Detecting long-term trends in water quality parameters using remote sensing techniques

Estuarine systems have undergone extensive alteration as a result of anthropogenic activities. Detecting the magnitude of alteration and anticipating future change are crucial for managing these systems, but challenging because they require long-term records of chemical and biological water quality, which are not widely available. Moderate resolution remote sensing imagery is a rich and temporally extensive source of information about ecological systems and may be useful for detecting past and predicting future changes in estuarine ecosystems. I evaluated the use of moderate resolution Landsat-5 TM imagery for estimating three indicators of water quality: Secchi depth (SDD), chlorophyll-a concentration (Chl-a), and dissolved organic carbon (DOC). Reflectance and in situ data were collected within seven days of satellite overpass and used to build calibration models for SDD, Chl-a, and DOC in the Hudson River Estuary, New York. The accuracy of model estimates was evaluated using a validation dataset and water quality indicators were mapped for the period 2005-2008. The correlation between predicted and observed values was highest for SDD and Chl-a (r=0.62 and 0.41, resp.) and lowest for DOC (r=0.26). The root mean squared error between predicted and observed values was 20.24 cm for SDD, 0.49 ug/L for Chl-a) and 0.24 mg/L for DOC. While predictive maps indicate that turbidity decreased and chlorophyll-a concentration increased with distance downstream in 2005, there were no apparent spatial gradients for these parameters by 2008. Further analysis suggests that discrepancies between predicted and observed values were likely due to asynchronous collection of satellite and in situ data that reduce the sensitivity of models to the dynamic nature of estuarine systems. Overall, these findings suggest a strong potential for Landsat TM imagery to be used to estimate SDD and Chl-a for this area, whereas higher resolution sensor and synchronous satellite and in situ data may be needed to improve the accuracy of satellite-based DOC estimates for the Hudson River.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Natural Res & Env Sciences

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2014

Author dc:creator

• Larkin, Jinna

Contributors dc:contributor

• Fraterrigo, Jennifer M.

• Chlorophyll-a
• Dissolved organic carbon
• Landsat Thematic Mapper (TM)
• Estuaries
• Hudson River
• Secchi Disk Depth
• Akaike information criterion (AIC)

Statement dc:rights

• Copyright 2014 Jinna Larkin

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/49576

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/49576