Identifying total phosphorus spectral signal in a tropical estuary lagoon using a hyperspectral sensor and its application to water quality modeling

The San Jose Lagoon (SJL) is located in the northern coast of Puerto Rico and is classified as a eutrophic water body. Site access difficulties establish the need for other monitoring alternatives to measure nutrients contamination. The Hyperion is a hyperspectral remote sensing imaging sensor which could provide adequate spectral resolution to monitor TP in such water system. In February, May, and August, 2006 the Hyperion collected hyperspectral groups of data from the SJL. Water quality field sampling and manual radio spectrometer data was concurrently obtained. Spectral indices were obtained from single, combined, and log combined bands statistical correlations, which were used to identify total phosphorus concentrations. A reflectance determination coefficient of 0.49 was obtained from the 467 to 529 nanometers bands ratio values, from which a polynomial algorithm was derived and used to produce a total phosphorus distribution map. In 1995 the U.S. Army Corps of Engineers (USCOE) developed the CH3D-WES and CE-QUAL-ICM hydrodynamic model for the SJL. The USCOE's model was updated with 2006 climate data. The water quality values obtained from the USCOE model compare well with the results obtained from our research. Color water quality maps were produced by the model and images with total phosphorus distribution, which provides the possible variability in its concentrations at different time periods. While it is recognized that TP has no unique reflectance spectral signal (which can be confused with reflectance from other water constituents) this study demonstrates the application of total phosphorus possible spectral indexes to monitor its content in eutrophic tropical lagoons though a hyperspectral sensor. Its use in water quality model validation provides an additional tool to identify point and non-point source pollutants which otherwise might not be detected using traditional procedures. Nonetheless, the use of the Hyperion sensor for small water systems may need to be further evaluated due its less spatial resolution which causes additional errors in the retrieved spectral data. The algorithm produced by this research may be used to improve the calibration procedure of water quality models for the detection of TP in tropical lakes and lagoons after validation with corresponding field data.