An　integrated　air　quality　simulation　and　optimization　approach　was　proposed　to　improve　the　accuracy　of　the　high-resolution　PM10　emission　inventories　developed　through　regression　models.　A　case　study　of　Tangshan　region　in　northern　China　was　presented　as　an　example.　A　linear　programming　optimization　model　was　developed　to　minimize　the　mean　error　between　simulated　and　monitored　PM10　concentrations　in　different　counties.　A　transfer　coefficient　matrix　was　used　to　represent　source　receptor　relationships　and　was　developed　through　running　a　MM5-CMAQ　air　quality　model.　The　results　revealed　that　the　proposed　simulation　optimization　approach　could　decrease　the　error　of　the　PM10　emission　inventory　from　17.0　to　7.9%　at　the　regional　level,　and　from　31.44　to　14.17%　at　the　county　level.　Accuracy　improvement　ranged　from　0.39　to　61.44%　for　the　study　counties　in　Tangshan.　The　correlation　coefficient　between　the　estimated　PM10　emissions　and　the　monitored　PM10　concentration　in　various　counties　was　also　increased　from　0.82　to　0.91.　Together　with　the　regression　models,　the　simulation　optimization　method　provides　a　promising　and　effective　framework　for　developing　high-accuracy　and　high-resolution　air　pollutant　emission　inventories.　(c)　2012　Elsevier　Ltd.　All　rights　reserved.