This　paper　focuses　on　the　abnormal　nodes　detection　of　poisonous　gas　in　wireless　sensor　networks,　namely,　finding　these　nodes　whose　concentrations　are　higher　than　the　threshold.　In　order　to　detect　abnormal　nodes,　we　had　better　collect　sensory　data　from　all　nodes.　However,　this　strategy　requires　much　more　energy　consumption,　so　we　should　try　to　wake　up　these　nodes　near　the　abnormal　filed.　Based　on　this　observation,　we　propose　a　novel　energy-efficient　method　to　wake　them　up.　The　main　idea　is　to　let　abnormal　nodes　send　out　control　packets　to　activate　their　one-hop　neighbor　nodes;　then,　neighbor　nodes　continue　detecting,　and　finally,　all　abnormal　nodes　send　information　to　the　sink　node　through　the　shortest　paths.　Thereafter,　we　further　propose　to　handle　these　information　in　the　sink　node,　including　extracting　boundary　nodes,　drawing　isolines,　and　estimating　the　location　of　leakage　source.　To　extract　boundary　nodes,　we　divide　all　abnormal　nodes　into　different　intervals　in　an　ascending　or　descending　order,　and　then　find　two　nodes　with　minimum　and　maximum　in　each　interval,　so　these　nodes　are　regarded　as　boundary　nodes.　As　to　the　second　point,　we　reuse　the　wide-adopted　interpolation　methods　to　draw　isolines,　such　as　cubic,　nearest,　and　invdist.　Besides,　we　use　interpolation　to　find　the　coordinate　of　the　peak,　and　then,　it　is　deemed　to　be　the　leakage　source.　The　experimental　results　show　that　our　proposed　method　is　feasible.