A GOCI-based three band model is proposed for chlorophyll-a concentration estimation based on the classical three band model. The model was built based on 289 in-situ measured chlorophyll-a concentration and hyperspectral spectrums-simulated GOCI spectrums, and was compared with MERIS-based three band model and GOCI band ratio model. At last, the model was validated using several GOCI images and an independent in-situ sampling dataset. The results showed that: (1) For the current dataset, the ratio of aph (680) and aph (660) was relatively stable. (2) The GOCI-based three band algorithm had a similar performance with MERIS-based three band algorithm in the modeling dataset. The R2 value of the GOCI-based three band model was 0. 809, which was a little lower than that of the MERIS-based three band model (R2 = 0. 820), but was obviously higher than that of GOCI band ratio model (R2 = 0. 450). (3) The performance of GOCI-based three band model in the validation dataset was similar with that in the modeling dataset, which was close to that of the MERIS-based three band model, and significantly better than that of GOCI band ratio model. (4) The GOCI image data validation indicated that GOCI band ratio model would clearly underestimate chlorophyll-a concentration in Taihu Lake. The spatial difference of chlorophyll-a concentration that yielded by the band ratio model was not clear. Compared with the widely used band ratio algorithm, the GOCI-based three band algorithm has higher stability, better accuracy, and stronger potential in application.