for the 2000 growing season.
Additional soil electroconductivity, protein content and oil content
obtained through NIR methods, and fatty acid profiles for the 2000
soybean crop are to be obtained, as well as additional GPS and field
spectrometry data.
2.Model Development
Datasets created from the electroconductivity, protein, yield, soil,
and vegetation index data are currently being statistically analyzed.
Models are being created to establish how well various vegetation
indicies collected throughout the season perform in predicting electroconductivity,
soybean oil and protein content, soybean yield, and soil type. Thus
far, linear stepwise regression techniques have been applied to
the datasets to aide in determining which vegetation indicies from
particular dates are the largest contributors to the model’s
predictive value. Also, electroconductivity and soil type are being
investigated as to better understand how they may affect soybean
yield and soybean oil and protein content levels. We are looking
at data mining (knowledge discovery in databases, or KDD) methods
for analyzing the immense amount of data.
3.Business Plan Development
Part of the Synergy concept is to examine the needs of the user
community for the determination of whether an effective delivery
system for satellite data & derived information can be developed.
The first of these tasks is to understand the spatial distribution
of key crops in the United States. Our goal is to develop the strategy,
or business plan, for an InfoMART (which we call the Missouri AgriMart).
The AgriMart would be an interface to EOSDIS where farmers or growers
could subscribe, or register their ‘parcel’ of land,
and, throughout the growing season they could receive information
products derived from NASA EOS satellite imagery. The interface
to the information would be web-based and the information available
in a format that could be viewed on screen, or downloaded and imported
into a variety of commercial-off-the-shelf (COTS) crop management
software systems.
One of the first steps in the business plan development is to determine
the economic feasibility of acquiring satellite imagery for regions
of the country growing the following agricultural commodities: soybean,
corn, wheat, millet, and sorghum. Data were extracted by county
from the 1997 USDA NAAS Census of Agriculture and include bushels
harvested, number of farms, and land under cultivation. Criteria
are presently being set up to determine the minimum data needs to
deliver the AgriMart Landsat 7 derived information products to farmers
at monthly or bi-mothly intervals nationwide for each of these commodities.
Cost structures are being determined for the number of scenes needed
and likelihood of at least one cloud free scene per month over a
given area.
Preliminary estimates are as follows:
Preliminary estimates for imagery needed are on a crop-by-crop
basis and do not include overlap between the crops (i.e., overlapping
geographic ranges between corn & soybean) and potential resale
value of the acquired imagery as a VAR (Landsat 7 is public domain
data once ‘purchased’, among other issues.
4.Field Trials
A series of field trials were set up at the University of Missouri’s
agricultural experiment farm known as Bradford Farm. The field was
planted with approximately six rows of soybeans in late May. Each
row contained six varieties of soybeans and rows alternated between
Roundup Ready (a Genetically Modified Organism or GMO) and non-Roundup
Ready varieties. The varieties were randomized over six rows. Field
spectroscopy work begins the week of June 12. Field work includes
the collection of spectral signature of the 12 varieties (6 Roundup
Ready and 6 non-Roundup Ready) on a weekly basis. Leaf collection
will alternate every two weeks, where samples will be collected
and analyzed under full illumination in order to minimize effects
of atmospheric absorption.
Issues:
A heads up: We will be inquiring every 16 days, or so, to get the
most current Landsat 7 image of our Bootheel study area. As noted
in previous emails and the proposal we will be acquiring approximately
8 images through this year’s growing season.
|
