An automatic system for rainfall signal recognition from tipping bucket gage strip charts

A system aimed at recognizing automatically the rainfall signal recorded by tipping bucket gages in paper strip charts is presented. The system consists of three modules. The first two modules automatically extract the rainfall signal from the digitized strip chart image: they are based not only on image analysis techniques, but also on the mechanical features of the recording instrument and on considerations arising from the examination of the strip charts. Thus the proposed system results much more robust and efficient than general purpose software designed for image vectorialization. The third module is a graphical user interface that permits verification of the results furnished by the previous modules and to interactively change those parts requiring correction, if any. The system, successfully tested on a large set of strip charts, represents a very powerful tool for automatic recovering of ancient rainfall data. Thus it can be employed to extract long series of high resolution rainfall signals from which several hydrological threads would certainly benefit.