Estimating　accurately　airborne　pollutant　emissions　source　information　(source　strength　and　location)　is　important　for　achieving　effective　air　pollution　management　or　adequate　emergency　responses　to　accidents.　Inversion　method　is　one　of　the　useful　tools　to　identify　the　source　parameters.　The　atmospheric　dispersion　scheme　has　been　proven　to　be　the　key　to　determining　the　source　inversion　performance　by　influencing　the　accuracy　of　the　dispersion　models.　Modifying　the　atmospheric　dispersion　scheme　is　an　important　potential　method　to　improve　the　inversion　performance,　but　this　has　not　been　studied　previously.　To　fill　this　gap,　a　novel　approach　for　parameter　sensitivity　analysis　combined　with　an　optimization　method　was　proposed　to　improve　the　source　inversion　performance　by　optimizing　empirical　scheme.　The　dispersion　coefficients　sigma(y)　and　sigma(z)　of　the　typical　BRIGGS　scheme　under　different　atmospheric　dispersion　conditions　were　optimized　and　used　for　air　pollutant　dispersion　and　source　inversion.　The　results　showed　that　the　prediction　performance　of　the　air　pollutant　concentrations　was　greatly　improved　with　statistical　indices　IFBI　and　NMSE　decreased　by　0.22　and　2.07,　respectively;　FAC2　and　R　increased　by　0.10,　and　0.08,　respectively.　For　source　inversion,　the　results　of　the　significance　analysis　suggested　that　the　accuracy　in　the　source　strength　and　location　parameter　(x0)　were　both　significantly　improved　by　similar　to　271%　(relative　deviation　reduced　from　60.0%　to　16.2%)　and　similar　to　121%　(absolute　deviation　reduced　from　27.6　to　12.5　m).　The　improvement　of　source　strength　inversion　accuracy　was　more　significant　under　unstable　atmospheric　conditions　(stability　class　A,　B,　and　C);　the　mean　absolute　relative　deviation　was　reduced　by　97.5%.　These　results　can　help　to　obtain　more　accurate　source　information　and　to　provide　reliable　reference　for　air　pollution　managements　or　emergency　response　to　accidents.　This　study　provides　a　novel　and　versatile　approach　to　improve　estimation　performance　of　pollutant　emission　sources　and　enhances　our　understanding　of　source　inversion.　(C)　2021　Elsevier　Ltd.　All　rights　reserved.