People who take a walk in urban parks, including or adjacent to a water body such as a river, a pond, or a lake, usually suffer from mosquito bites in summer and early autumn. The insects can negatively affect the health and mood of these visitors. Prior studies about the effects of landscape composition on the abundance of mosquitos have usually taken stepwise multiple linear regression protocols to look for the landscape variables that can significantly affect the abundance of mosquitos. However, those studies have largely overlooked the nonlinear effects of landscape plants on the abundance of mosquitos. In the present study, we compared the multiple linear regression (MLR) with the generalized additive model (GAM) based on the trapped mosquito abundance data obtained by using photo-catalytic CO2-baited lamps placed at the Xuanwu Lake Park, a representative subtropical urban scenic spot. We measured the coverage of trees, shrubs, forbs, proportion of hard paving, proportion of water body, and coverage of aquatic plants within a distance of 5 m from each lamp's location. We found that MLR and GAM both detected the significant influences of the coverage of terrestrial plants on the abundance of mosquitos, but GAM provided a better fit to the observations by relaxing the limitation of the linear relationship hypothesis by MLR. The coverage of trees, shrubs, and forbs accounted for 55.2% of deviance, and the coverage of shrubs had the greatest contribution rate among the three predictors, accounting for 22.6% of the deviance. The addition of the interaction between the coverage of trees and that of shrubs largely enhanced the goodness of fit, and it increased the explained deviance of the GAM from 55.2% to 65.7%. The information in this work can be valuable for the planning and design of landscape plants to reduce the abundance of mosquitos at special urban scenic points.
Keywords: Xuanwu Lake Park; coverage of plants; generalized additive models; multiple linear regression; nonlinear effects.