Comparison of Creel Survey Methods on Missouri River Reservoirs

Abstract

When creel survey methods such as noncontact aerial designs are applied to large water bodies, they are often biased because of scheduling difficulties and flight reductions made to reduce cost. An aerial creel survey has been used to estimate fishing pressure on South Dakota's Missouri River reservoirs since 1975. We compared fishing pressure estimates from aerial and bus route designs and examined the effect of count frequency on variance and point estimates. A total of 71,400 (observed and unadjusted) and 94,298 angler-hours were estimated on Lake Francis Case during June 2000 by use of the aerial and bus route methods, respectively. Aerial boat and shoreline fishing pressure estimates were 65,845 angler-hours (95% confidence interval (CI) = 25,770) and 5,554 angler-hours (95% CI = ±2,436), respectively. Bus route boat and shore fishing pressure estimates were 82,671 angler-hours (95% CI = ±14,878) and 11,627 angler-hours (95% CI = ±6,618), respectively. Analyses indicated that reservoir zone, day type, and type of fishing were significant main effects (P < 0.001), which supported spatial and temporal stratification in the sampling design. Within-strata comparisons of fishing pressure estimates from the two surveys did not reveal any significant differences (0.09 < P < 0.99). Although the bus route method led to 22,898 more angler-hours of fishing pressure than the aerial method, the bus route method estimated total fishing pressure with higher precision because the day-to-day variation in fishing pressure estimates was comparatively lower than that of the aerial estimates. The aerial and bus route methods both estimated shoreline fishing pressure with lower confidence than for boat fishing pressure; this was probably the result of overall low shoreline fishing pressure and high day-to-day shoreline fishing pressure variability. Bootstrap analysis and nonlinear regression indicated breakpoints in variance estimates between six and seven counts for both survey methods. Therefore, a minimum of eight counts per month was required to produce stable variance and point estimates with an acceptable level of confidence for either survey design.