Ocean Color Remote Sensing Algorithms And Red Tides Detection In The High Frequency Red Tides Areas Of The East China Sea

This dissertation is mainly concerned about ocean color remote sensing algorithms and red tides detection in the high frequency red tides areas of the East China Sea.Based on multi-year in situ measurements, relationships between inherent optical properties and sea water constitutes concentrations have been analyzed. After getting knowledge of sea water spectral properties, we develop a model for simulating water-leaving remote sensing reflectance, with sea water constitutes concentrations as model input. Compared with in situ measurements, simulation spectra are similar in figure but lower in value. The relative mean absolute discrepancy attains 66.7% at 551nm band. However, performances of the model have been improved largely after model coefficients being optimized. The root mean square errors and the relative mean absolute discrepancies reduce much. The correlative coefficients have been improved and figures of simulation spectra are much close to in situ measurements.Based on radiative transfer simulations and Artificial Neural Network (ANN) techniques, an atmospheric correction algorithm is established for MODIS in red tides areas of the East China Sea. The ANN models are successfully applied to simulated data, and not sensitive to the random appending errors. Appling the algorithm to MODIS data, we use in situ measurements to validate retrieval performances. The results show that the retrieval is close to in situ measurements in figure. However, the discrepancy is partially large in value.Based on analysis of spectral properties on the study areas, a semi analysis algorithm has been developed, which retrieves chlorophyll-a concentrations from water-leaving radiance. Compared with in situ measurements, the relative mean absolute discrepancy is 38.34%, which shows that the algorithm acquires good results.An algorithm to retrieve sea water constitutes concentrations base on ANN has also been established. Compared with the semi-analysis algorithm, the ANN algorithm has attained better performances, and the relative mean absolute discrepancy to retrieve chlorophyll-a concentrations is 28.9%.To retrieve sea water constitutes concentrations directly from MODIS data based on ANN has been concerned. Simulation data have been used to validate performances of the algorithm. The results show that the ANN model performs well and is not sensitive to random appending errors.A multi-band red tides detection method using MODIS data has been released. The method has been successfully applied to the high frequency red tides areas of the East China Sea. The results show that the method can remove disturbing of suspended sediments effectively, and qualify possible locations as well as extract large-scale red tides message.