1 October 2003
6-8 Case Study 6: A SPOT Survey of Wild Rice in Northern Minnesota
Subject Area: Agriculture
Purpose: To estimate the percentage of wild rice in a wetland environment
(1) A vegetation survey of natural wild rice surrounding three neighboring lakes 200
miles (518 km) south of St. Paul, Minnesota, was conducted to provide a base map for pol-
lutant and water level monitoring. The study presented here utilized standard supervised
classification, based on ground-truth, of high-resolution SPOT data. Wild rice is a natural
marsh grass that is sensitive to water level changes and to changes in phosphorous concen-
trations; increases in phosphorous and water levels can significantly destroy wild rice com-
munities. This is of concern as this important grass is a staple in the Chippewa Indian diet
and is consumed by migratory birds.
(2) The researchers in this study were tasked with mapping and estimating the acreage
of wild rice surrounding three lakes in Minnesota. Three spectral classes where developed
with the use of a supervised classification to delineate the grass and its varying substrate.
b. Description of Methods. Ground truth data were collected simultaneously with SPOT
over flight. The ground truth data included information regarding vegetation and substrate
type as well as the sites corresponding UTM (global position in the Universal Transverse
Mercator coordinate system).
c. Field Work. In the field, 18 ground control points (GCPs) were collected for rectifi-
cation of the SPOT image and an additional 132 ground truth points were collected for the
supervised classification algorithm. This data collection coincided with the SPOT over
d. Sensor System. SPOT was chosen for its optimal detection of vegetation in the pres-
ence of inorganic ground cover (i.e., water). Vegetation absorbs both red and blue radiation,
while reflecting green and near infrared (NIR) because of chlorophyll production. This
matched well with the spectrum data provided by SPOT (which maintains green, red, and
NIR bands among others).
e. Study Results.
(1) Prior to the classification process, it had been predicted that the wild rice would
dominate one spectral class, as wild rice is spectrally distinct from other vegetation. Open-
ings in the grass canopy, however, contributed to the spectral mixing observed in the image
scene. Three spectrally distinct populations were noted, likely because of the heterogeneity
of the background reflectance, varying crop canopy, and varying water content in the sub-