Massive　data　transmission　between　distributed　data　centers　is　the　major　efficiency　bottleneck　of　geospatial　workflow.　Although　many　data　placement　methods　have　been　proposed　to　overcome　this　problem,　few　researches　have　considered　the　impact　of　the　structure　of　the　workflow.　In　this　paper,　we　define　the　problem　of　data　placement　for　data-intensive　geospatial　workflow　aiming　to　minimize　the　data　transfer　time.　An　algorithm　called　ant　colony　optimization　based　data　placement　of　data-intensive　geospatial　workflow　(ACO-DPDGW)　is　proposed　to　handle　this　problem.　By　taking　advantage　of　the　node　vector　to　represent　the　traditional　workflow　model,　the　ants　could　place　datasets　and　tasks　in　appropriate　data　centers　according　to　the　combination　of　pheromone　information　and　heuristic　information,　when　they　visit　the　nodes　randomly.　To　prevent　premature　convergence,　a　variable　neighborhood　search　operation　is　embedded　into　ACO-DPDGW.　The　experiments　show　that　our　algorithm　can　reduce　data　transfer　volume　and　data　transfer　time　even　as　the　numbers　of　datasets,　tasks,　and　data　centers　increase.