Notes: Hirtle & Rencz (2003)

This paper is the earliest (circa 2003) of my collection that will be used for my paper that will be handed in for April exploring the implementation of satellite imagery and GIS programs in research for analyzing Dissolved Organic Carbon (DOC).  For this reason, I will post the important aspects that I believe will be helpful for my paper. I have also included the abstract, and the proper APA citation for your viewing pleasure. All of the citations can be found in the reference section of Hirtle & Renez 2003.

_______________________________________________________

Hirtle, H., & Rencz, A. (2003). The relation between spectral reflectance and dissolved organic carbon in lake water: Kejimkujik National Park, Nova Scotia, Canada. International Journal of Remote Sensing, 24(5), 953-967. doi: 10.1080/01431160210154957

Abstract. The ability to predict dissolved organic carbon (DOC) concentrations based on spectral reflectance of lake water was examined in Kejimkujik National Park. Spectral reflectance from both ground and satellite remote sensing platforms were used to create regression models for the prediction of DOC with r² values of 0.94 and 0.72 respectively. The location of the peak wavelength of the ground spectral measurements and a cluster analysis of the satellite measurements both separated the lakes into two distinct groups with different DOC concentrations. An analysis of the potential sources of DOC identified three variables important for the prediction of DOC concentrations within the lake, flushing rate and the area of both deciduous forest and open area within the watershed (r² =0.41). As DOC concentrations are related to mercury concentrations (r²=0.86) these models could be used to assist in the identification of lakes that are sensitive to mercury pollution (Hirtle & Renez 2003). 
______________________________________________________

This paper examines the high concentration of mercury found within the lakes in the Kejimkujik Park in Nova Scotia. Specifically, the paper tries to determine the possibility of predicting DOC in a lake using remote sensing procedures. Firstly, the project aimed to predict DOC using water spectra from the ground and correlates them with the spectra given from the satellite remote sensing images.  Secondly, the project aimed to correlate potential sources of DOC to predict the concentration within the lake.

A relationship between mercury concentrations and DOC has been found, as the existence of DOC in a lake results in increased mercury in the water system over time.  The relationship between mercury and dissolved organic matter has been examined both within park lakes, as well as in other places around the world (examine study done by Clair et al. 1998). This is an important correlation to make, as examining DOC concentrations for early identification of mercury pollution could prevent harmful damage of the water columns.

For background information on DOC and what causes the dark colour in lakes, we can look towards the Hirtle & Renez paper:

"DOC is the fraction of total organic carbon in water that passes through a 0.45mm pore-diameter filter (Wetzel 1983). DOC concentrations in lakes generally range from 0.5–30ppm (Driscoll et al. 1995, Watras et al. 1995, Cai et al. 1999) but can reach 30–50ppm in wetlands and marshes (Choi et al. 1998). It is a mixture of simple substances such as sugars, fatty acids and alkanes, and of complex polymeric molecules. These have a wide range of molecular weights and are often referred to as humic acids. DOC also contains fulvic acids, which are water-soluble, natural organic substances of low molecular weight. Fulvic acids are chelating agents that can bind and hold metal ions in solution (Driscoll et al. 1995, Gergel et al. 1999). Tannins are a type of fulvic acid, which consist of any group of phenolic compounds that are produced by metabolism in trees and plants and are part of the degradation- resistant materials formed during the decomposition of vegetation. Tannins are found in any water where large quantities of vegetation have decayed and can impart a faintly yellowish to brown colour to the water.

DOC is produced both autochthonously (within the water column) and alloch- thonously (from terrestrial sources) (Hessen and Tranvik 1998). Autochthonous DOC is often colourless, composed of humic acids, and breaks down to carbon dioxide within 48 hours (Gergel et al. 1999). Allochthonous DOC has humic and fulvic acid components, which are products of the degradation of vegetative lignin and cellulose. Fulvic acids are more stable than humic acids and can remain in circulation for months to years. These sources are responsible for imparting a stained colour to the lake water (Wetzel 1983)." (Hirtle & Renez 2003).

Obviously, the differing colors due to the DOC deposits in the lake impact the optical properties of the water in comparison with clear, unstained lake water (Bukata and Jerome 1997).  In optical terms, the color impacts the scattering and absorption of light, which would in turn impact the spectral signature that is given off by the lake water. Specifically, DOC absorbs light across the UV (280–400nm) and visible (400–700nm) portions of the electromagnetic spectrum, which reduces the potential for reflectance (Witte et al. 1982). Light absorption decreases with increasing wavelength no matter the concentration of DOC (Schindler et al. 1992). The high relative reflectance in the red and yellow wavelengths leaves the water appearing stained (Wetzel 1983).

So, in the end, this paper finds that the two ground spectral measures provide two useful methods of analysis are useful for the identification of DOC concentrations in lakes. The first method (predict DOC using water spectra from the ground and correlates them with the spectra given from the satellite remote sensing images) required little analysis, and separated the lakes examined into two DOC classes including ‘low’ (with DOC concentrations of 2.8 to 4.4 ppm) and ‘high’ (with DOC from 4.6 to 9.2 ppm). In turn, separating the lakes into two classes could actually more easily identify certain lakes with a greater vulnerability for Hg accumulation, and in turn reduce the amount of field sampling. Additionally, the second method (correlate potential sources of DOC to predict the concentration within the lake using spectral data), created a regression equation with a strong r² value, which demonstrates how it is a suitable predictor for DOC concentrations in the  Kejimkujik Park area with DOC concentrations ranging from 2.9 to 13.1 ppm. The authors also propose that this data, and/or models could be useful in other regions that share similar ranges of DOC concentrations.

The Cluster Analysis groups the two lakes into classes with the similar peak wavelength locations in the spectral signature. Results were found to be the same for both raw and corrected data, which indicates that using satellite data is as simple and useful as using ground measurements. I found this especially interesting about the results from this project was that predictive ability was decreased with atmospheric corrected data. This would be interesting to explore (or read other studies that explore this), as corrected data are standard for temporal analysis in the remote sensing analysis realm. However, atmospheric correction algorithms can be costly and easily influenced by human error when using the wrong inputs. In this case, two regression equations were used, but the one that used the raw image data equation had the highest r² value (0.72) in this scenario and would therefore give the best prediction of DOC concentrations. In this case, the predictive ability of the regression line is less strong, however doesn’t rely on field work which makes it easier to work with.

The final, and least strong relationship used land cover classifications to predict DOC concentrations in the lakes in Kejimkujik Park.  In this case, the data’s regression equation had an r² equal to 0.41. The authors note that although showing the least strong relationship, this method helped to identify the terrestrial sourced DOC. In this case, three variables were found to impact the DOC concentrations the most, being the flushing rate, the area of deciduous forest, and the open area. 

________________________________________________________