Estimating the water quality condition of river and lake water in the Midwestern United States from its spectral characteristics
This study focuses on developing/calibrating remote sensing algorithms for water quality retrieval in Midwestern rivers and lakes. In the first part of this study, the spectral measurements collected using a hand-held spectrometer as well as water quality observations for the Wabash River and its tributary the Tippecanoe River in Indiana were used to develop empirical models for the retrieval of chlorophyll (chl) and total suspended solids (TSS). A method for removing sky and sun glint from field spectra for turbid inland waters was developed and tested. Empirical models were then developed using a subset of the field measurements with the rest for model validation. Spectral characteristics indicative of waters dominated by different inherent optical properties (IOPs) were identified and used as the basis of selecting bands for empirical model development. The second part of this study focuses on the calibration of an existing bio-geo-optical model for studying the spatial variability of chl, non-algal particles (NAP), and colored dissolved organic matter (CDOM) in episodic St. Joseph River plumes in southern Lake Michigan. One set of EO-1 Hyperion imagery and one set of boat-based spectrometer measurements were successfully acquired to capture episodic plume events. Coincident water quality measurements were also collected during these plume events. A database of inherent optical properties (IOPs) measurements and spectral signatures was generated and used to calibrate a bio-geo-optical model. Finally, a comprehensive spectral-biogeochemical database was developed for the Wabash River and its tributaries in Indiana by conducting field sampling of the rivers using a boat platform over different hydrologic conditions during summer 2014. In addition to the various spectral measurements taken by a handheld field spectrometer, this database includes corresponding in situ measurements of water quality parameters (chl, NAP, and CDOM), nutrients (TN, TP, dissolved organic carbon (DOC)), water-column IOPs, water depths, substrate types and bottom reflectance spectra. The temporal variability of water quality parameters and nutrients in the rivers was analyzed and studied. A look-up table (LUT) based spectrum matching methodology was applied to the collected observations in the database to simplify the retrieval of water quality parameters and make the data accessible to a wider range of end users. It was found that the ratio of the reflectance peak at the red edge (704 nm) with the local minimum caused by chlorophyll absorption at 677 nm was a strong predictor of chl concentrations (coefficient of determination ( R2) = 0.95). The reflectance peak at 704 nm was also a good predictor for TSS estimation (R2 = 0.75). In addition, we also found that reflectance within the NIR wavelengths (700–890 nm) all showed strong correlation (0.85–0.91) with TSS concentrations and generated robust models. Field measured concentrations of NAP and CDOM at 67% of the sampled sites in the St Joseph River plume fall within one standard deviation of the retrieved means using the spectrometer measurements and the calibrated bio-geo-optical model. The percentage of sites within one standard deviation (88%) is higher for the estimation of chl concentrations. Despite the dynamic nature of the observed plume and the time lag during field sampling, 77% of sampled sites were found to have field measured chl and NAP concentrations falling within one standard deviation of the Hyperion derived values. The spatial maps of water quality parameters generated from the Hyperion image provided a synoptic view of water quality conditions. Analysis highlights that concentrations of NAP, chl, and CDOM were more than three times higher in conjunction with river outflow and inside the river plumes than in ambient water. It is concluded that the storm-initiated plume is a significant source of sediments, carbon and chl to Lake Michigan. The temporal variability of water quality parameters and nutrients in the Wabash River was closely associated with hydrologic conditions, while no significant correlations existed between these parameters and streamflow for the Tippecanoe River, probably due to the two upstream reservoirs. The poor relationship between CDOM and DOC indicates that most DOC in the rivers was from human sources such as wastewater. It was also found that the source of water (surface runoff or combined sewer overflows (CSO)) to a river, water temperature, and nutrients are important factors controlling instream concentrations of phytoplankton. The LUT retrieved chl and NAP concentrations were in good agreement with field measurements with slopes close to 1.0. The average estimation errors for NAP and chl were within 4.1% and 37.7%, respectively, of independently obtained lab measurements. The CDOM levels were not well estimated and the LUT retrievals for CDOM showed large variability, probably due to the small data range collected in this study and the insensitivity of remote sensing reflectance, Rrs, to CDOM change. (Abstract shortened by ProQuest.)
Cherkauer, Purdue University.
Agricultural engineering|Water Resource Management|Remote sensing
Off-Campus Purdue Users:
To access this dissertation, please log in to our