Roofing materials are considered as a major source of urban runoff contamination. Today, in the context of the European Water Directive (2000/60 CE), an accurate evaluation of contaminant flows from roofs is thus required at the city scale, and therefore the development of assessment tools is needed. However, at the city scale, the important diversity of roofing materials represents a difficult task for the quantification. In fact, the most restrictive step when estimating contaminant flows for large-scale territories is the estimation of the roofing materials surfaces area, for which no direct data exists. This study focuses on describing the method to evaluate gutter and valley lengths at the city scale. The proposed approach relies uses a statistical framework. The city was divided into homogenous urban areas. Then, a stratified random sampling technique was employed. The reliability of the methodology has been proved for three different urban areas (drained by sewer system) with known roofing materials distribution and known gutter and valley lengths. The obtained results show that a sample representing R=4% of the urban area leads to an absolute error less than 2% for gutter and valley length values.