USGS - science for a changing world

Northern Prairie Wildlife Research Center

  Home About NPWRC Our Science Staff Employment Contacts Common Questions About the Site

Improving Prairie Pond Counts with Aerial Video
and Global Positioning Systems

Results: Errors in Video Interpretation of Ponds

The Canadian Wildlife Service recorded 1,194 ponds during ground BGS of the 12 air-ground segments. We excluded 77 (6.4%) of these observations from analysis because of lack of comparative data (incomplete video coverage, or unknown or inaccurate location of ponds on BGS photographs). Composition of the 1,117 ponds used in analysis was 13 % Type I, 65% Type III, 8 % Type IV, and 2% Type V wetlands. Artificial wetlands and streams (n = 11) accounted for 10 and 1% of the ponds. Wetlands with vestigial, recessional, and intermediate water levels accounted for 15, 47, and 24% of ponds, respectively. Full, flooded or overflowing wetlands accounted for 15% of ponds.

The distribution of ponds among wetland type and water level from the video interpretation differed from that of ground BGS(X2 = 38.03, 14 df, P < 0.001) (Table 2). The proportion of ponds from video interpretation in Type I wetlands with vestigial water (4.0%) differed from the proportion classified as Type I wetlands with vestigial water from ground BGS (8.1%) (Z = 3.82, P < 0.001). The proportion of ponds from video interpretation in Type III wetlands with vestigial water (4.2%) differed from the proportion classified as Type III with vestigial water from ground BGS (6.4%)(Z = 2.2, P = 0.026).

Table 2. Percentage of ponds by wetland type and water level from ground Aerial Waterfowl Breeding Ground Population and Habitat Surveys in North America (n = 1,106) and percentage of ponds from video interpretation identified as correct (n = 880) on 12 air-ground segments in southern Saskatchewan in May 1992.


Water levela		 Wetland typeb
I III IV and V Artificial Total
Vestigial
Ground 8.14c 6.42c 0.09 0.27 14.92
Video 4.09c 4.20c 0.11 0.11 8.52
Recessional
Ground 4.70 36.89 2.17 2.80 46.56
Video 4.32 36.70 2.61 2.50 46.14
Intermediate
Ground 0.72 16.73 4.34 1.72 23.51
Video 0.80 18.98 5.45 2.16 27.39
Full, flooded, or overflowing
Ground 0 5.97 3.25 5.79 15.01
Video 0 7.16 3.98 6.82 17.95
Total
Ground 13.56 66.00 9.86 10.58 100.00
Video 9.20 67.05 12.16 11.59 100.00
a Classified according to Anonymous (1977).
b Classified according to Shaw and Fredine (1956).
c Signifies difference (P less than or equal to 0.05) between ground and video proportions within a cell.

Omission error for the 12 air-ground segments ranged from 0.05 to 0.36 with a mean of 0.21 (Table 3). Ponds in Type I and Type III wetlands accounted for 92% of omission errors (Table 4). Forty-six percent (69 of 150) of ponds in Type I wetlands were not detected on video. Ponds in wetland basins with vestigial and recessional water levels accounted for 88% of omission errors. Fifty-four percent (90 of 166) of ponds in wetland basins with a vestigial water level were not detected on video. Ponds in Type I wetlands with vestigial and recessional water levels accounted for 30% of the omission errors but only 13% of the total number of ponds. Ponds in farmyards or ponds where a change in water level between the time of the video survey and ground BGS was suspected accounted for the 0.05 (8 of 168) omission error for ponds in wetland basins with full or greater water levels. Omission error was 0.14 (66 of 459) for comparison of the video interpretation of ponds with our roadside survey of ponds.

Table 3. Omission and commission error for video interpretation of water presence and pond characteristics on 12 air-ground segments in southern Saskatchewan in May 1992 that are part of the Aerial Waterfowl Breeding Ground Population and Habitat Surveys in North America.


Air-ground
segment

Omission
(proportion)

Commission
(proportion)

Ponds with
vestigial
water levela (%)

 Type I and III
pondsb with
vestigial and
recessional
water level (%)
Shamrock 0.05 0.30 15 85
Hendon 0.09 0.23 18 64
Earl Grey 0.09 0.27 7 42
Cymric 0.11 0.07 8 48
St. Gregor 0.17 0.12 22 60
Hanley 0.21 0.07 13 60
Peterson 0.22 0.10 11 50
Pleasantdale 0.25 0.33 9 25
Craik 0.29 0.13 20 67
Nut Mountain 0.33 0.40 31 72
Tichfield 0.33 0.51 31 81
Marquis 0.36 0.18 29 51
Mean 0.21 0.22    
SE 0.03 0.04    
a Classified according to Anonymous (1977).
b Classified according to Shaw and Fredine (1956).

Table 4. Frequency of omission errors for video interpretation by wetland type and water level from ground Aerial Waterfowl Breeding Ground Population and Habitat Surveys on 12 air-ground segments in southern Saskatchewan in May 1992.


Water levela		 Wetland typeb
I III IV Artificial Stream Total
Vestigial no. 54 34 0 2 0 90
% 23.79 14.98   0.88   39.65
Row % 60.00 37.78   2.22    
Column % 78.26 24.29   13.33    
Recessional no. 14 85 1 9 1 110
% 6.17 37.44 0.44 3.96 0.44 48.46
Row % 12.73 77.27 0.91 8.18 0.91  
Column % 20.29 60.71 50.00 60.00 100.00  
Intermediate no. 1 18 0 0 0 19
% 0.44 7.93       8.37
Row % 5.26 94.74        
Column % 1.45 12.86        
Full, flooded, or
overflowing no.		 0 3 1 4 0 8
%   1.32 0.44 1.76   3.52
Row %   37.50 12.50 50.00    
Column %   2.14 50.00 26.67    
Total 69 140 2 15 1 227
  30.40 61.67 0.88 6.61 0.44 100.00
a Classified according to Anonymous (1977).
b Classified according to Shaw and Fredine (1956).

Commission error for the 12 air-ground segments ranged from 0.07 to 0.51 with a mean of 0.22 (Table 3). We detected 277 ponds that were identified as commission errors. Fifty-one percent (n = 141) of commission errors occurred at wetland basins included in ground BGS. Twenty-two percent (n = 60) of commission errors occurred at locations not included in ground BGS, but identified as wetlands on maps prepared by the U.S. Fish and Wildlife Service National Wetland Inventory, or interpreted as wetland basins on 1:12,000 scale color-infrared photographs acquired in May 1991 or May 1992. The remaining 73 commission errors resulted from interpretation of roadside ditches (29%), tree shadows (23%), building shadows (19%), dark areas in agricultural fields (11%), shrubs (5%), and unknown features (12%) as ponds.

Forty-eight ponds were not counted in ground BGS and video interpretation because of errors in the determining transect boundaries. Twenty-five percent (n = 12) of boundary errors were ponds counted by ground BGS but not included in video interpretation. The remaining 75% (n = 36) of boundary errors were ponds counted on video but not counted in the ground BGS.

Previous Section -- Methods: Statistical Analyses
Return to Contents
Next Section -- Results: Relationships of Pond Counts on Air-Ground Segments

Accessibility FOIA Privacy Policies and Notices

Take Pride in America logo USA.gov logo U.S. Department of the Interior | U.S. Geological Survey
URL: http://www.npwrc.usgs.gov/resource/methods/pndcount/error.htm
Page Contact Information: Webmaster
Page Last Modified: Saturday, 02-Feb-2013 06:02:39 EST
Menlo Park, CA [caww54]