As　the　energy　of　nodes　in　Wireless　Sensor　Networks　(WSNs)　is　generally　constrained,　it　is　urgent　to　develop　an　efficient　data　gathering　algorithm.　Recently,　Low　Rank　approximation　is　deeply　studied　and　successfully　applied　in　WSNs　data　recovery,　in　which　a　subset　of　nodes　is　randomly　selected　for　sensing　the　environmental　data,　such　as　temperature　and　humidity.　But　this　random　sampling　solution　generally　results　in　uneven　energy　consumption　of　nodes　and　shortens　the　life　time　of　networks.　In　this　paper,　we　propose　an　efficient　data　gathering　method　based　on　the　low　rank　approximation,　in　which　the　spatial　distribution　of　the　selected　nodes　and　the　networks　energy　consumption　are　considered　simultaneously.　The　sensing　nodes　selection　is　modeled　as　a　multi-objective　0-1　programming　problem　which　is　resolved　by　the　Particle　Swarm　Optimization　(PSO)　with　high　efficiency.　In　the　optimization,　the　Gini　index　is　adopted　to　measure　the　spatial　distribution　of　the　selected　nodes　and　energy　consumption　of　the　networks.　By　the　presented　temporal　constraint　low　rank　approximation,　the　complete　sensing　data　is　recovered　from　the　sensing　data　of　the　selected　node　at　the　sink　node.　We　perform　our　method　both　on　the　real　and　simulated　wireless　networks.　The　experimental　results　demonstrate　our　model　is　practical　and　efficient.　©　2013　IEEE.