Total　suspended　matter　(TSM)　is　an　important　indicator　to　evaluate　water　quality,　and　is　also　one　of　the　key　parameters　for　ocean　color　remote　sensing　inversion.　The　Ocean　and　Land　Color　Instrument　(OLCI)　is　a　new　generation　of　ocean　water　color　sensor　with　well　spectral　and　spatio-temporal　resolution.　This　paper　adopted　CatBoost,　Random　Forest　and　multiple　regression　methods　to　establish　the　TSM　concentration　inversion　model　based　on　OLCI　data　and　in-situ　observations,　and　the　validation　dataset　was　used　to　evaluate　the　model　accuracy.　The　results　showed　that　the　CatBoost　model　had　the　highest　accuracy　with　RMSE　of　2.76　mg•L-1,　MAPE　of　23.67%,　and　R2　of　0.89.　Finally,　the　CatBoost　model　was　applied　to　the　time-series　OLCI　images　to　obtain　the　distribution　of　TSM　concentration　in　Fujian　coastal　waters.　The　results　indicated　that　the　spatial　and　temporal　variation　of　TSM　concentration　was　significant,　and　the　general　pattern　presents　that　the　near-shore　is　higher　than　the　far-shore,　north　region　is　higher　than　south　region,　estuaries　and　harbors　are　higher　than　other　region,　spring　is　higher　than　summer.　This　study　provides　a　new　method　for　retrieving　TSM　concentration,　and　further　proves　the　good　water　color　inversion　ability　of　OLCI　images,　which　can　provide　an　effective　remote　sensing　data　for　water　quality　monitoring　in　the　Fujian　Province.　©　2020,　Science　Press.　All　right　reserved.