Accurate estimation of the floating green tide biomass has great significance to make the emergency response plan and realize the highly efficient allocation of resources. Relationships among the biomass per unit area of the floating green tide, spectral characteristics and the most frequently used indices for the green tide detection are analyzed with the in-situ data of floating green tide biomass and reflectance. Based on the analysis the estimation models are established and verified. Results show that the near infrared reflectance is strongly correlated with the biomass per unit area of green tide (correlation coefficient R≈0.8); strong correlation exits between the biomass per unit area and the frequently used indices, the reflectance peak value is located at 960 nm and 1060 nm (R>0.7); estimation models of the floating green tide with exponent form based on R960/R670 and R1060/R670 (R670，R960，R1060 are the reflectance at 670,960,1060 nm, respectively) have the highest accuracy (R2≈0.9, relative error RPE≈27%). Research results can provide references for the floating green tide estimation on the sea surface by the remote sensing technology.