A laboratory experiment was conducted to obtain a floating algae index (FAI) of the floating macroalgae (Ulva prolifera), corresponding to various values of biomass per unit area (BPA). A piecewise empirical model was used to fit the statistical relationships between BPA and FAI, corresponding to FAI ≤ 0.2 (BPA ≤ 1.81kg/m2) and FAI ˃ 0.2 (BPA ˃ 1.81 kg/m2). Spectral mixing derived results show that a linear relationship between FAI and BPA is maintained when the BPA of endmembers is less than 1.81 kg/m2. However, when the BPA of the endmembers exceeds 1.81 kg/m2, there is substantial uncertainty in the optical remote estimation of biomass. Although the MODIS-derived FAI of Ulva prolifera is often less than 0.2, it is very difficult to determine whether the FAI results from low BPA (≤ 1.81kg/m2) of the endmembers, or from a low area ratio including high BPA (˃ 1.81 kg/m2), due to pixel mixing. If it is assumed that the unit biomass distribution of pure endmembers is a standard Gaussian distribution, then the uncertainty in the biomass estimation of Ulva prolifera from MODIS data can be expressed. This results in the uncertainty of ~36% in total biomass estimation, ~43% of which was contributed by a few pixels (10% of total pixels) with high FAI (˃ 0.05). The uncertainty in BPA caused by high FAI (˃ 0.05) pixels is about 7.2 times that for low FAI (≤ 0.05) pixels. In future research, the spatial distribution characteristics of the FAI of pure endmembers need to be considered in order to improve the accuracy of optical remote estimation of floating Ulva prolifera.