Northern Prairie Wildlife Research Center
We found evidence of change in water conditions between dates of video and BGS when comparing video images with aerial photographs, and when comparing our roadside survey with ground BGS. The difference, 0.07, in the estimate of omission error (0.14) calculated from the comparison of ponds from video interpretation with our roadside survey of ponds and the estimate (0.21) calculated from the comparison of ponds from video interpretation with ponds from ground BGS indicates error due to a change in ponds between time of videography and ground BGS. We also found evidence of a small number of errors in the ground BGS from these comparisons. Mistaken identification of ponds by the BGS ground crew may have caused these errors. Further, some ponds identified as commission errors may have been correctly interpreted. Wetland basins can be absent from the list of wetland basins in ground BGS because they were dry at the time basins were delineated on photographs. This can occur for Type I and III wetlands that are completely cultivated during a drought and may not be immediately incorporated into the list of basins when water returns to the prairies unless the pond is visible from a delineated wetland basin. A similar situation can occur for wetland basins along transect boundaries where ponds may intersect transect boundaries in only some years.
Different criteria for counting ponds was another source of disagreement between the number of ponds counted during video and ground BGS. Roadside ditches are not counted in ground or aerial BGS. However, ground BGS does include Type I ponds adjacent to roads. Differentiating between a roadside ditch and a Type I pond adjacent to the roadside is difficult from video. Six percent of commission errors occurred adjacent to the roadside at the Peterson, Hendon, Nut Mountain, and Pleasantdale air-ground segments, where we did not conduct roadside surveys following video survey.
Using a training set prior to interpretation and carefully defined interpretation criteria will enable even inexperienced interpreters to perform a relatively accurate interpretation of ponds. Type I and III wetlands with vestigial and recessional water levels can be difficult to detect even for experienced interpreters. The potential existence of these ponds is a factor contributing to commission error. Tree shadows in aspen parkland vegetation and moist soil depressions in agricultural fields can be confused with Type I and III wetlands with vestigial and recessional water levels.
A useful relationship was found between pond counts from interpreting video and pond counts from ground BGS. The departure from an exact linear relationship is due to omission and commission errors in video interpretation, lack of compensation of 1 type of error for the other, and problems comparing video interpretation of ponds with ground BGS of ponds. Video interpretation did not appear to systematically over- or underestimate the pond count from ground BGS. The average number of ponds from video interpretation and ground BGS were not different. The aerial BGS pond count tended to underestimate number of ponds counted from the video interpretation of air-ground and transect segments.