Evaluating water quality monitoring with hyperspectral imagery

Water quality is an important indicator of the health of an environmental system. Traditionally, water quality analysis has involved directly sampling areas in question. This method is difficult, if not impractical, to apply to large areas where data needs to be taken frequently. Remote sensing offers the possibility of covering a large spatial area with a high temporal frequency. It also provides a spatial distribution of the constituents which direct sampling cannot economically accomplish. Spatial distributions provide deeper insight into many of the hydrologic and biological processes that are directly affected by the concentrations of water constituents. The drawback of remote sensing data, however, is that these physical quantities must be derived with little or no physical sampling of the water. Fortunately, there are many indicators of water quality, one of which is the spectral character of the light leaving the water surface. This project attempts to monitor water quality through the use of radiative transfer models that simulate water optical properties for varying levels of constituents. These simulated signals are then matched against those observed by an airborne instrument. The thrust of this project is to determine the accuracy and sensitivity of the water quality extraction algorithms to key input parameters by comparing against ground truth. Details of the techniques will be presented along with an analysis of the errors.