Near real-time mapping of floodwater mosquito breeding sites using aerial photographs

Sammanfattning: Summary: Floodwater mosquitoes occur in temporary flooded areas and can cause tremendous nuisance in the near and far surroundings. In the Nedre Dalälven region in Central Sweden, mosquito control is based on application of the biological larvicide Bacillus thuringiensis israelensis (BTI) by helicopter. Precise mapping of the floodwater mosquito breeding sites, a prerequisite for successful BTI treatments, is complicated by high and dense vegetation, inaccessibility, and time constraints. The aim of this study is to develop a method for quick and easy delineation of flooded areas to direct mosquito control treatments. Aerial photographs taken from helicopter were used for georeferencing (including test of two flight heights, different helicopter types and camera parameters) and for flood delineation (applying visual interpretation and unsupervised classification) during spring (May) and summer (August) conditions. The study was performed in the Norrån catchment west of the town of Österfärnebo during 2007 and 2008. All photographs were taken with a hand-held Pentax 10D digital camera and a 16-45 mm lens. The hand-held digital camera gave vertical images of sufficient quality for georeferencing in a GIS. Images taken from a flight height of 600 m and with wide-angle lens could easily be orientated and georeferenced. The best helicopter type for photography was the Eurocopter AS 350 B3 (also used for mosquito control). Flood delineation by visual interpretation, using image enhancement, provided accurate information about location and extent of flooded areas in May with low and sparse vegetation. However, in August with high and dense vegetation, the vegetated flooded areas were difficult to distinguish from vegetated non-flooded areas. Also, one of the study areas had no sharp edge between flooded and dry areas but rather a transition zone. Similarly, unsupervised classification showed better results (based on accuracy assessments) for images from May than from August. Finally, the assumed flooded area to be included for mosquito control was digitized based on enhanced images and classification results. In comparison to the actual mosquito control polygons, the boundaries digitized from classifications were most similar. Areas for mosquito control treatments can include small dry parts in order to keep rather straight borderlines which makes helicopter flight movements less time-consuming. In conclusion, the best method was taking photographs in 8-bit jpg, georeferencing them for use in GIS, and to use a combination of unsupervised classification and on-screen digitizing of the boundaries of flooded areas. This method was used in an inaccessible area during an actual mosquito control treatment in 2009 and provided valuable information about slightly larger flooded areas than initially assumed. Further improvements might be achieved by using the near-infrared spectrum for water detection, and using fuzzy classification in areas with transition zones between flooded and dry sites.