Estimating range-wide population trends of western burrowing owls (Athene cunicularia) requires standardized survey protocols that correct for detection bias in environments that support large owl populations. High concentrations of owls exist in irrigated agroecosystems within the southwest United States, yet little is known about the factors that affect detection bias during owl surveys in these systems. I used closed-population capture-recapture models to evaluate 4 factors that could affect the probability of a surveyor detecting an owl activity center (i.e., nest burrow) during visual surveys where owls are the focal object and analyzed the relationship (linear or curvilinear) between specific factors and detection probability. I recorded 1,199 detections of owls from 132 capture-recapture surveys within 12 sites of the Imperial Valley agroecosystem in California, USA between 16 April and 20 May 2006. I also conducted 96 time budget surveys throughout the day and used mixed linear models to evaluate the effect of each factor on probability of an owl activity center being available for detection (i.e., ≥1 owls above ground) during surveys. Model selection results indicated that detection probability was influenced by ambient air temperature interacting with wind speed. Detection probability followed a curvilinear relationship that resembled bell-shaped curve along a temperature gradient, with the maximum detection probability shifting as a function of wind speed. At low temperatures, detection probability declined with increased wind speed, but this relationship was reversed at high temperatures, producing a 3-dimensional pattern in detection probability characterized by a saddle-shaped hyperbolic paraboloid response surface. The probability of an activity center being available for detection declined curvilinearly with increased temperature and explained 51% of the variation in detection probability. Given the broad range of detection probabilities, correcting visual survey counts for detection bias is necessary for comparing population estimates among regions and through time. Survey designs intended to estimate abundance of owls in southwest agroecosystems should incorporate methods to estimate and correct for variation in detection probability that include measurements of ambient temperature and wind speed for use as covariates.