[Muti-model collaborative retrieval of chlorophyll a in Taihu lake based on data assimilation ]

Under the efforts of many scholars, large amount of remote retrieval models of water quality parameters have been developed. However, each model could only reflect the "true value" from one level because of the natural limitation of remote sensing. To get the relatively true value by combining various retrieval models, in this work, we developed a multi-model collaborative retrieval algorithm for retrieving the concentration of Chlorophyll a based on data assimilation. We measured water quality parameters and water reflectance spectra in Taihu Lake during 2006 to 2009. There were seven retrieve models established and six models were selected to participate in the multi-model collaborative retrieval algorithm. Then these selected models were combined to establish a multi-model for retrieving the concentration of Chlorophyll a. The results indicated: (1) the accuracy of multi-model retrieval algorithm was better than that of single-model retrieval method, with an optimal MAPE of only 22. 4% ; (2) with more models participating in the multi-model collaborative retrieval algorithm, the accuracy became better, the average MAPE was decreased from 25. 6% to 23. 4% , the average RMSE was decreased from 15. 082 μg.L-1 to 14. 575 μg.L-1, and the average correlation coefficient was improved from 0.91 to 0. 92; (3) the accuracy and errors of retrieval products could be effective evaluated through calculating the confidence interval, which makes possible the acquirement of spatial and temporal error distribution of Chlorophyll a concentration retrieval in Taihu Lake.