In salmonid species, such as Atlantic salmon (Salmo salar L.), the most frequent type of data set available related to adult escapements are redd counts. When collected over a broad spatio-temporal domain, redd counts data are of great interest for tracking the variation through time of the spatial distribution of the potential spawners. This is important for management purposes when the habitat quality is variable across river sections of a catchment or when the spatial distribution can vary depending on management actions or on environmental factors. However, long-term data sets are prone to changes in data collection methodology. In this article, we present a new hierarchical Bayesian modelling approach that allows both (i) to account for a change in the data collection procedure and (ii) to analyse the variation through time of the potential spawners’ spatial distribution. The value of the proposed approach is demonstrated by its application to the Atlantic salmon redd counts data collected in Allier (France) catchment from 1977 to 2011. The Allier can be divided into three main sections according to management and habitat considerations, and an important change occurred in the redd data collection in 1997: counts by foot or by boat were replaced by counts from a helicopter. A significant effect of this change on methodology is detected: less redds counted when using the helicopter counts. However, its explicit consideration in the modelling makes little difference with regard to the estimates of potential spawner abundance and their associated uncertainty