Accurate　identification　of　source　parameters　(source　strength　and　location)　of　sudden　air　pollution　accidents　(SAPAs)　is　important　for　implementation　of　adequate　responses.　However,　the　potential　impact　of　atmospheric　diffusion　conditions　on　source　parameter　identification　may　be　significant.　An　Inversion　model　that　combines　the　hybrid　particle　swarm　optimization　and　the　Nelder-Mead　simplex　search　method　(PSO-NM)　with　the　Gaussian　dispersion　model　was　proposed　to　identify　the　source　parameters　and　to　investigate　the　influences　of　different　atmospheric　conditions　on　the　identifications.　A　case　study　based　on　68　SO2　leakage　tests　from　the　Prairie　Grass　field　experiment　was　conducted.　The　source　strengths　and　locations　of　the　68　tests　were　estimated　by　the　combined　inversion　model.　The　results　indicated　that　the　inversion　model　can　effectively　get　accurate　and　robust　source　parameter　estimations.　The　average　absolute　value　of　relative　deviation　of　source　strength　was　13.8%　+/-　11.4%;　the　average　absolute　deviations　for　parameters　x(0),　y(0),　z(0)　and　the　total　distances　were　18.9　+/-　36.9　m,　2.7　+/-　5.2　m,　3.5　+/-　9.7　m　and　19.6　+/-　38.1　m,　respectively.　A　comprehensive　evaluation　method　was　also　proposed　for　analyzing　the　impacts　of　atmospheric　conditions　on　source　parameter　estimations.　The　results　showed　that　the　source　parameter　estimations　under　atmospheric　stability　classes　E　and　C　have　the　best　accuracy　and　robustness,　followed　by　stability　classes　A　and　D;　while　the　worst　occurred　under　atmospheric　stability　classes　B　and　F.　The　analysis　results　can　provide　scientific　support　for　the　formulation　or　adjustment　of　emergency　response　strategies　used　in　sudden　air　pollution　accidents.　The　new　inversion　model　proposed　is　a　supplement　to　the　methodology　of　inversing　source　parameters.