Accurate　source　inversion　for　sudden　air　pollution　accidents　(SAPAs)　relies　on　forward　dispersion　models,　and　the　dispersion　parameter　scheme　is　the　key　factor　for　modeling　performance.　However,　the　impacts　of　different　schemes　on　source　inversion　have　yet　to　be　investigated.　In　this　study,　the　performance　of　source　inversion　based　on　four　empirical　power-law　dispersion　schemes-BRIGGS,　SMITH,　Pasquill-Gifford　(P-G),　and　Chinese　National　Standard　(CNS)　schemes-was　evaluated　using　the　data　set　from　the　1956　Prairie　Grass　experiment.　The　inversion　performance　of　each　dispersion　scheme　was　assessed　with　both　single　(source　strength)　and　multiple　(source　strength　and　location)　unknown　source　parameters　under　six　different　atmospheric　stability　classes　(A,　B,　C,　D,　E,　and　F)　and　the　schemes　exhibited　substantially　different　performances.　With　a　single　unknown　source　parameter,　the　four　schemes　demonstrated　excellent　robustness.　The　estimation　results　of　BRIGGS,　P-G,　and　CNS　exhibited　comparable　accuracy　to　similar　average　absolute　relative　deviations　(ARD,　33.3-37.5%)　and　better　accuracy　than　SMITH.　The　CNS　scheme　performed　the　best　(37.6%)　in　unstable　atmospheric　conditions　(A,　B,　and　C),　P-G　and　BRIGGS　were　comparable　and　performed　the　best　(26.3%　and　25.4%,　respectively)　in　neutral　atmospheric　conditions　(D),　and　the　four　schemes　exhibited　similar　performances　(28.3-31.3%)　in　stable　atmospheric　conditions　(E　and　F).　When　location　was　also　unknown,　the　source　strength　accuracy　ranking　was　P-G　(31.6%)　＞　SMITH　(43.2%)　＞　CNS　(50.6%)　＞　BRIGGS　(80.1%);　however,　the　robustness　of　the　P-G,　SMITH,　and　CNS　schemes　was　poor.　Further　analysis　revealed　that　the　differences　in　the　schemes＇　inversion　errors　were　significantly　higher　than　those　of　algorithms　from　previous　studies,　which　implies　that　scheme　selection　had　a　greater　influence　on　source　inversion　than　the　algorithm　selection.　The　results　of　this　study　can　improve　our　understanding　of　the　factors　that　influence　source　inversion.