Effects of sampling design on the estimation of adult mosquito abundance

During 1994-5, Culex tarsalis comprised 75% of the 902,643 adult female mosquitoes collected by 63 dry-ice-baited Centers for Disease Control (CDC)-style traps operated biweekly in a uniform sampling grid that covered the southern Coachella Valley, Riverside County, California. The ln(y + 1) transformation successfully controlled the variance and normalized the distribution of catch size among trap nights. When tested by analysis of variance, abundance varied significantly among months, years, and trap sites. Although the trap by months interaction was not significant, female distribution changed seasonally as larval habitats shifted from wetlands along the Salton Sea to agriculture to managed duck marshes. Conditional simulations utilized subsets of trap sites to compare sampling designs that required no (uniform, random, and transect designs) or prior (best-estimate and stratified random designs) knowledge of mosquito spatial distribution. All designs provided similar information on population seasonal trends, but a stratified random design provided the most accurate and precise simulation. A uniform trap grid that employed every 2nd trap site subsequently was adopted by the Coachella Valley Mosquito and Vector Control District to provide information on focal changes in abundance indicative of missed or newly created larval habitats or control failures.