Detection of a reservoir water level using shape similarity metrics

Abstract

The matching between reservoirs' water edge and digital elevation model's (DEM) contour lines allowed determining the water level at the acquisition date of satellite images. A preliminary study was conducted on the Castello dam (Magazzolo Lake), between Alessandria della Rocca and Bivona (Agrigento, south-Italy). The accuracy assessment of the technique was than evaluated from the comparison between classified and reference objects using similarity metrics about the shape, theme, edge and position, through the plugin STEP of open source software GIS. Moreover, an independent GIS technique was implemented to evaluate the water level, based on a distances' array between existing contour lines and nodes extracted from vectorised classification images. Results have shown the potentiality of the techniques when applied on an ideal case; advantages and disadvantages when the images are characterized by clear sky, and limits when images are acquired during not ideal atmospheric conditions.