EM 1110-2-2907
1 October 2003
c. Field Work. Direct field observations were not necessary as a web-camera mounted on
a bridge provided the visual documentation of ice conditions in the river. At the Missouri
River site, web-cameras have been strategically placed in a variety of locations in the US by
ERDC/CRREL. To view the Missouri River images used in this study, as well as other river
web-camera images, go to http://webcam.crrel.usace.army.mil. Study sites without a web-
camera relied on District contacts for field information. At Red Lake River near Grand
Forks, North Dakota, field reconnaissance ice surveys were conducted by the Corps St.
Paul, Minnesota, District office.
d. Sensor Selection and Image Post-Processing.
(1) As stated above, RADARSAT SAR data was chosen for this study. Radar data
have already proven their utility in sea ice mapping and monitoring (Carsey, 1989). Radar
can aid in determining ice concentration, classification, ice motion monitoring, and ice fea-
ture changes. The study reviewed here adapted methods used to study large ice sheets to the
evaluation of smaller more temporal river ice.
(2) The acquired radar images were visually analyzed and classified using an unsu-
pervised classification to delineate open water, moving ice floes, and stationary ice covers.
The delineation of river channels was undertaken by two methods, described above (hand-
digitization and TM extraction and overlay).
e. Study Results. The following description summarizes the ice condition results stem-
ming from each river surveyed:
"In the Mississippi River imagery near St. Louis, Missouri, the wide channel width (5002000
meters) contributed to identifying river ice with RADARSAT imagery. In the 2002 image it was
determined that 30% of the channel had ice in the flow, and in the 2003 image, it was deter-
mined that there was 100% ice cover. Additionally, this ice cover was separated into forms of
ice; brash ice and border ice. In the 2003 image it is believed that the brash ice formed as a re-
sult of navigation ice-breaking activities.
(1) In the Missouri River imagery near Bismarck, North Dakota, the channel width
(4001000 m) was suitable, and river ice was determined from the RADARSAT imagery.
The 2002 image showed that 77% of the channel had ice in the flow, and in the 2003 image,
only 21% of the channel had ice. The 2003 imagery was acquired before full icing condi-
tions, and a small amount of ice was interpreted to exist.
(2) In the Red Lake River imagery near the confluence with the Red River of the
North at Grand Forks, North Dakota, the river channel is narrow (4075 m). The narrowness
of the channel limited the process of delineating the channel boundary on the imagery. As a
result of the narrow channel width, river ice was not determined by this process. However,
ice surveys were conducted by the US Army Corps of Engineers during the time of image
acquisitions, and an ice cover was recorded in both 2002 and 2003.
f. Conclusions. RADARSAT SAR data were able to detect ice on rivers with widths
ranging from 400 to 2000 m. Despite RADARSAT's 10-m resolution, this data set was un-
able to detect the present of ice on the narrower Red Lake River, with a width of 4075 m.
RADARSAT's overall suitability for detecting river ice and ice conditions was shown to be
6-9
Previous Page
