•  
  •  
 

Abstract

Ocean color optic remote sensing has been used to estimate various biogeochemical constituents to monitor ocean environmental changes in coastal regions and global open oceans. However, the most problematic issue in utilizing ocean color measurements is cloud coverage. Thus, this study introduces a neural network (NN) algorithm to derive Chlorophyll a (Chl) concentrations, which are unavailable due to the presence of clouds. Because ocean color remote sensing depends on cloud coverage, microwave measurements, including sea-surface temperature, cloud, water vapor, precipitation, and winds, were used as inputs for the NN algorithm. Accordingly, the NN algorithm was designed to predict Chl using five inputs of microwave measurements and geolocation data. The correlation coefficient and root mean square error between the predicted and remotely sensed Chl concentrations were about 0.89 and 0.30 mg/m3 , respectively. Therefore, the developed NN algorithm enabled us to obtain Chl concentration during cloudy days, and even typhoon passages, as demonstrated in this study. However, Chl concentrations along coastal regions could not be predicted based on the inputs of the NN algorithm. Microwave remote sensing could not measure inhomogeneous emissivity for areas that were partially ocean and land, such as near-coastal regions

COinS