Retrieval of Dissolved Inorganic Nitrogen from Multi-Temporal MODIS Data in Haizhou Bay

Abstract

It is important to map dissolved inorganic nitrogen (DIN) concentrations in order to accurately predict the outbreak of red tides. This study aims to determine the feasibility of retrieving DIN in Haizhou Bay, East China from multi-temporal MODIS satellite data and to assess the impact of DIN concentration level on its inversion accuracy. DIN was modeled from the reflectance of MODIS bands and their combinations both linearly and nonlinearly. It was found that individual MODIS bands are loosely correlated with DIN concentrations. Of the various combinations of multiple bands, bands 3 and 4 bear a highly positive correlation with DIN concentrations. In particular, (R₃ + R₄)/(R₃ − R₄) is a more accurate predictor of DIN than R₃ × R₄/(R₃ − R₄). All the regression models involving this exploratory variable have an R² value over 0.7 and a relative accuracy of around 70%. However, regression residuals are dependent of each other for DIN <70 μg/L. The removal of these observations from the regression analysis led to the establishment of the best model in the form of DIN_<70 = 7.403 (R₃ + R₄)/(R₃ − R₄) + 37.14 (R² = 0.87, n = 19). It improved the model accuracy to RMSE = 17.48 μg/L (15.8%) against the validation dataset. MODIS data allow accurate retrieval of DIN concentrations >70 μg/L in murky coastal waters. During the study period of 2004–2006, Haizhou Bay experienced a deteriorating trend in DIN in that low concentration areas decreased while areas of a high DIN concentration level expanded.

Keywords:

• DIN
• quantitative retrieval
• temporal evolution
• MODIS data
• Haizhou Bay

Acknowledgements

This research was financially supported by the National Science Foundation Project of China (No. 40606044) and the Open Fund of Jiangsu Provincial Key Laboratory of Coastal Wetland Bioresources and Environmental Protection (No. JLCBE09009). We are indebted to Dr. Jay Gao from the University of Auckland for his valuable suggestions on this research.