Rainfall being the result of complex atmospheric phenomena possesses a complex temporal and spatial structure. The intermittent yet organized nature of spatial rainfall not only makes quantitative precipitation forecasting (QPF) a challenging task but also renders QPF verification non-trivial. Currently, most common measures of QPF verification are: threat score, equitable threat score and bias score. Simple coefficient of correlation and root-mean-square error are also used quite frequently to delineate error growth curves to assess limits of predictability of precipitation. As debated in the literature, all the aforementioned traditional scores are fallacious for QPF verification. Even small displacement errors, rotation errors or amplitude errors can label a precipitation forecast as completely "useless." In this paper, we propose a novel measure called Precipitation Forecast Index (PFI) for QPF verification, based on the classical Hausdorff distance, the Universal Image Quality Index (an image similarity measure recently introduced in the image processing literature) and the concept of "surrogate" image. Comparing with the traditional measures on several simulated test images and real precipitation datasets (observed and forecast pairs) we demonstrate the potential of this newly proposed measure. Interestingly, PFI was able to consistently identify the "best" member in a multimodel ensemble precipitation forecast (SAMEX 98). This finding has immense potential in the case of ensemble forecasting
