The　objective　of　this　study　is　to　exploit　the　new　features　of　ALOS　PALSAR　dual　polarization　mode　data　and　to　develop　novel　classification　method　for　forest　mapping　in　heterogeneous　areas.　A　test　site　was　selected　in　Fujian　province　in　southeast　of　China.　Traditionally,　forest　is　detected　by　its　low　coherence,　low　temporal　variability　of　the　backscattering　intensity　and　mediate　backscattering　intensity.　However,　the　analyses　in　this　paper　indicate　that　it　is　not　possible　to　discriminate　forest　from　nonforest　by　any　single　PALSAR　feature　in　this　test　site.　After　examination　the　dependences　of　the　multitemporal　backscatter　intensity,　the　polarimetric　parameters　and　the　interferometric　coherence　on　different　land　cover　types,　a　hierarchical　classification　method　is　proposed　for　coastal　forest　and　hilly　forest　mapping.　The　forest　maps　are　validated　by　forest　inventory　data　and　SPOPT-5　images.　The　results　show　that　multitemporal　PALSAR　dual　polarization　data　can　accurate　maps　for　coastal　forest　in　flat　areas　using　the　proposed　method.　The　capability　to　map　forest　in　hilly　regions　is　still　limited.　©　2008　SPIE.