||In North West Africa, pelagic fisheries are an essential economic sector. However, the scientific community fails to perform satisfactory assessments of key pelagic species like sardinella, owing to a lack of relevant indices of abundance to tune the model. This paper provides an alternative acoustic index based on a semiquantitative modelisation of acoustic densities. Acoustic energy is split into binomial variables coding for null, low, medium, large, and very large densities. A multivariate geostatistical approach allows (i) mapping the spatial distribution of classes of densities and (ii) computing a new acoustic index of abundance for Sardinella aurita and Sardinella maderensis. We used the surveys of RV Fridtjof Nansen (1995-2006) and RV Al-Awam (2007-2010). Our results indicated that empirical spatial structures were highly stable over time for both between areas and surveys. Co-kriging maps also showed that sardinella had stable hot spots of distribution. The indices of abundance developed in the present study were tested in an assessment procedure and outperformed all the indices used routinely by the FAO-CECAF (Fishery Committee for the Eastern Central Atlantic) assessment working group.