1. Pairwise measures of neutral genetic differentiation are supposed to contain information about past and on-going dispersal events and are thus often used as dependent variables in correlative analyses to elucidate how neutral genetic variation is affected by landscape connectivity. However, spatial heterogeneity in the intensity of genetic drift, stemming from variations in population sizes, may inflate variance in measures of genetic differentiation and lead to erroneous or incomplete interpretations in terms of connectivity. Here, we tested the efficiency of two distance-based metrics designed to capture the unique influence of spatial heterogeneity in local drift on genetic differentiation. These metrics are easily computed from estimates of effective population sizes or from environmental proxies for local carrying capacities, and allow us to introduce the hypothesis of Spatial-Heterogeneity-in-Effective-Population-Sizes (SHNe). SHNe can be tested in a way similar to isolation-by-distance or isolation-by-resistance within the classical landscape genetics hypothesis-testing framework. 2. We used simulations under various models of population structure to investigate the reliability of these metrics to quantify the unique contribution of SHNe in explaining patterns of genetic differentiation. We then applied these metrics to an empirical genetic dataset obtained for a freshwater fish (Gobio occitaniae). 3. Simulations showed that SHNe explained up to 60% of variance in genetic differentiation (measured as Fst) in the absence of gene flow, and up to 20% when migration rates were as high as 0.10. Furthermore, one of the two metrics was particularly robust to uncertainty in the estimation of effective population sizes (or proxies for carrying capacity). In the empirical dataset, the effect of SHNe on spatial patterns of Fst was five times higher than that of isolation-by-distance, uniquely contributing to 41% of variance in pairwise Fst. Taking the influence of SHNe into account also allowed decreasing the signal-to-noise ratio, and improving the upper estimate of effective dispersal distance. 4. We conclude that the use of SHNe metrics in landscape genetics will substantially improve the understanding of evolutionary drivers of genetic variation, providing substantial information as to the actual drivers of patterns of genetic differentiation in addition to traditional measures of Euclidean distance or landscape resistance.