Some models for estimating chlorophyll concentration are developed in this study by taking Daya Bay as an experimental site and Landsat-5 TM （Thematic Mapper） data as information resources

and combining the survey data of chlorophyll concentration in the surface coastal water. On the base of studying the spectral characteristics of chlorophyll and the mechanism of surveying chlorophyll concentration using remote sensing data

taking 75 band compositions of TM1-TM4 bands as the subfactors and the chlorophyll concentration as the mother factor

factor analysis was carried out by using the Grey System Theory. According to the results of the factor analysis

5 band compositions

which have greater correlative degree to the mother factor

are selected for modeling. The error analysis shows that these models have high accuracy with the mean absolute relative error less than 11. 2% and relative standard error less than 6. 7%. Unlike the general way taking the ratio of TM2 or TM3 to TM1 as the best band compositions

this study shows that （TM3×TM4） is the best band composition for estimating chlorophyll concentration of the surface layer of coastal water

and the estimated accuracy can not be advanced by using the ratio of （TM3×TM4） to TM1 or TM2 or In（TM1+TM2） or In（TM1×TM2）.