An Artificial Neural Network Method for Detecting Red Tides with NOAA AVHRR Imagery
基于人工神经网络的赤潮卫星遥感方法研究

The Advanced Very High Resolution Radiometer (AVHRR)on board the TIROS-N/NOAA series of operational meteorological satellites has both visible and thermal infrared channels,which enable sea surface temperatures (SSTs) and other parameters related to water quality to be derived. An artificial neural network (ANN) method for detecting red tides with AVHRR imagery has been developed in this paper. The detection of red tides is based on the fact that 1) the seawater has higher concentration of phytoplankton pigments when red tides occur and 2) the occurrence of red tides in associated with sea surface temperature. The ANN method uses reflectivity in AVHRR visible channels and SSTs derived from AVHRR thermal infrared channels as inputs with five nodes in a single hidden layer to model the nonlinear transfer function between red tides and AVHRR data. The ANN method has been trained and tested using in situ and airborne measurements. The ANN method has been applied to detect red tide events occurred in the Bohai Sea of China in 1999.The results have illustrated good performance of the ANN method with a detection accuracy of 78.5%.