Estimating Observer Detection Probability and Flushing Probability of Upland Nesting Ducks

Abstract

Researchers commonly use vehicle-towed nest drags to locate duck nests in upland habitats, but the efficiency of this method has rarely been assessed. During a two-year study of duck nest survival in northeastern North Dakota, we assessed several potential sources of detection failure while using ATV-towed chain or cable-chain drags for nest searching. Our standard protocol involved teams of 2 ATV operators with no dedicated observer, therefore search-team members had to divide their attention between looking backwards to monitor the drag and looking forward to navigate, so some flushes may have gone undetected because both observers were looking away. We employed independent double-observer methodology to estimate the detection probability for each observer, and considered a variety of variables that might influence detection probabilities. We found that detection probability varied substantially among observers, ranging from 0.751 to 0.949, and that detection probability peaked in the middle of the 2020 nesting season and started out higher in 2021 with a gradual increase throughout the season. Although individual observers often failed to detect flushing females, we estimated that both observers working together detected 1957 out of 1993 (98.1%, SE = 0.2%) available flushing events. Based on our findings, we see no practical need to employ a third crew member as spotter because very few duck nests were missed due to failure to observe flushing hens. Another source of detection failure includes nesting females that failed to flush from the nest drag. To measure flushing probability, we used a 2-pass removal estimator based on repeated searches conducted ~ 10 minutes apart at 45 nesting fields during 2020 and 2021. The leading crew found 761 nests and the trailing crew found 141 nests. Estimated flushing probability was 0.707 (95% CRI 0.636 - 0.770) per search event. Variation between species was apparent with blue-winged teal having the lowest flushing probability at 0.62 and gadwall having the highest at 0.85. This means that the apparent composition of the nesting population based on nests found is not the same as the actual composition, with gadwall being overrepresented and blue-winged teal being underrepresented, relative to other species. Flushing probability also varied among study sites ranging from 0.52 to 0.87. There was little evidence that flushing probability differed by year, search date, or between cable-chain vs. chain only nest drags. In order to consider additional sources of variation, future studies should collect additional data for any variable that could have the possibility of causing a hen to sit tight and not flush. Focusing on increasing the flushing probability is more beneficial to increasing the number of found nests than increasing observer detection probability.