BACKGROUND: Gene drives based on CRISPR-Cas9 technology are increasingly being considered as tools for reducing the capacity of mosquito populations to transmit malaria, and one of the most promising options is driving endonuclease genes that reduce the fertility of female mosquitoes. In particular, there is much interest in constructs that target the conserved mosquito doublesex (dsx) gene such that the emergence of functional drive-resistant alleles is unlikely. Proof of principle that these constructs can lead to substantial population suppression has been obtained in population cages, and they are being evaluated for use in sub-Saharan Africa. Here, we use simulation modelling to understand the factors affecting the spread of this type of gene drive over a one million-square kilometre area of West Africa containing substantial environmental and social heterogeneity. RESULTS: We found that a driving endonuclease gene targeting female fertility could lead to substantial reductions in malaria vector populations on a regional scale. The exact level of suppression is influenced by additional fitness costs of the transgene such as the somatic expression of Cas9, and its deposition in sperm or eggs leading to damage to the zygote. In the absence of these costs, or of emergent drive-resistant alleles that restore female fertility, population suppression across the study area is predicted to stabilise at ~ 95% 4 years after releases commence. Small additional fitness costs do not greatly affect levels of suppression, though if the fertility of females whose offspring transmit the construct drops by more than ~ 40%, then population suppression is much less efficient. We show the suppression potential of a drive allele with high fitness costs can be enhanced by engineering it also to express male bias in the progeny of transgenic males. Irrespective of the strength of the drive allele, the spatial model predicts somewhat less suppression than equivalent non-spatial models, in particular in highly seasonal regions where dry season stochasticity reduces drive efficiency. We explored the robustness of these results to uncertainties in mosquito ecology, in particular their method of surviving the dry season and their dispersal rates. CONCLUSIONS: The modelling presented here indicates that considerable suppression of vector populations can be achieved within a few years of using a female sterility gene drive, though the impact is likely to be heterogeneous in space and time.