Northern Prairie Wildlife Research Center
Each video image was displayed on the monitor, the transect center line aligned with the road defining the center line of the transect, and the location of boundaries adjusted to 200 m on each side of the road. We compared transect boundary locations (computed using the nominal video scale) with location of transect boundaries on 1:12,000 scale maps or aerial photographs for several video images of an air-ground segment. If transect boundary locations did not correspond to those on the map, we determined an average correction for the video scale before beginning interpretation.
We used interactive image thresholding to identify potential ponds. The analyst then used tone, size, shape, texture, pattern, and association (Estes et al. 1983) to decide if a feature was a pond and then marked the pond with the computer mouse. We counted multiple ponds within a single wetland basin as separate ponds. Wetland basins joined by water were counted as a single pond. We counted a wetland basin divided by a road as 2 ponds if both parts contained water. We counted ponds that intersected a transect boundary in any part. Streams containing water were counted as separate ponds each time they entered the transect. We did not attempt to determine wetland type or water level, or distinguish between natural and artificial wetlands. The output of the analysis was the number of ponds and number of wetland basins with greater than or equal to 1 pond.
Video images for the air-ground segments were interpreted by an observer who had > 10 years', experience interpretating near infrared images and 2 summers of experience with prairie wetlands. All ponds detected within the air-ground segment were counted.
Video images for the 30 transect segments were interpreted by an observer who had < 1 year of experience with image interpretation and prairie wetlands. We used the criteria for counting ponds described above except that we subjectively excluded Type I wetlands and counted ponds from only 1 side of the transect segments so pond counts could be compared with pond counts from aerial BGS.
We evaluated partitioning of video images among transect segments by visually locating the start and end of each transect segment in files created using GPS latitude and longitude estimates in the Prepare Video Disk program. The GPS estimates of latitude and longitude without differential correction were sufficiently accurate to determine the video image that contained the start and end of transect segments.