Understanding　the　spectral　characteristics　of　moso　bamboo　leaves　damaged　by　Pantana　phyllostachysae　Chao　can　provide　theoretical　guidance　for　developing　applicable　and　effective　technologies　to　monitor　the　ecological　safety　of　the　bamboo　forest.　Compared　with　the　traditional　multispectral　data,　hyperspectral　remote　sensing　can　sense　the　subtle　changes　of　host　spectrum　among　different　severity　of　Pantana　phyllostachysae　Chao.　However,　the　related　researches　were　still　rare,　and　the　spectral　change　mechanism　of　the　host　needs　to　be　further　summarized.　Therefore,　this　study　analyzed　the　spectral　differences　among　healthy,　damaged　and　off-yearmoso　bamboo　leaves　based　on　552　field　measured　spectrums.　The　characteristic　variables　that　can　act　as　indicators　of　leaves　health　status　were　selected.　Finally,　the　model　for　monitoring　the　damage　of　leaves　caused　by　Pantana　phyllostachysae　Chao　was　established　using　the　XGBoost　algorithm.　The　results　show　that:　(1)　with　the　increase　of　pest　damage,　the　reflectance　of　damaged　leaves　gradually　appeared　＂green　low　and　red　high＂　in　visible-band,　while　the　reflectance　noticeably　decreased　in　near-infrared　band,　and　the　reflectance　of　damaged　leaves　in　shortwave　infrared　band　was　significantly　higher　than　that　of　healthy　leaves,　especially　in　the　two　typical　water　vapor　absorption　bands　(1　450　and　1　940　nm);　(2)　the　reflectance　of　off-year　leaves　in　visible　and　near　infrared　bands　was　significantly　higher　than　healthy　and　damaged　leaves;　(3)　the　spectral　characteristics　of　indentation-only　leaves　only　slightly　changed　in　comparison　with　healthy　leaves,　while　the　red　band　reflectance　of　leaves　with　red-brown　disease　spots　increased　to　some　extent,　andthe　leaves　with　gray-white　disease　spots　completely　lost　the　basic　spectral　characteristics　of　vegetation;　(4)　according　to　the　feature　importance　score　determined　by　the　XGBoost　algorithm,　the　contribution　of　each　characteristic　variable　was　PRI＞FDVI576,　717　＞　NPCI＞DSWI＞　VOG　1＞　RVSI＞　NDWI;　(5)　the　overall　average　accuracy　of　the　model　to　detect　the　damage　by　the　pest　was　74.　39%,　and　the　accuracy　for　healthy,　mild　damaged,　severe　damaged,　off-year,　and　moderate　damaged　leaves　was　94.　55%,　74.　93%,　84.　12%,　71.　10%,　and　33.　48%　respectively.